Skip to Content

Rule

National Emission Standards for Hazardous Air Pollutants: Final Standards for Hazardous Air Pollutants for Hazardous Waste Combustors (Phase I Final Replacement Standards and Phase II)

Document Details

Information about this document as published in the Federal Register.

Published Document

This document has been published in the Federal Register. Use the PDF linked in the document sidebar for the official electronic format.

Start Preamble Start Printed Page 59402

AGENCY:

Environmental Protection Agency (EPA).

ACTION:

Final rule.

SUMMARY:

This action finalizes national emission standards (NESHAP) for hazardous air pollutants for hazardous waste combustors (HWCs): hazardous waste burning incinerators, cement kilns, lightweight aggregate kilns, industrial/commercial/institutional boilers and process heaters, and hydrochloric acid production furnaces. EPA has identified HWCs as major sources of hazardous air pollutant (HAP) emissions. These standards implement section 112(d) of the Clean Air Act (CAA) by requiring hazardous waste combustors to meet HAP emission standards reflecting the performance of the maximum achievable control technology (MACT).

The HAP emitted by HWCs include arsenic, beryllium, cadmium, chromium, dioxins and furans, hydrogen chloride and chlorine gas, lead, manganese, and mercury. Exposure to these substances has been demonstrated to cause adverse health effects such as irritation to the lung, skin, and mucus membranes, effects on the central nervous system, kidney damage, and cancer. The adverse health effects associated with exposure to these specific HAP are further described in the preamble. For many HAP, these findings have only been shown with concentrations higher than those typically in the ambient air.

This action also presents our decision regarding the February 28, 2002 petition for rulemaking submitted by the Cement Kiln Recycling Coalition, relating to EPA's implementation of the so-called omnibus permitting authority under section 3005(c) of the Resource Conservation and Recovery Act (RCRA). That section requires that each permit issued under RCRA contain such terms and conditions as permit writers determine to be necessary to protect human health and the environment. In that petition, the Cement Kiln Recycling Coalition requested that we repeal the existing site-specific risk assessment policy and technical guidance for hazardous waste combustors and that we promulgate the policy and guidance as rules in accordance with the Administrative Procedure Act if we continue to believe that site-specific risk assessments may be necessary.

DATES:

The final rule is effective December 12, 2005. The incorporation by reference of Method 0023A into § 63.14 is approved by the Director of the Federal Register as of December 12, 2005.

ADDRESSES:

The official public docket is the collection of materials that is available for public viewing at the Office of Air and Radiation Docket and Information Center (Air Docket) in the EPA Docket Center, Room B-102, 1301 Constitution Ave., NW., Washington, DC.

Start Further Info

FOR FURTHER INFORMATION CONTACT:

For more information concerning applicability and rule determinations, contact your State or local representative or appropriate EPA Regional Office representative. For information concerning rule development, contact Michael Galbraith, Waste Treatment Branch, Hazardous Waste Minimization and Management Division, (5302W), U.S. EPA, 1200 Pennsylvania Avenue, NW., Washington DC 20460, telephone number (703) 605-0567, fax number (703) 308-8433, electronic mail address galbraith.michael@epa.gov.

End Further Info End Preamble Start Supplemental Information

SUPPLEMENTARY INFORMATION:

Regulated Entities

The promulgation of the final rule would affect the following North American Industrial Classification System (NAICS) and Standard Industrial Classification (SIC) codes:

CategoryNAICS codeSIC codeExamples of potentially regulated entities
Any industry that combusts hazardous waste as defined in the final rule
5622114953Incinerator, hazardous waste
3273103241Cement manufacturing, clinker production
3279923295Ground or treated mineral and earth manufacturing
32528Chemical Manufacturers
32429Petroleum Refiners
33133Primary Aluminum
33338Photographic equipment and supplies
488, 561, 56249Sanitary Services, N.E.C.
42150Scrap and waste materials
42251Chemical and Allied Products, N.E.C
512, 541, 561, 81273Business Services, N.E.C.
512, 514, 541, 71189Services, N.E.C.
92495Air, Water and Solid Waste Management

This table is not intended to be exhaustive, but rather provides a guide for readers regarding entities likely to be regulated by this action. This table lists examples of the types of entities EPA is now aware could potentially be regulated by this action. Other types of entities not listed could also be affected. To determine whether your facility, company, business, organization, etc., is regulated by this action, you should examine the applicability criteria in Part II of this preamble. If you have any questions regarding the applicability of this action to a particular entity, consult the person listed in the preceding FOR FURTHER INFORMATION CONTACT section.

Abbreviations and Acronyms Used in This Document

acfm  actual cubic feet per minute

Btu  British thermal units

CAA Clean Air Act

CFR Code of Federal Regulations

DRE destruction and removal efficiency

dscf dry standard cubic foot

dscm dry standard cubic meter Start Printed Page 59403

EPA Environmental Protection Agency

FR Federal Register

gr/dscf grains per dry standard cubic foot

HAP hazardous air pollutant(s)

ICR Information Collection Request

kg/hr kilograms per hour

kW-hour kilo Watt hour

MACT Maximum Achievable Control Technology

mg/dscm milligrams per dry standard cubic meter

MMBtu million British thermal unit

ng/dscm nanograms per dry standard cubic meter

NESHAP national emission standards for HAP

ng nanograms

POHC principal organic hazardous constituent

ppmv parts per million by volume

ppmw parts per million by weight

Pub. L. Public Law

RCRA Resource Conservation and Recovery Act

SRE system removal efficiency

TEQ toxicity equivalence

μg/dscm micrograms per dry standard cubic meter

U.S.C. United States Code

Table of Contents

Part One: Background and Summary

I. What Is the Statutory Authority for this Standard?

II. What Is the Regulatory Development Background of the Source Categories in the Final Rule?

A. Phase I Source Categories

B. Phase II Source Categories

III. How Was the Final Rule Developed?

IV. What Is the Relationship Between the Final Rule and Other MACT Combustion Rules?

V. What Are the Health Effects Associated with Pollutants Emitted by Hazardous Waste Combustors?

Part Two: Summary of the Final Rule

I. What Source Categories and Subcategories Are Affected by the Final Rule?

II. What Are the Affected Sources and Emission Points?

III. What Pollutants Are Emitted and Controlled?

IV. Does the Final Rule Apply to Me?

V. What Are the Emission Limitations?

VI. What Are the Testing and Initial Compliance Requirements?

A. Compliance Dates

B. Testing Requirements

C. Initial Compliance Requirements

VII. What Are the Continuous Compliance Requirements?

VIII. What Are the Notification, Recordkeeping, and Reporting Requirements?

IX. What Is the Health-Based Compliance Alternative for Total Chlorine, and How Do I Demonstrate Eligibility?

A. Overview

B. HCl-Equivalent Emission Rates

C. Eligibility Demonstration

D. Assurance that the 1-Hour HCl-Equivalent Emission Rate Will Not Be Exceeded

E. Review and Approval of Eligibility Demonstrations

F. Testing Requirements

G. Monitoring Requirements

H. Relationship Among Emission Rates, Emission Rate Limits, and Feedrate Limits

I. Changes

X. Overview on Floor Methodologies

Part Three: What Are the Major Changes Since Proposal?

I. Database

A. Hazardous Burning Incinerators

B. Hazardous Waste Cement Kilns

C. Hazardous Waste Lightweight Aggregate Kilns

D. Liquid Fuel Boilers

E. HCl Production Furnaces

F. Total Chlorine Emissions Data Below 20 ppmv

II. Emission Limits

A. Incinerators

B. Hazardous Waste Burning Cement Kilns

C. Hazardous Waste Burning Lightweight Aggregate Kilns

D. Solid Fuel Boilers

E. Liquid Fuel Boilers

F. Hydrochloric Acid Production Furnaces

G. Dioxin/Furan Testing for Sources Not Subject to a Numerical Standard

III. Statistics and Variability

A. Using Statistical Imputation to Address Variability of Nondetect Values

B. Degrees of Freedom when Imputing a Standard Deviation Using the Universal Variability Factor for Particulate Matter Controlled by a Fabric Filter

IV. Compliance Assurance for Fabric Filters, Electrostatic Precipitators, and Ionizing Wet Scrubbers

V. Health-Based Compliance Alternative for Total Chlorine

Part Four: What Are the Responses to Major Comments?

I. Database

A. Revisions to the EPA's Hazardous Waste Combustor Data Base

B. Use of Data from Recently Upgraded Sources

C. Correction of Total Chlorine Data to Address Potential Bias in Stack Measurement Method

D. Mercury Data for Cement Kilns

E. Mercury Data for Lightweight Aggregate Kilns

F. Incinerator Database

II. Affected Sources

A. Area Source Boilers and Hydrochloric Acid Production Furnaces

B. Boilers Eligible for the RCRA Low Risk Waste Exemption

C. Mobile Incinerators

III. Floor Approaches

A. Variability

B. SRE/Feed Methdology

C. Air Pollution Control Technology Methodologies for the Particulate Matter Standard and for the Total Chlorine Standard for Hydrochloric Acid Production Furnaces

D. Format of Standards

E. Standards Can Be No Less Stringent Than the Interim Standards

F. How Can EPA's Approach to Assessing Variability and its Ranking Methodologies be Reasonable when they Result in Standards Higher than the Interim Standards?

IV. Use of Surrogates

A. Particulate Matter as Surrogate for Metal HAP

B. Carbon Monoxide/Hydrocarbons and DRE as Surrogates for Dioxin/Furan

C. Use of Carbon Monoxide and Total Hydrocarbons as Surrogate for Non-Dioxin Organic HAP

V. Additional Issues Relating to Variability and Statistics

A. Data Sets Containing Nondetects

B. Using Statistical Imputation to Address Variability of Nondetect Values

C. Analysis of Variance Procedures to Assess Subcategorization

VI. Emission Standards

A. Incinerators

B. Cement Kilns

C. Lightweight Aggregate Kilns

D. Liquid Fuel Boilers

E. General

VII. Health-Based Compliance Alternative for Total Chlorine

A. Authority for Health-Based Compliance Alternatives

B. Implementation of the Health-Based Standards

C. National Health-Based Standards for Cement Kilns.

VIII. Implementation and Compliance

A. Compliance Assurance Issues for both Fabric Filters and Electrostatic Precipitators (and Ionizing Wet Scrubbers)

B. Compliance Assurance Issues for Fabric Filters

C. Compliance Issues for Electrostatic Precipitators and Ionizing Wet Scrubbers

D. Fugitive Emissions

E. Notification of Intent to Comply and Compliance Progress Report

F. Startup, Shutdown, and Malfunction Plan

G. Public Notice of Test Plans

H. Using Method 23 Instead of Method 0023A

I. Extrapolating Feedrate Limits for Compliance with the Liquid Fuel Boiler Mercury and Semivolatile Metal Standards

J. Temporary Compliance with Alternative, Otherwise Applicable MACT Standards

K. Periodic DRE Testing and Limits on Minimum Combustion Chamber Temperature for Cement Kilns

L. One Time Dioxin and Furan Test for Sources Not Subject to a Numerical Limit for Dioxin and Furan

M. Miscellaneous Compliance Issues

IX. Site-Specific Risk Assessment under RCRA

A. What Is the Site-Specific Risk Assessment Policy?

B. Why Might SSRAs Continue To Be Necessary for Sources Complying With Phase 1 Replacement Standards and Phase 2 Standards?

C. What Changes Are EPA Finalizing With Respect To the Site-Specific Risk Assessment Policy?

D. How Will the New SSRA Regulatory Provisions Work?

E. What Were Commenters' Reactions to EPA's Proposed Decision Not to Provide National Criteria for Determining When an SSRA Is or Is Not Necessary?Start Printed Page 59404

F. What Are EPA's Responses to the Cement Kiln Recycling Coalition's Comments on the Proposal and What is EPA's Final Decision on CKRC's Petition?

X. Permitting

A. What is the Statutory Authority for the RCRA Requirements Discussed in this Section?

B. Did Commenters Express any Concerns Regarding the Current Permitting Requirements?

C. Are There Any Changes to the Proposed Class 1 Permit Modification Procedure?

D. What Permitting Approach Is EPA Finalizing for New Units?

E. What Other Permitting Requirements Were Discussed In the Proposal?

Part Five: What Are the CAA Delegation Clarifications and RCRA State Authorization Requirements?

I. Authority for this Rule.

II. CAA Delegation Authority.

III. Clarifications to CAA Delegation Provisions for Subpart EEE.

A. Alternatives to Requirements.

B. Alternatives to Test Methods.

C. Alternatives to Monitoring.

D. Alternatives to Recordkeeping and Reporting.

E. Other Delegation Provisions

IV. RCRA State Authorization and Amendments To the RCRA Regulations.

Part Six: Impacts of the Final Rule

I. What Are the Air Impacts?

II. What Are the Water and Solid Waste Impacts?

III. What Are the Energy Impacts?

IV. What Are the Control Costs?

V. What Are the Economic Impacts?

A. Market Exit Estimates

B. Waste Reallocations

VI. What Are the Social Costs and Benefits of the Final Rule?

A. Combustion Market Overview

B. Baseline Specification

C. Analytical Methodology and Findings—Social Cost Analysis

D. Analytical Methodology and Findings—Benefits Assessment

Part Seven: How Does the Final Rule Meet the RCRA Protectiveness Mandate?

I. Background

II. Evaluation of Protectiveness

Part Eight: Statutory and Executive Order Reviews

I. Executive Order 12866: Regulatory Planning and Review

II. Paperwork Reduction Act

III. Regulatory Flexibility Act

IV. Unfunded Mandates Reform Act of 1995

V. Executive Order 13132: Federalism

VI. Executive Order 13175: Consultation and Coordination with Indian Tribal Governments

VII. Executive Order 13045: Protection of Children from Environmental Health Risks and Safety Risks

VIII. Executive Order 13211: Actions Concerning Regulations that Significantly Affect Energy Supply, Distribution, or Use

IX. National Technology Transfer and Advancement Act

X. Executive Order 12898: Federal Actions to Address Environmental Justice in Minority Populations and Low-Income Populations

XI. Congressional Review

Part One: Background and Summary

I. What Is the Statutory Authority for This Standard?

Section 112 of the Clean Air Act requires that the EPA promulgate regulations requiring the control of HAP emissions from major and certain area sources. The control of HAP is achieved through promulgation of emission standards under sections 112(d) and (in a second round of standard setting) (f).

EPA's initial list of categories of major and area sources of HAP selected for regulation in accordance with section 112(c) of the Act was published in the Federal Register on July 16, 1992 (57 FR 31576). Hazardous waste incinerators, Portland cement plants, clay products manufacturing (including lightweight aggregate kilns), industrial/commercial/institutional boilers and process heaters, and hydrochloric acid production furnaces are among the listed 174 categories of sources. The listing was based on the Administrator's determination that these sources may reasonably be anticipated to emit one or more of the 186 listed HAP in quantities sufficient to designate them as major sources.

II. What Is the Regulatory Development Background of the Source Categories in the Final Rule?

Today's notice finalizes standards for controlling emissions of HAP from hazardous waste combustors: incinerators, cement kilns, lightweight aggregate kilns, boilers, process heaters [1] , and hydrochloric acid production furnaces that burn hazardous waste. We call incinerators, cement kilns, and lightweight aggregate kilns Phase I sources because we have already promulgated standards for those source categories. We call boilers and hydrochloric acid production furnaces Phase II sources because we intended to promulgate MACT standards for those source categories after promulgating MACT standards for Phase I sources. The regulatory background of Phase I and Phase II source categories is discussed below.

A. Phase I Source Categories

Phase I combustor sources are regulated under the Resource Conservation and Recovery Act (RCRA), which establishes a “cradle-to-grave” regulatory structure overseeing the safe treatment, storage, and disposal of hazardous waste. We issued RCRA rules to control air emissions from hazardous waste burning incinerators in 1981, 40 CFR Parts 264 and 265, Subpart O, and from cement kilns and lightweight aggregate kilns that burn hazardous waste in 1991, 40 CFR Part 266, Subpart H. These rules rely generally on risk-based standards to assure control necessary to protect human health and the environment, the applicable RCRA standard. See RCRA section 3004 (a) and (q).

The Phase I source categories also are subject to standards under the Clean Air Act. We promulgated standards for Phase I sources on September 30, 1999 (64 FR 52828). This final rule is referred to in this preamble as the Phase I rule or 1999 final rule. These emission standards created a technology-based national cap for hazardous air pollutant emissions from the combustion of hazardous waste in these devices. The rule regulates emissions of numerous hazardous air pollutants: dioxin/furans, other toxic organics (through surrogates), mercury, other toxic metals (both directly and through a surrogate), and hydrogen chloride and chlorine gas. Where necessary, Section 3005(c)(3) of RCRA provides the authority to impose additional conditions on a source-by-source basis in a RCRA permit if necessary to protect human health and the environment.

A number of parties, representing interests of both industrial sources and of the environmental community, sought judicial review of the Phase I rule. On July 24, 2001, the United States Court of Appeals for the District of Columbia Circuit granted portions of the Sierra Club's petition for review and vacated the challenged portions of the standards. Cement Kiln Recycling Coalition v. EPA, 255 F. 3d 855 (D.C. Cir. 2001). The court held that EPA had not demonstrated that its calculation of MACT floors met the statutory requirement of being no less stringent than (1) the average emission limitation achieved by the best performing 12 percent of existing sources and, for new sources, (2) the emission control achieved in practice by the best controlled similar source for new sources. 255 F.3d at 861, 865-66. As a remedy, the court, after declining to rule on most of the issues presented in the industry petitions for review, vacated the “challenged regulations,” stating that: “[W]e have chosen not to reach the bulk of industry petitioners’ claims, and leaving the regulations in place during remand would ignore petitioners' potentially meritorious challenges.” Id. Start Printed Page 59405at 872. Examples of the specific challenges the Court indicated might have merit were provisions relating to compliance during start up/shut down and malfunction events, including emergency safety vent openings, the dioxin/furan standard for lightweight aggregate kilns, and the semivolatile metal standard for cement kilns. Id. However, the Court stated, “[b]ecause this decision leaves EPA without standards regulating [hazardous waste combustor] emissions, EPA (or any of the parties to this proceeding) may file a motion to delay issuance of the mandate to request either that the current standards remain in place or that EPA be allowed reasonable time to develop interim standards.” Id.

Acting on this invitation, all parties moved the Court jointly to stay the issuance of its mandate for four months to allow EPA time to develop interim standards, which would replace the vacated standards temporarily, until final standards consistent with the Court's mandate are promulgated. The interim standards were published on February 13, 2002 (67 FR 6792). EPA did not justify or characterize these standards as conforming to MACT, but rather as an interim measure to prevent adverse consequences that would result from the regulatory gap resulting from no standards being in place. Id. at 6793, 6795-96; see also 69 FR at 21217 (April 20, 2004). EPA also entered into a settlement agreement, enforceable by the Court of Appeals, to issue final standard conforming to the Court's mandate by June 14, 2005. That date has since been extended to September 14, 2005.

B. Phase II Source Categories

Phase II combustors—boilers and hydrochloric acid production furnaces—are also regulated under the Resource Conservation and Recovery Act (RCRA) pursuant to 40 CFR Part 266, Subpart H, and (for reasons discussed below) are also subject to the MACT standard setting process in section 112(d) of the CAA. We delayed promulgating MACT standards for these source categories pending reevaluation of the MACT standard-setting methodology following the Court's decision to vacate the standards for the Phase I source categories. We also have entered into a judicially enforceable consent decree with Sierra Club that requires EPA to promulgate MACT standards for the Phase II sources by June 14, 2005, since extended to September 14, 2005—the same date that (for independent reasons) is required for the replacement standards for Phase I sources.

III. How Was the Final Rule Developed?

We proposed standards for HWCs on April 20, 2004 (69 FR 21197). The public comment period closed on July 6, 2004. In addition, on February 4, 2005, we requested certain key commenters to comment by email on a limited number of issues arising from public comments on the proposed rule. The comment period for those issues closed on March 7, 2005.

We received approximately 100 public comment letters on the proposed rule and the subsequent direct request for comments. Comments were submitted by owner/operators of HWCs, trade associations, state regulatory agencies and their representatives, and environmental groups. Today's final rule reflects our consideration of all of the comments and additional information we received. Major public comments on the proposed rule along with our responses, are summarized in this preamble.

IV. What Is the Relationship Between the Final Rule and Other MACT Combustion Rules?

The amendments to the Subpart EEE, Part 63, standards for hazardous waste combustors apply to the source categories that are currently subject to that subpart—incinerators, cement kilns, and lightweight aggregate kilns that burn hazardous waste. Today's final rule, however, also amends Subpart EEE to establish MACT standards for the Phase II source categories—those boilers and hydrochloric acid production furnaces that burn hazardous waste.

Generally speaking, you are an affected source pursuant to Subpart EEE if you combust, or have previously combusted, hazardous waste in an incinerator, cement kiln, lightweight aggregate kiln, boiler, or hydrochloric acid production furnace. You continue to be an affected source until you cease burning hazardous waste and initiate closure requirements pursuant to RCRA. Affected sources do not include: (1) Sources exempt from regulation under 40 CFR part 266, subpart H, because the only hazardous waste they burn is listed under 40 CFR 266.100(c); (2) research, development, and demonstration sources exempt under § 63.1200(b); and (3) boilers exempt from regulation under 40 CFR part 266, subpart H, because they meet the definition of small quantity burner under 40 CFR 266.108. See § 63.1200(b).

If you never previously combusted hazardous waste, or have ceased burning hazardous waste and initiated RCRA closure requirements, you are not subject to Subpart EEE. Rather, EPA has promulgated separate MACT standards for sources that do not burn hazardous waste within the following source categories: commercial and industrial solid waste incinerators (40 CFR Part 60, Subparts CCCC and DDDD); Portland cement manufacturing facilities (40 CFR Part 63, Subpart LLL); industrial/commercial/institutional boilers and process heaters (40 CFR Part 63, Subpart DDDDD); and hydrochloric acid production facilities (40 CFR Part 63, Subpart NNNNN). In addition, EPA considered whether to establish MACT standards for lightweight aggregate manufacturing facilities that do not burn hazardous waste, and determined that they are not major sources of HAP emissions. Thus, EPA has not established MACT standards for lightweight aggregate manufacturing facilities that do not burn hazardous waste.

Note that non-stack emissions points are not regulated under Subpart EEE.[2] Emissions attributable to storage and handling of hazardous waste prior to combustion (i.e., emissions from tanks, containers, equipment, and process vents) would continue to be regulated pursuant to either RCRA Subpart AA, BB, and CC and/or an applicable MACT that applies to the before-mentioned material handling devices. Emissions unrelated to the hazardous waste operations may be regulated pursuant to other MACT rulemakings. For example, Portland cement manufacturing facilities that combust hazardous waste are subject to both Subpart EEE and Subpart LLL, and hydrochloric acid production facilities that combust hazardous waste may be subject to both Subpart EEE and Subpart NNNNN.[3] In these instances Subpart EEE controls HAP emissions from the cement kiln and hydrochloric acid production furnace stack, while Subparts LLL and NNNNN would control HAP emissions from other operations that are not directly related to the combustion of hazardous waste (e.g., clinker cooler emissions for cement production facilities, and hydrochloric acid product transportation and storage for hydrochloric acid production facilities).

Note that if you temporarily cease burning hazardous waste for any reason, you remain an affected source and are still subject to the applicable Subpart Start Printed Page 59406EEE requirements. However, even as an affected source, the emission standards or operating limits do not apply if: (1) Hazardous waste is not in the combustion chamber and you elect to comply with other MACT (or CAA section 129) standards that otherwise would be applicable if you were not burning hazardous waste, e.g., the nonhazardous waste burning Portland Cement Kiln MACT (Subpart LLL); or (2) you are in a startup, shutdown, or malfunction mode of operation.

V. What Are the Health Effects Associated With Pollutants Emitted by Hazardous Waste Combustors?

Today's final rule protects air quality and promotes the public health by reducing the emissions of some of the HAP listed in Section 112(b)(1) of the CAA. Emissions data collected in the development of this final rule show that metals, hydrogen chloride and chlorine gas, dioxins and furans, and other organic compounds are emitted from hazardous waste combustors. The HAP that would be controlled with this rule are associated with a variety of adverse health affects. These adverse health effects include chronic health disorders (e.g., irritation of the lung, skin, and mucus membranes and effects on the blood, digestive tract, kidneys, and central nervous system), and acute health disorders (e.g., lung irritation and congestion, alimentary effects such as nausea and vomiting, and effects on the central nervous system). Provided below are brief descriptions of risks associated with HAP that are emitted from hazardous waste combustors.

Antimony

Antimony occurs at very low levels in the environment, both in the soils and foods. Higher concentrations, however, are found at antimony processing sites, and in their hazardous wastes. The most common industrial use of antimony is as a fire retardant in the form of antimony trioxide. Chronic occupational exposure to antimony (generally antimony trioxide) is most commonly associated with “antimony pneumoconiosis,” a condition involving fibrosis and scarring of the lung tissues. Studies have shown that antimony accumulates in the lung and is retained for long periods of time. Effects are not limited to the lungs, however, and myocardial effects (effects on the heart muscle) and related effects (e.g., increased blood pressure, altered EKG readings) are among the best-characterized human health effects associated with antimony exposure. Reproductive effects (increased incidence of spontaneous abortions and higher rates of premature deliveries) have been observed in female workers exposed in an antimony processing facilities. Similar effects on the heart, lungs, and reproductive system have been observed in laboratory animals.

EPA assessed the carcinogenicity of antimony and found the evidence for carcinogenicity to be weak, with conflicting evidence from inhalation studies with laboratory animals, equivocal data from the occupational studies, negative results from studies of oral exposures in laboratory animals, and little evidence of mutagenicity or genotoxicity.[4] As a consequence, EPA concluded that insufficient data are available to adequately characterize the carcinogenicity of antimony and, accordingly, the carcinogenicity of antimony cannot be determined based on available information. However, the International Agency for Research on Cancer in an earlier evaluation, concluded that antimony trioxide is “possibly carcinogenic to humans” (Group 2B).

Arsenic

Chronic (long-term) inhalation exposure to inorganic arsenic in humans is associated with irritation of the skin and mucous membranes. Human data suggest a relationship between inhalation exposure of women working at or living near metal smelters and an increased risk of reproductive effects, such as spontaneous abortions. Inorganic arsenic exposure in humans by the inhalation route has been shown to be strongly associated with lung cancer, while ingestion or inorganic arsenic in humans has been linked to a form of skin cancer and also to bladder, liver, and lung cancer. EPA has classified inorganic arsenic as a Group A, human carcinogen.

Beryllium

Chronic inhalation exposure of humans to high levels of beryllium has been reported to cause chronic beryllium disease (berylliosis), in which granulomatous (noncancerous) lesions develop in the lung. Inhalation exposure to high levels of beryllium has been demonstrated to cause lung cancer in rats and monkeys. Human studies are limited, but suggest a causal relationship between beryllium exposure and an increased risk of lung cancer. We have classified beryllium as a Group B1, probable human carcinogen, when inhaled; data are inadequate to determine whether beryllium is carcinogenic when ingested.

Cadmium

Chronic inhalation or oral exposure to cadmium leads to a build-up of cadmium in the kidneys that can cause kidney disease. Cadmium has been shown to be a developmental toxicant in animals, resulting in fetal malformations and other effects, but no conclusive evidence exists in humans. An association between cadmium exposure and an increased risk of lung cancer has been reported from human studies, but these studies are inconclusive due to confounding factors. Animal studies have demonstrated an increase in lung cancer from long-term inhalation exposure to cadmium. EPA has classified cadmium as a Group B1, probable carcinogen.

Chlorine gas

Chlorine is an irritant to the eyes, the upper respiratory tract, and lungs. Chronic exposure to chlorine gas in workers has resulted in respiratory effects including eye and throat irritation and airflow obstruction. No information is available on the carcinogenic effects of chlorine in humans from inhalation exposure. A National Toxicology Program (NTP) study showed no evidence of carcinogenic activity in male rats or male and female mice, and equivocal evidence in female rats, from ingestion of chlorinated water. The EPA has not classified chlorine for potential carcinogenicity. In the absence of specific scientific evidence to the contrary, it is the Agency's policy to classify noncarcinogenic effects as threshold effects. RfC development is the default approach for threshold (or nonlinear) effects.

Chromium

Chromium may be emitted in two forms, trivalent chromium (chromium III) or hexavalent chromium (chromium VI). The respiratory tract is the major target organ for chromium VI toxicity for inhalation exposures. Bronchitis, decreases pulmonary function, pneumonia, and other respiratory effects have been noted from chronic high does exposure in occupational settings due to chromium VI. Limited human studies suggest that chromium VI inhalation exposure may be associated with complications during pregnancy and childbirth, while animal studies have not reported reproductive effects from inhalation exposure to chromium VI. Human and animal studies have clearly established that inhaled chromium VI is Start Printed Page 59407a carcinogen, resulting in an increased risk of lung cancer. EPA has classified chromium VI as a Group A, human carcinogen.

Chromium III is less toxic than chromium VI. The respiratory tract is also the major target organ for chromium III toxicity, similar to chromium VI. Chromium III is an essential element in humans, with a daily intake of 50 to 200 micrograms per day recommended for an adult. The body can detoxify some amount of chromium VI to chromium III. EPA has not classified chromium III with respect to carcinogenicity.

Cobalt

Cobalt is a relatively rare metal that is produced primarily as a by-product during refining of other metals, especially copper. Cobalt has been widely reported to cause respiratory effects in humans exposed by inhalation, including respiratory irritation, wheezing, asthma, and pneumonia. Cardiomyopathy (damage to the heart muscle) has also been reported, although this effect is better known from oral exposure. Other effects of oral exposure in humans are polycythemia (an abnormally high number of red blood cells) and the blocking of uptake of iodine by the thyroid. In addition, cobalt is a sensitizer in humans by any route of exposure. Sensitized individuals may react to inhalation of cobalt by developing asthma or to ingestion or dermal contact with cobalt by developing dermatitis. Cobalt is as a vital component of vitamin B12, though there is no evidence that intake of cobalt is ever limiting in the human diet.

A number of epidemiological studies have found that exposures to cobalt are associated with an increased incidence of lung cancer in occupational settings. The International Agency for Research on Cancer (part of the World Health Organization) classifies cobalt and cobalt compounds as “possibly carcinogenic to humans” (Group 2B). The American Conference of Governmental Industrial Hygienists has classified cobalt as a confirmed animal carcinogen with unknown relevance to humans (category A3). An EPA assessment concludes that under EPA's cancer guidelines, cobalt would be considered likely to be carcinogenic to humans.[5]

Dioxins and Furans

Exposures to 2,3,7,8-tetrachlorodibenzo-p-dioxin (2,3,7,8-TCDD) and related compounds at levels 10 times or less above those modeled to approximate average background exposure have resulted in adverse non-cancer health effects in animals. This statement is based on assumptions about the toxic equivalent for these compounds, for which there is acknowledged uncertainty. These effects include changes in hormone systems, alterations in fetal development, reduced reproductive capacity, and immunosuppression. Effects that may be linked to dioxin and furan exposures at low dose in humans include changes in markers of early development and hormone levels. Dioxin and furan exposures are associated with altered liver function and lipid metabolism changes in activity of various liver enzymes, depression of the immune system, and endocrine and nervous system effects. EPA in its 1985 dioxin assessment classified 2,3,7,8-TCDD as a probable human carcinogen. The International Agency for Research on Cancer (IARC) concluded in 1997 that the overall weight of the evidence was sufficient to characterize 2,3,7,8-TCDD as a known human carcinogen.[6] In 2001 the U.S. Department of Health and Human Services National Toxicology Program in their 9th Report on Carcinogens classified 2,3,7,8-TCDD as a known human carcinogen.[7]

The chemical and environmental stability of dioxins and their tendency to accumulate in fat have resulted in their detection within many ecosystems. In the United States and elsewhere, accidental contamination of the environment by 2,3,7,8-TCDD has resulted in deaths in many species of wildlife and domestic animals.[8] High residues of this compound in fish have resulted in closing rivers to fishing. Laboratory studies with birds, mammals, aquatic organisms, and other species have demonstrated that exposure to 2,3,7,8-TCDD can result in acute and delayed mortality as well as carcinogenic, teratogenic, mutagenic, histopathologic, immunotoxic, and reproductive effects, depending on dose received, which varied widely in the experiments.[9]

Hydrogen chloride/hydrochloric acid

Hydrogen chloride, also called hydrochloric acid, is corrosive to the eyes, skin, and mucous membranes. Chronic (long-term) occupational exposure to hydrochloric acid has been reported to cause gastritis, bronchitis, and dermatitis in workers. Prolonged exposure to low concentrations may also cause dental discoloration and erosion. No information is available on the reproductive or developmental effects of hydrochloric acid in humans. In rats exposed to hydrochloric acid by inhalation, altered estrus cycles have been reported in females and increased fetal mortality and decreased fetal weight have been reported in offspring. EPA has not classified hydrochloric acid for carcinogenicity. In the absence of specific scientific evidence to the contrary, it is the Agency's policy to classify noncarcinogenic effects as threshold effects. RfC development is the default approach for threshold (or nonlinear) effects.

Lead

Lead can cause a variety of effects at low dose levels. Chronic exposure to high levels of lead in humans results in effects on the blood, central nervous system, blood pressure, and kidneys. Children are particularly sensitive to the chronic effects of lead, with slowed cognitive development, reduced growth and other effects reported. Reproductive effects, such as decreased sperm count in men and spontaneous abortions in women, have been associated with lead exposure. The developing fetus is at particular risk from maternal lead exposure, with low birth weight and slowed postnatal neurobehavioral development noted. Human studies are inconclusive regarding lead exposure and cancer, while animal studies have reported an increase in kidney cancer from lead exposure by the oral route. EPA has classified lead as a Group B2, probable human carcinogen.

Manganese

Health effects in humans have been associated with both deficiencies and excess intakes of manganese. Chronic exposure to low levels of manganese in the diet is considered to be nutritionally essential in humans, with a recommended daily allowance of 2 to 5 milligrams per day (mg/d). Chronic Start Printed Page 59408exposure to high levels of manganese by inhalation in humans results primarily in central nervous system effects. Visual reaction time, hand steadiness, and eye-hand coordination were affected in chronically-exposed workers. Impotence and loss of libido have been noted in male workers afflicted with manganism attributed to inhalation exposures. EPA has classified manganese in Group D, not classifiable as to carcinogenicity in humans.

Mercury

Mercury exists in three forms: elemental mercury, inorganic mercury compounds (primarily mercuric chloride), and organic mercury compounds (primarily methyl mercury). Each form exhibits different health effects. Various sources may release elemental or inorganic mercury; environmental methyl mercury is typically formed by biological processes after mercury has precipitated from the air.

Chronic exposure to elemental mercury in humans also affects the central nervous system, with effects such as increased excitability, irritability, excessive shyness, and tremors. The EPA has not classified elemental mercury with respect to cancer.

The major effect from chronic exposure to inorganic mercury is kidney damage. Reproductive and developmental animal studies have reported effects such as alterations in testicular tissue, increased embryo resorption rates, and abnormalities of development. Mercuric chloride (an inorganic mercury compound) exposure has been shown to result in forestomach, thyroid, and renal tumors in experimental animals. EPA has classified mercuric chloride as a Group C, possible human carcinogen.

Nickel

Nickel is an essential element in some animal species, and it has been suggested it may be essential for human nutrition. Nickel dermatitis, consisting of itching of the fingers, hand and forearms, is the most common effect in humans from chronic exposure to nickel. Respiratory effects have also been reported in humans from inhalation exposure to nickel. No information is available regarding the reproductive of developmental effects of nickel in humans, but animal studies have reported such effects, although a consistent dose-response relationship has not been seen. Nickel forms released from industrial boilers include soluble nickel compounds, nickel subsulfide, and nickel carbonyl. Human and animal studies have reported an increased risk of lung and nasal cancers from exposure to nickel refinery dusts and nickel subsulfide. Animal studies of soluble nickel compounds i.e., nickel carbonyl) have reported lung tumors. The EPA has classified nickel refinery subsulfide as a Group A, human carcinogen and nickel carbonyl as a Group B2, probable human carcinogen.

Organic HAP

Organic HAPs include halogenated and nonhalogenated organic classes of compounds such as polycyclic aromatic hydrocarbons (PAHs) and polychlorinated biphenyls (PCBs). Both PAHs and PCBs are classified as potential human carcinogens, and are considered toxic, persistent and bioaccumulative. Organic HAP also include compounds such as benzene, methane, propane, chlorinated alkanes and alkenes, phenols and chlorinated aromatics. Adverse health effects of HAPs include damage to the immune system, as well as neurological, reproductive, developmental, respiratory and other health problems.

Particulate Matter

Atmospheric particulate matter (PM) is composed of sulfate, nitrate, ammonium, and other ions, elemental carbon, particle-bound water, a wide variety of organic compounds, and a large number of elements contained in various compounds, some of which originate from crustal materials and others from combustion sources. Combustion sources are the primary origin of trace metals found in fine particles in the atmosphere. Ambient PM can be of primary or secondary origin.

Exposure to particles can lead to a variety of serious health effects. The largest particles do not get very far into the lungs, so they tend to cause fewer harmful health effects. Fine particles pose the greatest problems because they can get deep into the lungs. Scientific studies show links between these small particles and numerous adverse health effects. Epidemiological studies have shown a significant correlation between elevated PM levels and premature mortality. Other important effects associated with PM exposure include aggravation of respiratory and cardiovascular disease (as indicated by increased hospital admissions, emergency room visits, absences from school or work, and restricted activity days), lung disease, decreased lung function, asthma attacks, and certain cardiovascular problems. Individuals particularly sensitive to PM exposure include older adults and people with heart and lung disease.

This is only a partial summary of adverse health and environmental effects associated with exposure to PM. Further information is found in the 2004 Criteria Document for PM (“Air Quality Criteria for Particulate Matter,” EPA/600/P-99/002bF) and the 2005 Staff Paper for PM (EPA, “Review of the National Ambient Air Quality Standards for Particulate Matter, Policy Assessment of Scientific and Technical Information: OAQPS Staff Paper,” (June 2005)).

Selenium

Selenium is a naturally occurring substance that is toxic at high concentrations but is also a nutritionally essential element. Studies of humans chronically exposed to high levels of selenium in food and water have reported discoloration of the skin, pathological deformation and loss of nails, loss of hair, excessive tooth decay and discoloration, lack of mental alertness, and listlessness. The consumption of high levels of selenium by pigs, sheep, and cattle has been shown to interfere with normal fetal development and to produce birth defects. Results of human and animal studies suggest that supplementation with some forms of selenium may result in a reduced incidence of several tumor types. One selenium compound, selenium sulfide, is carcinogenic in animals exposed orally. We have classified elemental selenium as a Group D, not classifiable as to human carcinogenicity, and selenium sulfide as a Group B2, probable human carcinogen.

Part Two: Summary of the Final Rule

I. What Source Categories and Subcategories Are Affected by the Final Rule?

Today's rule promulgates standards for controlling emissions of HAP from hazardous waste combustors: incinerators, cement kilns, lightweight aggregate kilns, boilers, and hydrochloric acid production furnaces that burn hazardous waste. A description of each source category can be found in the proposed rule (see 69 FR at 21207-08).

Hazardous waste burning incinerators, cement kilns, and lightweight aggregate kilns are currently subject to 40 CFR part 63, subpart EEE, National Emission Standards for Hazardous Air Pollutants (NESHAP). Today's rule revises the emissions limits and certain compliance and monitoring provisions of subpart EEE for these Start Printed Page 59409source categories. The definitions of hazardous waste incinerator, hazardous waste cement kiln, and hazardous waste lightweight aggregate kiln appear at 40 CFR 63.1201(a).

Boilers that burn hazardous waste are also affected sources under today's rule. The rule uses the RCRA definition of a boiler under 40 CFR 260.10 and includes industrial, commercial, and institutional boilers as well as thermal units known as process heaters. Hazardous waste burning boilers will continue to comply with the emission standards found under 40 CFR part 266, subpart H (i.e., the existing RCRA rules) until they demonstrate compliance with the requirements of 40 CFR part 63, subpart EEE, and, for permitted sources, subsequently remove these requirements from their RCRA permit.

Finally, hydrochloric acid production furnaces that burn hazardous waste are affected sources under today's rule. These furnaces are a type of halogen acid furnace included in the definition of “industrial furnace” defined at § 260.10. Hydrochloric acid production furnaces that burn hazardous waste will continue to comply with the emission standards found under 40 CFR part 266, subpart H, until they demonstrate compliance with 40 CFR part 63, subpart EEE, and, for permitted sources, subsequently remove these requirements from their RCRA permit.

II. What Are the Affected Sources and Emission Points?

Today's rule apply to each major and area source incinerator, cement kiln, lightweight aggregate kiln, boiler, and hydrochloric acid production furnace that burns hazardous waste.[10] We note that only major source boilers and hydrochloric acid production furnaces are subject to the full suite of subpart EEE emission standards.[11] The emissions limits apply to each emission point (e.g., stack) where gases from the combustion of hazardous waste are discharged or otherwise emitted into the atmosphere. For facilities that have multiple combustion gas discharge points, the emission limits generally apply to each emission point. A cement kiln, for example, could be configured to have dual stacks where the majority of combustion gases are discharged though the main stack and other combustion gases emitted through a separate stack, such as an alkali bypass stack. In that case, the emission standards would apply separately to each of these stacks.[12]

III. What Pollutants Are Emitted and Controlled?

Hazardous waste combustors emit dioxin/furans, sometimes at high levels depending on the design and operation of the emission control equipment, and, for incinerators, depending on whether a waste heat recovery boiler is used. All hazardous waste combustors can also emit high levels of other organic HAP if they are not designed, operated, and maintained to operate under good combustion conditions.

Hazardous waste combustors can also emit high levels of metal HAP, depending on the level of metals in the waste feed and the design and operation of air emissions control equipment. Hazardous waste burning hydrochloric acid production furnaces, however, generally feed and emit low levels of metal HAP.

All of these HAP metals (except for the volatile metal mercury) are emitted as a portion of the particulate matter emitted by these sources. Hazardous waste combustors can also emit high levels of particulate matter, except that hydrochloric acid production furnaces generally feed hazardous wastes with low ash content and consequently emit low levels of particulate matter. A majority of particulate matter emissions from hazardous waste combustors are in the form of fine particulate. Particulate emissions from incinerators and liquid fuel-fired boilers depend on the ash content of the hazardous waste feed and the design and operation of air emission control equipment. Particulate emissions from cement kilns and lightweight aggregate kilns are not significantly affected by the ash content of the hazardous waste fuel because uncontrolled particulate emissions are attributable primarily to fine raw material entrained in the combustion gas. Thus, particulate emissions from kilns depends on operating conditions that effect entrainment of raw material, and the design and operation of the emission control equipment.

IV. Does the Final Rule Apply to Me?

The final rule applies to you if you own or operate a hazardous waste combustor—an incinerator, cement kiln, lightweight aggregate kiln, boiler, or hydrochloric acid production facility that burns hazardous waste. The final rule does not apply to a source that meets the applicability requirements of § 63.1200(b) for reasons explained at 69 FR at 21212-13.

V. What Are the Emission Limitations?

You must meet the emission limits in Tables 1 and 2 of this preamble for your applicable source category and subcategory. Standards are corrected to 7 percent oxygen. As noted at proposal, we previously promulgated requirements for carbon monoxide, total hydrocarbon, and destruction and removal efficiency standards under subpart EEE for incinerators, cement kilns, and lightweight aggregate kilns. We view these standards as unaffected by the Court's vacature of the challenged regulations in its decision of July 24, 2001. We are therefore not re-promulgating and reopening consideration of these standards in today's final rule, but are summarizing these standards in Tables 1 and 2 for reader's convenience.[13] See 69 FR at 21221, 21248, 21261 and 21274.

Liquid fuel boilers equipped with dry air pollution control devices are subject to different dioxin/furan emission standards than liquid fuel boilers that are not equipped with dry air pollution control devices.[14] Liquid fuel boilers processing hazardous waste with a heating value less than 10,000 BTU/lb must comply with the emission concentration-based standards (expressed as mass of total HAP emissions per volume of stack gas emitted) for mercury, semivolatile metals, low volatile metals, and total chlorine. Liquid fuel boilers processing hazardous waste with heating values greater than 10,000 BTU/lb must comply with thermal emissions-based standards (expressed as mass of HAP emissions attributable to the hazardous waste per million BTU input from the hazardous waste) for those same pollutants. Low volatile metal standards for liquid fuel boilers apply only to emissions of chromium, whereas the low volatile metal standard for the other source categories applies to the combined emissions of chromium, arsenic, and beryllium. Semivolatile metal standards apply to the combined emissions of lead and cadmium.

For any of the source categories except hydrochloric acid production Start Printed Page 59410furnaces, you may elect to comply with an alternative to the total chlorine standard under which you would establish site-specific, health-based emission limits for hydrogen chloride and chlorine based on national exposure standards. This alternative chlorine standard is discussed in part two, section IX and part four, section VII.

Incinerators and liquid and solid fuel boilers may elect to comply with an alternative to the particulate matter standard that would limit emissions of all the semivolatile metal HAPs and low volatile metal HAPs. Under this alternative, the numerical emission limits for semivolatile metal and low volatile metal emission HAP are identical to the limitations included in Tables 1 and 2. However, for semivolatile metals, the alternative standard applies to the combined emissions of lead, cadmium, and selenium; for low volatile metals, the standard applies to the combined emissions of chromium, arsenic, beryllium, antimony, cobalt, manganese, and nickel. See § 63.1219(e).

Table 1.—Summary of Emission Limits for Existing Sources

IncineratorsCement kilnsLightweight aggregate kilnsSolid fuel-fired boilers 1Liquid fuel-fired boilers 1Hydrochloric acid production furnaces 1
Dioxin/Furans (ng TEQ/dscm)0.20 or 0.40 and temperature control < 400°F at APCD inlet 60.20 or 0.40 and temperature control < 400°F at APCD inlet0.20 or rapid quench below 400°F at kiln exitCO or HC and DRE standard as a surrogate0.40 for dry APCD sources; CO or HC and DRE standard as surrogate for othersCO or HC and DRE standard as surrogate.
Mercury130 μg/dscmHazardous waste feed restriction of 3.0 ppmw and 120 μg/dscm MTEC 11; or 120 μg/dscm total emissions120 hazardous waste MTEC 11 feed restriction or 120 μg/dscm total emissions11 μg/dscm4.2E-5lb/MMBtu 2,5 or 19 μg/dscm 2; depending on BTU content of hazardous waste 13Total chlorine standard as surrogate.
Particulate Matter0.013 gr/dscf 80.028 gr/dscf and 20% opacity 120.025 gr/dscf0.030 gr/dscf 80.035 gr/dscf 8Total chlorine standard as surrogate.
Semivolatile Metals (lead + cadmium)230 μg/dscm7.6 E-4 lbs/MMBtu 5 and 330 μg/dscm 33.0E-4 lb/MMBtu 5 and 250 μg/dscm 3180 μg/dscm8.2 E-5 lb/MMBtu 2,5 or 150 μg/dscm 2; depending on BTU content of hazardous waste 13Total chlorine standard as surrogate.
Low Volatile Metals (arsenic + beryllium + chromium)92 μg/dscm2.1 E-5 lbs/MMBtu 5 and 56 μg/dscm 39.5E-5 lb/MMBtu 5 and 110 μg/dscm 3380 μg/dscm1.26E-4 lbMMBtu 4,5 or 370 μg/dscm 4; depending on BTU content of hazardous waste 13Total chlorine standard as surrogate.
Total Chlorine (hydrogen chloride + chlorine gas)32 ppmv 7120 ppmv 7600 ppmv 7440 ppmv 75.08E-2 lb/MMBtu 5,7 or 31 ppmv 7; depending on BTU content of hazardous waste 13150 ppmv or 99.923% system removal efficiency.
Carbon Monoxide (CO) or Hydrocarbons (HC)100 ppmv CO 9 or 10 ppmv HCSee Note # 10 below100 ppmv CO 9 or 20 ppmv HC(2) 100 ppmv CO 9 or 10 ppmv HC
Destruction and Removal Efficiency99.99% for each principal organic hazardous pollutant. For sources burning hazardous wastes F020, F021, F022, F023, F026, or F027, however, 99.9999% for each principal organic hazardous pollutant.
Notes:
1 Particulate matter, semivolatile metal, low volatile metal, and total chlorine standards for solid and liquid fuel boilers apply only to major sources. Particulate matter, semivolatile and low volatile metal standards for hydrochloric acid production furnaces apply only to major sources, although area sources must still comply with the surrogate total chlorine standard to control mercury emissions.
2 Standard is based on normal emissions data, and is therefore expressed as an annual average emission limitation.
3 Sources must comply with both the thermal emissions and emission concentration standards.
4 Low volatile metal standard for liquid fuel-fired boilers is for chromium only.
5 Standards expressed as mass of pollutant contributed by hazardous waste per million BTU contributed by the hazardous waste.
6 APCD means “air pollution control device”.
7 Sources may elect to comply with site-specific risk-based emission limits for hydrogen chloride and chlorine gas
8 Sources may elect to comply with an alternative to the particulate matter standard.
9 Sources that elect to comply with the CO standard must demonstrate compliance with the HC standard during the comprehensive performance test that demonstrates compliance with the destruction and removal efficiency requirement.
10 Kilns without a bypass: 20 ppmv HC or 100 ppmv CO 9. Kilns with a bypass/mid-kiln sampling system: 10 ppmv HC or 100 ppmv CO9 in the bypass duct, mid-kiln sampling system or bypass stack.
11 MTEC means “maximum theoretical emission concentration”, and is equivalent to the feed rate divided by gas flow rate
12 The opacity standard does not apply to a source equipped with a bag leak detection system under 63.1206(c)(8) or a particulate matter detection system under 63.1206(c)(9).
13 Emission concentration-based standards apply to sources processing hazardous waste with energy content less than 10,000 BTU/lb; thermal emission standards apply to sources processing hazardous waste with energy content greater than 10,000 btu/lb.
Start Printed Page 59411

Table 2.—Summary of Emission Limits for New or Reconstructed Sources

IncineratorsCement kilnsLightweight aggregate kilnsSolid fuel boilers 1Liquid fuel boilers 1Hydrochloric acid production furnaces 1
Dioxin/Furans (ng TEQ/dscm)0.11 for dry APCD and/or WHB 5 sources; 0.20 for other sources0.20 or 0.40 and temperature control <400 °F at APCD inlet0.20 or rapid quench <400 °F at kiln exitCO or HC and DRE standard as a surrogate0.40 for sources with dry APCD; CO or HC and DRE standard as a surrogate for other sourcesCO or THC and DRE standard as a surrogate.
Mercury8.1 μg/dscmHazardous waste feed restriction of 1.9 ppmw and 120 μg/dscm MTEC 10; or 120 μg/dscm total emissions120 hazardous waste MTEC 10 feed restriction or 120 μg/dscm total emissions11 μg/dscm1.2E-6 lb/MMBtu 24 or 6.8 μg/dscm 2; depending on BTU content of hazardous waste 12TCl as surrogate.
Particulate matter (gr/dscf)0.0015 70.0023 and 20% opacity 110.00980.015 70.0087 7TCl as surrogate.
Semivolatile Metals (lead + cadmium)10 μg/dscm6.2E-5 lb/MMBtu 4 and 180 μg/dscm3.7 E-5 lb/MMBtu 4 and 43 μg/dscm180 μg/dscm6.2 E-6 lb/MMBtu 24 or 78 μg/dscm 2; depending on BTU content of hazardous waste 12TCl as surrogate.
Low Volatile Metals (arsenic + beryllium + chromium)23 μg/dscm1.5E-5 lb/MMBtu 4 and 54 μg/dscm3..3E-5 lb/MMBtu 4 and 110 μg/dscm190 μg/dscm1.41E-5lb/MMBtu 34 or 12 μg/dscm 3; depending on BTU content of hazardous waste 12TCl as surrogate.
Total Chlorine (Hydrogen chloride + chlorine gas)21 ppmv 686 ppmv 6600 ppmv 673 ppmv 65.08E-2 lb/MMBtu 46 or 31 ppmv 6; depending on BTU content of hazardous waste 1225 ppmv or 99.987% SRE.
Carbon monoxide (CO) or Hydrocarbons (HC)100 ppmv CO 8 or 10 ppmv HCSee note #9 below100 ppmv CO 8 or 20 ppmv HC100 ppmv CO 8 or 10 ppmv HC
Destruction and Removal Efficiency99.99% for each principal organic hazardous pollutant. For sources burning hazardous wastes F020, F021, F022, F023, F026, or F027, however, 99.9999% for each principal organic hazardous pollutant.
Notes:
1 Particulate matter, semivolatile metal, low volatile metal, and total chlorine standards for solid and liquid fuel boilers apply only to major sources. Particulate matter, semivolatile and low volatile metal standards for hydrochloric acid production furnaces apply only to major sources, although area sources must still comply with the surrogate total chlorine standard to control mercury emissions.
2 Standard is based on normal emissions data, and is therefore expressed as an annual average emission limitation.
3 Low volatile metal standard for liquid fuel-fired boilers is for chromium only. Arsenic and beryllium are not included in the low volatile metal total for liquid fuel-fired boilers.
4 Standards expressed as mass of pollutant contributed by hazardous waste per million BTU contributed by the hazardous waste.
5 APCD means “air pollution control device”, WHB means “waste heat boiler”.
6 Sources may elect to comply with risk-based emission limits for hydrogen chloride and chlorine gas.
7 Sources may elect to comply with an alternative to the particulate matter standard.
8 Sources that elect to comply with the CO standard must demonstrate compliance with the THC standard during the comprehensive performance test that demonstrates compliance with the destruction and removal efficiency requirement.
9 Greenfield kilns without a bypass: 20 ppmv HC or 100 ppmv CO 8 and 50 ppmv HC. Greenfield kilns with a bypass/mid kiln sampling system: Main stack standard of 50 ppmv HC and 10 ppmv HC or 100 ppmv CO 8 in the bypass duct, mid-kiln sampling system or bypass stack. Greenfield kilns with a bypass/mid-kiln sampling system: 10 ppmv HC or 100 ppmv CO 8 in the bypass duct, mid-kiln sampling system or bypass stack; Non-greenfield kilns without a bypass: 20 ppmv HC or 100 ppmv CO 8. A greenfield kiln is a kiln whose construction commenced after April 19, 1996 at a plant site where a cement kiln (whether burning hazardous waste or not) did not previously exist.
10 MTEC means “maximum theoretical emission concentration”, and is equivalent to the feed rate divided by gas flow rate.
11 The opacity standard does not apply to a source equipped with a bag leak detection system under 63.1206(c)(8) or a particulate matter detection system under 63.1206(c)(9).
12 Emission concentration-based standards apply to sources processing hazardous waste with energy content less than 10,000 BTU/lb; thermal emission standards apply to sources processing hazardous waste with energy content greater than 10,000 btu/lb.

VI. What Are the Testing and Initial Compliance Requirements?

The testing and initial compliance requirements we promulgate today for solid fuel boilers, liquid fuel boilers, and hydrochloric acid production furnaces are identical to those that are applicable to incinerators, cement kilns, and lightweight aggregate kilns at §§ 63.1206, 63.1207, and 63.1208. We note, however, that today's final rule revises some of these requirements as they apply to all or specific HWCs (e.g., one-time dioxin/furan test for sources not subject to a numerical dioxin/furan standard; dioxin/furan stack test method; hydrogen chloride and chlorine stack test methods)

We also discuss compliance and testing dates for incinerators, cement kilns, and lightweight aggregate kilns as well. Even though we are not repromulgating the compliance and testing requirements for those source categories, those sources must demonstrate compliance with the replacement emission standards promulgated today. Start Printed Page 59412

A. Compliance Dates

The time-line for testing and initial compliance requirements is as follows:

1. The compliance date is October 14, 2008; [15]

2. You must submit a comprehensive performance test plan to the permitting authority for review and approval 12 months prior to commencing the test.

3.You must submit an eligibility demonstration for the health-based compliance alternative to the total chlorine emission standard 12 months before the compliance date if you elect to comply with § 63.1215;

4. You must place in the operating record a Documentation of Compliance by the compliance date identifying the operating parameter limits that, using available information, you have determined will ensure compliance with the emission standards;

5. For boilers and hydrochloric acid production furnaces, you must commence the initial comprehensive performance test within 6 months after the compliance date;

6. For incinerators, cement kilns, and lightweight aggregate kilns, you must commence the initial comprehensive performance test within 12 months after the compliance date;

7. You must complete the initial comprehensive performance test within 60 days of commencing the test; and

8. You must submit a Notification of Compliance within 90 days of completing the test documenting compliance with emission standards and continuous monitoring system requirements.

B. Testing Requirements

All hazardous waste combustors must commence the initial comprehensive performance test under the time lines discussed above. The purpose of the comprehensive performance test is to document compliance with the emission standards of the final rule and establish operating parameter limits to maintain compliance with those standards. You must also conduct periodic comprehensive performance testing every five years.

If your source is subject to a numerical dioxin/furan emission standard (i.e., incinerators, cement kilns, lightweight aggregate kilns that comply with the 0.2 ng TEQ/dscm standard, and liquid fuel boilers equipped with a dry air pollution control device), you must conduct a dioxin/furan confirmatory performance test no later than 2.5 years after each comprehensive performance test (i.e., midway between comprehensive performance tests). If your source is not subject to a numerical dioxin/furan emission standard (e.g., solid fuel boilers, lightweight aggregate kilns that comply with the 400 °F temperature limit at the kiln exit, liquid fuel boilers equipped with wet or no air pollution control system, and hydrochloric acid production furnaces), you must conduct a one-time dioxin/furan test to enable the Agency to evaluate the effectiveness of the carbon monoxide/hydrocarbon standard and the destruction and removal efficiency standard in controlling dioxin/furan emissions for those sources. Previous dioxin/furan emission tests may be used to meet this requirement if the combustor operated under the conditions required by the rule and if design and operation of the combustor has not changed since the test in a manner that could increase dioxin/furan emissions. The Agency will use those emissions data when reevaluating the MACT standards under CAA section 112(d)(6), when determining whether to develop residual risk standards for these sources pursuant to section 112(f)(2), and when determining whether the source's RCRA Permit is protective of human health and the environment.

You must use the following stack test methods to document compliance with the emission standards: (1) Method 29 for mercury, semivolatile metals, and low volatile metals; and (2) Method 26/26A, Methods 320 or 321, or ASTM D 6735-01 for hydrogen chloride and chlorine; [16] (3) either Method 0023A or Method 23 for dioxin/furans; and (4) either Method 5 or 5i for particulate matter.

C. Initial Compliance Requirements

The initial compliance requirements for solid fuel boilers, liquid fuel boilers, and hydrochloric acid production furnaces include: [17]

1. You must place in the operating record a Documentation of Compliance by the compliance date identifying the operating parameter limits that, using available information, you have determined will ensure compliance with the emission standards;

2. You must develop and comply with a startup, shutdown, and malfunction plan;

3. You must install an automatic waste feed cutoff system that links the operating parameter limits to the waste feed cutoff system;

4. You must control combustion system leaks;

5. You must establish and comply with an operator training and certification program;

6. You must establish and comply with an operation and maintenance plan;

7. If your source is equipped with a baghouse, you must install either a bag leak detection system or a particulate matter detection system; [18] and

8. If your source is equipped with an electrostatic precipitator or ionizing wet scrubber, you must either establish site-specific control device operating parameter limits which limits are linked to the automatic waste feed cutoff system, or install a particulate matter detection system and take corrective measures when the alarm level is exceeded.

VII. What Are the Continuous Compliance Requirements?

The continuous compliance requirements for solid fuel boilers, liquid fuel boilers, and hydrochloric acid production furnaces are identical to those applicable to incinerators, cement kilns, and lightweight aggregate kilns. See § 63.1209. We note, however, that today's final rule revises some of these requirements as they apply to all or specific HWCs (e.g., bag leak detection system requirements; optional particulate matter detection system requirements; compliance assurance for thermal emissions-based standards).

You must use carbon monoxide or hydrocarbon continuous emissions monitors (as well as an oxygen continuous emissions monitor to correct the carbon monoxide or hydrocarbon values to 7% oxygen) to ensure compliance with the carbon monoxide or hydrocarbon emission standards.

You must also establish limits (as applicable) on the feedrate of metals, chlorine, and ash, key combustor operating parameters, and key operating Start Printed Page 59413parameters of the air pollution control device based on operations during the comprehensive performance test. You must continuously monitor these parameters with a continuous monitoring system.

VIII. What Are the Notification, Recordkeeping, and Reporting Requirements?

The notification, recordkeeping, and reporting requirements that we promulgate today for solid fuel boilers, liquid fuel boilers, and hydrochloric acid production furnaces are identical to those that are applicable to incinerators, cement kilns, and lightweight aggregate kilns. See §§ 63.1210 and 63.1211. We note, however, that today's final rule revises some of these requirements as they apply to all or specific HWCs.

You must submit notifications including the following to the permitting authority in addition to those required by the NESHAP General Provisions, subpart A of 40 CFR part 63:

1. Notification of changes in design, operation, or maintenance (§ 63.1206(b)(5)(i));

2. Notification of performance test and continuous monitoring system evaluation, including the performance test plan and continuous monitoring system performance evaluation plan (§ 63.1207(e));

3. Notification of compliance, including results of performance tests and continuous monitoring system evaluations (§§ 63.1210(b), 63.1207(j); 63.1207(k), and 63.1207(l)); and

4. Various notifications if you request or elect to comply with alternative requirements at § 63.1210(a)(2).

You must submit the following reports to the permitting authority in addition to those required by the NESHAP General Provisions, subpart A of 40 CFR part 63:

1. Startup, shutdown, and malfunction plan, if you elect to comply with § 63.1206(c)(2)(ii)(B));

2. Excessive exceedances report (§ 63.1206(c)(3)(vi)); and

3. Emergency safety vent opening reports (§ 63.1206(c)(4)(iv)).

Finally, you must keep records documenting compliance with the requirements of Subpart EEE. Recordkeeping requirements are prescribed in § 63.1211(b), and include requirements under the NESHAP General Provisions, subpart A of 40 CFR

IX. What Is the Health-Based Compliance Alternative for Total Chlorine, and How Do I Demonstrate Eligibility?

A. Overview

The rule allows you to establish and comply with health-based compliance alternatives for total chlorine for hazardous waste combustors other than hydrochloric acid production furnaces in lieu of the MACT technology-based emission standards established under §§ 63.1216, 63.1217, 63.1219, 63.1220, and 63.1221. See § 63.1215. To identify and comply with the limits, you must:

(1) Identify a total chlorine emission rate for each on-site hazardous waste combustor. You may select total chlorine emission rates as you choose to demonstrate eligibility for the health-based limits, except the total chlorine emission rate limits for incinerators, cement kilns, and lightweight aggregate kilns cannot result in total chlorine emission concentrations exceeding the Interim Standards provided by §§ 63.1203, 63.1204, and 63.1205;[19]

(2) Calculate the HCl-equivalent emission rate for the total chlorine emission rates you select, considering long-term exposure and using Reference Concentrations (RfCs) as the health threshold metric. This emission rate is called the annual average HCl-equivalent emission rate;

(3) Perform an eligibility demonstration to determine if your annual average HCl-equivalent emission rate meets the national exposure standard (i.e., Hazard Index not exceeding 1.0 considering the maximum annual average ambient concentration of hydrogen chloride and chlorine at an off-site receptor location which concentrations are attributable to all on-site hazardous waste combustors) and thus is below the annual average HCl-equivalent emission rate limit;

(4) Calculate the HCl-equivalent emission rate for the total chlorine emission rates you select, considering short-term exposure and using acute Reference Exposure Levels (aRELs) as the health threshold metric. This emission rate is called the 1-hour average HCl-equivalent emission rate.

(5) Determine whether your 1-hour HCl-equivalent emission rate may exceed the national exposure standard (i.e., Hazard Index not exceeding 1.0 considering the maximum 1-hour average ambient concentration of hydrogen chloride and chlorine at an off-site receptor location which concentrations are attributable to all on-site hazardous waste combustors) and thus may exceed the 1-hour average HCl-equivalent emission rate limit when complying with the annual average HCl-equivalent emission rate limit, absent an hourly rolling average limit on the feedrate of total chlorine and chloride.

(6) Submit your eligibility demonstration, including your determination of whether the 1-hour average HCl-equivalent emission rate limit may be exceeded absent an hourly rolling average limit on the feedrate of total chlorine and chloride, for review and approval;

(7) Document during the comprehensive performance test the total chlorine system removal efficiency for each combustor and use this system removal efficiency to calculate chlorine feedrate limits. Also, document that total chlorine emissions during the test do not exceed the 1-hour average HCl-equivalent emission rate limit during any run of the test. In addition, establish operating limits on the emission control device based on operations during the comprehensive performance test; and

(8) Comply with the requirements for changes in the design, operation, or maintenance of the facility which could affect the HCl-equivalent emission rate limits or system removal efficiency for total chlorine, and changes in the vicinity of your facility over which you do not have control (e.g., new receptors locating proximate to the facility).

B. HCl-Equivalent Emission Rates

You must express total chlorine emission rates (lb/hr) from each on-site hazardous waste combustor, including hydrochloric acid production furnaces [20] , as an annual average HCl-equivalent emission rate and a 1-hour average HCl-equivalent emission rate. See § 63.1215(b). The annual average HCl-equivalent emission rate equates chlorine emission rates to hydrogen chloride (HCl) emission rates using Reference Concentrations (RfCs) as the health risk metric for long-term exposure. The 1-hour average HCl-equivalent emission rate equates chlorine emission rates to HCl emission rates using 1-hour Reference Exposure Start Printed Page 59414Levels (aRELs) as the health risk metric for acute exposure.

To calculate HCl-equivalent emission rates, you must apportion total chlorine emissions (ppmv) between chlorine and HCl using the volumetric ratio of chlorine to hydrogen chloride (Cl2/HCl).

  • To calculate the annual average HCl-equivalent emission rate (lb/hr) and the emission rate limit, you must use the historical average Cl2/HCl volumetric ratio from all regulatory compliance tests and the gas flowrate (and other relevant parameters) from the most recent RCRA compliance test or MACT performance test.
  • To calculate the 1-hour average HCl-equivalent emission rate (lb/hr) and emission rate limit, you must use the highest Cl2/HCl volumetric ratio from all regulatory compliance tests and the gas flowrate from the most recent RCRA compliance test or MACT performance test.
  • If you believe that the Cl2/HCl volumetric ratio for one or more historical compliance tests is not representative of the current ratio, you may request that the permitting authority allow you to screen those ratios from the analysis of historical ratios.
  • If the permitting authority believes that too few historical Cl2/HCl ratios are available to establish a representative average ratio and a representative maximum ratio, the permitting authority may require you to conduct periodic testing to establish representative ratios.
  • You must include the Cl2/HCl volumetric ratio demonstrated during each performance test in your data base of historical Cl2/HCl ratios to update the ratios for subsequent calculations of the annual average and 1-hour average HCl-equivalent emission rates (and emission rate limits).

C. Eligibility Demonstration

You must perform an eligibility demonstration to determine whether the total chlorine emission rates you select for each on-site hazardous waste combustor meet the national exposure standard (i.e., the Hazard Index of 1.0 cannot be exceeded at an off-site receptor location considering maximum annual average ambient concentrations attributable to all on-site hazardous waste combustors and the RfCs for HCl and chlorine) using either a look-up table analysis or a site-specific compliance demonstration.[21] Eligibility for the health-based total chlorine standard is determined by comparing the annual average HCl-equivalent emission rate for the total chlorine emission rate you select for each combustor to the annual average HCl-equivalent emission rate limit.

The annual average HCl-equivalent emission rate limit is the HCl-equivalent emission rate, determined by equating the toxicity of chlorine to HCl using RfCs as the health risk metric for long-term exposure, which ensures that maximum annual average ambient concentrations of HCl equivalents do not exceed a Hazard Index of 1.0, rounded to the nearest tenths decimal place (0.1) and considering all on-site hazardous waste combustors. See § 63.1215(b)(2).

Your facility is eligible for the health-based compliance alternatives for total chlorine if either: (1) The annual average HCl-equivalent emission rate for each on-site hazardous waste combustor is below the HCl-equivalent emission rate limit determined from the appropriate value for the emission rate limit in the applicable look-up table and the proration procedure for multiple combustors discussed below; or (2) the annual average HCl-equivalent emission rate for each on-site hazardous waste combustor is below the annual average HCl-equivalent emission rate limit you calculate based on a site-specific compliance demonstration.

1. Look-Up Table Analysis

Look-up tables for the eligibility demonstration are provided as Tables 1 and 2 to § 63.1215. Table 1 presents annual average HCl-equivalent emission rate limits for sources located in flat terrain. For purposes of this analysis, flat terrain is terrain that rises to a level not exceeding one half the stack height within a distance of 50 stack heights.

Table 2 presents annual average HCl-equivalent emission rate limits for sources located in simple elevated terrain. For purposes of this analysis, simple elevated terrain is terrain that rises to a level exceeding one half the stack height, but that does not exceed the stack height within a distance of 50 stack heights.

If your facility is not located in either flat or simple elevated terrain, you must conduct a site-specific compliance demonstration.

To determine the annual average HCl-equivalent emission rate limit for a source from the look-up table, you must use the stack height and stack diameter for your hazardous waste combustors and the distance between the stack and the property boundary. If any of these values for stack height, stack diameter, and distance to nearest property boundary do not match the exact values in the look-up table, you must use the next lowest table value. If you have more than one hazardous waste combustor on site, you must adjust the emission rate limits provided by the tables such that the sum of the ratios for all combustors of the adjusted emission rate limit to the emission rate limit provided by the table cannot exceed 1.0. See § 63.1215 (c)(3)(v).

2. Site-Specific Compliance Demonstration

You may use any scientifically-accepted peer-reviewed risk assessment methodology for your site-specific compliance demonstration to calculate an annual average HCl-equivalent emission rate limit for each on-site hazardous waste combustor. An example of one approach for performing the demonstration for air toxics can be found in the EPA's “Air Toxics Risk Assessment Reference Library, Volume 2, Site-Specific Risk Assessment Technical Resource Document,” which may be obtained through the EPA's Air Toxics Web site at http://www.epa.gov/​ttn/​atw.

To determine the annual average HCl-equivalent emission rate limit for each on-site hazardous waste combustor, your site-specific compliance demonstration must, at a minimum: (1) estimate long-term inhalation exposures through the estimation of annual or multi-year average ambient concentrations; (2) estimate the inhalation exposure for the actual individual most exposed to the facility's emissions from hazardous waste combustors, considering locations where people reside and where people congregate for work, school, or recreation; (3) use site-specific, quality-assured data wherever possible; (4) use health-protective default assumptions wherever site-specific data are not available, and: (5) contain adequate documentation of the data and methods used for the assessment so that it is transparent and can be reproduced by an experienced risk assessor and emissions measurement expert.

To establish the annual average HCl-equivalent emission rate limit for each combustor, you may apportion as you elect among the combustors the annual average HCl-equivalent emission rate limit for the facility, which limit ensures that the RfC-based Hazard Index of 1.0 is not exceeded. Start Printed Page 59415

D. Assurance That the 1-Hour HCl-Equivalent Emission Rate Will Not Be Exceeded

The long-term, RfC-based Hazard Index will always be higher than the short-term, aREL-based Hazard Index for a constant HCl-equivalent emission rate because the health threshold levels for short-term exposure are orders of magnitude higher than the health threshold levels for long-term exposure.[22] Even though maximum 1-hour average ambient concentrations are substantially higher than maximum annual average concentrations, the higher short-term ambient concentrations do not offset the much higher health threshold levels for short-term exposures. Thus, the long-term, RfC-based Hazard Index will always govern regarding whether a source can make an eligibility demonstration. Accordingly, eligibility for the health-based emission limits is based solely on whether a source can comply with the annual average HCl-equivalent emission rate limit.

Nonetheless, some sources may have highly variably chlorine feedrates (and corresponding highly variable HCl-equivalent emission rates) such that they may feed chlorine at very high levels for short periods of time and still remain in compliance with the chlorine feedrate limit established to ensure compliance with the annual average HCl-equivalent emission rate limit.[23] To ensure that the 1-hour HCl-equivalent emission rate limit will not be exceeded during these periods of peak emissions, you must establish a 1-hour average HCl-equivalent emission rate and 1-hour average HCl-equivalent emission rate limit for each combustor and consider site-specific factors including prescribed criteria to determine if the 1-hour average HCl-equivalent emission rate limit may be exceeded absent an hourly rolling average limit on chlorine feedrate. If the 1-hour average HCl-equivalent emission rate limit may be exceeded, you must establish an hourly rolling average feedrate limit on chlorine.

You must calculate the 1-hour average HCl-equivalent emission rate from the total chlorine emission rate you select for each source.

You must establish the 1-hour average HCl-equivalent emission rate limit for each affected source using either a look-up table analysis or site-specific analysis. Look-up tables are provided for 1-hour average HCl-equivalent emission rate limits as Table 3 and Table 4 to this section. Table 3 provides limits for facilities located in flat terrain. Table 4 provides limits for facilities located in simple elevated terrain. You must use the Tables to establish emission rate limits in the same manner as you use Tables 1 and 2 to establish annual average HCl-equivalent emission rate limits.

If you conduct a site-specific analysis to establish a 1-hour average HCl-equivalent emission rate limit, you must follow the risk assessment procedures you used to establish an annual average HCl-equivalent emission rate limit. The 1-hour HCl-equivalent emission rate limit, however, is the emission rate than ensures that the Hazard Index associated with maximum 1-hour average exposures is not greater than 1.0.

You must consider criteria including the following to determine if a source may exceed the 1-hour HCl-equivalent emission rate limit absent an hourly rolling average chlorine feedrate limit: (1) The ratio of the 1-hour average HCl-equivalent emission rate based on the total chlorine emission rate you select for each hazardous waste combustor to the 1-hour average HCl-equivalent emission rate limit for the combustor; and (2) the potential for the source to vary total chlorine and chloride feedrates substantially over the averaging period for the feedrate limit you establish to ensure compliance with the annual average HCl-equivalent emission rate limit.

If you determine that a source may exceed the 1-hour average HCl-equivalent emission rate limit, you must establish an hourly rolling average chlorine feedrate limit as discussed below in Section G.

You must include the following information in your eligibility demonstration to document your determination whether an hourly rolling average feedrate limit is needed to maintain compliance with the 1-hour HCl-equivalent emission rate limit: (1) Determination of the Cl2/HCl volumetric ratio established for 1-hour average HCl-equivalent emission rate determinations as provided by § 63.1215(b)(6)(ii); (2) determination of the 1-hour average HCl-equivalent emission rate calculated from the total chlorine emission rate you select for the combustor; (3) determination of the 1-hour average HCl-equivalent emission rate limit; (4) determination of the ratio of the 1-hour average HCl-equivalent emission rate to the 1-hour HCl-equivalent emission rate limit for the combustor; and (5) determination of the potential for the source to vary chlorine feedrates substantially over the averaging period for the long-term feedrate limit (i.e., 12-hours, or up to annually) established to maintain compliance with the annual average HCl-equivalent emission rate limit.

E. Review and Approval of Eligibility Demonstrations

The permitting authority will review and approve your eligibility demonstration. Your eligibility demonstration must contain, at a minimum, the information listed in § 63.1215(d)(1).

1. Review and Approval for Existing Sources

If you operate an existing source, you must submit the eligibility demonstration to your permitting authority for review and approval not later than 12 months prior to the compliance date. You must also submit a separate copy of the eligibility demonstration to: U.S. EPA, Risk and Exposure Assessment Group, Emission Standards Division (C404-01), Attn: Group Leader, Research Triangle Park, North Carolina 27711, electronic mail address REAG@epa.gov.

Your permitting authority should notify you of approval or intent to disapprove your eligibility demonstration within 6 months after receipt of the original demonstration, and within 3 months after receipt of any supplemental information that you submit. A notice of intent to disapprove your eligibility demonstration will identify incomplete or inaccurate information or noncompliance with prescribed procedures and specify how much time you will have to submit additional information or to comply with the MACT total chlorine standards. If your eligibility demonstration is disapproved, the permitting authority may extend the compliance date of the total chlorine standard to allow you to make changes to the design or operation of the combustor or related systems as quickly as practicable to enable you to achieve compliance with the MACT standard for total chlorine.

If your permitting authority has not approved your eligibility demonstration by the compliance date, and has not issued a notice of intent to disapprove your demonstration, you may nonetheless begin complying, on the compliance date, with the annual average HCl-equivalent emission rate limits you present in your eligibility demonstration.

If your permitting authority issues a notice of intent to disapprove your eligibility demonstration after the Start Printed Page 59416compliance date, the authority will identify the basis for that notice and specify how much time you will have to submit additional information or to comply with the MACT total chlorine standards. The permitting authority may extend the compliance date of the total chlorine standard to allow you to make changes to the design or operation of the combustor or related systems as quickly as practicable to enable you to achieve compliance with the MACT standard for total chlorine.

2. Review and Approval for New and Reconstructed Sources

The procedures for review and approval of eligibility demonstrations applicable to existing sources discussed above also apply to new or reconstructed sources, except that the date you must submit the eligibility demonstration is as discussed below.

If you operate a new or reconstructed source that starts up by April 12, 2007, or a solid fuel-fired boiler or liquid fuel-fired boiler that is an area source that increases its emissions or its potential to emit such that it becomes a major source of HAP before April 12, 2007, you must either: (1) Submit an eligibility demonstration for review and approval by April 12, 2006 and comply with the HCl-equivalent emission rate limits and operating requirements you establish in the eligibility demonstration; or (2) comply with the final total chlorine emission standards under §§ 63.1216, 63.1217, 63.1219, 63.1220, and 63.1221, by October 12, 2005, or upon startup, whichever is later, except for a standard that is more stringent than the standard proposed on April 20, 2004 for your source. If a final standard is more stringent than the proposed standard, you may comply with the proposed standard until October 14, 2008, after which you must comply with the final standard.

If you operate a new or reconstructed source that starts up on or after April 12, 2007, or a solid fuel-fired boiler or liquid fuel-fired boiler that is an area source that increases its emissions or its potential to emit such that it becomes a major source of HAP on or after April 12, 2007, you must comply with either of the following. You may submit an eligibility demonstration for review and approval 12 months prior to startup. Alternatively, you may comply with the final total chlorine emission standards under §§ 63.1216, 63.1217, 63.1219, 63.1220, and 63.1221 upon startup. If the final standard is more stringent than the standard proposed for your source on April 20, 2004, however, and if you start operations before October 14, 2008, you may comply with the proposed standard until October 14, 2008, after which you must comply with the final standard.

F. Testing Requirements

You must comply with the requirements for comprehensive performance testing under § 63.1207.

1. Test Methods for Stack Gas Containing Alkaline Particulate

If you operate a cement kiln or a combustor equipped with a dry acid gas scrubber, you must use EPA Method 320/321 or ASTM D 6735-01, or an equivalent method, to measure hydrogen chloride, and the back-half (caustic impingers) of Method 26/26A, or an equivalent method, to measure chlorine.

2. Test Methods for Stack Gas Containing High Levels of Bromine or Sulfur

If you operate an incinerator, boiler, or lightweight aggregate kiln and your feedstreams contain bromine or sulfur during the comprehensive performance test at the levels indicated below, you must use EPA Method 320/321 or ASTM D 6735'01, or an equivalent method, to measure hydrogen chloride, and Method 26/26A, or an equivalent method, to measure chlorine and hydrogen chloride combined. You must determine your chlorine emissions to be the higher of: (1) The value measured by Method 26/26A, or an equivalent method; or (2) the value calculated by the difference between the combined hydrogen chloride and chlorine levels measured by Method 26/26a, or an equivalent method, and the hydrogen chloride measurement from EPA Method 320/321 or ASTM D 6735-01, or an equivalent method.

These procedures apply if you feed during the comprehensive performance test bromine at a bromine/chlorine ratio in feedstreams greater than 5 percent by mass, or sulfur at a sulfur/chlorine ratio in feedstreams greater than 50 percent by mass.[24]

Finally, you should precondition the M26/26A filter for one hour prior to beginning the performance test to minimize the potential for a low bias caused by adsorption/absorption of hydrogen chloride on the filter.

G. Monitoring Requirements

You must establish and comply with limits on the same operating parameters that apply to sources complying with the MACT standard for total chlorine under § 63.1209(o), except that feedrate limits on total chlorine and chloride must be established as described below.

1. Feedrate Limit to Ensure Compliance with the Annual Average HCl-Equivalent Emission Rate Limit

For sources subject to the feedrate limit for total chlorine and chloride under § 63.1209(n)(4) to ensure compliance with the semivolatile metals standard, the feedrate limit (and averaging period) for total chlorine and chloride to ensure compliance with the annual average HCl-equivalent emission rate limit is the same as required by that paragraph. Thus, the chlorine feedrate limit is the average of the run averages during the comprehensive performance test, and is established as a 12-hour rolling average.

That chlorine feedrate limit cannot exceed the numerical value (i.e., not considering the averaging period) of the feedrate limit that ensures compliance with the annual average HCl-equivalent emission rate limit, however. Therefore, the numerical value of the total chlorine and chloride feedrate limit must not exceed the value you calculate as the annual average HCl-equivalent emission rate limit (lb/hr) divided by [1 − system removal efficiency]. You must calculate a total chlorine system removal efficiency for each test run of the comprehensive performance test as [1 − total chlorine emission rate (g/s)/chlorine feedrate (g/s)], and calculate the average system removal efficiency of the test run averages. If your source does not control total chlorine, you must assume zero system removal efficiency. If emissions during the comprehensive performance test exceed the annual average HCl-equivalent emission rate limit, eligibility for the health-based emission limits is not affected. This is because the emission rate limit is an annual average limit. Compliance is based on a 12-hour rolling average chlorine feedrate limit (rather than an (up to) an annual averaging period) for sources subject to the 12-hour rolling average feedrate limit for total chlorine and chloride under § 63.1209(n)(4) to ensure compliance with the semivolatile metals standard given that the more stringent feedrate limit (i.e., the feedrate limit with the shorter averaging period) would apply.

For sources exempt from the feedrate limit for total chlorine and chloride under § 63.1209(n)(4) because they comply with § 63.1207(m)(2) (which allows compliance with the semivolatile metals emission standard absent emissions testing by assuming all metals fed are emitted), the feedrate limit for total chlorine and chloride to ensure Start Printed Page 59417compliance with the annual average HCl-equivalent emission rate must be established as follows:

  • You must establish an average period for the feedrate limit that does not exceed an annual rolling average;
  • You must calculate a total chlorine system removal efficiency for each test run of the comprehensive performance test as [1 − total chlorine emission rate (g/s)/chlorine feedrate (g/s)], and calculate the average system removal efficiency of the test run averages. If your source is not equipped with a control system that consistently and reproducibly controls total emissions (e.g., wet or dry scrubber), you must assume zero system removal efficiency. If emissions during the comprehensive performance test exceed the annual average HCl-equivalent emission rate limit, eligibility for emission limits under this section is not affected. The emission rate limit is an annual average limit and compliance is based on an annual average feedrate limit on total chlorine and chloride (or a shorter averaging period if you so elect under paragraph (g)(2)(ii)(A) of this section); and
  • You must calculate the feedrate limit for total chlorine and chloride as the annual average HCl-equivalent emission rate limit (lb/hr) divided by [1 − system removal efficiency] and comply with the feedrate limit on the averaging period you establish.

2. Feedrate Limit To Ensure Compliance With the 1-Hour Average HCl-Equivalent Emission Rate Limit

You must establish an hourly rolling average feedrate limit on total chlorine and chloride to ensure compliance with the 1-hour average HCl-equivalent emission rate limit unless you determine that the hourly rolling average feedrate limit is waived as discussed under Section D above. If required, you must calculate the hourly rolling average feedrate limit for total chlorine and chloride as the 1-hour average HCl-equivalent emission rate limit (lb/hr) divided by [1 − system removal efficiency] using the system removal efficiency demonstrated during the comprehensive performance test.

H. Relationship Among Emission Rates, Emission Rate Limits, and Feedrate Limits

We summarize here the relationship among: (1) the total chlorine emission rate you select in your eligibility demonstration; (2) the annual average and 1-hour average HCl-equivalent emission rates you present in your eligibility demonstration; (3) the annual average and 1-hour average emission rate limits you present in your eligibility demonstration; (4) performance test emission rates for total chlorine and HCl-equivalent emissions; and (5) long-term and hourly rolling average chlorine feedrate limits.

1. Total Chlorine Emission Rate, Annual Average HCl-Equivalent Emission Rate, and Annual Average HCl-Equivalent Emission Rate Limit

For the eligibility demonstration, you must select a total chlorine emission concentration (ppmv) for each combustor, determine the Cl2/HCl volumetric ratio, calculate the annual average HCl-equivalent emission rate (lb/hr), and document that the emission rate does not exceed the annual average HCl-equivalent emission rate limit.

You select a total chlorine (i.e., HCl and chlorine combined) emission concentration (ppmv) for each hazardous waste combustor expressed as chloride (Cl(-)) equivalent. For incinerators, cement kilns, and lightweight aggregate kilns, this emission concentration cannot exceed the Interim Standards for total chlorine. You then determine the average Cl2/HCl volumetric ratio considering all historical regulatory emissions tests and apportion total chlorine emissions between Cl2 and HCl accordingly. You use these apportioned volumetric emissions to calculate the Cl2 and HCl emission rates (lb/hr) using the average gas flowrate (and other relevant parameters) for the most recent RCRA compliance test or MACT performance test for total chlorine. Finally, you use these Cl2 and HCl emission rates to calculate an annual average HCl-equivalent emission rate, which cannot exceed the annual average HCl-equivalent emission rate limit that you establish as discussed below.

To establish the annual average HCl-equivalent emission rate limit, you may either use Tables 1 or 2 in § 63.1215 to look-up the limit, or conduct a site-specific risk analysis. Under the site-specific risk analysis option, the annual average HCl-equivalent emission rate limit would be the highest emission rate that the risk assessment estimates would result in a Hazard Index not exceeding 1.0 for the actual individual most exposed to the facility's emissions considering off-site locations where people reside and where people congregate for work, school, or recreation.

If you have more than one on-site hazardous waste combustor, and if you use the look-up tables to establish the annual average HCl-equivalent emission rate limits, the sum of the ratios for all combustors of the annual average HCl-equivalent emission rate to the annual average HCl-equivalent emission rate limit cannot not exceed 1.0. This will ensure that the RfC-based Hazard Index of 1.0 is not exceeded, a principle criterion of the eligibility demonstration.

If you use site-specific risk analysis to demonstrate that a Hazard Index of 1.0 is not exceeded, you would generally identify for each combustor the maximum annual average HCl-equivalent emission rate that the risk assessment estimates would result in an RfC-based Hazard Index of 1.0 at any off-site receptor location (i.e., considering locations where people reside and where people congregate for work, school, or recreation.[25] This emission rate would be the annual average HCl-equivalent emission rate limit for each combustor.

2. 1-Hour Average HCl-Equivalent Emission Rate and Emission Rate Limit

As discussed in Section D above, you must determine in your eligibility demonstration whether the 1-hour HCl-equivalent emission rate limit may be exceeded absent an hourly rolling average chlorine feedrate limit. To make this determination, you must establish a 1-hour average HCl-equivalent emission rate and a 1-hour average HCl-equivalent emission rate limit.

You calculate the 1-hour average HCl-equivalent emission rate from the total chlorine emission rate, established as discussed above, using the equation in § 63.1215(b)(3).

You establish the 1-hour average HCl-equivalent emission rate limit by either using Tables 3 or 4 in § 63.1215 to look-up the limit, or conducting a site-specific risk analysis. Under the site-specific risk analysis option, the 1-hour average HCl-equivalent emission rate limit would be the highest emission rate that the risk assessment estimates would result in an aREL-based Hazard Index not exceeding 1.0 at any off-site receptor location (i.e., considering locations where people reside and where people congregate for work, school, or recreation).

3. Performance Test Emissions

During the comprehensive performance test, you must demonstrate a system removal efficiency for total chlorine as [1 − TCl emitted (lb/hr)/chlorine fed (lb/hr)]. During the test, however, the total chlorine emission rate you select for each combustor and the annual average HCl-equivalent Start Printed Page 59418emission rate limit can exceed the levels you present in the eligibility demonstration. This is because those emission rates are annual average rates and need not be complied with over the duration of three runs of the performance test, which may be nominally only 3 hours.

The 1-hour average HCl-equivalent emission rate limit cannot be exceeded during any run of the comprehensive performance test, however. This limit is based on an aREL Hazard Index of 1.0; an exceedance of the limit over a test run with a nominal 1-hour duration would result in a Hazard Index of greater than 1.0.

4. Chlorine Feedrate Limits

To maintain compliance with the annual average HCl-equivalent emission rate limit, you must establish a long-term average chlorine feedrate limit. In addition, if you determine under § 63.1205(d)(3) that the 1-hour average HCl-equivalent emission rate may be exceeded (i.e., because your chlorine feedrate may vary substantially over the averaging period for the long-term chlorine feedrate limit), you must establish an hourly rolling average chlorine feedrate limit.

Long-Term Chlorine Feedrate Limit. The chlorine feedrate limit to maintain compliance with the annual average HCl-equivalent emission rate is either: (1) The chlorine feedrate during the comprehensive performance test if you demonstrate compliance with the semivolatile metals emission standard during the test (see § 63.1209(o)); or (2) if you comply with the semivolatile metals emission standard under § 63.1207(m)(2) by assuming all metals in the feed to the combustor are emitted, the HCl-equivalent emission rate limit divided by [1 − system removal efficiency] where you demonstrate the system removal efficiency during the comprehensive performance test.

If you establish the chlorine feedrate limit based on the feedrate during the performance test to demonstrate compliance with the semivolatile metals emission standard, the averaging period for the feedrate limit is a 12-hour rolling average. If you establish the chlorine feedrate limit based on the system removal efficiency during the performance test, the averaging period is up to an annual rolling average. See discussion in Part Four, Section VII.B of this preamble.

If you comply with the semivolatile metals emission standard under § 63.1207(m)(2), however, the long-term chlorine feedrate limit is based on the system removal efficiency during the comprehensive performance test rather than the feedrate during the performance test. This is because the averaging period for this chlorine feedrate limit (that ensures compliance with the annual average HCl-equivalent emission rate limit) is up to an annual rolling average. See § 63.1215(g)(2). Thus, the chlorine feedrate, and total chlorine emissions, can be higher than the limit during the relatively short duration of the comprehensive performance tests.

Hourly Rolling Average Chlorine Feedrate Limit. If you determine under § 63.1205(d)(3) that the 1-hour average HCl-equivalent emission rate limit may be exceeded, you must establish an hourly rolling average chlorine feedrate limit. That feedrate limit is established as the 1-hour HCl-equivalent emission rate limit divided by [1 − system removal efficiency]. The hourly rolling average chlorine feedrate limit is not established based on feedrates during the performance test because performance test feedrates may be substantially lower than the feedrate needed to ensure compliance with the 1-hour average HCl-equivalent emission rate. Note, however, that the hourly rolling average feedrate limit cannot be exceeded during any run of the comprehensive performance test. This chlorine feedrate limit is based on the 1-hour average HCl-equivalent emission rate limit, which is based on an aREL Hazard Index of 1.0. Thus, an exceedance of the hourly rolling average feedrate limit (and the 1-hour lHCl-equivalent emission rate limit) over a test run with a nominal 1-hour duration would result in a Hazard Index of greater than 1.0.

I. Changes

Your requirements will change in response to changes that affect the HCl-equivalent emission rate or HCl-equivalent emission rate limit for a source.

1. Changes Over Which You Have Control

Changes That Affect HCl-Equivalent Emission Rate Limits. If you plan to change the design, operation, or maintenance of the facility in a manner that would decrease the annual average or 1-hour average HCl-equivalent emission rate limit (e.g., reduce the distance to the property line; reduce stack gas temperature; reduce stack height), prior to the change you must submit to the permitting authority a revised eligibility demonstration documenting the lower emission rate limits and calculations of reduced total chlorine and chloride feedrate limits.

If you plan to change the design, operation, or maintenance of the facility in a manner than would increase the annual average or 1-hour average HCl-equivalent emission rate limit, and you elect to increase your total chlorine and chloride feedrate limits, prior to the change you must submit to the permitting authority a revised eligibility demonstration documenting the increased emission rate limits and calculations of the increased feedrate limits prior to the change.

Changes That Affect System Removal Efficiency. If you plan to change the design, operation, or maintenance of the combustor in a manner than could decrease the system removal efficiency, you are subject to the requirements of § 63.1206(b)(5) for conducting a performance test to reestablish the combustor's system removal efficiency. You also must submit a revised eligibility demonstration documenting the lower system removal efficiency and the reduced feedrate limits on total chlorine and chloride.

If you plan to change the design, operation, or maintenance of the combustor in a manner than could increase the system removal efficiency, and you elect to document the increased system removal efficiency to establish higher feedrate limits on total chlorine and chloride, you are subject to the requirements of § 63.1206(b)(5) for conducting a performance test to reestablish the combustor's system removal efficiency. You must also submit a revised eligibility demonstration documenting the higher system removal efficiency and the increased feedrate limits on total chlorine and chloride.

2. Changes Over Which You Do Not Have Control

If you use site-specific risk assessment in lieu of the look-up tables to establish the HCl-equivalent emission rate limit, you must review the documentation you use in your eligibility demonstration every five years from the date of the comprehensive performance test and submit for review and approval with the comprehensive performance test plan either a certification that the information used in your eligibility demonstration has not changed in a manner that would decrease the annual average HCl-equivalent emission rate limit, or a revised eligibility demonstration. Examples of changes beyond your control that may decrease the annual average HCl-equivalent emission rate limit (or 1-hour average HCl-equivalent emission rate limit) are construction of residences at a location exposed to higher ambient Start Printed Page 59419concentrations than evaluated during your previous risk analysis, or a reduction in the RfCs or aRELs.

If, in the interim between the dates of your comprehensive performance tests, you have reason to know of changes that would decrease the annual average HCl-equivalent emission rate limit, you must submit a revised eligibility demonstration as soon as practicable but not more frequently than annually.

If you determine that you cannot demonstrate compliance with a lower annual average HCl-equivalent emission rate limit (dictated by a change over which you do not have control) during the comprehensive performance test because you need additional time to complete changes to the design or operation of the source or related systems, you may request that the permitting authority grant you additional time to make those changes as quickly as practicable.

X. Overview on Floor Methodologies

The most contentious issue in the rulemaking involved methodologies for determining MACT floors, namely, which sources are best performing, and what is their level of performance. Superficially, these questions have a ready answer: the best performers are the lowest emitters as measured by compliance tests, and those tests fix their level of performance. But compliance tests are snapshots which do not fully capture sources' total operating variability. Since the standards must be met at all times, picking lowest compliance test data to set the standard results in standards best performing sources themselves would be unable to meet at all times.

To avoid this impermissible result, EPA selected approaches that reasonably estimate best performing sources' total variability. Certain types of variability can be quantified statistically, and EPA did so here (using standard statistical approaches) in all of the floor methodologies used in the rule. There are other components of variability, however, which cannot be fully quantified, but nonetheless must be accounted for in reasonably estimating best performing sources' performance over time. EPA selected ranking methodologies which best account for this total variability.

Where control of the feed of HAP is feasible and technically assessable (the case for HAP metals and for total chlorine), EPA used a methodology that ranked sources by their ability to best control both HAP feed and HAP emissions. This methodology thus assesses the efficiency of control of both the HAP inputs to a hazardous waste combustion unit, and the efficiency of control of the unit's outputs. This methodology reasonably selects the best performing (and for new sources, best controlled) sources, and reasonably assesses their level of performance. When HAP feed control is not feasible, notably where HAP is contributed by raw material and fossil fuel inputs, EPA determined best performers and their level of performance using a methodology that selects the lowest emitters using the best air pollution control technology. This methodology reasonably estimates the best performing sources' level of performance, and better accounts for total variability in emissions levels of the best performing sources.

EPA carefully examined approaches selecting lowest emitters as best performers. Examination of other test conditions from the same best performing sources shows, however, that this approach results in standards not achievable even by the best performers. Indeed, in order to meet such standards, even “best performing” sources (lowest emitting in individual tests) would have to add additional air pollution control technology. EPA views this result as an end run around the section 112(d)(2) beyond-the-floor process, because floor standards would force industry-wide technological changes without consideration of the factors (cost and energy in particular) which Congress mandated for consideration when establishing beyond-the-floor standards.

Part Three: What Are the Major Changes Since Proposal?

I. Database

A. Hazardous Burning Incinerators

Five incinerators have been removed from the database because they have initiated or completed RCRA closure.[26] Two incinerators have been added to the list of sources used to calculate the floor levels.[27] Emissions data from source 3015 has been excluded for purposes of calculating the particulate matter floor because the source was processing an atypical waste stream from a particulate matter compliance perspective. See part four, section I.F. We have excluded the most recent mercury and dioxin/furan emissions data from source 327, and have instead used data from an older test condition to represent this source's emissions because the source encountered problems with its carbon injection system during the most recent test. See part four, section I.F. Emissions data from source 3006 has been excluded for purposes of calculating the semivolatile metal standard because this source did not measure cadmium emissions during its emissions test. See part four, section I.F. We have added mercury emissions data from source 901 (DSSI) to the incinerator mercury database because this source (which is otherwise subject to standards for liquid fuel boilers) is burning a waste which is unlike that burned by any other liquid fuel boiler with respect to mercury concentration and waste provenance, but typical of waste burned by incinerators with respect to those factors. See part four, section VI.D.1. This change correspondingly affects the liquid fuel boiler standard by removing that data from the liquid fuel boiler database.

B. Hazardous Waste Cement Kilns

1. Use of Emissions Data From Ash Grove Cement Company

The emissions data from Ash Grove Cement Company, which operates a recently constructed preheater/precalciner kiln located in Chanute, Kansas, are considered when calculating MACT floors for new hazardous waste burning cement kilns. In the proposal, we did not consider their emissions data in the floor analyses for existing sources because Ash Grove Cement used the data to demonstrate compliance with the new source interim standards, and did not address the data for purposes of new source standards. See 69 FR at 21217 n. 35. Consistent with our position on use of post-1999 emissions data, we are including Ash Grove Cement's emissions data in the floor analyses for new sources. See also Part Four, Section I.B of the preamble.

2. Removal of Holcim's Emissions Data From EPA's HWC Data Base

Following cessation of hazardous waste operations in 2003, we are removing all emissions data from both wet process cement kilns at Holcim's Holly Hill, South Carolina, plant from our hazardous waste combustor data base. This is consistent with our approach in both this rule and the 1999 rule to base the standards only on performance of sources that actually are operating (i.e., burning hazardous waste). See also Part Four, Section I.A and 64 FR at 52844. Start Printed Page 59420

3. Use of Mercury Data

As discussed below, we are using a commenter-submitted dataset as the basis of the mercury standards for existing and new cement kilns. This comprehensive dataset documents the day-to-day levels of mercury in hazardous waste fired to all cement kilns for a three year period covering 1999 to 2001. We have determined that the commenter-submitted data are more representative than data used at proposal. See Part Four, Section I.D of the preamble for our rationale.

C. Hazardous Waste Lightweight Aggregate Kilns

We are incorporating mercury data submitted by a commenter into the MACT floor analysis for existing and new lightweight aggregate kilns. These data document the day-to-day levels of mercury in hazardous waste fired to lightweight aggregate kilns located at Solite Corporation's Arvonia plant between October 2003 and June 2004. We have determined that the commenter-submitted data are more representative than the data used at proposal. See Part Four, Section I.E of the preamble for our rationale.

D. Liquid Fuel Boilers

In the proposed rule, we classified liquid fuel boilers as one category. The final rule classifies them into two for purposes of the mercury, semivolatile metals, chromium, and total chlorine standards: one for liquid fuel boilers burning lower heating value hazardous waste (hazardous waste with a heating value less than 10,000 Btu/lb), and another for liquid fuel boilers burning higher heating value hazardous waste (hazardous waste with a heating value of 10,000 Btu/lb or greater).

We also made other, minor changes to the data base because some sources have initiated closure, were misclassified as other sources in the proposed rule, or were inadvertently not considered in the floor calculations although the sources' test reports were in the docket at proposal.

E. HCl Production Furnaces

Six of the 17 hydrochloric acid production furnaces have ceased burning hazardous waste since proposal. Consequently, we do not use emissions data from these sources to establish the final standards. All six of these sources were equipped with waste heat recovery boilers and had relatively high dioxin/furan emissions. In addition, we reclassified source #2020 as a boiler based on comments received at proposal.

F. Total Chlorine Emissions Data Below 20 ppmv

We corrected all the total chlorine measurements in the data base that were below 20 ppmv to account for potential systemic negative biases in the Method 0050 data in response to comments on the proposed rule. See the discussion in Part Four, Section I.C.1 below.

To account for the bias, we corrected all total chlorine emissions data that were below 20 ppmv to 20 ppmv. We accounted for within-test condition emissions variability for the corrected data by imputing a standard deviation that is based on a regression analysis of run-to-run standard deviation versus emission concentration for all data above 20 ppmv. This approach of using a regression analysis to impute a standard deviation is similar to the approach we used to account for total variability (i.e., test-to-test and within test variability) of PM emissions for sources that use fabric filters.

II. Emission Limits

A. Incinerators

The changes in the incinerator standards for existing sources since proposal are:

StandardProposed limitFinal limit
Dioxin/Furans (ng TEQ/dscm)Sources with dry air pollution control systems or waste heat boilers: 0.28; For others: 0.2 or 0.4 and temperature control at inlet of air pollution control device < 400 °FFor all sources, 0.20 or 0.40 and temperature control < 400 °F at the air pollution control device inlet.
Particulate Matter (gr/dscf)0.0150.013.
Semivolatile Metals (μg/dscm)59230.
Low Volatile Metals (μg/dscm)8492.
Total Chlorine (ppmv)1.532.
Alternative to the particulate matter standard: Combined emissions of lead, cadmium and selenium (μg/dscm)59230.
Alternative to the particulate matter standard: Combined emissions of arsenic, berrylium, chrome, antimony, cobalt, manganese, and nickel (μg/dscm)8492.

The changes in the incinerator standards for new sources since proposal are:

StandardProposed limitFinal limit
Particulate Matter (gr/dscf)0.00070.0015
Mercury (μg/dscm)8.08.1
Semivolatile Metals (μg/dscm)6.510
Low Volatile Metals (μg/dscm)8.923
Total Chlorine (ppmv)0.1821
Alternative to the particulate matter standard: Combined emissions of lead, cadmium and selenium (μg/dscm)6.510
Alternative to the particulate matter standard: Combined emissions of arsenic, berrylium, chrome, antimony, cobalt, manganese, and nickel (μg/dscm)8.923
Start Printed Page 59421

Hazardous Waste Burning Cement Kilns

The changes in the standards for existing cement kiln since proposal are:

StandardProposed limitFinal limit
Mercury (μg/dscm)64 1Both 3.0 ppmw 2 and either 120 μg/dscm (stack emissions) or 120 μg/dscm (expressed as a hazardous waste MTEC) 3.
Particulate matter0.028 gr/dscf0.028 gr/dscf and 20% opacity 4.
Semivolatile metals4.0E-04 lb/MMBtu 57.6E-04 lb/MMBtu 5and 330 μg/dscm.
Low volatile metals1.4E-05 lb/MMBtu 52.1E-05 lb/MMBtu 5and 56 μg/dscm.
Total chlorine (ppmv) 6110120.
1 The proposed mercury standard was an annual limit.
2 Feed concentration of mercury in hazardous waste as-fired.
3 HW MTEC means maximum theoretical emissions concentration of the hazardous waste and MTEC is defined at § 63.1201(a).
4 The opacity standard does not apply to a source equipped with a bag leak detection system under § 63.1206(c)(8) or a particulate matter detection system under § 63.1206(c)(9).
5 Standard is expressed as mass of pollutant stack emissions attributable to the hazardous waste per million British thermal unit heat input of the hazardous waste.
6 Combined standard, reported as a chloride (Cl(-)) equivalent.

The changes in the standards for new cement kilns since proposal are:

StandardProposed limitFinal limit
Mercury (μg/dscm)35 1Both 1.9 ppmw 2 and either 120 μg/dscm (stack emissions) or 120 μg/dscm (expressed as a hazardous waste MTEC) 3.
Particulate matter0.0058 gr/dscf0.0023 gr/dscf and 20% opacity 4.
Semivolatile metals6.2E-05 lb/MMBtu 56.2E-05 lb/MMBtu 5and 180 μg/dscm.
Low volatile metals1.4E-05 lb/MMBtu 51.5E-05 lb/MMBtu 5and 54 μg/dscm.
Total chlorine (ppmv) 67886.
1 The proposed mercury standard was an annual limit.
2 Feed concentration of mercury in hazardous waste as-fired.
3 HW MTEC means maximum theoretical emissions concentration of the hazardous waste and MTEC is defined at § 63.1201(a).
4 The opacity standard does not apply to a source equipped with a bag leak detection system under § 63.1206(c)(8) or a particulate matter detection system under § 63.1206(c)(9).
5 Standard is expressed as mass of pollutant stack emissions attributable to the hazardous waste per million British thermal unit heat input of the hazardous waste.
6 Combined standard, reported as a chloride (Cl(-)) equivalent.

C. Hazardous Waste Burning Lightweight Aggregate Kilns

The changes in the standards for existing lightweight aggregate kilns since proposal are:

StandardProposed limitFinal limit
Dioxins and furans (ng TEQ/dscm)0.400.20 or rapid quench of the flue gas at the exit of the kiln to less than 400 °F.
Mercury (μg/dscm)67 1120 μg/dscm (stack emissions) or 120 μg/dscm (expressed as a hazardous waste MTEC) 2.
Semivolatile metals3.1E-04 lb/MMBtu 3and 250 μg/dscm3.0E-04 lb/MMBtu 3and 250 μg/dscm.
1 The proposed mercury standard was an annual limit.
2 HW MTEC means maximum theoretical emissions concentration of the hazardous waste and MTEC is defined at § 63.1201(a).
3 Standard is expressed as mass of pollutant stack emissions attributable to the hazardous waste per million British thermal unit heat input of the hazardous waste.

The changes in the standards for new lightweight aggregate kilns since proposal are:

StandardProposed limitFinal limit
Dioxins and furans (ng TEQ/dscm)0.400.20 or rapid quench of the flue gas at the exit of the kiln to less than 400 °F.
Start Printed Page 59422
Particulate matter0.0099 gr/dscf0.0098 gr/dscf.
Mercury (μg/dscm)67 1120 μg/dscm (stack emissions) or 120 μg/dscm (expressed as a hazardous waste MTEC) 2.
Semivolatile metals2.4E-05 lb/MMBtu 3and 43 μg/dscm3.7E-05 lb/MMBtu 3and 43 μg/dscm.
1 The proposed mercury standard was an annual limit.
2 HW MTEC means maximum theoretical emissions concentration of the hazardous waste and MTEC is defined at § 63.1201(a).
3 Standard is expressed as mass of pollutant stack emissions attributable to the hazardous waste per million British thermal unit heat input of the hazardous waste.

D. Solid Fuel Boilers

The changes in the solid fuel boiler standards for existing sources since proposal are:

StandardProposed limitFinal limit
Mercury (μg/dscm)1011
Semivolatile Metals (μg/dscm)170180
Low Volatile metals (μg/dscm)210380
Alternative to the particulate matter standard: Combined emissions of lead, cadmium and selenium (μg/dscm)170180
Alternative to the particulate matter standard: Combined emissions of arsenic, beryllium, chromium, antimony, cobalt, manganese, and nickel (μg/dscm)210380

The changes in the solid fuel boiler standards for new sources since proposal are:

StandardProposed limitFinal limit
Mercury (μg/dscm)1011
Semivolatile Metals (μg/dscm)170180
Low Volatile metals (μg/dscm)210380
Alternative to the particulate matter standard: Combined emissions of lead, cadmium and selenium (μg/dscm)170180

E. Liquid Fuel Boilers

We redefined the liquid fuel boiler subcategory into two separate boiler subcategories based on the heating value of the hazardous waste they burn: Those that burn waste below 10,000 Btu/lb, those that burn hazardous waste with a heating value of 10,000 Btu/lb or greater. See Part Four, Section VI.D.2 of today's preamble for a complete discussion.

The additional changes to the liquid fuel boiler standards for existing sources since proposal are:

StandardProposed limitFinal limit
HW Fuel < 10,000 Btu/lbHW Fuel ≥ 10,000 Btu/lb
Mercury (lb/MM Btu)3.7E-619 μg/dscm4.2E-5
Particulate matter (gr/dscf)0.0320.035
Semivolatile metals (lb/MM Btu)1.1E-5150 μg/dscm8.2E-5
Chromium (lb/MM Btu)1.1E-4370 μg/dscm1.3E-4
Total chlorine (Lb/MM Btu)2.5E-231 ppmv5.1E-2
Alternative to the particulate matter standard: Combined emissions of lead, cadmium and selenium (lb/MM Btu)1.1E-5150 μg/dscm8.2E-5
Alternative to the particulate matter standard: Combined emissions of arsenic, beryllium, chromium, antimony, cobalt, manganese, and nickel (lb/MM Btu)1.1E-4370 μg/dscm1.3E-4

The changes in the liquid fuel boiler standards for new sources since proposal are: Start Printed Page 59423

StandardProposed limitFinal limit
HW fuel < 10,000 Btu/lbHW fuel > 10,000 Btu/lb
Dioxin and Furan, dry APCD (ng TEQ/dscm)0.015 or temp control <400F for dry APCD0.40
Mercury (lb/MM Btu)3.8E-76.8 μg/dscm1.2E-6
Particulate matter (gr/dscf)0.00760.0087
Semivolatile metals (lb/MM Btu)4.3E-678 μg/dscm6.2E-6
Chromium (lb/MM Btu)3.6E-512 μg/dscm1.4E-5
Total chlorine (lb/MM Btu)7.2E-431 μg/dscm5.1E-2
Alternative to the particulate matter standard: Combined emissions of lead, cadmium and selenium (lb/MM Btu)4.3E-678 μg/dscm 16.2E-6 1
Alternative to the particulate matter standard: Combined emissions of arsenic, beryllium, chromium, antimony, cobalt, manganese, and nickel (lb/MM Btu)3.6E-512 μg/dscm 21.4E-5 2
1 New or reconstructed liquid fuel boilers that process residual oil or liquid feedstreams that are neither fossil fuel nor hazardous waste and that operate pursuant to the alternative to the particulate matter standard must comply with the alternative emission concentration standard of 4.7 μg/dscm, which is applicable to lead, cadmium and selenium emissions attributable to all feedstreams (hazardous and nonhazardous).
2 New or reconstructed liquid fuel boilers that process residual oil or liquid feedstreams that are neither fossil fuel nor hazardous waste that operate pursuant to the alternative to the particulate matter standard must comply with the alternative emission concentration standard of 12 μg/dscm, which is applicable to arsenic, beryllium, chrome, antimony, cobalt, manganese, and nickel emissions attributable to all feedstreams (hazardous and nonhazardous).

F. Hydrochloric Acid Production Furnaces

The changes in the hydrochloric acid production furnace standards for existing sources since proposal are:

StandardProposed limitFinal limit
Dioxin and Furans0.4 ng TEQ/dscmCarbon Monoxide/Total Hydrocarbons and DRE standards as surrogates.
Total chlorine14 ppmv or 99.9927% system removal efficiency150 ppmv or 99.923% system removal efficiency.

The changes in the hydrochloric acid production furnace standards for new sources since proposal are:

StandardProposed limitFinal limit
Dioxin and Furans0.4 ng TEQ/dscmCarbon Monoxide/Total Hydrocarbons and DRE standards as surrogates
Total chlorine1.2 ppmv or 99.9994% system removal efficiency25 ppmv or 99.987% system removal efficiency

G. Dioxin/Furan Testing for Sources Not Subject to a Numerical Standard

Today's final rule requires that all sources not subject to a numerical dioxin and furan standard perform a one time test to determine their dioxin and furan emissions. See the discussion in Part Four, Section VII.L.

In the proposed rule, this requirement was limited to solid fuel boilers and those liquid fuel boilers with a wet or no air pollution control system. The final rule expands this requirement to include hydrochloric acid production furnaces and those lightweight aggregate kilns that elect to comply with the temperature limit at the kiln exit in lieu of the 0.20 ng TEQ/dscm dioxin/furan standard. Those sources are not subject to a numerical dioxin/furan standard under the final rule for reasons explained in Volume III of the Technical Support Document, Sections 12 and 15. We note that sources not subject to a numerical dioxin/furan emission standard are subject to the carbon monoxide or hydrocarbon standards and the DRE standard as surrogates.

We are making no changes to the implementation of this requirement. See the proposed rule at 69 FR at 21307 for more information.

III. Statistics and Variability

A. Using Statistical Imputation To Address Variability of Nondetect Values

In the final rule, we use a statistical approach to impute the value of nondetect emissions and feedrate measurements to avoid dampening of the variability of data sets when nondetect measurements are assumed to be present at the detection limit.

At proposal, we assumed that nondetects (i.e., HAP levels in stack emissions below the level of detection of the applicable analytic method) are invariably present at the detection limit. Commenters on the proposed rule stated, however, that assuming nondetects are present at the detection limit dampens emissions variability—a consideration necessary to reasonably ascertain sources' performance over time. This could have significant practical consequence for those data sets (such as the data base for liquid fuel boilers) dominated by nondetected values. We agree with these commenters, and instead of making the arbitrary assumption that all nondetected values are identical (which Start Printed Page 59424in fact is highly unlikely), we are using a statistical methodology to impute the value of nondetect measurements.

The imputation approach assigns a value for each nondetect measurement in a data set within the possible range of values that results in maximizing the 99th percentile upper prediction limit for the data set. For example, the possible range of values for a measurement that is 100% nondetect is between zero and the detection limit.

On February 4, 2005 we distributed a direct request for comments on the imputation approach to major stakeholders. We respond to the comments we received in Part Four, Section IV.D of today's notice.

B. Degrees of Freedom When Imputing a Standard Deviation Using the Universal Variability Factor for Particulate Matter Controlled by a Fabric Filter

The use of the universal variability factor to impute a standard deviation for particulate emissions from sources controlled with a fabric filter takes advantage of the empirical observation that the standard deviation of particulate emissions from sources is positively correlated to the average particulate emissions of sources. Based on this observation, we use regression analysis to determine the best fitting curve to explain the relationship of average value to standard deviation.

In the final rule, we use the actual sample size, rather than an assumed sample size of nine used at proposal, to determine the degrees of freedom for the t-statistic to calculate the floor using the standard deviation imputed from the universal variability factor for particulate matter controlled by a fabric filter.

At proposal, we used eight degrees of freedom to identify the t-statistic to account for within-test condition variability (i.e., run-to-run variability) for standard deviations imputed from the universal variability factor regression.[28] This is because, on average, about three test conditions with nine individual test runs are associated with each source used to develop the regression curve.

A commenter states, however, that this approach can dramatically understate variability when imputing a standard deviation for a source with only three runs because the t-statistic is substantially higher for 2 degrees of freedom than 8 degrees of freedom.

We agree with the commenter. Moreover, using the actual number of runs to identify the t-statistic rather than assuming nine runs is appropriate given that the true test condition average is less certain for sources with only three runs, and thus there is less certainty in the imputed standard deviation. The higher t-statistic associated with a three-run data set reflects this uncertainty.

In addition, we include emissions data classified as “normal” in the regression analysis for the final rule. At proposal, we used only data classified as CT (i.e., highest compliance test condition in a test campaign) or IB (i.e., a compliance test condition that achieved lower emissions than another compliance test condition in the test campaign). We conclude that normal data (i.e., emissions data that were not used to establish operating limits and thus do not reflect variability in controllable operating parameters) should also be considered in the regression analysis because particulate matter emissions are relatively insensitive to baghouse inlet loading and operating conditions.[29] Including normal emissions in the analysis provides additional data to better quantify these devices' performance variability.

IV. Compliance Assurance for Fabric Filters, Electrostatic Precipitators, and Ionizing Wet Scrubbers

The final rule provides additional requirements to clarify how you determine the duration of periods of operation when the alarm set point has been exceeded for a bag leak detection system or a particulate matter detection system:

1. You must keep records of the date, time, and duration of each alarm, the time corrective action was initiated and completed, and a brief description of the cause of the alarm and the corrective action taken.

2. You must record the percent of the operating time during each 6-month period that the alarm sounds.

3. In calculating the operating time percentage, if inspection of the fabric filter, electrostatic precipitator, or ionizing wet scrubber demonstrates that no corrective action is required, no alarm time is counted.

4. If corrective action is required, each alarm shall be counted as a minimum of 1 hour.

The final rule also establishes revised procedures for establishing the alarm set point if you elect to use a particulate matter detector system in lieu of site-specific operating parameter limits for compliance assurance for sources equipped with electrostatic precipitators and ionizing wet scrubbers. The rule explicitly allows you to maximize controllable operating parameters during the comprehensive performance test to account for variability by, for example, detuning the APCD or spiking ash. To establish the alarm set-point, you may either establish the set-point as the average of the test condition run average detector responses during the comprehensive performance test or extrapolate the detector response after approximating the correlation between the detector response and particulate matter emission concentrations. You may extrapolate the detector response up to a response value that corresponds to 50% of the particulate matter emission standard or 125% of the highest particulate matter concentration used to develop the correlation, whichever is greater. To establish an approximate correlation of the detector response to particulate matter emission concentrations you should use as guidance Performance Specification-11 for PM CEMS (40 CFR Part 60, Appendix B), except that you need conduct only 5 runs to establish the initial correlation rather than a minimum of 15 runs required by PS-11.

The final rule also notes that an exceedance of a detector response that corresponds to the particulate matter emission standard is not evidence that the standard has been exceeded because the correlation is an approximate correlation used for the purpose of compliance assurance to determine when corrective measures must be taken. The correlation, however, does not meet the requirements of PS-11 for compliance monitoring.

In addition, if you elect to use a particulate matter detection system in lieu of site-specific control device operating parameter limits on the electronic control device, the ash feedrate limit for incinerators and boilers under § 63.1209(m)(3) is waived. The ash feedrate limit is waived because the particulate matter detection system continuously monitors relative particulate matter emissions and the alarm set point provides reasonable assurance that emissions will not exceed the standard.[30]

Start Printed Page 59425

Finally, you must submit an excessive exceedance notification within 30 days of the date that the alarm set-point is exceeded more than 5 percent of the time during any 6-month block period of time, or within 30 days after the end of the 6-month block period, whichever is earlier. The proposed rule would have required you to submit that notification within 5 days of the end of the 6-month block period.

V. Health-Based Compliance Alternative for Total Chlorine

The final rule includes the following major changes to the proposed health-based compliance alternative for total chlorine:

(1) You must use 1-hour Reference Exposure Levels (aRELs) rather than 1-hour acute exposure guideline levels (AEGL-1) as the acute health risk threshold metric when calculating 1-hour HCl-equivalent emission rates;

(2) You must establish a long-term average chlorine feedrate limit (i.e., 12 hour rolling average or an (up to) annual rolling average) as the annual average HCl-equivalent emission rate limit divided by [1 − system removal efficiency]. You establish the total chlorine system removal efficiency during the comprehensive performance test. The proposed rule would have required you to establish the long-term average chlorine feedrate limit as the average of the test run averages of the comprehensive performance test.[31]

(3) At proposal, we requested comment on whether and how to establish a short-term chlorine feedrate limit to ensure that the acute exposure Hazard Index of 1.0 is not exceeded. See 69 FR at 21304. We conclude for the final rule that a 1-hour rolling average feedrate limit may be needed for some situations (i.e., if chlorine feedrates can vary substantially during the averaging period for the long-term feedrate limit and potentially result in an exceedance of the 1-hour average HCl-equivalent emission rate limit). Accordingly, although your eligibility for the health-based compliance alternatives is based on annual average HCl-equivalent emissions, you must determine considering prescribed criteria whether your 1-hour HCl-equivalent emission rate may exceed the national exposure standard (i.e., Hazard Index not exceeding 1.0 considering the maximum 1-hour average ambient concentration of hydrogen chloride and chlorine at an off-site receptor location[32] ) and thus may exceed the 1-hour average HCl-equivalent emission rate limit absent an hourly rolling average limit on the feedrate of chlorine. If the acute exposure standard may be exceeded, you must establish an hourly rolling average chlorine feedrate limit as the 1-hour HCl-equivalent emission rate limit divided by [1 − system removal efficiency]. You establish the system removal efficiency during the comprehensive performance test.

(4) When calculating HCl-equivalent emission rates, rather than partitioning total chlorine emissions between chlorine and HCl (i.e., the Cl2/HCl volumetric ratio) based on the comprehensive performance test as proposed, you must establish the Cl2/HCl volumetric ratio used to calculate the annual average HCl-equivalent emission rate based on the historical average ratio from all regulatory compliance tests. You must establish the Cl2/HCl volumetric used to calculate the 1-hour average HCl-equivalent emission rate as the highest of the historical ratios from all regulatory compliance tests. The rule allows you to exclude ratios from historical compliance tests where the emission data may not be representative of the current Cl2/HCl ratio for reasons such as changes to the design or operation of the combustor or biases in measurement methods. The rule also explicitly allows the permitting authority to require periodic emissions testing to obtain a representative average and maximum ratio;

(5) The look-up table analysis has been refined by presenting annual average and 1-hour HCl-equivalent emission rate limits as a function of stack height, stack diameter, and distance to property line. In addition, separate look-up tables are presented for flat terrain and simple elevated terrain;

(6) The proposed rule required approval of the eligibility demonstration before you could comply with the alternative health-based emission limits for total chlorine. Under the final rule, if your permitting authority has not approved your eligibility demonstration by the compliance date, and has not issued a notice of intent to disapprove your demonstration, you may nonetheless begin complying, on the compliance date, with the annual average HCl-equivalent emission rate limits you present in your eligibility demonstration. In addition, if your permitting authority issues a notice of intent to disapprove your eligibility demonstration, the authority will identify the basis for that notice and specify how much time you will have to submit additional information or to comply with the MACT total chlorine standards. The permitting authority may extend the compliance date of the total chlorine standards to allow you to make changes to the design or operation of the combustor or related systems as quickly as practicable to enable you to achieve compliance with the MACT total chlorine standards;

(7) We have revised the approach for determining chlorine emissions if you feed bromine or sulfur during the comprehensive performance test at levels higher than those specified in § 63.1215(e)(3)(ii)(B). Under the final rule, you must use EPA Method 320/321 or ASTM D 6735'01, or an equivalent method, to measure hydrogen chloride, and Method 26/26A, or an equivalent method, to measure chlorine and hydrogen chloride. You must determine your chlorine emissions to be the higher of: (1) The value measured by Method 26/26A, or an equivalent method; or (2) the value calculated by difference between the combined hydrogen chloride and chlorine levels measured by Method 26/26a, or an equivalent method, and the hydrogen chloride measurement from EPA Method 320/321 or ASTM D 6735-01, or an equivalent method; and

(8) The proposed rule would have required you to conduct a new comprehensive performance test if you planned to make changes to the facility that would lower the annual average HCl-equivalent emission rate limit. Under the final rule, you would be required to conduct a performance test as a result of a planned change only for a change to the design, operation, or maintenance of the combustor that could affect the system removal efficiency for total chlorine if the change could reduce the system removal efficiency, or if the change would increase the system removal efficiency and you elect to increase the feedrate limits on total chlorine and chloride.Start Printed Page 59426

Part Four: What Are the Responses to Major Comments?

I. Database

A. Revisions to the EPA's Hazardous Waste Combustor Data Base

Comment: Several commenters identify sources which have ceased operations as a hazardous waste combustor and should be removed from EPA's data base.

Response: We agree with commenters that data and information from sources no longer burning hazardous waste should not be included in our hazardous waste combustor data base and should not be used to calculate the MACT standards. We consider any source that has initiated RCRA closure procedures and activities as a source that is no longer burning hazardous waste. This data handling decision is consistent with the approach we used in the 1999 final rule. See 64 FR at 52844. As we stated in that rule, ample emissions data remain to support calculating the MACT standards without using data from sources that no longer burn hazardous waste.

As a result, we removed the following former hazardous waste combustors from the data base: the Safety-Kleen incinerator in Clarence, New York, the Dow Chemical Company incinerators in Midland, Michigan, and LaPorte, Texas, the two Holcim wet process cement kilns in Holly Hill, South Carolina, the Dow Chemical Company liquid fuel-fired boiler in Freeport, Texas, the Union Carbide liquid fuel-fired boilers in Hahnville, Louisiana, and Texas City, Texas, and six Dow Chemical Company hydrochloric production furnaces in Freeport, Texas.

We are retaining, however, Solite Corporation's lightweight aggregate facility in Cascade, Virginia, in the data base. Even though the facility recently initiated RCRA closure procedures, this data handling decision differs from those listed in the preceding paragraph because Solite Corporation provided this new information in February 2005 while information on the other closures was reported or available to us in 2004. Because we cannot continually adjust our data base and still finalize this rulemaking by the court-ordered deadline, we stopped making revisions to the data base in late 2004. Additional facility changes after that date, like Solite Corporation's Cascade facility closure, simply could not be incorporated.

Comment: One commenter identifies a source in EPA's data base that should be classified as a boiler instead of a hydrochloric acid production furnace.

Response: We agree with the commenter. In today's rule, Dow Chemical Company's boiler F-2820, located in Freeport, Texas, is reclassified in our data base as a boiler. This source is identified as unit number 2020 in our data base.

B. Use of Data From Recently Upgraded Sources

Comment: Many commenters recommend that EPA remove from the data base (or not consider for standards-setting purposes) emissions data from sources that upgraded their emissions controls to comply with the promulgated emission standards of either the 1999 rule or the 2002 interim standards. Several commenters also state that any emissions data that were obtained or used to demonstrate compliance with the promulgated standards of 1999 or 2002 should not be used for standard-setting purposes by the Agency. That is, EPA must evaluate the source category as it existed at the beginning of the rule development process and not after emissions controls are later added to comply with the 1999 or 2002 standards. Several commenters also state that EPA is only partly correct in claiming that the interim standards are not MACT standards because the interim standards were established and considered to be MACT until the Court issued its opinion in July 2001. Until that time, sources proceeded to upgrade their facilities to achieve the standards promulgated in 1999. The rationale for these recommendations is threefold: (1) Use of the data unfairly ignores the MACT-driven reductions already achieved by some sources; (2) it is contrary to sound public policy to use data from upgraded facilities to “ratchet down” the MACT floors to a level more stringent because these sources would not have increased their level of performance but for the legal obligation to comply with the standards; and (3) EPA's reliance on National Lime Ass'n v. EPA, 233 F.3d 625, 640 (D.C. Cir. 2000), for the proposition that the motivation for a source's performance is legally irrelevant in developing MACT floor levels is misplaced because that case involved the initial MACT standard setting process, and not a subsequent rule.

One commenter agrees with EPA's proposed position and states that use of data from sources that have upgraded is not only appropriate, but also required by the Clean Air Act. This commenter states that the actual performance of sources that have upgraded their emissions equipment—to meet the 1999 standards or for any reason—is reflected only by the most recently generated emissions data for the source. Thus, the Clean Air Act requires EPA to use the most recently generated data available to it and precludes the Agency from using older, out-of-date performance data.

EPA also received several comments stating that the language of section 112(d)(3)(A) of the Clean Air Act informs how the Agency should consider emissions data from sources that conducted testing after that 1999 rule was promulgated. One commenter states that the only data which should not be used in calculating the MACT floors are from sources that are subject to lowest achievable emission rates (LAER). Thus, the commenter states, Congress considered the possibility of significant and recent upgrades, and concluded that EPA should use up-to-date data to reflect source's performance, but must exclude certain sources from the floor calculation if their upgrades were of a specific degree and were accomplished within a specific period of time. Another commenter states that Congress did not intend to pile technology upon technology as confirmed by section 112(d)(3)(A) that specifically excludes sources that implemented LAER from consideration when establishing section 112(d) standards. Thus, the commenter states, considering data from sources that have upgraded violates both the language and intent of the Clean Air Act. Another commenter states that, while Congress no doubt contemplated that EPA should use all available emissions information in setting initial MACT standards, neither the statute nor the legislative history suggest that follow-up MACT rulemakings require the use of data reflecting compliance efforts with previous MACT standards or interim standards.

Response: As proposed, EPA maintains its position on use of post-1999 emissions data. The statute indicates that EPA is to base MACT floors on performance of sources “for which the Administrator has emissions information.” Section 112(d)(3)(A); CKRC, 255 F. 3d at 867. There can be no dispute that post-1999 performance data in EPA's possession fits this description. We also reiterate that the motivation for the control reflected in data available to us is irrelevant. See 69 FR at 21217-218. We further agree with those commenters who pointed out that Congress was explicit when it wanted certain emissions information (i.e., sources operating pursuant to a LAER standard) excluded from consideration in establishing floors. There is, of course, no such enumerated exception Start Printed Page 59427for sources that have upgraded their performance for other reasons.

We also do not agree with those commenters arguing (with respect to the standards for the Phase 1 sources (incinerators, cement kilns, and lightweight aggregate kilns)) in effect that the present rulemaking involves revision of an existing MACT standard. If this were indeed a revision of a MACT standard under section 112(d)(6), then EPA would not redetermine floor levels. See 70 FR at 20008 (April 15, 2005). However, EPA has not to date promulgated valid MACT floors or valid MACT standards for these sources. The 1999 standards do not reflect MACT, as held by the CKRC court. The interim standards likewise do not reflect MACT, but were designed to prevent a regulatory gap and were described as such from their inception. 67 FR at 7693 (Feb. 13, 2002); see also Joint Motion of all Parties for Stay of Issuance of Mandate in case no. 99-1457 (October 19, 2001), pp. 11-12 (“The Parties emphasize that the contemplated interim rule is in the nature of a remedy. It would not respond to the Court's mandate regarding the need to demonstrate that EPA's methodology reasonably predicts the performance of the average of the best performing twelve percent of sources (or best-performing source). EPA intends to address those issues in a subsequent rule, which will necessarily require a longer time to develop, propose, and finalize.”) EPA consequently believes that it is adopting in this rule the initial section 112(d) MACT standards for hazardous waste burning incinerators, cement kilns, and lightweight aggregate kilns, and that the floor levels for existing sources are based, as provided in section 112(d)(3), on performance of those sources for which EPA has “emissions information.”

However, we disagree with the comment that we must make exclusive use of the most recent information from hazardous waste combustion sources. There is no such restriction in section 112(d)(3). EPA has exhaustively examined all of the data in its possession for all source categories covered by this rule, and determined (and documented) which data are suitable for evaluating sources' performance.

C. Correction of Total Chlorine Data to Address Potential Bias in Stack Measurement Method

Comment: Several commenters state that EPA's proposed total chlorine standards of 1.5 ppm for existing incinerators and 0.18 ppm for new incinerators are based on biased data of indeterminate quality and are unachievable. Commenters assert that Method 26A and its RCRA equivalent, SW 846 Method 0050, have a negative bias at concentrations below 20 ppmv when used on stacks controlled with wet scrubbers. Commenters cite two recurring situations when this bias is likely to occur: (1) hydrogen chloride dissolving in condensed moisture in the sampling train; and (2) hydrogen chloride reacting with alkaline compounds from the scrubber water that are collected on the filter ahead of the impingers.

Commenters are particularly concerned about the negative bias associated with stack gas containing substantial water vapor. Commenters note that EPA found in a controlled laboratory study by Steger [33] that the bias is between 17 and 29 percent at stack gas moisture content of 7 to 9 percent. This stack gas moisture is much less than the nominal 50% moisture contained in some hazardous waste combustor stacks according to the commenters. Commenters believe this is why EPA's Method 0050, which was used to gather most of the data in the HWC MACT data base, states in Section 1.2 that “this method is not acceptable for demonstrating compliance with HCl emission standards less than 20 ppm.”

Moreover, commenters state that the procedures in Method 0050 to address the negative bias caused by condensed moisture were not followed for many RCRA compliance tests. The method uses an optional cyclone to collect moisture droplets, and requires a 45 minute purge of the cyclone and sampling train to recover hydrogen chloride from water collected by the cyclone and any condensed moisture in the train. The cyclone is not necessary if the stack gas does not contain water droplets. According to commenters, the cyclone and subsequent purge were often not used in the presence of water droplets because a potential low bias below 20 ppmv was irrelevant when demonstrating compliance with emission standards on the order of 100 ppmv. There was no need for the extra complexity and expense of using a cyclone and train purge given the purpose of the test. Although the data were acceptable for their intended purpose, commenters conclude that the data are not useful for establishing standards below 20 ppmv.

For these reasons, commenters suggest that EPA not consider total chlorine measurements below 20 ppmv when establishing the standards.

Response: For the reasons discussed below, we corrected all total chlorine measurements in our data base for all source categories that were below 20 ppmv to 20 ppmv to establish the total chlorine floors. Moreover, to address run-to-run variability given that all runs for several data sets are now corrected to 20 ppmv, we impute a run standard deviation based on a regression analysis of run standard deviation versus total chlorine concentration for sources with total chlorine measurements greater than 20 ppmv. This is the same approach we used to impute variability from sources using fabric filters when determining the particulate matter MACT floors.

Effect of Moisture Vapor. Commenters imply that stack gas with high levels of gas phase water vapor will inherently be problematic, particularly at emissions less than 20 ppmv. There is no basis for claiming that water vapor, per se, causes a bias in SW-846 Method 0050 or its equivalent, Method 26A. Condensed moisture (i.e., water droplets), however, can cause a bias because it can dissolve hydrogen chloride in the sampling train and prevent it from being captured in the impingers if the sampling train is not properly purged. Water droplets can potentially be present due to entrainment from the wet scrubber, condensation in cooler regions of the stack along the stack walls, and entrainment from condensed moisture dripping down the stack wall across the inlet duct opening.

Although Method 0050 addresses the water droplet issue by use of a cyclone and 45 minute purge, the Steger paper (Ibid.) concludes that a 45 minute purge is not adequate to evaporate all water collected by the cyclone in stacks with a total moisture content (vapor and condensed moisture) of 7 to 9%. At those moisture levels, Steger documented the negative bias that commenters reference. Steger's recommendation was to increase the heat input to the sample train by increasing the train and filter temperature from 120C (248F) to 200C (392F). We agree that increasing the probe and filter temperature will provide a better opportunity to evaporate any condensed moisture, but another solution to the problem is to require that the post-test purge be run long enough to evaporate all condensed moisture. That is the approach used by Method 26A, which EPA promulgated after Method 0050, and which sources must use to demonstrate compliance with the final standards. Method 26A uses an extended purge time rather than Start Printed Page 59428elevating the train temperature to address condensed moisture because that approach can be implemented by the stack tester at the site without using nonstandard equipment.

We attempted to quantify the level of condensed moisture in the Steger study and to compare it to the levels of condensed moisture that may be present in hazardous waste combustor stack gas. This would provide an indication if the bias that Steger quantified with a 45 minute purge might also be applicable to some hazardous waste combustors. We conclude that this comparison would be problematic, however, because: (1) given the limited information available in the Steger paper, it is difficult to quantify the level of condensed moisture in his gas samples; and (2) we cannot estimate the levels of condensed moisture in hazardous waste combustor stack gas because, even though condensed moisture may have been present during a test, method protocol is to report the saturation moisture level only (i.e., the amount of water vapor present), and not the total moisture content (i.e., both condensed and vapor phase moisture).

We can conclude, however, that, if hazardous waste combustor stack gas were to contain the levels of condensed moisture present in the gas that Steger tested, the 45 minute purge required by Method 0050 would not be sufficient to avoid a negative bias. We also conclude that this is potentially a practical issue and not merely a theoretical concern because, as commenters note, hazardous waste combustors that use wet scrubbers are often saturated with water vapor that will condense if the flue gas cools.

Data from Wet Stacks When a Cyclone Was Not Used. Commenters state that Method 0050 procedures for addressing water droplets (adequate or not, as discussed above) were not followed in many cases because a low bias below 20 ppmv was not relevant to demonstrating compliance with standards on the order of 100 ppmv. We do not know which data sets may be problematic because, as previously stated, the moisture concentration reported was often the saturation (vapor phase only) moisture level and not the total (vapor and liquid) moisture in the flue gas. We also have no documentation that a cyclone was used—even in situations where the moisture content was documented to be above the dew point. We therefore conclude that all data below 20 ppmv from sources controlled with a wet scrubber are suspect and should be corrected.

Potential Bias Due to Filter Affinity for Hydrogen Chloride. Studies by the American Society of Testing and Materials indicate that the filter used in the Method 0050 train (and the M26/26A trains) may adsorb/absorb hydrogen chloride and cause a negative bias at low emission levels. (See ASTM D6735-01, section 11.1.3 and “note 2” of section 14.2.3) This inherent affinity for hydrogen chloride can be satisfied by preconditioning the sampling train for one hour. None of the tests in our database were preconditioned in such a manner.

We are normally not concerned about this type of bias because we would expect the bias to apply to all sources equally (e.g., wet or dry gas) and for all subsequent compliance tests. In other words, we are ordinarily less concerned if a standard is based on biased data, as long as the means by which the standard was developed and the means of compliance would experience identical bias.

However, we did correct the wet gas measurements below 20 ppmv to address the potential low bias caused by condensed moisture. This correction would also correct for any potential bias caused by the filter's inherent affinity for hydrogen chloride. This results in a data set that is partially corrected for this issue—sources with wet stacks would be corrected for this potential bias while sources with dry stacks would not be corrected. To address this unacceptable mix of potentially biased and unbiased data (i.e., dry gas data biased due to affinity of filter for hydrogen chloride and wet gas data corrected for condensed moisture and affinity of filter for hydrogen chloride), we also correct total chlorine measurements from dry gas stacks (i.e., sources that do not use wet scrubbers).

Deposition of Alkaline Particulate on the Filter. Commenters are also concerned that hydrogen chloride may react with alkaline compounds from the scrubber water droplets that are collected on the filter ahead of the impingers. Commenters suggest this potential cause for a low bias at total chlorine levels below 20 ppmv is another reason not to use measurements below 20 ppmv to establish the standards.

Although alkaline particulate deposition on the method filter causing a negative bias is a much greater concern for sources that have stack gas containing high levels of alkaline particulate (e.g., cement kilns, sources equipped with dry scrubbers), we agree with commenters that this may be of concern for all sources equipped with wet scrubbers. Our approach to correct all data below 20 ppmv addresses this concern.

Decision Unique to Hazardous Waste Combustors. We note that the rationale for our decision to correct total chlorine data below 20 ppmv to account for the biases discussed above is unique to the hazardous waste combustor MACT rule. Some sources apparently did not follow Method 0050 procedures to minimize the low bias caused by condensed moisture for understandable reasons. Even if sources had followed Method 0050 procedures to minimize the bias (i.e., cyclone and 45 minute purge) there still may have been a substantial bias because of insufficient purge time, as Steger's work may indicate. We note that the total chlorine stack test method used by sources other than hazardous waste combustors—Method 26A—requires that the cyclone and sampling train be purged until all condensed moisture is evaporated. We believe it is necessary to correct our data below 20 ppmv data because of issues associated exclusively with Method 0050 and how it was used to demonstrate compliance with these sources.

Determining Variability for Data at 20 ppmv. Correcting those total chlorine data below 20 ppmv to 20 ppmv brings about a situation identical to the one we confronted with nondetect data. See Part Four, Section V.B. below. The MACT pool of best performing source(s) for some data sets is now comprised of largely the same values. This has the effect of understating the variability associated with these data.

To address this concern, we took an approach similar to the one we used to determine variability of PM emissions for sources equipped with a fabric filter. In that case, we performed a linear regression on the data, charting variability against emissions, and used the variability that resulted from the linear regression analysis as the variability for the sources average emissions. In this case, most or all of the incinerator and liquid fuel boiler sources in the MACT pool have average emissions at or near 20 ppmv. We therefore performed a linear regression on the total chlorine data charting average test condition results above 20 ppmv against the variability associated with that test condition. The variability associated with 20 ppmv was the variability we used for incinerator and liquid fuel boiler data sets affected by the 20 ppmv correction.

We also considered using the statistical imputation approach we used for nondetect values. See discussion in Section IV.B below. The statistical imputation approach for correcting data below 20 ppmv without dampening variability would involve imputing a value between the reported value and 20 Start Printed Page 59429ppmv because the “true” value of the biased data would lie in this interval. This approach would be problematic, however, given that many of the reported values were much lower than 20 ppmv; our statistical imputation approach would tend to overestimate the run to run variability. Consequently, we conclude that a regression analysis approach is more appropriate. A regression analysis is particularly pertinent in this situation because: (1) We consider data above 20 ppmv used to develop the regression to be unbiased; and (2) all the corrected data averages for which we are imputing a standard deviation from the regression curve are at or near 20 ppmv. Thus, any potential concern about downward extrapolation from the regression would be minimized.

We note that, although a regression analysis is appropriate to estimate run-to-run variability for the corrected total chlorine data, we could not use a linear regression analysis to address variability of nondetect values. To estimate a standard deviation from a regression analysis, we would need to know the test condition average emissions. This would not be feasible, however, because some or all of the run measurements for a test condition are nondetect. In addition, we are concerned that a regression analysis would not accurately estimate the standard deviation at low emission levels because we would have to extrapolate the regression downward to levels where we have few measured data (i.e., data other than nondetect). Moreover, the statistical imputation approach is more suitable for handling nondetects because the approach calculates the run-to-run variability by taking into account the percent nondetect for the emissions for each run.[34] A regression approach would be difficult to apply particularly in the case of test conditions containing partial nondetects or a mix of detect and nondetect values. Given these concerns with using a regression analysis to estimate the standard deviation of test conditions with runs that have one or more nondetect (or partial nondetect) measurements, we conclude that the statistical imputation approach best assures that the calculated floor levels account for run-to-run emissions variability.

Compliance with the Standards. The final standards are based on data that were corrected to address specific issues concerning these data. See the above discussion regarding stack gas moisture, filter affinity for hydrogen chloride, and alkaline compound reactions with hydrogen chloride in the sampling train.

Sources must demonstrate compliance using a stack test method that also addresses these issues. Sources with wet stacks must use Method 26A and follow those procedures regarding the use of a cyclone and the purging of the system whenever condensed moisture may be present in the sampling system.

Finally, all sources—those with either wet or dry gas—should precondition the sampling train for one hour prior to beginning the test to satisfy the filter's affinity for hydrogen chloride. The permitting authority will ensure that sources precondition the sample train (under authority of § 63.1209(g)(2)) when they review and approve the performance test plan.

D. Mercury Data for Cement Kilns

Comment: Several commenters state that EPA's data base of mercury emissions data (and associated feed concentrations of mercury in the hazardous waste) are unrepresentative and unsuitable for use in determining MACT standards for cement kilns. These comments are supported by an extensive amount of data submitted by the cement manufacturing industry including three years of data documenting day-to-day levels of mercury in hazardous waste fuels fired to all 14 hazardous waste burning cement kilns.[35] The commenters recommend that EPA use the commenter-submitted data as the basis for assessing cement kilns' performance for control of mercury because it is the most complete and representative data available to EPA.

Response: We agree that the commenter-submitted mercury data are more representative than those we used at proposal. First, these data represent a significantly larger and more comprehensive dataset compared to the one used to support the proposed mercury standard. The commenter-submitted data document the day-to-day levels of mercury in hazardous waste fired to all cement kilns for a three year period covering 1999 to 2001. In total, approximately 20,000 measurements of the concentration of mercury in hazardous waste are included in the dataset. When considered in whole, these data describe the performance (and variability thereof) of all cement kilns for the three year period because each measurement represents the mercury concentration in the burn tank used to fire the kiln over the course of a day's operation (or longer period).[36] In comparison, the data used to support the proposed floor level consisted of a much smaller dataset of approximately 50 test conditions representing a snapshot of performance somewhere in the range of normal operations, with each test condition representing a relatively short period of time (e.g., several hours).[37] As discussed at proposal, we were concerned regarding the representativeness of this smaller dataset. See 69 FR at 21251. In addition, the commenter-submitted dataset allows us to better evaluate the only mercury control technique used by existing hazardous waste burning cement kilns—controlling the feed concentration of mercury in the hazardous waste. The commenters have demonstrated convincingly that the mercury dataset used at proposal does not properly show the range of performance and variability in performance these cement kilns actually experience, while the significantly more robust dataset submitted by commenters does illustrate this variability. Thus, we conclude the larger commenter-submitted dataset is superior to EPA's smaller testing dataset.

We note that our MACT floor analysis of the commenter-submitted dataset to determine which sources are the best performers and to identify a mercury standard for cement kilns is discussed in the background document.[38] Additional discussion of issues related to the mercury standard for cement kilns is found in Part Four, Section VI.B of the preamble.

Start Printed Page 59430

E. Mercury Data for Lightweight Aggregate Kilns

Comment: One commenter, an owner and operator of seven of the nine operating lightweight aggregate kilns, states that the mercury dataset used by EPA at proposal is a limited and unrepresentative snapshot of performance of their seven kilns. To support their position that the snapshot emissions data are unrepresentative, the commenter submitted eight months of data documenting levels of mercury in hazardous waste fuels fired to their lightweight aggregate kilns.[39]

Response: We agree with the commenter that their mercury data submission is more representative than those used at proposal. As discussed in a notice for public comment sent directly to certain commenters,[40] the commenter-submitted dataset documents the day-to-day levels of mercury in hazardous waste fuels fired to Solite Corporation's Arvonia kilns between October 2003 and June 2004. The dataset consists of over 310 measurements of the concentration in mercury in hazardous waste. Each measurement represents the mercury concentration of the burn tank used to fire the kiln over the course of a day's operation (or longer period). In comparison, the data used to support the proposed floor level consisted of a smaller dataset of 15 test conditions.

The nature of the mercury data submitted by the commenter is the same as we received for the cement kiln category discussed in the preceding section. For similar reasons, we accept the more comprehensive commenter-submitted dataset as one that better shows the range of performance and variability in performance for these lightweight aggregate kilns. One notable difference, however, is that the commenter submitted mercury data only for its company (representing seven of nine lightweight aggregate kilns). Thus, we received no data documenting day-to-day levels of the concentration of mercury in hazardous waste fuels for the other two lightweight aggregate kilns owned by a different company. For these two lightweight aggregate kilns, we continue to use available data available in our database.[41]

Comment: One commenter opposes the use of the commenter-submitted mercury data because EPA would be uncritically accepting a limited and select data set from a commenter with a direct interest in the outcome of its use. Instead, the commenter suggests EPA use its section 114 authority to obtain all data that are available, not just the data selected by that commenter.

Response: We disagree that we uncritically accepted the commenter-submitted mercury data. The reason the commenter submitted data collected between October 2003 and June 2004 is that the facility was, prior to October 2003, in the process of upgrading its on-site analysis equipment. One outcome of this laboratory upgrade was its capability to detect mercury in hazardous waste at lower concentrations. Prior to the upgrade, the facility's on-site laboratory was capable of detecting mercury in the hazardous waste at a concentration of approximately 2 ppmw, which is a level such that the vast majority of measurements would neither be detected nor useful for identifying best performers and their level of performance.[42] The June 4, 2004 cutoff date represents a practicable date that measurements could still be incorporated into the commenter's public comments to the proposed rule, which were submitted on July 6, 2004. Finally, the commenter provided all waste fuel measurements during this period and states reliably that no measurements made during this period were selectively excluded.[43]

We also reject the commenter's suggestion that we use our authority under section 114 of the Clean Air Act to obtain additional hazardous waste mercury concentration data from the facility. There is no obligation for us to gather more performance data, given that the statute indicates that we are to base floor levels on performance of sources “for which the Administrator has emissions information.” Section 112(d)(3)(A); CKRC, 255 F. 3d at 867. In addition, given our concerns about the usefulness of measurements with high detection limits discussed above, the collection of additional data prior to the laboratory upgrade would not be productive. When balanced against the expenditure of significant resources, both in time and level of effort, to collect several more months of data, we conclude that obtaining additional mercury measurements is unnecessary because the available eight months of data—including over 310 individual measurements—represent a significant amount of data that we judge to be adequately reflective of the source's performance and variability in performance.

F. Incinerator Database

Comment: Commenters state that many of the top performers (e.g., 3011, 3015, 3022, 349) dilute emission concentrations in the stack by burning natural gas to initiate reactive waste (e.g., explosives, inorganic hydrides) or to decontaminate inert material. Commenters do not believe these units should be considered “representative” of the overall incinerator source category and should not be used to establish standards for incinerators combusting primarily organic wastes.

Response: Source 3022 has closed and has been removed from the database. Emission data from source #3015 (ICI explosives) has been excluded for purposes of calculating the particulate matter floor because the test report indicates this source was primarily feeding scrap metal, which we conclude to be an atypical waste stream from a particulate matter compliance perspective.[44]

The sources identified by the commenter are among the best performing sources in two instances. Source 3011 is the second ranked best performer for the particulate matter standard. This source is among the best performers for particulate matter because it uses a state-of-the art baghouse that is equipped with Teflon coated bags. There is no evidence to suggest that this source was diluting its particulate matter emissions. We acknowledge that we do not have ash feed data for the test conditions that were used in the particulate matter standard analysis. However, this source had the third and fourth highest metal feed control levels among all the sources used in the MACT analysis for the semivolatile and low volatile metal Start Printed Page 59431standards.[45] We therefore conclude that it is appropriate to include this source in the MACT analysis that determines the relevant best performers for particulate matter.

Source 349 is the eighth ranked (out of 11) best performer for the particulate matter standard. We acknowledge that the ash feed level for this source is lower than most incinerators equipped with baghouses. However, particulate matter emissions from sources equipped with baghouses are not significantly affected by the ash inlet loading to the baghouse.[46] This is further supported by the fact that this source is ranked eighth among the best performers. We conclude source 349 is a best performer not because of its relatively low ash feed level, but rather because it is equipped with a well designed and operated baghouse. It is therefore appropriate to include this source in the MACT analysis.

Comment: Commenters state that source 341 should not be considered in the MACT analysis because it is a small laboratory waste burner that processes only 900 lbs/hr of waste. Commenters claim that more than 80 percent of the waste profile is non-hazardous waste.

Response: We approached this comment by asking if it would be appropriate to create a separate subcategory for source 341. We conclude it is not necessary to subcategorize hazardous waste incinerators based on the size of combustion units. This is because the ranking factors used to identify the relevant best performing sources are normalized in order to remove the influence that combustion unit size would otherwise have when identifying best performing sources. See part 4 section III.D below. Air pollution control system types (a ranking factor for particulate matter) are generally sized to match the corresponding volumetric gas flow rate in order to achieve a given control efficiency. The size of the combustor therefore does not influence a source's ability to achieve a given control efficiency. System removal efficiency and hazardous waste feed control MTECs (ranking factors used by the SRE/Feed methodology as described in part 4 section III.B below) are also not influenced by the size of the combustor.[47]

Emission limitations are similarly normalized to remove the influence of combustion unit size by expressing the standards as emission concentration limits rather than as mass emission rate limits. See section III.D. This is illustrated in the following example. Assume there are two cement kilns side by side with similar designs, the only difference being one is twice the size of the other, producing twice as much clinker. They both have identical types of air pollution control systems (the larger source is equipped with a larger control device that is appropriately sized to accommodate the larger volumetric gas flow rates and achieves the same control efficiency as the smaller control device). If we were to assess performance based on HAP mass emission rates (e.g., pounds per hour), the smaller source would be the better performer because its mass emission rates would be half of the mass emission rate of the larger source, even though they both are achieving the same back-end control efficiency. Emission concentrations, on the other hand, are calculated by dividing the HAP mass emission rate (e.g., pounds per hour) by the volumetric gas flowrate (e.g., cubic feet per hour). In the above example, both sources would have identical HAP emission concentrations (the larger source has twice the mass emission rate, but twice the volumetric gas flow rate), accurately reflecting their identical control efficiency. Emission concentrations normalize the size of each source by accounting for volumetric gas flowate, which is directly tied to the amount of raw material each source processes (and subsequently the amount of product that is produced). This is a reason we point out that normalization eliminates the need to create subcategories based on unit size. See part four section III.D.

Further, it would be difficult to determine an appropriate minimum size cutoff in which to base such a subcategorization determination. Such a subcategorization scheme could also yield nonsensical floor results, as was the case when we assessed subcategorizing commercial incinerators and on-site incinerators.[48]

We have identified source 341 as the best performing source for particulate matter and low volatile metals. It is the single best performing source for these standards because it is equipped with a state-of-the-art baghouse.[49] This source, which simultaneously feeds hazardous and nonhazardous wastes, conducted several emission tests that reflected different modes of operation. The amount of nonhazardous waste that was processed in the combustion unit varied across test conditions. We could not ascertain the exact amount of hazardous waste processed in the test condition that was used in the MACT analysis for low volatile metals because the test report stated the wastes that were processed were a mixture of hazardous and nonhazardous wastes, although we estimate that at least 26% of the waste processed was nonhazardous.[50] We note that we are aware of several other incinerators that processed nonhazardous waste at levels greater than 26 percent during their emission tests. We therefore do not believe this to be atypical operation that warrants subcategoriztion.

Moreover, the fact that this source was feeding nonhazardous wastes does not result in atypically low hazardous waste low volatile metal feed control levels, as evidenced by the relative feed control ranking for this source of thirteenth among the 26 sources assessed in the MACT analysis. It also has the highest normalized hazardous waste feed control level among the best performing sources, and has the fifth best low volatile metal system removal efficiency among those same 26 sources. We repeat that this source is being identified as the best performing source primarily because it is equipped with a highly efficient baghouse, not because it is feeding low levels of HAP metals attributable to its hazardous waste.

Furthermore, this source is not the lowest emitting source in the database. There are two sources with similar, but slightly lower low volatile metal compliance test emissions (one commercial incinerator and one onsite, non-commercial incinerator). This provides further evidence that the Start Printed Page 59432emissions from this source appropriately represent emissions of a relevant best performing source.

Regarding the particulate matter standard, source 341 does not have atypically low ash feed rates as compared to other sources equipped with baghouses. Out of the nine best performing particulate matter sources for which we have ash feed information, this source ranks fourth (a ranking of one is indicative of the lowest ash feed rate). Nonetheless, as previously discussed, particulate matter emissions from sources equipped with baghouses are not significantly affected by the ash inlet loading to the baghouse. We note that particulate matter emissions from the second and third best performing source are not significantly different from this source, providing further evidence that this source is representative of the range of emissions exhibited by other well designed and operating incinerators equipped with baghouses.[51]

Comment: Commenters state that sources 3018 and 3019 are identified as best performers for mercury emissions for incinerators. After evaluating the trial burn plans for these sources, the commenter believes the data should not be used to calculate the MACT floor because the spiking rate for mercury was extremely low for a compliance test. The ranking for feedrate is therefore unrepresentative. The commenter suggests that these test results should be characterized as “normal”.

Response: We have verified that the emission tests performed for sources 3018 and 3019 reflect the upper range of mercury emissions that are not to be exceeded by these sources, and that their spiked mercury feed rates were back-calculated from a risk assessment. We therefore conclude that we properly characterized these emissions as compliance test emissions data because they reflect the emissions resulting from the upper bound of hazardous waste mercury feedrates from these sources.[52] Consequently, these data are properly included with the other data used to calculate floor standards for mercury for incinerators.

Comment: Commenters state the trial burn plan for sources 3018 and 3019 describes these units to be of similar design. Thus the difference in results between these two similar sources is indicative of additional variability above and beyond the run-to-run variability and should be assessed if the data are deemed usable at all.

Response: We conclude both of these sources are in fact unique sources that should be assessed as individual sources for purposes of the MACT analysis. Although these sources are of similar design, we do not believe they are identical, in part because: (1) The facility itself conducted separate emission tests for the two units (rather than trying to avail itself of the ‘data in lieu’ option, which could save it the expense of a second compliance test, the obvious inference being that the source or regulatory official regards the two units as different); and (2) discussions with facility representatives indicated these units are similar, but not identical.[53] As a result, it would be inappropriate to assess emissions variability by combining the emissions of these two sources into one test condition given they are not identical units.

Comment: Commenters state that emissions data from source 327 should not be used to calculate dioxin/furan and mercury floors because they claim the carbon injection system did not appear to function properly during the test.

Response: We agree with the commenters. We have determined that this source encountered problems with its carbon injection system during the emissions test from which the data were obtained and subsequently used in EPA's proposed MACT analysis. We have also verified that this source did not establish operating parameter limits for the carbon injection system as a result of this test.[54] We therefore have excluded this mercury and dioxin data from the MACT analysis, and have instead used emissions data from an older test condition to represent this source's emissions.

Comment: Commenters state that the emissions data from source 3006 were based on a miniburn to determine how close the unit was to achieving the interim MACT standards. The commenter questions whether these data should be used for purposes of calculating MACT standards.

Response: The fact that a source conducts a voluntary emissions test (e.g., a miniburn) to determine how close it is operating to upcoming emission standards does not necessarily lead us to conclude that the emission data are inappropriate for purposes of calculating MACT standards. However, since proposal, we have determined that this source did not measure cadmium emissions during this emissions test. As a result, we conclude the semivolatile metal emissions data from this source should not be used in the MACT standard calculation for semivolatile metals because the data do not represent the source's combined emissions of lead and cadmium.

II. Affected Sources

A. Area Source Boilers and Hydrochloric Acid Production Furnaces

Comment: Five commenters state that the area sources subject to the proposed rule are negligible contributors to 112(c)(6) HAP emissions and should not be subject to major source standards for 112(c)(6) HAP. Commenters note that requiring compliance with MACT for 112(c)(6) HAP and RCRA for other toxic pollutants is more complicated and burdensome for sources than complying only with RCRA. Although an area source can choose to become regulated as a major source in order to reduce some RCRA requirements, they would become subject to more onerous emissions limits under Subpart EEE and the other MACT requirements.

One of these commenters states that subjecting an area source to major source standards under 112(c)(6) sends a negative message to industry that EPA does not value emissions reduction and/or chemical substitution, or other methods used by area sources to achieve that status. EPA is no longer providing any incentive for sources to take such difficult yet environmentally beneficial steps to become an area source. Imposing Title V permitting requirements on an entire facility that operates as an area source of hazardous air pollutants (HAPs) will impose an unfair and undue burden on the facility.

Another of these commenters states that section 112(c)(6) requires in pertinent part that EPA list categories and subcategories of sources assuring that sources accounting for not less than 90% of the aggregate emissions of each pollutant (specified in 112(c)(6)) are subject to standards under Section 112(d)(2) or (d)(4). In 1998, EPA published a notice identifying the list of source categories accounting for the section 112(c)(6) HAP emissions and to be regulated under section 112(d) to meet the 90% requirement. (63 FR 17838) At the time, EPA acknowledged that MACT standards for a number of the source categories had not yet been promulgated, and stated that when the Start Printed Page 59433regulations for each of those categories are developed, EPA will analyze the data specific to those sources and determine, under Section 112(d), in what manner requirements will be established. EPA also stated that:

“Some area categories may be negligible contributors to the 90% goal, and as such pose unwarranted burdens for subjecting to standards. These trivial source categories will be removed from the listing as they are evaluated since they will not contribute significantly to the 90% goal.” (63 FR 17841)

The commenter believes the “two or fewer” area source boilers identified by EPA in the present rulemaking are “negligible contributors” to the 90% goal and therefore, should not be required to adopt the same MACT emission limitations and requirements as major sources of the 112(c)(6) pollutants. The commenter believes EPA's decision to subject area source boilers and hydrochloric acid production furnaces is incorrect, unsupported by the administrative record, and therefore arbitrary and capricious.

One commenter states that, if EPA regulates area sources, it should significantly reduce the administrative burden for area sources by: exempting them from Title V provisions for Subpart EEE requirements; exempting them from compliance with the General Provisions of 63 Subpart A; limiting them to a one-time comprehensive performance test; or limiting other applicable requirements.

Response: We continue to believe that boiler and hydrochloric acid furnace area sources warrant regulation under the major source MACT standards for mercury, dioxin/furan, carbon monoxide/hydrocarbons, and destruction and removal efficiency pursuant to section 112(c)(6).

As discussed at proposal (69 FR at 21212), section 112(c)(6) of the CAA requires EPA to list and promulgate section 112(d)(2) or (d)(4) standards (i.e., standards reflecting MACT) for categories and subcategories of sources emitting seven specific pollutants. Five of those listed pollutants are emitted by boilers and hydrochloric acid production furnaces: mercury, 2,3,7,8-tetrachlorodibenzofuran, 2,3,7,8-tetrachlorodibenzo-p-dioxin, polycyclic organic matter, and polychlorinated biphenyls.

As discussed below, EPA must assure that source categories accounting for not less than 90 percent of the aggregated emissions of each enumerated pollutant are subject to MACT standards (and of course is not prohibited from requiring more than 90 percent of aggregated emissions to be controlled by MACT standards). Congress singled out the pollutants in section 112(c)(6) as being of “'specific concern”' not just because of their toxicity but because of their propensity to cause substantial harm to human health and the environment via indirect exposure pathways (i.e., from the air through other media, such as water, soil, food uptake, etc.). Furthermore, these pollutants have exhibited special potential to bioaccumulate, causing pervasive environmental harm in biota and, ultimately, human health risks.

Section 112(c)(6) of the CAA requires EPA to list categories and subcategories of sources of seven specified pollutants to assure that sources accounting for not less than 90 percent of the aggregate emissions of each such pollutant are subject to standards under CAA section 112(d)(2) or 112(d)(4). In 1998, EPA issued the list of source categories pursuant to section 112(c)(6), and that list is published at 63 Fed. Reg. 17838, 17849, Table 2 (April 10, 1998).

In the 1998 listing, EPA identified the following three subcategories of the HWC source category that emit one or more of the seven section 112(c)(6) pollutants: (1) Hazardous waste incinerators—(emit mercury, dioxin, furans, polycyclic organic matter (POM) and polychlorinated biphenyls (PCBs)); (2) Portland cement manufacture: hazardous waste kilns—(emit mercury, dioxin, furans, and POM); and (3) lightweight aggregate kilns: hazardous waste kilns—(emit dioxin, furans, and mercury). These three subcategories are all subject to today's rule, which is issued pursuant to CAA section 112(d)(2). As explained below, the HWC NESHAP effectively controls emissions of the identified section 112(c)(6) pollutants from the identified subcategories. Accordingly, EPA considers the sources in these three subcategories as being “subject to standards” for purposes of section 112(c)(6).

Specifically, with regard to hazardous waste-burning incinerators, cement kilns, and lightweight aggregate kilns, EPA is adopting in this final rule MACT standards for mercury and dioxins/furans. EPA has already adopted MACT standards for control of POM and PCBs emitted by these sources in the 1999 rule, which standards were not reopened or reconsidered in this rulemaking. These standards are the CO/HC standards, which in combination with the Destruction Removal Efficiency (DRE) requirement, assure that these sources operate continuously under good combustion conditions which inhibit formation of POM and PCBs as combustion by-products, or destroy these HAP if they are present in the wastes being combusted.[55] See discussion in Part Four, Sections V.A and V.B of this preamble.

The HWC NESHAP also applies to hazardous waste-burning boilers and hydrochloric acid production furnaces. In particular, for these boilers and furnaces, this rule addresses emissions of dioxin/furan, mercury, POM and PCBs either through specific numeric standards for the identified HAP, or through standards for surrogate pollutants which control emissions of the identified HAP.

We estimate that approximately 620 pounds of mercury are emitted annually in aggregate from hazardous waste burning boilers in the United States.[56] Also, we estimate that hazardous waste burning boilers and hydrochloric acid production furnaces emit in aggregate approximately 2.3 and 0.2 grams TEQ per year of dioxin/furan, respectively. Controlling emissions of these HAP from area sources consequently reduces emissions of these HAP through application of MACT standards. We note that only major source boilers and hydrochloric acid furnaces are subject to the full suite of subpart EEE emission standards.[57] Section 112(c)(3) of the CAA requires us to subject area sources to the full suite of standards applicable to major sources if we find “a threat of adverse effects to human health or the environment” that warrants such action. We cannot make this finding for area source boilers and halogen acid production furnaces. 69 FR at 21212. Consequently, as proposed, area sources in these categories would be subject only to the MACT standards for mercury, dioxin/furan, and polycyclic Start Printed Page 59434organic matter and polychlorinated biphenyls (through the surrogate standards for carbon monoxide/hydrocarbons and destruction and removal efficiency) to control the HAP enumerated in section 112(c)(6). RCRA standards under Part 266, Subpart H for particulate matter, metals other than mercury, and hydrogen chloride and chlorine gas would continue to apply to these area sources unless an area source elects to comply with the major source standards in lieu of the RCRA standards. See § 266.100(b)(3) and the revisions to §§ 270.22 and 270.66.

Commenters refer to the “two or fewer” potential area source boilers we identified at proposal as “negligible contributors” and, therefore, conclude that these area sources should not be subject to major source standards for emission of these HAPs. Commenters did not quantify the amount of emissions from area sources, and did not even identify how many area sources are at issue. We do not know how many boilers and hydrochloric acid furnaces are area sources. We apparently underestimated the number given that four companies commented on the proposed rule saying that area sources should not be subject to major source standards for mercury, dioxin/furan, PCBs, and polycyclic organic matter, and one of those companies indicates it operates multiple area sources. Consequently, we continue to believe that area sources in these categories may have the potential to emit more than negligible levels of these HAP.

We also note that the major source standards are tailored to minimize the compliance burden for sources that emit low levels of HAP. Commenters raise concerns about applying the major source standards for HAP enumerated in section 112(c)(6) to liquid fuel boiler area sources. The emission standard compliance burden for liquid fuel boilers that have the potential to emit only low levels of mercury, dioxin/furan, and polycyclic organic matter is minimal. For example, sources that emit low levels of mercury because their feedstreams have low levels of mercury can elect to comply with the mercury emission standard by documenting that the mercury in feedstreams will not exceed the standard assuming zero removal by emission control equipment. We note that 75% of the liquid fuel boilers in our data base, and the two boilers cited by commenters, do not have emission control devices.

The compliance burden for the major source standards for dioxin/furan and for the surrogates to control other polycyclic organic matter—carbon monoxide/hydrocarbons and destruction and removal efficiency (DRE)—should also be minimal for area source liquid fuel boilers. The dioxin/furan standard applicable to the 90% of liquid fuel boilers with wet or no air pollution control equipment is compliance with the carbon monoxide/hydrocarbon standard and the DRE standard. Liquid fuel boilers already comply with these same standards under RCRA. The surrogate standards to control other polycyclic organic matter are also the carbon monoxide/hydrocarbon and DRE standards. Finally, we note that the DRE requirement under Subpart EEE is less burdensome than the DRE requirement under RCRA. Under Subpart EEE, a source needs to conduct a one-time only DRE test, provided that design and operation does not change in a manner than could adversely affect DRE. Under RCRA, the DRE test must be conducted each time the RCRA permit is renewed.

The incremental compliance burden associated with the other Subpart EEE major source requirements, such as the operations and maintenance plan, the startup, shutdown, and malfunction plan, operator training, and the automatic waste feed cutoff system should also be minimal for liquid fuel boilers without an emission control device. In addition, most of the requirements are either identical to or very similar to requirements under RCRA with which these area sources are already complying.[58]

B. Boilers Eligible for the RCRA Low Risk Waste Exemption

Comment: Several commenters state that EPA should exempt those boilers that qualify as Low Risk Waste Exemption (LRWE) burners under the RCRA Boiler and Industrial Furnace Rule at § 266.109 from the MACT particulate matter and destruction and removal efficiency (DRE) standards because EPA has not: (1) Made a demonstration that the data used to provide the exemption to low risk burners under RCRA is no longer valid; or (2) established in the affirmative that regulating these units will provide any benefit to human, health and the environment. Commenters believe that regulating LRWE units under Subpart EEE is unnecessary and inconsistent with RCRA subtitle C and more importantly, appears to be controlling LRWE units for control's sake.

Commenters also state that EPA has not properly addressed the requirements of CAA section 112(n)(7) regarding the inconsistency between the requirements for Low Risk Waste Exempt (LRWE) units under RCRA and those of Subpart EEE. The purported purpose of section 112(n)(7) is to allow EPA to avoid imposing additional emission limitations on a source category subcategory when such limitations would be unnecessary and duplicative.

In addition, commenters state that the costs associated with this MACT are much more than improved feed control or better back-end control. This proposed rule also requires substantial dollar investment in improved data acquisition, computer controls and recordkeeping systems, performance testing, training, development of plans, and other regulatory requirements.

Response: Boilers and hydrochloric acid production furnaces that currently qualify for the RCRA § 266.109 low risk waste exemption are not exempt from Subpart EEE under the final rule.

The Administrator does not have the authority under CAA section 112(d) to exempt sources that comply with RCRA § 266.109. Indeed, there is no necessary connection between the two provisions, since one is technology-based and the other is risk-based. CAA section 112(d)(2) requires the Administrator to establish technology-based emission standards, standards that require the maximum degree of reduction in emissions that is deemed achievable. Although section 112(d)(4) gives the Administrator the authority to establish health-based emission standards in lieu of the MACT standards for pollutants for which a health threshold has been established, we cannot use that authority to develop health-based standards for sources that comply with RCRA § 266.109 because those sources emit HAP for which a health threshold has not been established.

The final rule complies fully with CAA section 112(n)(7) by coordinating applicability of the RCRA and CAA requirements and precluding dual requirements. For example, RCRA requirements that are duplicative of MACT requirements will be removed from the RCRA operating permit when the permitting authority issues a certification of compliance after the source submits a Notification of Compliance.

We also note that the MACT standards are tailored to impose Start Printed Page 59435minimal burden on sources that have low emissions of HAP. The particulate matter emission standard and associated testing can be waived (similar to the § 266.109 exemption) for boilers that elect to document that emissions of total metal HAP do not exceed the limits provided by § 63.1206(b)(14). Hydrochloric acid production furnaces are not subject to a particulate matter emission standard.

The compliance burden with the destruction and removal efficiency (DRE) standard is also minimal given that it is a one-time test, provided that the source does not change its design or operation in a manner that would adversely affect DRE. In addition, the compliance burden for sources with low levels of metals in their feedstreams is minimal. Sources can document compliance with the metals emission standards by assuming all metals in the feed are emitted (i.e., by assuming zero system removal efficiency). Under this procedure, boilers burning relatively clean wastes are not required to conduct a performance test to document compliance with the metals emission standards.

Further, we note that the MACT standard to control organic HAP emissions other than dioxin/furan is the same as the RCRA standard—demonstrating good combustion conditions by complying with a carbon monoxide standard of 100 ppmv.

Finally, we note that the ancillary requirements under MACT (e.g., personnel training; operating and maintenance plan; startup, shutdown, and malfunction plan) should not pose substantially higher costs than similar requirements under RCRA. See response to comment in Section A above. To the extent that compliance costs increase, we have accounted for those costs in our estimates of the cost of the final rule.[59]

C. Mobile Incinerators

Comment: A mobile incinerator used as a directly-fired thermal desorption unit at a Superfund remediation site should not be an affected source under this rule.

Response: EPA is not determining or changing the applicability of any hazardous waste burning unit under today's rule. A combustion unit that treats hazardous waste and meets the definition of incinerator at 40 CFR 260.10 is an affected source under this rule. 40 CFR part 63 also defines a source as any building, structure, facility, or installation which emits or may emit any air pollutant. A mobile incinerator at a remediation site meets this definition.

Comment: One commenter states that a subcategory with different standards must be created for mobile incinerators, or the standards for incinerators must be calculated using actual emissions data from mobile units.

Response: EPA did not have any emissions data from mobile incinerators in the database for the proposed rule. That data base was developed over many years with ample opportunity for public comment. We developed a data base for incinerators to support the 1996 proposed rule (61 FR 17358) and noticed that data base for public comment on January 7, 1997 (64 FR 52828). We updated that data base in July 2002, and noticed the revised data base for public comment (67 FR 44452). We used that revised data base to support the proposed rule. We did not receive comments providing data for mobile incinerators as a result of either public notice.

One commenter on the proposed rule provided a summary of emissions data from one test at a mobile incinerator. The commenter suggested that the data support its view that its mobile incinerator is unique and that EPA should consider subcategorizing incinerators according to mobile incinerators versus other incinerators. We analyzed these data and conclude that the final standards are readily achievable by this source. Moreover, as explained elsewhere, EPA's approach to assess the need for subcategorization is to apply a statistical test to determine whether the emissions data are statistically different from the remaining group. Given that owners and operators of mobile incinerators have not provided emissions data prior to proposal, and that the commenter provides summarized data for only one mobile incinerator (which also indicate that the source can achieve the emission standards in the final rule); we are not compelled to gather additional information, particularly given our time constraints to promulgate the final rule under a court-ordered deadline.

Comment: In support of subcategorizing mobile incinerators, commenters state that mobile thermal treatment systems are substantially different from hazardous waste incinerators. They are much smaller in size, firing capacity rate, refractory lining, and operating temperatures. Most of them treat contaminated soil, so have very high particulate feedrate loading with high ash content, rapid kiln rotation rate, and counter-current flow design like cement kilns. This results in high particulate matter emissions. They operate only for a short duration at a site (usually less than 6 months), and have no flexibility with regard to their waste feed.

Response: We recognize that there is variability between various sources' with regard to size, capacity, operating temperatures etc., and so we applied a statistical test to assess the need of subcategorization, as has been discussed above. The emissions data provided by the commenter also indicate the source can achieve the final standards. The soil entrained in desorber off-gases of mobile incinerators has a relatively large particle size, and is very easy to capture with conventional particulate control systems (such as a fabric filter) used by the incinerators.

Comment: Since mobile incinerators are relocated from site to site, the new source standard should not apply based on the erection date of the mobile unit.

Response: We are not changing the applicability of a new or reconstructed source designation in this rulemaking. The relocation issue is addressed in the definition of “construction” in 40 CFR Section 63.2, which states: “Construction does not include the removal of all equipment comprising an affected source from an existing location and the reinstallation of such equipment at a new location * * *” (emphasis added). Therefore, the relocation of an existing Subpart EEE affected source, such as a mobile incinerator, would not result in that mobile incinerator becoming a “new” source. Keep in mind also that the relocation exemption only applies to affected sources. If a mobile incinerator is relocated from an R&D facility (where the unit is not an affected source per Table 1 to Section 63.1200) to a location where the mobile incinerator would become an affected source, the relocation exemption within the definition of “construction” would not apply and the mobile incinerator would be a “new” source. Also, with regard to leased sources, the owner/operator of the facility is responsible for all affected sources operating at his/her facility regardless of whether the sources are owned or leased. The owner or operator should obtain from the leasing company all relevant information pertaining to the affected source in order to be able to demonstrate that the affected source is operating in compliance with the appropriate standards.

III. Floor Approaches

In this section we discuss comments addressing methodologies used in this rule for determining MACT floors. We address comments relating both to Start Printed Page 59436general, overarching issues and to the specific methodologies used in the rule. Our most important point is that the methodologies EPA selected reasonably estimate the performance of the best performing sources by best accounting for these sources' total variability.

A. Variability

1. Authority To Consider Emissions Variability

Comment: Many commenters concur with our approach to account for emissions variability while several commenters believe that our approach does not adequately account for emissions variability. See discussions on separate topics below. One commenter, however, states that use of variability factors (however derived) is inherently unlawful and arbitrary and capricious. The commenter notes that, because floors for existing sources must reflect the “average” emission level achieved by the relevant best performing sources, they cannot reflect any worse levels of performance from the best performers. Indeed, the argument is that the Clean Air Act already accounts for variability by requiring EPA to base existing source floors on the average emission level achieved by the best performing sources.

The commenter continues by stating that EPA has added variability factors both to each individual source's performance and to the collective performance of the alleged best performers, in each case purporting to find an emission level that the individual or group would meet ninety-nine times out of 100 future emission tests. Thus, EPA ignores sources' measured performance in favor of the theoretical worst performance that might ever be expected from them. By looking to the best performers' worst performance rather than their average performance, EPA would set weaker floors than the Clean Air Act allows.

In addition, the commenter notes that EPA's approach to account for emissions variability is arbitrary and capricious because EPA never explains why it chose the 99th percentile for its variability adjustments rather than some other percentile.

Finally, the commenter notes that EPA appears to indicate that its variability analysis would either be applied to variation between sources or would affect EPA's statistical analysis of the variation between sources. The commenter states that any attempt by EPA to add a variability factor to adjust for intersource variability is unlawful and arbitrary and capricious.

Response: Our response explains our approach to estimating best performing sources' variability and addresses the following issues: (1) Considering the variability in each source's performance is necessary to identify the best performing sources and their level of performance; (2) EPA reasonably considered variability in ranking sources to identify the best performers and in considering the range of best performing sources' performance over time to identify an emission level that the average of those sources can achieve; (3) considering variability at the 99th percentile level is reasonable; (4) considering intersource variability by pooling run-to-run variability is appropriate; and (5) compliance test conditions do not fully reflect all of best performing sources' performance variability.

a. Variability Must Be Considered. Variability in each source's performance must be considered at the outset in identifying the best performing sources. This is simply another way of saying that best performers are those that perform best over time (i.e. day-in, day-out), a reasonable approach. This approach not only reasonably reflects the statutory language, but also furthers the ultimate objective of section 112 which is to reduce risk from exposure to HAP. Since most of the risk from exposure to emissions from this source category is associated with chronic exposure to HAP (see Part 1 section VI above), assessing a source's performance over time by accounting for variability is reasonable and appropriate.

For similar reasons, variability must be considered in ascertaining these sources' level of performance. Floors for existing sources must reflect “the average emission limitation achieved by the best performing 12 percent” of sources, and for new sources, must reflect “the emission control that is achieved in practice by the best controlled source.” Section 112 (d) (3). EPA construes these requirements as meaning achievable over time, since sources are required to achieve the standards at all times. This interpretation has strong support in the case law. See Sierra Club v. EPA, 167 F. 3d 658, 665 (D.C. Cir. 1999), stating that “EPA would be justified in setting the floors at a level that is a reasonable estimate of the performance of the ‘best controlled similar unit’ under the worst reasonably foreseeable circumstances. It is reasonable to suppose that if an emissions standard is as stringent as ‘the emissions control that is achieved in practice’ by a particular unit, then that particular unit will not violate the standard. This only results if ‘achieved in practice’ is interpreted to mean ‘achieved under the worst foreseeable circumstances'; see also National Lime Ass'n v. EPA, 627 F. 2d 416, 431 n. 46 (D.C. Cir. 1980) (where a statute requires that a standard be ‘achievable,’ it must be achievable under “the most adverse circumstances which can reasonably be expected to recur”);

The court has further indicated that EPA is to account for variability in assessing sources' performance for purposes of establishing floors, and stated that this assessment may require EPA to make reasonable estimates of performance of best performing sources. CKRC, 255 F. 3d at 865-66; Mossville Environmental Action Now v. EPA, 370 F. 3d 1232, 1242 (D.C. Cir. 2004)(maximum daily variability must be accounted for when establishing MACT floors).[60] Indeed, EPA's error in CKRC was not in estimating best performing sources' variability, but in using an unreasonable means of doing so. CKRC, 255 F. 3d at 866; Mossville, 370 F. 3d at 1241.

Since the emission standards in today's rule must be met at all times, the standards need to account for performance variability that could occur on any single day of these sources' operation (assuming proper design and operation). See Mossville, 370 F. 3d at 1242 (upholding MACT floor because it was established at a level that took into account sources' long term performance, not just performance on individual days). Moreover, since EPA's database consists of single data points (because there are no continuous emission monitors for HAPs in stack emissions), EPA must of necessity estimate long-term performance, including daily maximum performance, from this limited set of short term data.

b. EPA Reasonably Considered Variability in Ranking Sources to Identify the Best Performers and in Considering the Range of Best Performing Sources' Performance Over Time to Identify an Emission Level that the Average of Those Sources Can Achieve. (1) Selecting Best Performing Sources. Each of the floor methodologies used in the rule considers various factors in ranking which sources are the best performing. For each methodology, we therefore consider the quantifiable variability of Start Printed Page 59437the ranking factors in determining which are the best performing sources. 69 FR at 21230-31. Specifically, we assess run-to-run variability (normally the only type of variability which we can quantify) of the factors used under each methodology to rank best performers. Where SRE/Feed is the ranking methodology, we thus assess run-to-run variability of hazardous waste HAP feedrate and of system removal efficiency. Where ranking is based on sources' emissions (the straight emissions methodology), we assess the run-to-run variability of emission levels. Where we use the air pollution control device methodology for ranking, we assess the run-to-run variability of emissions of the lowest-emitting sources (as we do for straight emissions) using the best air pollution control devices. For hydrochloric acid production furnaces, we assess the run-to-run variability of total chlorine system removal efficiency. Id.[61]

To account for run-to-run variability in these ranking factors, we rank sources by the 99th percentile upper prediction limit (UPL99). The UPL99 is an estimate of the value that the source would achieve in 99 of 100 future tests if it could replicate the operating conditions of the compliance test. Id. at 21231.

(2). Assessing the Best Performers' Level of Performance Over Time. Once we identify the best performing sources, we need to consider their emissions variability to establish a floor level that the average of the best performing sources can achieve day-in, day-out. There are two components of emissions variability that must be considered: run-to-run variability and test-to-test variability. Run-to-run emissions variability encompasses variability in individual runs comprising the compliance tests, and includes uncertainties in correlation of monitoring parameters and emissions, and imprecision of stack test methods and laboratory analyses. See 69 FR at 21232.[62] Test-to-test emissions variability is the variability that exists between multiple compliance tests conducted at different times and includes the variability in control device collection efficiency caused by testing at different points in the maintenance cycle of the emission control device [63] , and the variability caused by other uncontrollable factors such as using a different stack testing crew or different analytical laboratory, and by different weather conditions (e.g., ambient moisture and temperature) that may affect measurements.

We are able to quantify run-to-run variability. We do so by applying a 99th percentile modified upper prediction limit to the averaged emissions of the best performing sources. Id. at 21233 and Technical Support Document Volume III section 7.2. The modified upper prediction limit accounts for run-to-run variability of the best performers by pooling their run variance (i.e., within-test condition variability).[64] See Chemical Manufacturer's Ass'n v EPA, 870 F. 2d 177, 228 (5th Cir. 1989) (upholding use of a variability factor derived, as here, by pooling the performance variability of the best performing plants). Using this approach, we ensure that the average of the best performing sources will be able to achieve the floor in 99 of 100 future performance tests, assuming these best performing sources could replicate their performance when attempting to operate under identical conditions to those used for the compliance test establishing the source as best performing. As just noted, we call this value the modified UPL 99.

The only instance in which we are able to quantify test-to-test variability (as noted above, the other significant component of total operating variability) is for fabric filters (baghouses) when used to control emissions of particulate matter. The modified UPL 99 in these instances reflects not only run-to-run variability, but test-to-test variability as well. That total variability is expressed by the Universal Variability Factor which is derived from analyzing long-term variability in particulate matter emissions for best performing sources across all of the source categories sources that are equipped with fabric filters. 69 FR at 21233. See also the discussion below in Section III.A.2.

Test-to-test variability must be accounted for in other instances as well, however. It follows that if the performance of most efficient fabric filters varies over time relative to particulate matter emissions, then so does their performance relative to the non-mercury metal HAP emissions. We also believe that particulate matter emissions variability from sources equipped with back-end controls other than fabric filters also exists, and is furthermore likely to be higher than what was calculated for fabric filters because there are more uncertainties associated with the correlations between operating parameter limits and control efficiency for these devices.[65] Again, it clearly follows that if the performance of these other control devices varies relative to particulate matter emissions (perhaps even more than what has already been quantified for fabric filters), then so does their performance relative to the non-mercury metal HAP emissions.

Although we cannot quantify this test-to-test variability, we can document its existence and its significance. We conducted two parallel analyses examining all situations where we had multiple test conditions for the sources ranked as best performing performing (examining separate pools for best performing sources under both the straight emissions and SRE/feed ranking methodologies). These analyses showed that these sources' emissions do in fact vary over time, sometimes significantly. In many instances sources had poorer system removal efficiencies and higher emission levels than those in the compliance test used to identify the source as best performing. We further projected that in many instances these best performing sources would not achieve their own UPL 99, the statistically determined prediction limit which captures 99 out of 100 future three-run test averages for the source, if they were to operate at the poorer system removal efficiency of its earlier test and used the federate of its later (best-performing) compliance test. This is significant because the UPL 99 reflects all of a source's run-to-run Start Printed Page 59438variability. Failure to meet the UPL 99 thus shows both that further variability exists, namely test-to-test variability, and that it is a significant component of total variability. We obtained similar results when we projected best performing sources' performance based on each of these sources' overall system removal efficiency obtained by pooling the removal efficiencies of all of its tests. In many instances, moreover, these projected levels exceeded floor levels calculated by using the straight emissions approach, which ranks best performers as those with the lowest emission levels. This point is discussed further in Section III.B below. EPA's analysis is set out in detail in chapters 16 and 17 of Volume III of the Technical Support Document.[66]

EPA's conclusion is that total variability includes both run-to-run and test-to-test variability, and that both must be accounted for in determining which are the best performing sources and what are their levels of performance over time. As explained in the following Sections B and C, EPA has accordingly adopted floor methodologies which account for this total variability either quantitatively or qualitatively. The approach advocated by the commenter simply ignores that variability exists. Since this approach is contrary to both fact and law, EPA is not adopting it.

c. Quantifying Run-to-Run Variability at the 99th Percentile Level Is Reasonable. We selected the 99% prediction limit to ensure a reasonable level “ namely the 99th percentile—of achievability for sources designed and operated to achieve emission levels equal to or better than the average of the best performing sources.[67] Because of the randomness of the emission values, there is an associated probability of the average of the best performing sources, and similarly designed and operated sources, not passing the performance test conducted under the same conditions.[68] At a 99% confidence level, the average of the best performing sources could expect to achieve the floor in 99 of 100 future performance tests conducted under the same conditions as its performance test.. The commenter thus sharply mischaracterizes a 99% confidence level as the worst performance of a best performing source.: the level in fact assumes identical operating conditions as those of the performance test.

EPA routinely establishes not-to-exceed standards (daily maximum values which cannot be exceeded in any compliance test) using the 99% confidence level. National Wildlife Federation v. EPA, 286 F. 3d 554, 572 (D.C. Cir. 2002).[69] At a confidence level of only 97% for example, the average of the best performing sources could expect to achieve the floor in only 97 of 100 future performance tests.

We note that the choice of a confidence level is not a choice regarding the stringency of the emission standard. Although the numerical value of the floor increases with the confidence level selected it only appears to become less stringent. If EPA selected a lower confidence interval, we would necessarily adjust the standard downward due to the expectation that a source would not be expected to achieve the standard for uncontrollable reasons a larger per cent of the time. We would then have to account in some manner for this inability to achieve the standard. See Weyerhaeuser v. Costle, 590 F. 2d 1011, 1056-57 (D.C. Cir. 1978) (also upholding standards established at 99 % confidence level). The governing issue is what level of confidence should the average of the best performing sources, and similarly designed and operated sources, have of passing the performance test demonstrating compliance with the standard. We believe that the 99% confidence level is a confidence level within the range of values we could have reasonably selected.[70]

d. Considering Intersource Variability by Pooling Run-to-Run Variability is Appropriate. The commenter believes that any attempt by EPA to add a variability factor to adjust for intersource variability is unlawful and arbitrary and capricious. We see no statutory prohibition in considering intersource run-to-run variability of the best performing sources (which is all our floor calculation does, by considering the pooled run-to-run variability of the best performing sources). Section 112(d)(3) states that MACT floors are to reflect the “average emission limitation achieved” but does not specify any single method of ascertaining an average. Considering the average run-to-run variability among the group of best performing sources is well within the language of the provision (and was upheld in CMA, as noted above; see 870 F. 2d at 228). The commenter's further argument that ‘average' can only mean average of emission levels achieved in performance tests is inconsistent with the holding in Mossville, 370 F. 3d at 1242, that EPA must account for variability in developing MACT floors and that individual performance tests do not by themselves account for such variability.

We believe that it is reasonable and necessary to account for intersource variability of the best performing sources by taking the pooled average of the best performing sources' run-to-run variability. This is an aspect of identifying the average performance of those sources. Emissions data for each best performing source are random in nature, and this random nature is characterized by a stochastic distribution. The stochastic distribution is defined by its central tendency (average value) and the amount of dispersion from the point of central tendency (variance or standard deviation). Consequently, to define the performance of the average of the best performing sources, we must consider the average of the average emissions for the best performing sources as well as the pooled variance for those sources. Hence, we must consider intersource variability to identify the floor—the average performance of the best performing sources.

The commenter further states that EPA's attempt to adjust for intersource variability is unlawful, arbitrary, and capricious. EPA set floors at the 99th percentile worst emission level that it believed any source within the group of best performers could achieve, according to the commenter. The 99th percentile worst performance that could be expected from a source within the best performers is, simply put, not the average performance of the sources in that group, according to the commenter.

The commenter misunderstands our approach to calculate the floor—the floor is not the 99th percentile highest emission level that any best performing source could achieve. The floor for Start Printed Page 59439existing sources is calculated as the 99th percentile modified upper prediction limit of the average of the best performing sources. It represents the average of the best performing sources' emissions levels plus the pooled within-test condition variance of the best performing sources. The floor for existing sources is not the highest 99th percentile upper prediction limit for any best performing source as the commenter states.

e. Why isn't Total Variability Already Accounted for by Compliance Test Conditions?

Comment: One commenter states that EPA's use of variability factors along with worst-case data is unlawful and arbitrary and capricious. EPA has stated that its use of worst case “compliance” data accounts for variability. EPA admits that compliance data reflect special worst case conditions created artificially for the purpose of obtaining lenient permit limits, according to the commenter. EPA provides no reason whatsoever to believe that a source would continue to operate under such conditions even one percent of the time. Thus, the commenter concludes, by applying a 99 percent variability factor to compliance test data, EPA ensures that the adjusted data do not accurately reflect the performance of any source. Accordingly, EPA's use of a variability factor is unlawful.

The commenter also states that, to increase compliance data with the reality that sources will not be operating under the worst case conditions except during permit setting tests, the Agency's use of a variability factor with compliance data is arbitrary and capricious.

Response: All but two standards in the final rule are based on compliance test data—when sources maximized operating parameters that affect emissions to reflect variability of those parameters and to achieve emissions at the upper end of the range of normal operations. Use of these data is appropriate both because they are data in EPA's possession for purposes of section 112(d)(3) and because these data help account for best performing sources' operating variability. CKRC, 255 F. 3d at 867.

The main thrust of the comment is that total variability is accounted for by the conditions of the performance test, so that making further adjustments to allow for additional variability is improper. The commenter believes that the floor should be calculated simply as the average emissions of the best performing sources and that this floor would encompass the range of operations of the average of the best performing sources. We disagree.

The compliance test is designed to mirror the outer end of the controllable variability occurring in normal operations. These controllable factors include the amount of HAP fed to a source in hazardous waste, and controllable operating parameters on pollution control equipment (such as power input to ESPs, or pressure drop across wet scrubbers, factors which are reflected in the parametric operating limits written into the source's permit and which are based on the results of the compliance testing). However, this is plainly not all of the variability a source experiences. Other components of run-to-run variability, including variability relating to measuring (both stack measurements and measurements at analytic laboratories) are not reflected, for example. Nor is test-to-test variability reflected, notably the point in the maintenance cycle that testing is conducted and the variability associated with those inherently differing test conditions even though the source attempts to replicate the test conditions (e.g., measurement variability attributable to use of a different test crew and analytical laboratory and different weather conditions such as ambient temperature and moisture). Other changes that occur over time are due to a wide variety of factors related to process operation, fossil fuels, raw materials, air pollution control equipment operation and design, and weather. Sampling and analysis variations can also occur from test to test (above and beyond those accounted for when assessing within-test variability) due to differences in emissions testing equipment, sampling crews, weather, and analytical laboratories or laboratory technicians.

Thus, there is some need for a standard to account for this additional variability, and not simply expect for a single performance test to account for it. The analyses in Sections 16 and 17 of Volume III of the Technical Support Document confirm these points.

Moreover, the best performing sources (and the average of the best performers) must be able to replicate the compliance test if they are to be able to continue operating under their full range of normal operations. It is thus no answer to say that the best performing sources could operate under a more restricted set of conditions in subsequent performance tests and still demonstrate compliance, so that there is no need to assure that results of initial performance tests can be replicated. To do so would no longer allow the best performing sources (and thus the average of the best performing sources) to operate under their full range of normal operations, and thus impermissibly would fail to account for their total variability.

As discussed throughout this preamble, emissions variability—run-to-run and test-to-test variability—is real and must be accounted for if a best performing source is to be able to replicate the emissions achieved during the initial compliance test. We consequently conclude that we must account for variability in establishing floor levels, and that merely considering the average of compliance test data fails to do so. We have therefore quantified run-to-run variability using standard statistical methodologies, and accounted for test-to-test variability either by quantifying it (in the case of fabric filter particulate matter removal performance) or accounting for it qualitatively (in the case of the SRE/feed ranking methodology).

Comment: The commenter notes that if EPA believes that single performance test results do not accurately capture source's variability, the solution is to gather more data, not to avoid using a straight emissions methodology. EPA cannot use this as an excuse for basing floor levels on a chosen technology rather than the performance of the best performing sources.

Response: There is no obligation for EPA to gather more performance data, since the statute indicates that EPA is to base floor levels on performance of sources “for which the Administrator has emissions information.” Section 112(d)(3)(A); CKRC, 255 F. 3d at 867 (upholding EPA's decision to use the compliance test data in its possession in establishing MACT standards). Indeed, the already-tight statutory deadlines for issuing MACT standards would be even less feasible if EPA took further time in data gathering. EPA notes further that because particulate matter continuous emission monitors are not widely used, even further data gathering would be limited to snapshot, single performance test results, still leaving the problem of estimating variability from a limited data set.[71] See also Sierra Club v. EPA, 167 F. 3d at 662 (“EPA typically has wide latitude in determining the extent of data-gathering necessary to solve a problem”).

Thus, EPA has no choice but to assess best performers and their level of performance on the basis of limited amounts of data per source. As explained in the previous response to Start Printed Page 59440comments, EPA has selected a methodology that reasonably do so.

EPA notes further that it has carefully examined those instances where there are multiple test conditions (usually compliance tests conducted at different times) for sources ranked as best performing. This analysis confirms EPA's engineering judgment that total variability is not fully encompassed in the single test condition results used to identify these sources as best performing, and that without taking this additional variability into account, best performing sources would be unable to achieve the floor standard reflecting their own performance in those single test conditions.[72]

2. Universal Variability Factor for Particulate Emissions Controlled with a Fabric Filter

Comment: One commenter states that, in calculating the universal variability factor (UVF) to account for total variability—test-to-test variability and within-test variability—for sources controlling particulate matter with a fabric filter, it appears that EPA considered the variability of sources that are not best performing sources. If so, EPA has contravened the law.

The commenter also states that EPA's attempt to use a variability factor derived from an analysis of variability of multiple sources is unlawful. If EPA considers variability at all, it must consider the relevant source's variability.

Response: We developed the particulate matter UVF for sources equipped with a fabric filter using data from best performing sources only.[73]

It is reasonable to aggregate particulate matter emissions data across source categories for all best performing sources equipped with a fabric filter because the relationship between standard deviation and emissions of particulate matter is not expected to be impacted by the source category type.[74] Rather, particulate emissions from fabric filters are a function of seepage (i.e., migration of particles through the filter cake) and leakage (i.e., particles leaking through pores, channels, or pinholes formed as the filter cake builds up). The effect of seepage and leakage on emissions variability should not vary across source categories.[75] Put another way, fabric filter particulate matter reduction is relatively independent of inlet loadings to the fabric filter. 69 FR 21233. This is confirmed by the fact that there are no operating parameters that can be readily changed to increase emissions from fabric filters, id., so control efficiencies reflected in test conditions from different source types will still accurately reflect fabric filter control efficiency.

3. Test-to-Test Variability

Comment: Several commenters state that EPA seems to have ignored test-to-test variability resulting from changes that occur over time such as: normal and natural changes in a wide variety of factors related to process operation, fuels, raw materials, air pollution control equipment operation and design, and differences in emissions testing equipment, sampling crews, weather, analytical laboratories or laboratory technicians. All these sources of variation are expected in that they are typical and are not aberrations. In addition, there are unexpected sources of variability that occur in real-world operations, which also must be accommodated according to commenters.

Commenters state that using compliance test data and assessing within-test condition variability (i.e., run variance) do not fully account for test-to-test variability and thus understates total variability. Consequently, the average of the best performing sources may not be able to achieve the same emission level under a MACT performance test when attempting to operate under the same conditions as it did during the compliance test EPA used to establish the floor. Even though sources generally operated at the extreme high end of the range of normal operations during the compliance tests EPA uses to establish the standards, the average of the best performing sources would need to operate under those same compliance test conditions to establish the same operating envelope—the operating envelope needed to ensure the source can operate under the full range of normal emissions.

Response: We agree with commenters that we have not quantified test-to-test variability when establishing the floors for standards other than particulate matter where a best performing source uses a fabric filter. We are able to quantify only within-test variability (i.e., run-to-run variability) for the other floors, which is only one component of total variability. This is one reason we use the SRE/Feed approach wherever possible rather than a straight emissions approach to rank the best performing sources to calculate the floor—the SRE/Feed ranking approach derives floors that better estimate the levels of best performing sources' performance. See also discussion in Part Four, Section III.A, and the discussion below documenting that test-to-test variability can be substantial.

Comment: One commenter states that EPA should use the universal variability factor (UVF) that accounts for total variability for particulate matter controlled with a fabric filter to derive a correction factor to account for the missing test-to-test variability component of variability for semivolatile metals and low volatile metals. The commenter then suggests that the within-test variability for semivolatile and low volatile metals be adjusted upward by the correction factor to correct for the missing test-to-test variability component.

The commenter focused on cement kilns and compared the total variability imputed from the UVF for the three cement kiln facilities used to establish the UVF to the within-test variability (i.e., run variance) for each facility. The commenter determined that, on average for the three facilities, total variability was a factor of 4.2 higher than within-test variability. Because semivolatile and low volatile metals are also controlled with a fabric filter, the commenter suggested that the total variability of particulate matter could be used as an estimate of the total variability for semivolatile and low volatile metals. Thus, the commenter suggested that the within-test condition variability for semivolatile and low volatile metals be increased by a factor of 4.2 to account for total variability when calculating floors.

Response: As stated throughout this preamble, we believe that there is variability in addition to within-test condition (i.e., run-to-run) variability that we cannot quantify—that we refer to as test-to-test variability. We also do not believe this test-to-test variability is captured by compliance test operating conditions as discussed above, and thus establishing the floor using emissions data representing the extreme high end of the range of normal emissions does not account for test-to-test variability. We disagree, however, with the commenter's attempts to quantify the remaining test-to-test variability for floors other than particulate matter Start Printed Page 59441where all best performing sources are equipped with fabric filters.

We generally agree with the commenter's approach for extracting the test-to-test component of variability using the UVF curve for particulate matter controlled with a fabric filter.[76] The commenter has documented that for cement kilns, test-to-test variability of particulate emissions controlled with a fabric filter is on average a factor of 4.2 higher than within-test variability.

We believe the commenter's suggestion to adopt this correction factor to semivolatile and low volatile metals is technically flawed and for several reasons would present statistical difficulties. First, total variability for semivolatile metals and low volatile metals controlled with a fabric filter can be different from the total variability of particulate matter controlled with a fabric filter because: (1) The test methods are different (i.e., Method 5 for particulate matter and Method 29 for metals) and thus sample extraction and analysis methods differ; (2) the factors that affect partitioning of particulate matter to combustion gas (i.e., entrainment) are different from the factors that affect semivolatile metal partitioning to the combustion gas (i.e., metal volatility); and (3) the volatility of semivolatile metals is affected by chlorine feedrates.

Second, adopting a variability factor applicable to fabric filters for use on electrostatic precipitators [77] is problematic because both test-to-test and within-test variability of these emission control devices can be vastly different. Factors that affect emissions variability for sources equipped with a fabric filter include: (1) Bag wear and tear due to thermal degradation and chemical attack; and (2) variability in flue gas flowrate. Factors that affect emissions variability for sources equipped with an electrostatic precipitator are different (see discussion in Section III.B above) and include: variations in particle loading and particle size distribution, erosion of collection plates, and variation in fly ash resistivity due to changes atmospheric moisture and in sulfur feedrate (e.g. different type of coal).

Finally, the approach raises several difficult statistical questions including: (1) What is the appropriate number of runs to use to identify the degrees of freedom and the t-statistic in the floor calculations (e.g., should we use the number of runs available for metals emissions for the source or the number of runs available for particulate matter emissions from which the correction factor is derived); and (2) should we use a generic correction factor for all source categories or calculate source category-specific or source-specific correction factors.

For these reasons, we believe the approach we use for quantifying baghouse particulate matter collection variability is not readily transferable to other types of control devices and other HAP. We therefore are not applying a quantified correction factor in the final rule but rather are using a MACT ranking methodology that qualitatively accounts for total emission variability, notably test-to-test variability.

B. SRE/Feed Methdology

1. Description of the Methodology

As proposed, we are using the System Removal Efficiency (SRE)/Feed approach to determine the pool of best performing sources for those HAP whose emissions can be controlled in part by controlling the hazardous waste feed of the HAP—that is, controlling the amount of HAP in the hazardous waste fed to the source. These are HAP metals and chlorine. Our basic approach is to determine the sources in our database with the lowest hazardous waste feedrate of the HAP in question (semi-volatile metals, low volatile metals, mercury, or chlorine), and the sources with the best system removal efficiency for the same HAP. The system removal efficiency is a measure of the percentage of HAP that is removed prior to being emitted relative to the amount fed to the unit from all inputs (hazardous waste, fossil fuels, raw materials, and any other input). The pool of best performing sources are those with the best combination of hazardous waste feedrate and system removal efficiency as determined by our ranking procedure, separate best performer pools being determined for each HAP in question (SVM, LVM, mercury, and chlorine), reflecting the variability inherent in each of these ranking factors (see A.2.a.(1) above). We then use the emission levels from these sources to calculate the emission level achieved by the average of the best performing sources, as also explained in the previous section. This is the MACT floor for the HAP from the source type. For new sources, we use the same methodology but select the emission level (adjusted statistically to account for quantifiable variability) of the source with the best combined ranking. A more detailed description of the methodology is found in Volume III of the Technical Support Document, section 7.3.

This methodology provides a reasonable estimate of the best performing sources and their level of performance for HAP susceptible to hazardous waste feed control. As required by section 112(d)(2), EPA has considered measures that reduce the volume of emissions through process changes, or that prevent pollutant release through capture at the stack, and assessed how these control measures are used in combination. Section 112(d)(2)(A), (C) and (E). Hazardous waste feed control is clearly a process change that reduces HAP emissions; air pollution control systems collect pollutants at the stack. These are the best systems and measures for controlling HAP emissions from hazardous waste combustors. 69 FR at 21226. In considering these factors, EPA has necessarily considered such factors as design of different air pollution control devices, waste composition, pollution control operator training and behavior, and use of pollution control devices and methodologies in combination. CKRC, 255 F. 3d at 864-65 (noting these as factors, in addition to a particular type of air pollution control device, that can influence pollution control performance); 69 FR at 21223 n. 47 (system removal efficiency measures all internal control mechanisms as well as back-end emission control device performance).

EPA also believes that this methodology reasonably estimates the best performing sources' level of performance by accounting for these sources' total variability, including their performance over time. The methodology quantifies run-to-run variability. See 69 FR at 21232-33. It does not quantify test-to-test variability because we are unable to do so for these pollutants. (See sections A. 2.a.(2) and 3 above.) Although all variability must be accounted for when calculating floors, the only definitive way to accurately quantify this test-to-test emissions variability is through evaluation of long-term continuous emissions monitoring data, which do not presently exist. We believe, however, that SRE/Feed methodology provides some margin for estimating this additional, non-quantifiable variability. This is illustrated in the technical support document (volume III section 17), which clearly shows that the straight emissions approach underestimates (indeed, fails to account Start Printed Page 59442for) lower emitting sources' long-term emissions variability. These lower emitting sources that would otherwise not meet the floor levels on individual days under the straight emission approach would be able (or otherwise are more capable) to do so under the SRE/feed approach.

EPA further believes that the SRE/Feed methodology appropriately accounts for design variability that exists across sources for categories, like those here, which consist of a diverse and heterogeneous mixture of sources. This is especially true of incinerators and boilers, for which there are smaller on-site units that are located at widely varying industrial sectors that essentially combust single, or multiple wastestreams that are specific to their industrial process, and off-site commercial units dealing with many different wastes of different origins and HAP metal and chlorine composition. EPA believes that these variations are best encompassed in the SRE/Feed approach, rather than with a subcategorization scheme that could result in anomalous floor levels because there are fewer sources in each source subcategory from which to assess relative performance.[78] See Mossville, 370 F. 3d at 1240 (upholding floor methodology involving reasonable estimation, rather than use of emissions data, when sources in the category have heterogeneous emission characteristics due to highly variable HAP concentrations in feedstocks).

Use of the SRE/Feed approach also avoids basing the floor standards on a combination of the lowest emitting low feeding sources and the lowest emitting high feeding sources. For example, the five lowest emitting incinerators for semivolatile metals that would comprise the MACT pool using a straight emissions methodology include three sources that are the first, second, and fourth lowest feeding sources among all the incinerators.[79] The other two best performing incinerators have the first and second best system removal efficiencies (and the highest two metal feedrates). It is noteworthy that the highest feed control level among these best performing sources is over three orders of magnitude higher than the feed control level of the lowest feeding best performing source.[80] Establishing limits dominated by both superior feed control sources and back-end controlled sources would result in floor levels that are not reflective of the range of emissions exhibited by either low feeding sources or high feeding sources and would more resemble new source standards for both of these different types of combustors. Such floors could lead to situations, for example, where commercial sources could find it impracticable to achieve the standards without reducing the overall scope of their operations (since the standard could operate as a direct constraint on the amount of hazardous waste that could be fed to the device, in effect depriving a combustion source of its raw material). Similarly, low feeding sources that cannot achieve this floor level may be required to add expensive back-end control equipment that would result in minimal emission reductions, likely forcing the smaller on-site source to cease hazardous waste treatment operations and to instead send the waste to a commercial treatment unit.

The inappropriateness of a straight emissions-based approach for feed controlled pollutants for commercial hazardous waste combustors is further highlighted by the fact that several commercial hazardous waste combustors that are achieving the design level of the particulate matter standard are not achieving the semivolatile and/or low volatile metals straight emissions based design level, and, in some instances, floor level.[81] This provides further evidence that low feeding sources are in fact biasing some of the straight emissions-based floors to the extent that even the sources with the most efficient back-end control devices would be incapable of achieving the emission standards calculated on a straight emission basis.

These results are inconsistent with the intent of the section 112 (d) (see 2 Legislative History at 3352 (House Report) stating that MACT is not intended to drive sources out of business). Standards that could force commercial sources to reduce the overall scope of their operations are also inconsistent with requirements and objectives of the Resource Conservation and Recovery Act to require treatment of hazardous wastes before the wastes can be land disposed, and to encourage hazardous waste treatment. RCRA sections 3004 (d), (e), (g) and 1003 (a) (6); see also section 112 (n) (7) of the CAA, stating that section 112 (d) MACT standards are to be consistent with RCRA subtitle C emission standards for the same sources to the maximum extent practicable (consistent with the requirements of section 112 (d)); moreover, EPA doubts that a standard which precludes effective treatment mandated by a sister environmental statute must be viewed as a type of best performance under section 112 (d). The SRE/Feed methodology avoids this result by always considering hazardous waste feed control in combination with system removal efficiency and according equal weight to both means of control in the ranking process.

It is also important to emphasize what the SRE/Feed methodology does not evaluate: Feed control of HAP in fossil fuel or raw material inputs to these devices. Emission reduction of these HAP are controllable by back-end pollution control devices which remove a given percentage of pollutants irrespective of their origin and is assured by the system removal efficiency portion of the methodology, as well as through the particulate matter standard (see section IV.A below). Feed control of these inputs is not a feasible means of control, however. HAP content in raw materials and fossil fuel can be highly variable, and so cannot even be replicated by a single source. Raw material and fossil fuel sources are also normally proprietary, so other sources would not have access to raw material and fossil fuel available (in its performance test) to a source with low HAP fossil fuel and raw material inputs. Such sources would thus be unable to duplicate these results. Moreover, there are no commercial-scale pretreatment processes available for removing or reducing HAP content in raw materials or fossil fuels to these units. See technical support document volume III section 17.5 and 25; see also 69 FR at 21224 and n. 48.

2. Why Aren't the Lowest Emitters the Best Performers?

Some commenters nonetheless argue that best performing sources can only mean sources with the lowest HAP Start Printed Page 59443emissions, and that the SRE/Feed methodology is therefore flawed because it does not invariably select lowest emitters as best performers.[82] The statute does not compel this result. There is no language stating that lowest emitting sources are by definition the best performers. The floor for existing sources is to be based on the average emission limitation achieved by the “best performing” 12 per cent of sources. Section 112(d)(3)(A). This language does not specify how “best performing” is to be determined: by means of emission level, emission control efficiency, measured over what period of time, etc. See Sierra Club v. EPA, 167 F. 3d at 661 (language of floor requirement for existing sources “on its own says nothing about how the performance of the best units is to be calculated”). Put another way, this language does not answer the question of which source is the better performing: one that emits 100 units of HAP but also feeds 100 units of that HAP, or one that emits 101 units of the HAP but feeds 10,000 units. See 69 FR at 21223. Moreover, new source floors are to be based on the performance of the “best controlled” similar source achieved in practice. Section 112(d)(3). “Best controlled” can naturally be read to refer to some means of control such as system removal efficiency as well as to emission level.

Use of a straight emissions approach to identify floor levels can lead to arbitrary results. Most important, as explained above, it leads to standards which cannot be achieved consistently even by the best performing sources because operating variability is not accounted for. This is shown in section 17 of volume III of the technical support document. These analyses show that (a) emissions from these sources do in fact vary from test-to-test, and that no two snapshot emission test results are identical; (b) our statistical approach that quantifies within test, run-to-run variability underestimates the best performing sources' long term, test-to-test variability; [83] (c) best performing sources under the straight emissions approach advocated by the commenter (i.e. the lowest emitting sources) had other test conditions that did not achieve straight emission floor levels; (d) best performing sources under the straight emissions approach are projected, based on two separate analyses using reasonable assumptions, not to achieve the straight emissions floor standard based on these sources' demonstrated variations in system removal efficiencies over time (i.e., from test-to-test); and (e) SRE/feed methodology yields floor levels (i.e. the floor standards in the rule) that better estimate the emission levels reflecting the performance over time of the best performing sources. See Mossville, 370 F. 3d at 1242 (floor standard is reasonable because it accommodated best performing source's highest level of performance (i.e. its total variability), even though the level of the standard was higher than any individual measurement from that source).

As noted earlier, the straight emissions methodology can also limit operation of commercial units because the standard reflects a level of hazardous waste feed control which could force commercial units to burn less hazardous waste because such standards more resemble new source standards. The straight emissions methodology also arbitrarily reflects HAP levels in raw materials and fossil fuels, an infeasible means of control for any source.

Another arbitrary, and indeed impermissible, result of the straight emissions methodology is that in some instances (noted in responses below) the methodology results in standards which would force sources identified as best performing to install upgraded air pollution control equipment. This result undermines section 112 (d) (2) of the statute, by imposing what amounts to a beyond the floor standard without consideration of the beyond the floor factors: the cost of achieving those reductions, as well as energy and nonair environmental impacts.

Comment: The commenter states that because MACT floors must reflect the “actual performance” of the relevant best performing hazardous waste combusters, this means that the lowest emitters must be the best performers. The commenter cites CKRC v. EPA, 255 F. 3d at 862 and other cases in support.

Response: As explained in the introduction above, the statute does not specify that lowest emitters are invariably best performers. Nor does the caselaw cited by the commenter support this position. The D.C. Circuit has held repeatedly that EPA may determine which sources are best performing and may “reasonably estimate” the performance of the top 12 percent of these sources by means other than use of actual data. Mossville, 370 F. 3d at 1240-41 (collecting cases). In Mossville, sources had varying levels of vinyl chloride emissions due to varying concentrations of vinyl chloride in their feedstock. Individual measurements consequently did not adequately represent these sources' performance over time. Not-to-exceed permit limits thus reasonably estimated sources' performance, corroboration being that individual sources with the lowest long-term average performance occasionally came close to exceeding those permit limits. Id. at 1241-42. The facts are similar here, since our examination of best performing sources with multiple test conditions likewise shows instances where these sources would be unable to meet floors established based solely on lowest emissions (including their own). As here, EPA was not compelled to base the floor levels on the lowest measured emission levels.

Comment: The same commenter maintains that it is clear from the caselaw that MACT floors must reflect the relevant best performing sources' “actual performance”, and that this must refer to the emissions level it achieves.

Response: As just stated, the D.C. Circuit has repeatedly stated that EPA may make reasonable estimates of sources' performance in assessing both which sources are best performing and the level of their performance. The court has further indicated that EPA is to account for variability in assessing sources' performance for purposes of establishing floors, and this assessment may require that EPA make reasonable estimates of performance of best performing sources. CKRC, 255 F. 3d at 865-66; Mossville, 370 F. 3d at 1241-42. See discussion in A.1.a above.

Comment: The commenter generally maintains that EPA's floor approaches consider only the performance of back-end pollution control technology and so fail to capture other means of HAP emission control that otherwise would be captured if EPA were to assess performance based on the emission levels each source achieved.

Response: EPA agrees that factors other than end-of-stack pollution control can affect metal HAP and chlorine emissions. This is why EPA assesses performance for these HAP by considering combinations of system removal efficiency (which measures every element in a control system resulting in HAP reduction, not limited to efficiency of a control device), and hazardous waste HAP feed control. Standards for dioxins and other organic HAP (which have no hazardous waste feed control component) likewise assess every element of control. Start Printed Page 59444

EPA also accounts for the variability of HAP levels in the (essential) use of raw materials and fossil fuels by assessing performance of back-end control but not evaluating fuel/raw material substitution, which, as discussed later in the response to comments section, are infeasible means of control. Mossville, 370 F. 3d at 1241-42, is instructive on this point. The court held that the constant change in raw materials justified EPA's use of a regulatory limit to estimate a floor level. The reasonableness of this level was confirmed by showing that the highest individual data point of a best performing source was nearly at the level of the regulatory limit. Under the commenter's approach, the court would have had no choice but to hold that the level the source achieved in a single test result using ‘clean’ raw materials—i.e. the ‘level achieved’ in the commenter's language—dictated the floor level.

See part four, section III.C for EPA's response to this comment as it relates to the methodologies for the particulate matter standard and total chlorine standard for hydrochloric acid production furnaces.

Comment: The commenter notes that the SRE/Feed methodology does not account for all HAP emissions, failing to account for metal and chlorine feedrates in raw materials and fossil fuels.

Response: The methodology does not assess the effect of feed “control” of HAP levels in raw materials or fossil fuels which may be inputs to the combustion units. This is because such control may not be replicable by an individual source, or duplicable by any other source. See 69 FR at 21224 and n. 48; Sierra Club v. EPA, 353 F. 3d 976, 988 (“substitution of cleaner ore stocks was not * * * a feasible basis on which to set emission standards. Metallic impurity levels are variable and unpredictable both from mine to mine and within specific ore deposits, thereby precluding ore-switching as a predictable and consistent control strategy”).[84] EPA's methodology does account for HAP control of all inputs by assessing system removal efficiency, which measures reductions of HAPs in all inputs (including fossil fuel and raw materials) to a hazardous waste combustion unit. Further, nonmercury metal HAP emissions attributable to raw materials and fossil fuels are effectively controlled with the particulate matter standard, a standard that is based on the sources with best back-end control devices. The only element which is not controlled is what cannot be: HAP levels in feeds for which fuel or raw material switching is simply not an available option.

Comment: The commenter further maintains, however, that the means by which sources may be achieving levels of performance are legally irrelevant (citing National Lime Ass'n v. EPA, 233 F. 3d 625 , 634 and 640 (D.C. Cir. 2000)). The fact that sources with “cleaner” raw material and fossil fuel inputs may not intend to have resulting lower HAP emissions is therefore without legal bearing.

Response: The issue here is not one of intent. The Court, in National Lime, rejected the argument that sources' lack of intent to control a HAP did not preclude EPA from establishing a section 112(d) standard for that HAP. See 233 F. 3d at 640, rejecting the argument that HAP metal control achieved by use of back-end control devices (baghouses) could not be assessed by EPA because the sources used the back-end control devices to control emissions of particulate matter. The case did not consider the facts present here, where the issue is not a source's intent, but rather a means of control which involves happenstance (composition of HAP in raw materials and fossil fuel used the day the test was conducted) and so is neither replicable nor duplicable.

National Lime also held that EPA must establish a section 112(d) emission standard for every HAP emitted by a major source. 233 F. 3d at 634. EPA is establishing emission standards for all HAP emitted by these sources. In establishing these standards, EPA is not evaluating emission reductions attributable to the type of fossil fuel and raw material used in the performance tests, because this is not a “feasible basis on which to set emission standards.” Sierra Club, 353 F. 3d at 988.

EPA thus does not agree with this comment because the issue is not a source's intent but rather whether or not to assess emission reductions from individual test results which reflect an infeasible means of control.

Comment: The commenter maintains, however, that even if individual sources (including those in the pool of best performing sources) cannot reduce HAP concentrations in raw materials and fossil fuels, they may achieve the same reductions by adding back-end pollution control. Nothing in section 112(d)(3) says that sources have to use the means of achieving a level of performance that other best performing sources used.

Response: The thrust of this comment is essentially to impermissibly bypass the beyond-the-floor factors set out in section 112(d)(2) under the guise of adopting a floor standard. Suppose that EPA were to adopt a floor standard dominated by emission levels reflecting HAP concentrations present in a few sources' raw materials and fossil fuels during their test conditions. Suppose further that some sources have to upgrade their back-end control equipment to operate at efficiencies better than the average level demonstrated by the best performing sources, because test results based on fossil fuel and raw material levels are neither replicable nor duplicable. In this situation, EPA believes that it would have improperly adopted a beyond-the-floor standard because EPA would have failed to consider the beyond-the-floor factors (cost, energy, and nonair environmental impacts) set out in section 112(d)(2).[85]

Comment: EPA has not substantiated its claim that sources cannot switch fossil fuels or raw materials.

Response: At proposal we evaluated fuel switching and raw material substitution as beyond-the-floor technologies for cement kilns and lightweight aggregate kilns and stated these technologies would not be cost effective.[86] We also discussed why fuel switching is not an appropriate floor control technology for solid fuel-fired boilers. 69 FR at 21273. Upon further evaluation, we again conclude that fuel switching and raw material substitution are not floor control technologies and are not cost effective beyond-the-floor technologies for cement kilns, lightweight aggregate kilns, and solid fuel-fired boilers.[87]

Comment: EPA has failed to document the basis for its SRE ranking. Start Printed Page 59445Specifically, EPA has not stated how it measured sources' SREs, or how it knows those rankings are accurate.

Response: System removal efficiency is a parameter that is included in our database that is calculated by the following formula:

The HAP feedrate and emission data are components of the database that were extracted from emission test reports for each source. We use system removal efficiency for each relevant pollutant or pollutant group (e.g., semivolatile metals, low volatile metals, mercury, total chlorine) whenever the data allows us to calculate a reliable system removal efficiency. For example, we generally do not use system removal efficiencies that are based on normal emissions data because of the concern that normal feed data are too sensitive to sampling and measurement error. See 69 FR at 21224.[88]

The system removal efficiencies used in our ranking process are reliable and accurate because the feed and emissions data originate from compliance tests that demonstrate compliance with existing emission standards (primarily RCRA requirements). As such, the data are considered to have excellent accuracy and quality. RCRA trial burn and certification of compliance reports are typically reviewed in detail by the permitting authority. The compliance tests and test reports generally contain the use of various quality assurance procedures, including laboratory, method, and field blanks, spikes, and surrogate samples, all of which are designed to minimize sampling and analytical inaccuracies. EPA also noticed the data base for this rule for multiple rounds of comment and has made numerous changes in response to comment to assure accuracy of the underlying data. Thus, EPA concludes the calculated system removal efficiencies used in the ranking process are both reliable and accurate.

Comment: EPA's approach with regard to use of stack data is internally contradictory. EPA uses stack data in establishing floors, but does not use stack data to determine which performers are best. EPA has failed to explain this contradiction.

Response: Emission levels are used to calculate system removal efficiencies in order to assess each source's relative back-end control efficiency. Also, as explained in the introduction to this comment response section, the SRE/Feed methodology uses the stack emission levels of the sources using the best combinations of hazardous waste feed control and system-wide air pollution control (expressed as HAP percent removal over the entire system) to calculate the floors. The data are adjusted statistically to account for quantifiable forms of variability (run-to-run variability). This methodology reasonably selects best performing sources (for HAP amenable to these means of control), and reasonably estimates these sources' performance over time. As further stated in section B.2 above, using a straight emissions approach to identify best performers and their level of performance can lead to standards for these HAP that do not fully account for variability (including variability resulting from varying and/or uncontrollable amounts of HAP in raw materials and fossil fuels) and could force installation of de facto beyond-the-floor controls without consideration of the section 112(d)(2) beyond-the-floor factors.

EPA thus does not see the contradiction expressed by the commenter. Use of the straight emissions approach as advocated by the commenter would lead to standards that do not reasonably estimate sources' performance and which could not be achieved even by the best performers with individual test conditions below the average of the 12 percent of best performing sources. These problems would be compounded many-fold if the data were not normalized and adjusted to at least account for quantifiable variability, steps the commenter also opposes. EPA's use of emissions data (suitably adjusted) after identifying best performers through the ranking methodology avoids these problems and reasonably estimates best performers' level of performance.

Comment: The commenter rejects EPA's finding (69 FR at 21226) that individual test results in the data base do not fully express the best performing sources' performance. The commenter gives a number of reasons for its criticisms, which we answer in the following sequence of comments listed a though f.

a. Comment: The commenter states that EPA claims emission levels do not fully reflect variability in part because they are sometimes based on tests where the source was feeding low levels of HAP during the test. The commenter claims this is inconsistent with the fact that EPA preferentially uses worst-case emissions obtained from tests where the sources spiked their feedstreams with metals, and that the mere possibility that these emissions do not reflect test data from conditions where variability was not maximized does not mean those data fail to represent a source's actual performance. The commenter also states that “EPA's apparent suggestion that the best performing sources could not replicate the average performance of the sources with the lowest emissions is unsubstantiated and unexplained. Assuming that EPA accurately assesses a source's actual performance, the source can replicate that performance.”

Response: HAPs in raw materials and fossil fuels contribute to a source's emissions. EPA has concerns that a straight emissions approach to setting floors may not be replicable by the best performing sources nor duplicable by other non-best performing sources because of varying concentration levels of HAP in raw material and nonhazardous waste fuels. The best performing sources operated under compliance test conditions as the commenter suggests. However, raw material and nonhazardous fuel HAP concentrations for the best performing sources will change over time, perhaps due to a different source of fuel or raw material quarry location, which could affect their ability to achieve the floor level that was based on emissions obtained while processing different fossil fuel or raw materials. EPA takes sharp issue with the commenter's statement that a single performance test result is automatically replicable so long as it is measured properly in the first instance. This statement is incorrect even disregarding HAP contributions in raw materials and fossil fuels since, as noted previously in section A.2.e, there are many other sources of variability Start Printed Page 59446which will influence sources' performance over time (i.e., in subsequent performance tests).

A straight emissions approach for establishing semivolatile and low volatile metal floors may result in instances where the best performing sources would not be capable of achieving the standards if their raw material and nonhazardous waste fuel HAP levels change over time. For each cement kiln and lightweight aggregate kiln, we estimated the emissions attributable to these raw materials and fossil fuels assuming each source was operating with hazardous waste HAP feed and back-end control levels equivalent to the average of the best performing sources (the difference in emissions across sources only being the result of the differing HAP levels in the nonhazardous waste feeds). The analysis shows that emissions attributable to these nonhazarous waste feedstreams (raw materials and fossil fuels) varies across sources, and can be significant relative to the level of the straight emissions-based floor design level and floor, and therefore could inappropriately impact a source's ability to comply with the floor standard.[89]

b. Comment: The commenter states that EPA must consider contributions to emissions from raw materials and fossil fuels, that it is irrelevant if sources from outside the pool of best performing sources can duplicate emission levels reflecting “cleaner” raw materials and fossil fuels used by the best performing sources, and that sources unable to obtain such “cleaner” inputs may always upgrade other parts of their systems to achieve that level of performance.

Response: As previously discussed, EPA's methodology does account for HAP control of all inputs by assessing system removal efficiency, which measures reductions of HAPs from all inputs. Further, nonmercury metal HAP emissions attributable to raw materials and fossil fuels are effectively controlled with the particulate matter standard, a standard that is based on the sources with lowest emissions from best back-end control devices. We are not basing any standards on performance of sources not ranked as among the best performing.

c. Comment: The commenter disputes EPA's conclusions that failure of sources to meet all of the standards based on a straight emissions methodology at once shows that the methodology is flawed. The standards are not mutually dependent, so the fact that they are not achieved simultaneously is irrelevant. There is no reason a best performer for one HAP should be a best performer for other HAP.

Response: EPA agrees with this comment. On reflection, EPA believes that because all our standards are not technically interdependent (i.e., implementation of one emission control technology does not prevent the source from implementing another control technology), the fact that sources are not achieving all the standards simultaneously does not indicate a flaw in a straight emissions approach. See Chemical Manufacturers Ass'n, 870 F. 2d at 239 (best performing sources can be determined on a pollutant-by-pollutant basis so that different plants can be best performers for different pollutants).

d. Comment: Several commenters took the opposite position that EPA must assure that all existing source standards must be achievable by at least 6 percent of the sources, and that all new source standards must be achievable by at least one existing source.

Response: As discussed above, we are not obligated to establish a suite of floors that are simultaneously achievable by at least six percent of the sources because the standards are not technically interdependent. Nonetheless, the SRE/Feed methodology does result in existing floor levels (when combined with the other floor levels for sources in the source category) that are simultaneously achievable by at least six percent of the sources (or, for source categories that have fewer than 30 sources, by at least two or three sources).[90] However, for the new source standards, three of the source categories do not include any sources that are simultaneously achieving all the standards (incinerators, cement kilns, and lightweight aggregate kilns). Again, similar to existing sources, EPA is not obligated to establish a suite of new source floors that are simultaneously achievable by at least one existing source because these standards are not technically interdependent. We conclude that a new source can be designed (from a back-end control perspective) to achieve all the new source standards.[91]

e. Comment: The commenter criticizes EPA's discussion at 69 FR 21227-228 indicating that both hazardous waste feed control and back-end pollution control are superior means of HAP emission control and treatment standards should be structured to allow either method to be the dominant control mechanism.

Response: EPA is not relying on this part of the proposed preamble discussion as justification for the final rule, with the one exception noted in the response to the following comment.

f. Comment: Considerations of proper waste disposal policy are not relevant to MACT floor determinations. In any case, the possibility that some commercial waste combustors may upgrade their back-end pollution control systems to meet standards reflecting low hazardous waste HAP feedrates, or divert wastes to better-controlled units, is positive, not negative.

Response: As discussed in section B.1 above, there are instances where standards derived by using a straight emissions approach are based on a combination of lowest emitting low feeding sources and lowest emitting higher feeding sources. Resulting floor standards would thus reflect these low hazardous waste feedrates and could put some well-controlled commercial incinerators in the untenable situation of having to reduce the amount of hazardous waste that is treated at their source. Our database verifies that such an outcome is in fact realistic.[92]

This type of standard would operate as a direct constraint on the amount of hazardous waste that could be fed to the device, in effect depriving a combustion source of its raw material. In this instance, hazardous wastes could not be readily diverted to other units because the low feeding hazardous waste sources tend not to be commercial units. In these circumstances, there would be a significant adverse nonair environmental impact. Hazardous waste is required to be treated by Best Demonstrated Available Technology (BDAT) before it can be land disposed. RCRA sections 3004 (d), (e), (g), and (m); Hazardous Waste Treatment Council v. EPA, 866 F. 2d 355, 361 (D.C.Cir. 1990) (upholding Best Demonstrated Available Technology treatment requirement). Most treatment standards for organic pollutants in hazardous waste can only be achieved by combustion. Leaving some hazardous wastes without a Start Printed Page 59447treatment option is in derogation of these statutory requirements and goals, and calls into question whether a treatment standard that has significant adverse nonair environmental impacts must be viewed as best performing. See Portland Cement Ass'n v. Ruckelshaus, 486 F. 2d 375 , 386 (D.C. Cir. 1973); Essex Chemical Co. v. EPAEPA, 486 F. 2d 427, 439 (D.C. Cir. 1973). The commenter's statement that waste disposal policy is not relevant to the MACT standard-setting process is not completely correct, since section 112 (n) (7) of the Clean Air Act directs some accommodation between MACT and RCRA standards for sources combusting hazardous waste. Part of this accommodation is using a methodology to evaluate best performing sources that evaluates as best performers those using the best combination of hazardous waste feed control (among other things, an existing control measure under RCRA rules) and system-wide removal.

We assessed whether we could address this issue by subcategorizing commercial incinerators and on-site incinerators. Applying the straight emission approach to such a subcategorization scheme, however, yields anomalous results due to the scarcity of available and complete compliance test data from commercial incinerators. Calculated floor levels for semivolatile metals and low volatile metals for the commercial incinerator subcategory equate to 2,023 and 111 μg/dscm, respectively (both higher than the current interim standards).[93] We conclude that the SRE/Feed methodology better addresses this issue because it yields floor levels that better represent the performance of the best performing commercial incinerators and onsite incinerators alike by applying equal weights to hazardous waste feed control and back-end control in the ranking process.

EPA notes, however, that its choice of the SRE/Feed methodology is justified independent of considerations of adverse impact on hazardous waste treatment and disposal.

Comment: The commenter reiterates its comments with respect to floor levels for new sources.

Response: EPA's previous responses to comments apply to both new and existing source standards.

Comment: Two commenters recommend that EPA define the single best performing source as that source with the lowest aggregated SRE/Feed aggregated score (as proposed), as opposed to the source with the lowest emissions among the best performing existing sources (an approach on which we requested comment).

Response: We agree with the commenters because this is consistent with our methodology for defining best performers for existing sources and assessing their level of performance. We note, however, that with respect to the new source standards, we encountered two instances where the SRE/Feed methodology identified multiple sources with identical single best aggregated scores, resulting in a tie for the best performing source. This occurred for the mercury and low volatile metal new source standards for incinerators. In these instances, EPA applied a tie breaking procedure that resulted in selecting as the single best performing source as that source (of the tied sources) with the lowest emissions. We believe this is a reasonable interpretation of section112(d)(3), which states the new source standard shall not be less stringent than the emission control that is achieved in practice by the best controlled similar source (“source” being singular, not plural). Moreover, we believe use of the emission level as the tie-breaking criteria is reasonable, not only because it is a measure of control, but because we have already fully accounted for hazardous waste feedrate control and system removal efficiency in the ranking methodology. To choose either of these factors to break the tie would give that factor disproportionate weight.

C. Air Pollution Control Technology Methodologies for the Particulate Matter Standard and for the Total Chlorine Standard for Hydrochloric Acid Production Furnaces

At proposal, EPA used what we termed “air pollution control technology” methodologies to estimate floor levels for particulate matter from all source categories as a surrogate for non-mercury HAP metals, and for total chlorine from hydrochloric acid furnace production furnaces. 69 FR at 21225-226. Under this approach, we do not estimate emission reductions attributable to feed control, but instead assess the performance of back-end control technologies.[94] We are adopting the same methodologies for these HAP in the final rule. Because the details of the approaches differ for particulate matter and for total chlorine, we discuss the approaches separately below.

1. Air Pollution Control Device Methodology for Particulate Matter

Our approach to establishing floor standards for particulate matter raises three major issues.

The first issue is whether particulate matter is an appropriate surrogate for non-enumerated HAP metals from all inputs, and for all non-mercury HAP metals in raw material and fossil fuel inputs. This issue is discussed at section IV.A of this part, where we conclude that particulate matter is indeed a reasonable surrogate for these metal HAP.

The second issue is why EPA is not evaluating some type of feed control for the particulate matter floor. There are two potential types of feed control at issue: hazardous waste feed control of nonenumerated metals, and feed control of non-mercury HAP metals in raw material and fossil fuel inputs. With respect to feed control of non-enumerated metals in hazardous waste, as discussed in more detail in section IV.A of this part, we lack sufficient reliable data on non-enumerated metals to assess their feedrates in hazardous waste. In addition, there are significant questions about whether feedrates of the non-enumerated metals can be optimized along with SVM and LVM feedrates. We also have explained elsewhere why control of hazardous waste ash feedrate would be technically inappropriate, since it would not properly assess feed control of nonenumerated metals in hazardous waste. See also 69 FR at 21225.

We have also explained why we are not evaluating control of feedrates of HAP metals in raw materials and fossil fuels to hazardous waste combusters: it is an infeasible means of control. See section B of this part. We consequently are not evaluating raw material and fossil fuel ash feed control in determining the level of the various floors for particulate matter.

a. The methodology. The final issue is the means by which EPA is evaluating back-end control. Essentially, after determining (as just explained) that back-end control is the means of controlling non-mercury metal HAP and that particulate matter is a proper surrogate for these metals, EPA is using its engineering judgment to determine what the best type of air pollution control device (i.e., back-end control) is to control particulate matter (and, of course, the contained HAP metals). We then ascertain the level of performance by taking the average of the requisite number of sources (either 12 % or five, Start Printed Page 59448depending on the size of the source category) equipped with the best back-end control with the lowest emissions.[95] These floor standards are therefore essentially established using a straight emissions methodology. We have determined that baghouses (also termed fabric filters) are generally the best air pollution control technology for control of particulate matter, and that electrostatic precipitators are the next best.

b. Why not select the lowest emitters? Although sources with baghouses tended to have the lowest emission levels for particulate matter, this was not invariably the case. There are certain instances when sources controlled with electrostatic precipitators (or, in one instance, a venturi scrubber) had lower emissions in individual test conditions than sources we identified as best performing which were equipped with baghouses.[96] Under the commenter's approach, we must always use these lowest emitting sources as the best performers.

We again disagree. We do not know if these sources equipped with control devices other than baghouses with lower emissions in single test conditions would actually have lower emissions over time than sources equipped with baghouses because we cannot assess their uncontrollable emissions variability over time. Our data suggests that they likely are not better performing sources. We further conclude that our statistical procedures that account for these sources' within test, run-to-run emissions variability underestimates these sources long-term emissions variability. This is not the case for sources equipped with baghouses, where we have completely assessed, quantified, and accounted for long-term, test-to-test emissions variability through application of the universal variability factor.[97] The sources equipped with control devices other than baghouses with lower snapshot emissions data could therefore have low emissions in part because they were operating at the low end of the “uncontrollable” emissions variability profile for that particular snapshot in time. The basis for these conclusions, all of which are supported by our data, are found in section 16 of volume III of the technical support document.

We therefore conclude sources equipped with baghouses are the best performers for particulate matter control not only based on engineering judgment, but because we are able to reliably quantify their likely performance over time. The straight emissions methodology ignores the presence of long-term emissions variability from sources not equipped with baghouses, and assumes without basis that these sources are always better performing sources in instances where they achieved lower snapshot emissions relative to the emissions from baghouses, emissions that have notably already been adjusted to account for long-term emissions variability.

A straight emissions approach also results in inappropriate floor levels for particulate matter because it improperly reflects/includes low ash feed when identifying best performing sources for particulate matter. 69 FR at 21228. For example, the MACT pool of best performing liquid fuel boilers for particulate matter under the straight emissions approach includes eight sources, only one of which is equipped with a back-end control device. These sources have low particulate matter emissions solely because they feed low levels of ash. The average ash inlet loadings for these sources are well over two orders of magnitude lower than the average ash inlet loading for the best performing sources that we identify with the Air Pollution Control Technology approach. (Of course, since ash loadings are not a proper surrogate for HAP metals, these sources' emissions are lowest for particulate matter but not necessarily for HAP metals.) The straight emissions approach would yield a particulate matter floor level of 0.0025 gr/dscf (with a corresponding design level of 0.0015 gr/dscf). There is not one liquid fuel boiler that is equipped with a back-end control that achieved this floor level, much less the design level. The best performing source under the air pollution control technology approach, which is equipped with both a fabric filter and HEPA filter, did not even make the pool of best performing sources for the straight emissions approach. Yet this unit has an excellent ash removal efficiency of 99.8% and the lower emitting devices' removal efficiencies are, for the most part, 0% because they do not have any back-end controls. EPA believes that it is arbitrary to say that these essentially uncontrolled devices must be regarded as “best performing” for purposes of section 112(d)(3). We therefore conclude that a straight emissions floor would not be achievable for any source feeding appreciable levels of ash, even if they all were to upgrade with baghouses, or baghouses in combination with HEPA filters, and that a rote selection of lowest emitters as best performers can lead to the nonsensical result of uncontrolled units being classified as best performers.

Comment: Commenter claims end-of-stack control technology is not the only factor affecting emissions of particulate matter, stating that EPA admits that particulate matter emission levels are affected by the feedrate of ash. Accordingly, the performance of a source's end-of-stack control technology is not a reasonable estimate of that source's total performance.

Response: The particulate matter standard serves as a surrogate control for the non-enumerated metals in the hazardous waste streams (for all source categories), and all nonmercury metal HAP in the nonhazardous waste process streams (essentially, raw materials and fossil fuels) for cement kilns, lightweight aggregate kilns, and liquid fuel boilers. The commenter suggests that the APCD approach inappropriately ignores HAP feed control in the assessment of best performing sources. We conclude that it would not be appropriate to use a methodology that directly assesses feed control, such as the SRE/Feed methodology, to determine particulate matter floors. First, direct assessment of total ash feed control would inappropriately assess and seek to control (even though variability of raw material and fossil fuel inputs are uncontrollable) raw material and fossil fuel HAP input, as well as raw material and fossil fuel input. Controlling raw material and fossil fuel HAP input is infeasible, as previously discussed. It also inappropriately limits theses sources' feedstocks that are necessary for their associated production process.

Second, we do not believe that developing a floor standard based on hazardous waste feed control of nonenumerated metals (as opposed to feed control of these metals in raw material and fossil fuels) is appropriate or feasible. In part four, section IV.A, we explain that we lack the data to reliably assess direct feedrate of these metals in hazardous waste. In addition, we also discuss that it is unclear (the lack of certainty resulting from the sparse available data) that hazardous waste feed control of the nonenumerated metals is feasible. The majority of these metals are not directly regulated under existing RCRA requirements, so sources have optimized control of the other HAP Start Printed Page 59449metals, raising issues of whether simultaneous optimization of feed control of the remaining metals is feasible. Moreover, even if one were to conclude that hazardous waste feed control is feasible for the nonenumerated metal HAPs, hazardous waste ash feedrates are not reliable indicators of nonmercury metal HAP feed control levels and are therefore inappropriate parameters to assess in the MACT evaluation process. For example, a source could reduce its ash feed input by reducing the amount of silica in its feedstreams. This would not result in feed control or emission reductions of metal HAP.[98]

Finally, hazardous waste ash feed control levels do not significantly affect particulate matter emissions from cement kilns, lightweight aggregate kilns, and solid fuel-fired boilers because the majority of particulate matter that is emitted originates from the raw material and nonhazardous fuel. Hazardous waste ash feed control levels also do not significantly affect particulate matter emissions from sources equipped with baghouses because these control devices are not sensitive to particulate matter inlet loadings.[99]

Thus, even if one were to conclude that the nonenumerated metal HAPs are amenable to hazardous waste feed control, explicit use of ash feed control in a MACT methodology would not assure that each source's ability to control either nonmercury metal HAP or surrogate particulate matter emissions is assessed. The Air Pollution Control Device methodology identifies and assesses (with the surrogate particulate matter standard) the known technology that always assures metal HAP emissions are being controlled to MACT levels—that technology being back-end control.

Comment: Commenter claims the Air Pollution Control Device approach to calculate particulate matter floors is flawed because the performance of back-end control technology alone does not reflect the performance of the relevant best sources that otherwise would be reflected if EPA were to assess performance based on the emission levels each source achieved because, as EPA admits, it fails to account for the effect of ash feed rate.

Response: We explain above why the Air Pollution Control Technology approach properly identifies the relevant best performing sources for purposes of controlling non-mercury metal HAP (measured as particulate matter), irrespective of ash feed rates. Typically, this results in selecting the sources with the lowest particulate matter emission rates, the result the commenter advocates. This is because we evaluate sources with the best-performing (e.g. lowest emitting) baghouses, and particulate matter emissions from baghouses are not significantly affected by inlet particulate matter loadings. Where the pool of best performing sources includes sources operating some other type of back-end control device (because insufficient numbers of sources are equipped with baghouses to comprise 12% of sources, or five sources (depending on the size of the source category)), we again use the lowest particulate matter emission level from the sources equipped with second best technology. Although these data do not reflect test-to-test variability, they are the best remaining data in EPA's possession to estimate performance and EPA is therefore, as required by section 112 (d) (3) (A) and (B), using the data to fill out the requisite percentage of sources for calculating floors.

Comment: Commenter states that EPA has failed to demonstrate how it reasonably estimated the actual performance of each source's end-of-stack control technology because: (1) It failed to acknowledge that there can be substantial differences between the performance of different models of the same type of technology; and (2) it did not explain or support its rankings of pollution control devices.

Response: As discussed in sections 7.4 and 16.2 of volume III of the technical support document and C.1 of this comment response section, we rank associated back-end air pollution control device classes (e.g., baghouses, electrostatic precipitators, etc.), after assessing particulate matter control efficiencies from hazardous waste combustors that are equipped with the associated back-end control class. The data used to make this assessment are included in our database. We also evaluated particulate matter control efficiencies from other similar source categories that also use these types of control systems, such as municipal waste combustors, medical waste incinerators, sewage sludge combustors, coal-fired boilers, oil fired boilers, non-hazardous industrial waste combustors, and non-hazardous waste Portland cement kilns.[100]

After we assign a ranking score to each back-end control class, we determine the number of sources that are using each of these control technology classes. We then identify the MACT control technology or technologies to be those best ranked back-end controls that are being used by 12 percent of the sources (or used by five sources in instances where there are fewer than 30 sources). We then look only at those sources using MACT back-end control and rank order all these sources first by back-end control type, and second by emissions. For example, in instances where there is more than one MACT back-end control, we array the emissions from the sources equipped with the top ranked back-end controls from best to worst (i.e., lowest to highest), followed by the emissions from sources equipped with the second ranked back-end controls from best to worst, and so on. We then determine the appropriate number of sources to represent 12 percent of the source category (5 in instances where there are fewer than 30 sources). If 10 sources represented 12% of the sources in the source category, we would then select the emissions from best ranked 10 sources in accordance with this ranking procedure to calculate the MACT floor. This methodology results in selection of lowest emitters using best back-end air pollution control as pool of the best performing sources.

The commenter is correct that there can be differences between the performance of different models of the same type of technology. We are not capable of thoroughly assessing differences in designs of each air pollution control device in a manner that could be used in the MACT evaluation process, so that we would only select, for example, baghouses of a certain type. Each baghouse, for example, will be designed differently and thus will have different combinations of design aspects that may or may not make that baghouse better than other baghouses (e.g., bag types, air to cloth ratios, control mechanisms to collect accumulated filter cake and maintain optimum pressure drops). We also do not have detailed design information for each source's air pollution control system; such an assessment would therefore not be Start Printed Page 59450possible even if the information could be used to assess relative performance.

We instead account for this difference by selecting sources with the lowest emissions that are using the defined MACT back-end controls to differentiate the performance among those sources that are using that technology (the best performer being the source with the lowest emissions, as just explained). For example, in situations where more than 12% of the sources are using the single best control technology (e.g., more than 12% of incinerators use baghouses to control particulate matter), we use the emissions from the lowest emitting sources equipped with baghouses to calculate the MACT floor. In instances where there are two defined MACT technologies (i.e., 12% of sources do not use the single best control technology), we use all the emissions data from sources equipped with the best ranked control class, and then subsequently use only the lowest emissions from the sources equipped with the second ranked back-end controls.

Comment: EPA did not say how it picked the best performers if more than twelve percent used the chosen technologies. If EPA used emissions data to differentiate performance, the Agency is necessarily acknowledging that emissions data are a valid measure of sources' performance—in which case the Agency's claims to the contrary are arbitrary and capricious.

Response: We did use emissions data to select the pool of best performers where over 12% use the best type of emissions control technology, as explained in the previous response. Emissions data is obviously one means of measuring performance. EPA's position is that it need not be the exclusive means, in part because doing so leads to arbitrary results in certain situations. Our use of emission levels to rank sources that use the best particulate matter control (i.e., baghouses) does not lead to arbitrary results, however. First, we are assessing emission levels here as a means of differentiating sources using a known type of pollution control technology. More importantly, the adjusted emission levels from sources equipped with baghouses are the most accurate measures of performance because these emissions have been statistically adjusted to accurately account for long-term variability through application of the universal variability factor.

Comment: Commenter states that EPA, in its support for its Air Pollution Control Technology Approach used to calculate particulate matter floors, claims that an emissions-based approach would result in floor levels that “could not necessarily be achieved by sources using the chosen end-of-stack technology,” citing 69 FR at 21228. Commenter claims that it is settled law that standards do not have to be achievable through the use of any given control technology, and that it is also erroneous to establish floors at levels thought to be achievable rather than levels sources actually achieve.

Response: EPA is not establishing floor levels based on assuring the standards are achievable by a particular type of end-of-stack technology (or, for that matter, any end-of-stack technology). The floor levels in today's final rule reasonably estimate average performance of the requisite percent of best performing sources without regard for whether the levels themselves can be achieved by a particular means. Floor standards for particulate matter are based on the performance of those sources with the lowest emissions using the best back-end control technology (most often baghouses, and sometimes electrostatic precipitators). EPA uses this approach not to assure that the floors are achievable by sources using these control devices, but to best estimate performance of the best performing sources, including these sources' variability.

2. Total Chlorine Standard for Hydrochloric Acid Production Furnaces

We are adopting the methodology we proposed to estimate floor levels for total chlorine from hydrochloric acid production furnaces. 69 FR at 21225-226. As stated there, we are defining best performers as those sources with the best total chlorine system removal efficiency. We are not assessing a level of control attributable to control of chlorine in feedstocks because this would simply prevent these furnaces from producing their ultimate product. Further details are presented in responses below.

Comment: Basing the standard for hydrochloric acid production furnaces on the basis of system removal efficiency rather than chlorine emission reduction is impermissible. Even though these devices' purpose is to produce chlorinated product, the furnaces can use less chlorinated inputs. EPA's proposed approach is surreptitious, an impermissible attempt to assure that the standards are achievable by all sources using EPA's chosen technology, the approach already rejected in CKRC.

Response: EPA disagrees. There is nothing in the text of the statute that compels an approach that forces sources to produce less product to achieve a MACT floor standard. Yet this is the consequence of the comment. If standards were based on levels of chlorine in feedstock to these units, less product would be produced since there would be less chlorine to recover. EPA has instead reasonably chosen to evaluate best performing/best controlled sources for this source category by measuring the efficiency of the entire chlorine emission reduction system. Indeed, the situation here is similar to that in Mossville, where polyvinyl chloride production units fed raw materials containing varying amounts of vinyl chloride depending on the product being produced. This led to variable levels of vinyl chloride in plant emissions. Rather than holding that EPA must base a floor standard reflecting the lowest amount of vinyl chloride being fed to these units, the court upheld a standard estimating the amount of pollution control achievable with back-end control. 370 F. 3d at 1240, 1243. In the present case, as in Mossville, the standard is based on actual performance of back-end pollution control (although here EPA is assessing actual performance of the control technology rather than estimating performance by use of a regulatory limit, making the situation here a fortiorari from that in Mossville), and does not reflect “emission variations not related to technological performance”. 370 F. 3d at 1240.

It also should be evident that EPA is not establishing a standard to assure its achievability by a type of pollution control technology, as the commenter mistakenly asserts. The standard for total chlorine is based on the average of the best five sources “ best meaning those sources with greatest (most efficient) system removal efficiencies. EPA did not, as in CKRC, establish the standard using the highest emission limit achieved by a source operating a particular type of control.

Comment: The commenter generally maintains that EPA's methodology to determine total chlorine floors for hydrochloric acid production furnaces fails to capture other means of HAP emission control that otherwise would be captured if EPA were assess performance based on the emission levels each source achieved.

Response: As discussed above, the standard for total chlorine is based on the sources with the best system removal efficiencies. System removal efficiency encompasses all means of MACT floor control when assessing relative performance because: (1) Chlorine feed control is not a MACT floor technology for these sources; and (2) the measure of system removal efficiency accounts for every other controllable factor that can affect Start Printed Page 59451emissions (e.g., operating practices, worker training, proper maintenance, pollution control device type, etc).

D. Format of Standards

1. Thermal Emissions

EPA proposed, and is finalizing standards for HAP metals and chlorine (the HAPs amenable to hazardous waste feed control) emitted by energy recovery units (cement kilns, lightweight aggregate kilns, and liquid fuel boilers) expressed in terms of pounds of HAP attributable to the hazardous waste fuel per million british thermal units (BTUs) of hazardous waste fired. 69 FR at 21219-20. EPA received many comments on this issue to which we respond below and in the Response to Comment Document. Some initial discussion of the issue is appropriate, however.

a. Expressing Standards in Terms of a Normalizing Parameter is Reasonable. First, using a thermal emissions form of a standard is an example of expressing standards in terms of a normalizing parameter. EPA routinely normalizes emission standards either by expressing them as stack HAP concentrations or by expressing the standards in units of allowable mass emissions per amount of production or raw material processed. Emission concentration-based standards normalize the size of each source by accounting for volumetric gas flowrate, which is directly tied to the amount of raw material each source processes (and subsequently the amount of product that is produced). Metal and particulate matter emission standards for commercial and industrial solid waste incinerators are expressed in emission concentration format. See § 60.2105. The particulate matter standard for Portland cement kilns is expressed as mass of allowable emissions per mass of raw material processed. See § 63.1342. The particulate matter, mercury, and hydrogen chloride standards for nonhazardous waste industrial boilers are expressed as pounds of allowable emissions per million British thermal units (BTUs). See § 63.7500.

Technology-based standards typically normalize emissions because such a format assures equal levels of control across sources per amount of raw material that is processed, and allows EPA to equally assess source categories that comprise units that differ in size. By normalizing the emissions standard we better ensure the same percentage of emission reduction per unit of raw material processed by each source.[101] See Weyerhaeuser v. Costle, 590 F. 2d 1011, 1059 (D.C. Cir. 1978) (technology-based standards are typically expressed in terms of volume of pollutants emitted per volume of some type of unit of production).

There is no legal bar to this approach since the statute does not directly address the question of whether a source emitting 100 units of HAP per unit of production but 100 units of HAP overall is a better performer (or, for new sources, better controlled) than a source emitting 10 units of HAP per unit of production but emitting 101 units overall.[102] One commenter appeared to suggest that we should assess performance on mass feedrates and mass emission rates, without normalizing. Such an approach would yield nonsensical results because the best performing sources would more likely be the smallest sources in the source category (smaller sources generally have lower mass emission rates because they process less hazardous waste). This would likely yield emission standards that would not be achievable by the larger sources that more likely are better controlled sources based on a HAP removal efficiency basis.[103] Normalization by unit of production is another way of expressing unit size, so that normalizing on this basis is a reasonable alternative to subcategorization on a plant size-by-plant size basis. See section 112(d)(1) (size is an enumerated basis for subcategorizing).

b. Using Hazardous Waste Thermal Input as the Normalizing Parameter is Permissible and Reasonable. Normalization of standards based on thermal input is analogous. For energy recovery units (in this rule, kilns and most liquid fuel boilers), normalizing on the basis of thermal input uses a key feed input as the normalizing parameter, allowing comparison of units with different inputs rather than separately evaluating these units by size and type (see section 112(d)(1)). Again, this approach is legally permissible. The statute does not answer the question of which source is better performing, the source emitting 100 pounds of HAP per million BTUs hazardous waste but 100 pounds of HAP overall or the source emitting 10 pounds of HAP per million BTUs hazardous waste but emitting 101 pounds overall.

The approach also is reasonable. First, as with other standards expressed in normalized terms, by normalizing the emissions standard we ensure the same percentage of emission reduction per unit of raw material processed by each source, thus allowing meaningful comparison among sources. For example, emission concentration-based standards normalize the size of each source by accounting for volumetric gas flowrate, which is directly tied to the amount of raw material each source processes (and subsequently to the amount of product that is produced), and assures equal levels of control per amount of product. Normalization on the basis of HAP amount in hazardous waste per BTU level in the hazardous waste similarly assures equal levels of control across sources per amount of raw material that is processed. Here, the raw material is the hazardous waste fuel, expressed as units of energy. It is reasonable to regard a hazardous waste fuel as a raw material to an energy recovery device. Indeed, fuels are the only input to boilers, so fuels are necessarily such units' sole raw material.[104 105] Hazardous waste burning cement kilns and lightweight aggregate kilns produce a product in addition to recovered energy and so process other raw materials. However, the reason these units use hazardous waste as inputs is typically to recover usable energy from the wastes. Hence, the hazardous waste fuel is reasonably viewed as a raw material to these devices.

In this regard, we note that our choice of normalizing parameter essentially says that best performers with respect to hazardous waste fuel burned in energy recovery units are those using the lowest HAP feedrate (for metals and chlorine) per amount of energy Start Printed Page 59452recovered.[106] This approach accords well with the requirement in section 112(d)(2) that EPA take energy considerations into account in developing MACT, and also that the Agency consider front-end means of control such as input substitution (section 112(d)(2)(A)). In addition, our choice furthers the RCRA goal of encouraging properly conducted recycling and reuse (RCRA section 1003(b)(6)), which is of relevance here in that Congress directed EPA to consider the RCRA emission controls for hazardous waste combustion units in developing MACT standards for these units, and to ensure “to the maximum extent possible, and consistent with [section 112 ]” that section 112 standards are “consistent” with the RCRA scheme. CAA section 112(n)(7).[107] Conversely, emission concentration-based standards, the methodology that otherwise would be used to calculate emission concentration-based standards, may result in standards that are biased against sources that recover more energy from hazardous waste. This may discourage sources from recovering energy from hazardous waste because such standards do not normalize each source's allowable emissions based on the amount of hazardous waste it processes for energy recovery purposes. See 69 FR at 21219 and responses below.

Second, use of this normalizing parameter makes it much more likely that hazardous waste feed controls will be utilized by these devices as an aspect of emissions control. See section 112(d)(2)(A) (use of measures reducing the volume of pollutants emitted through “substitution of materials”); CKRC, 255 F. 3d at 865 (EPA to consider means of control in addition to back-end pollution control technology when establishing MACT floors). As explained in our discussion of the SRE/Feed methodology, the MACT floor level for metals and chlorine reflects the best combination of hazardous waste feedrate, and total HAP removal efficiency. See section III.B. However, if standards for energy recovery units are expressed in terms of mass of HAP per volume of stack gas, then it would be relatively easy for these energy recovery devices to achieve a standard, without decreasing concentrations of HAP in their hazardous waste fuels, by diluting the HAP contribution of hazardous waste with emissions from fossil fuel. A thermal emissions format prevents this type of dilution from happening because it ignores additions of stack gases attributable to burning fossil fuels. Weyerhaeuser, 590 F. 2d at 1059 (use of production of a unit as a normalizing parameter serves “the commendable purpose” of preventing plants from achieving emission limitations via dilution).

For example, assume there are two identical energy recovery units with identical back-end control devices (that reflect the performance of the average of the best performing sources). Source A fulfills 25% of its energy demand from the combustion of hazardous waste; source B fulfills 50% of its energy demand from the combustion of hazardous waste. Also assume that the hazardous waste for these two sources have equivalent energy contents. If these sources were required to comply with an emission concentration based-standard (e.g., μg/dscm), source A would be allowed to feed hazardous waste containing twice the metal content (on a mass concentration basis, e.g., ppm), and would be allowed to emit metal HAP at the same mass emission rate relative to source B. This is because this source is effectively diluting its emissions with the emissions that are being generated by the fossil fuels.[108] A thermal emissions standard format does not allow sources to dilute their emissions with the emissions from fossil fuel inputs because it directly regulates the emissions and feeds associated with the hazardous waste fuel. Under a thermal emissions format both sources would be required to feed hazardous waste with the same thermal feed concentrations (on a lb HAP per million BTU hazardous waste basis), and source A would be required to process hazardous waste with an equivalent concentration of metal HAP (on a mass basis) and also be required to emit half as much metal HAP (on a mass emission rate basis) relative to source B, because source A is processing half as much hazardous waste fuel, thus vindicating the hazardous waste feed control aspect of the standard (see also note below regarding the likelihood of sources using hazardous waste feed control). Further, the thermal feed concentration with which these sources must comply reflects the feed control of the average performance of the best performing sources (on a mass of HAP per million BTU basis). Such a requirement assures that these sources are processing the cleanest hazardous waste fuels to recover energy and are reducing HAP emissions to MACT levels.

We note that it would not be appropriate to express the emission standards for incinerators, hydrochloric acid production furnaces, and solid fuel boilers in terms of thermal emissions. As just explained, the choice of a normalizing parameter is fitted to the nature of the device to which it is applied in order to allow the most meaningful comparisons between devices of like type. We therefore conclude that a thermal emissions format (i.e., normalizing parameter) for incinerators is not appropriate because the primary function of incinerators is to thermally treat hazardous waste (as opposed to recovering energy from the hazardous waste). See 67 FR at 17362 (April 19, 1996). Our database indicates that most incinerators processed hazardous waste during their emissions tests that had, on average, heating values below 10,000 BTU/lb.[109] We have emission test hazardous waste heating value information for 62 incinerators in our database. Of these 62 sources, 40 sources processed hazardous waste with an average heating value of less than 10,000 BTU/lb. The other 22 sources processed hazardous waste with heating values greater than 10,000 BTU/lb in at least one test condition, although we note that 14 of these 22 sources also processed hazardous waste in different test conditions with heating values lower than 10,000 BTU/lb.[110]

We assessed whether we should subcategorize incinerators, similar to how we subcategorize liquid fuel boilers, based on the BTU content of the hazardous waste. Incinerators do recover energy from processing high BTU wastes. Some incinerators are equipped with waste heat boilers, and high BTU hazardous waste can displace fossil fuels that otherwise would have to be burned to thermally treat low BTU wastestreams. However, such energy recovery is considered to be a secondary product because their primary function is to thermally treat hazardous waste. A Start Printed Page 59453thermal emissions normalization approach for incinerators that combust hazardous wastes with heating values greater than 10,000 BTU/lb would therefore not be appropriate because the normalized parameter would not be tied to the primary production output that results from the processing of hazardous waste (i.e., treated hazardous waste). In confirmation, no commenters suggested that we apply a thermal emissions format to incinerators.

We also conclude that a thermal emission format is inappropriate for hydrochloric acid production furnaces. These devices recover chlorine, an essential raw material in the process, from hazardous waste. The classic normalizing parameter of amount of product (HCl) produced is therefore the obvious normalizing parameter for these sources. It is true that some hydrochloric acid production furnaces recover energy from high BTU hazardous wastes. See 56 FR at 7141/1 and 7141-42 (Feb. 21, 1991). Some sources are equipped with waste heat boilers, and high BTU wastes help sustain the combustion process, which is necessary to liberate the chlorine from the wastestreams prior to recovering the chlorine in the scrubbing systems. Again, energy recovery is not the primary function of these types of sources.[111] Hydrochloric acid production furnace hazardous waste heating values range from 1,100 to 11,000 BTU/lb (the median energy content for these sources is slightly above 6,000 BTU/lb). The range of hazardous waste heating contents from these sources is much lower than the ranges for cement kilns, lightweight aggregate kilns, and liquid fuel boilers, supporting the premise that energy recovery is of secondary importance. In addition, and critically, the hazardous waste that is processed in these units contains high concentrations of chlorine, confirming that the wastes serve as feedstock for hydrochloric acid production, even if the wastes also have energy value.[112] No commenters suggested that we apply a thermal emissions format to hydrochloric acid production furnaces.

We consider the processing of hazardous waste in solid fuel boilers to be more reflective of energy recovery (relative to incinerators and hydrochloric acid production furnaces) because these sources directly recover the heat that is released from the combustion of the waste streams. However, as stated at proposal, not all these sources are processing hazardous wastes for energy recovery. 69 FR at 21220. These boilers are generally not commercial units, and so tend to burn whatever hazardous wastes are generated at the facility where they are located. Heating values for this source category range from 1,300 to 10,500 BTU/lb, with a median value of 8,000 BTU/lb. We therefore conclude that thermal emission standards for these sources are not appropriate because most of these sources are processing hazardous waste with energy content lower than 10,000 BTU/lb. As discussed in section VI.D, we conclude that 10,000 BTU/lb is an appropriate level that distinguishes whether thermal emission standards or mass emission concentration-based standards are appropriate. We also note that no commenters suggested that we apply a thermal emissions format to solid fuel boilers.

Comment: Commenters state that thermal emission standards are inappropriate because sources burning hazardous waste with a higher energy content or higher percent hazardous waste firing rate (i.e., one that fulfills a greater percentage of its total energy demand from the hazardous waste) would be allowed to emit more HAP.

Response: Part of this comment would apply regardless of what normalizing parameter is used. Technology-based standards (including MACT standards) are almost always expressed in terms of some type of normalizing parameter, i.e., “X” amount of HAP may be emitted per unit of normalizing parameter. This allows a meaningful comparison between units of different size and production capacity. A consequence is that the overall mass of HAP emissions varies, but the rate of control remains constant per the normalizing unit. As explained in the introduction to this section, this approach is both routine and permissible.

Cement kilns, lightweight aggregate kilns, and liquid fuel boilers combust hazardous waste to recover valuable energy. Recovering energy is an integral part of their production process. As discussed at proposal, emission concentration-based standards (and the methodology that otherwise would be used to calculate emission concentration-based standards) may result in standards that are biased against sources that recover more energy from hazardous waste. 69 FR at 21219. This may discourage sources from recovering energy from hazardous waste because such standards do not normalize each source's allowable emissions based on the amount of hazardous waste it processes for energy recovery purposes. A source that fulfills 100 percent of its energy demand from hazardous waste would be required to limit its mass HAP emissions to the same levels as an identical source that satisfies, for example, only 10 percent of its energy demand from hazardous waste and 90% from coal. This would inappropriately discourage the safe recovery of energy from hazardous waste, and could in turn result in greater consumption of valuable fossil fuels that otherwise would be consumed.

Sources which fulfill a greater percentage of their energy demand from hazardous waste (either by processing hazardous wastes that are higher in energy content, or by simply processing more hazardous waste) will be allowed to emit more HAP (on a mass emission rate basis) than an identical source that satisfies less of its total energy demand from hazardous waste. This is appropriate because: (1) The source fulfilling a greater percentage of its energy demand from hazardous waste is processing more raw material than the other source (the raw material being the energy content of the waste); and (2) The source fulfilling a lower percentage of its energy demand requirements from hazardous waste would not be allowed to dilute its emissions with nonhazardous waste fuels, and we would thus assure that all sources implement hazardous waste feed control to levels consistent with MACT.[113] This Start Printed Page 59454was illustrated in the example provided in the introduction to this comment response section.

Similarly, two sources that combust hazardous waste with the same energy content and the same metal concentrations (on both a thermal concentration and mass-based concentration basis), but at different hazardous waste firing rates, would be required to achieve identical back-end control device operating efficiencies to comply with a thermal emissions-based standard. Holding these factors constant, thermal emission standards require sources to achieve identical percent reductions of the HAP that is processed within the combustor via removal with an air pollution control device. A thermal emission standard format is thus equally stringent for these sources on a percent HAP removal basis, irrespective of the amount of hazardous waste it processes for energy recovery, and better assures that sources burning smaller amounts of hazardous waste (from an energy recovery perspective) are also controlling emissions as well as the average of the best performing sources.

Sources processing higher energy content hazardous wastes would be allowed to feed hazardous wastes with higher metal and chlorine mass-based concentrations relative to other sources combusting lower energy content wastes. To illustrate this, assume there are two sources (named C and D) with identical back-end control systems and identical mass feedrates of hazardous waste. Also assume the hazardous waste of source C has twice the energy content as compared to the hazardous waste processed by source D. A thermal emission standard will allow Source C to feed a hazardous waste that has twice the metals concentration (as measured on a mass basis) as compared to source D, even though both sources would be required to comply with equivalent thermal feed rates limitations. Notably, however: (1) Source C is displacing (i.e., not using) twice as much valuable fossil fuel as the source with the lower energy content hazardous waste, and is feeding twice as much raw material—the raw material being energy content contained in the hazardous waste; (2) source C cannot exceed the feed control levels (expressed on a lbs of HAP per million BTU basis) that was achieved by the average of the best performing sources (assuming its back-end control efficiency is equivalent to the average performance demonstrated by the best performing sources); and (3) source D is required to have lower mass concentrations of metals in its hazardous waste because it is firing poorer quality hazardous waste fuel (from an energy recovery perspective) and because it is feeding less of the same raw material (measured by energy content). Thus, the thermal emissions format appropriately encourages and promotes the processing of clean, high energy content hazardous waste fuels (consistent with evaluating hazardous waste feed control as an aspect of MACT, and not just relying on control solely through use of back end technology), and does so equally for all sources because it normalizes the allowable emissions based on the amount of energy each source recovers from the hazardous waste. Put another way, source C in the above example is controlling HAP emissions to the same extent as the average of the best performing sources per every BTU of hazardous waste fuel it processes (as is source D).

We note that this is a hypothetical example. In practice the average energy content of hazardous waste processed at cement kilns does not vary significantly across sources. Cement kilns burn hazardous wastes with relatively consistent energy contents because that is what their production process necessitates. This is supported by our database and by comments received from the Cement Kiln Recycling Coalition.[114] Heating values of hazardous wastes processed at cement kilns during compliance tests (information which is included in our database) range from 10,300 to 17,600 BTU/lb, with a median value of 12,400 BTU/lb. We note that these are snapshot representations of hazardous waste heating content from these sources that originate from compliance tests. We also have long term average hazardous waste heating measurements from cement kilns indicating that the heating content of the hazardous wastes on average range from 9,900 to 12,200 BTU/lb, with a median value of 11, 500 BTU/lb. We thus conclude that the commenter's concern regarding sources being allowed to emit more HAP if they process hazardous waste with higher energy content is overstated for these sources.

Energy content of hazardous wastes processed in liquid fuel boilers and lightweight aggregate kilns varies more than energy content of hazardous wastes processed by cement kilns, and sources with higher energy content wastes would be allowed to emit more metals than identical sources burning identical volumes of lower energy content wastes (although the degree of control is identical per BTU of hazardous waste fuel processed).[115] Again, these are hypothetical examples. Each energy recovery unit will have an upper bound on the amount of energy it can process from the hazardous waste. Sources that process higher energy content hazardous wastes would not necessarily feed the same volume of hazardous waste as compared to sources processing lower energy content hazardous wastes because they cannot exceed the thermal capacity of their combustion unit. Under a thermal emission standard format, the mass emission rates that would be allowed for identical sources that fulfill 100 percent of their energy demand from hazardous waste and that have differing hazardous waste energy contents would be identical. Although the source with the higher energy content hazardous waste would have a higher allowable mass-based hazardous waste feed concentration, this source would have to process less hazardous waste (on a mass basis) to remain within its thermal capacity. This helps to ensure that its mass HAP emission rate is similar to other sources that process lower energy content hazardous waste.

One commenter's apparent concern with thermal emissions seems to center on an assertion that sources will intentionally blend nonhazardous, high heating value wastes or fuels with low energy, high metal bearing hazardous wastes in order to increase the energy content of these metal bearing wastes so that they will be subject to higher allowable emissions via thermal emission standards. We specifically address that comment later as it relates to commercial energy recovery units (lightweight aggregate kilns and cement kilns). We note here, however, that we do not consider that comment to be of practical concern for liquid fuel boilers Start Printed Page 59455because they do not engage in commercial fuel blending practices.

Comment: A commenter states that EPA's assessment of thermal emissions to identify the relevant best sources is inappropriate because thermal emissions are not emission levels, but rather a ratio of emissions to the heat content in a source's hazardous waste.

Response: This comment challenges the basic idea of normalization, since the comment would be the same regardless of the normalizing parameter being used. Thermal emissions are emission levels that are normalized to account for the amount of energy (i.e., raw material) these sources recover by processing hazardous waste. Similarly, a mass emission concentration (i.e., μg/dscm) is a ratio of the emissions to the volume of combustion gas that is generated, which normalize emissions to account for differences in the size of the combustion units (as well as differences in production capacity). This rulemaking assesses performance and expresses emission standards in both of these formats; both formats normalize the emissions so that we may better assess emission control efficiencies equally across sources based on the percent of HAP in the feed (whether thermal feed or feed normalized based on combustor size) [116] that is controlled or removed from the stack gas prior to being emitted into the atmosphere. As discussed above, technology-based standards have historically assessed performance after normalizing emissions based on the amount of raw material processed by the given industry sector. Thermal emissions normalize each source's emissions based on the amount of raw material (hazardous waste fuel) it processes, and are therefore appropriate to assess and identify the relevant best performers. Finally, as previously explained, this approach is consistent with both the language of section 112 (d) (2) and (3), and the purpose of these provisions.

Comment: A commenter states that EPA's assessment of thermal emissions to identify the relevant best sources is inappropriate because it ignores HAP emissions attributable to the nonhazardous fuel and raw material.

Response: Thermal emission standards do not directly control HAP emissions attributable to the fossil fuels and raw material, in the sense that we did not assess feed control of fossil fuels or raw materials. However, this issue is not related to our choice to use thermal content of hazardous waste as a normalizing parameter. Rather, the issue is whether feed control of fossil fuels and raw materials is a feasible means of control at all. We have determined that it is not, and that only back-end control (expressed as system removal efficiency) is feasible. Moreover, today's rule controls emissions from HAP in raw material and fossil fuels. All non-mercury metal HAP emissions attributable to fossil fuels or raw material are effectively and efficiently controlled to the level of the average of the best performing sources with the surrogate particulate matter standard, as well as the system removal efficiency component of the SRE/Feed methodology.

Comment: EPA has failed to document sources' actual feedrates. Feedrates are presented either as MTECs (where hazardous waste HAP feedrates are divided by gas flow rates) or as thermal feedrates, (where feedrate is expressed as the mass of HAP per million BTUs of hazardous waste fired). This is impermissible, since it does not measure actual feed levels.

Response: This comment essentially takes the position that it is legally impermissible to normalize standards, i.e., express standards on a common basis. EPA rejects this comment for the reasons stated in the introduction to this section.

Comment: A commenter states that an increasing number of fuel blenders are producing fuels with a minimum heating content and maximum metals content in order to maximize revenues because high metal bearing wastes command a higher revenue on the commercial waste market. The commenter states that thermal emission standards are not appropriate because they are based on the implicit assumption that energy recovery entails metals feed.

Response: Contrary to what the commenter suggests, the thermal emissions format will more likely discourage the alleged practice of fuel blenders producing fuels with a minimum heat content and maximum metals content because the standard limits the allowable metal emissions based on the amount of energy contained in the hazardous waste. Thus, a source with a lower energy waste would have to ensure that the mass concentration of metals is also lower to comply with the thermal emission formatted standard. The source would consequently emit less metals (on a mass basis) because of the lower metal mass concentration in the waste fuel. Thermal emission standards reflect the reality that the hazardous waste fuels that are currently processed safely and efficiently in energy recovery units to displace valuable fossil fuel do in fact contain metal HAP. From a feed control perspective, the thermal emissions format appropriately requires sources to process high energy content hazardous waste fuels that reflect the thermal feed control levels achieved by the average of the best performing sources, and does so equally for all sources because it normalizes the allowable emissions based on the amount of energy each source recovers from the hazardous waste.

Comment: A commenter states that EPA should be concerned that fuel blenders and kilns will use the thermal emission standard format to increase the allowable metals feedrates for their units. The commenter claims that sources could inappropriately convert non-hazardous waste fuel to hazardous waste fuel by simply putting coal in a bunker in which hazardous waste was once stored, or mixing nonhazardous waste fuel oil with hazardous waste. The commenter states that a facility with a low hazardous waste firing rate, and relatively low allowable emissions can become a facility with a high hazardous waste percent firing rate, with higher allowable emissions, simply by ‘creative’ use of the hazardous waste mixture rule. The commenter suggests that EPA clearly state that the hazardous waste thermal emission standards apply only to the hazardous waste portion of the fuel blend mixture. The commenter further suggests that EPA require fuel blenders to report the amount of nonhazardous waste fuel that is contained in the fuel blend, and that cement kilns use this to determine allowable metal feed rates based on the original hazardous waste energy content.

Response: We do not believe hazardous waste combustors will engage in the practice of redesignating their fossil fuels, i.e., coal, as hazardous wastes with creative use of the mixture rule in order to increase their allowable metal HAP emission rate. That would require large quantities of coal to be newly classified as hazardous waste. The coal, and the unit where the coal is stored, would subsequently become subject to all applicable subtitle C requirements, which include storage and closure/post closure requirements. We believe this disincentive will discourage this hypothetical practice.

Moreover, as previously discussed, today's rule does not allow cement kiln or lightweight aggregate kiln emissions to exceed the interim standards. The fact that we are issuing emission Start Printed Page 59456standards for some pollutants in the thermal emissions standard format will not encourage fuel blenders to send more metals to these commercial energy recovery sources because their allowable emission concentrations are, by definition, either equivalent to or more stringent than the current limitations with which they are complying. Thus, even if the fuel blenders and energy recovery units engaged in this practice, they could not emit more metals than they are currently allowed to emit. We therefore conclude that it is not necessary to promulgate complicated regulatory provisions that would increase the reporting and recordkeeping requirements of fuel blenders and energy recovery units in order to address a hypothetical scenario that likely would never occur.

Finally, we note that combustion of certain high HAP metal content wastes is already prohibited under RCRA rules. See 40 CFR 268.3. Such wastes remain prohibited from combustion even if they are mixed with fossil fuel so that the mixture has a higher energy content. U.S. v. Marine Shale Processors, 81 F. 3d 1361, 1366 (5th Cir. 1996) (an unrecyclable hazardous waste is not recycled when it is mixed with a usable non-waste and the mixture is processed). Thus, the dilution prohibition in § 268.3 serves as a further guard against the commenter's concern.

Comment: A commenter states that the thermal emissions format may be problematic because it is based on a flawed assumption that metal HAP from the cement kiln raw material and hazardous waste partition in equal proportions to the total stack gas emissions. The commenter believes that metal retention in the raw materials is higher than the hazardous waste, suggesting that thermal emission standards allow an arbitrary increase in allowable hazardous waste metals emissions. The commenter suggests that EPA require that compliance demonstrations be conducted only under conditions where the metals content in the hazardous waste is significantly higher than the metal content in the raw material to minimize this bias.

Response: The commenter has not provided any emissions data to support this claim, nor does the EPA know of data available that reaches this conclusion. We do not believe there is a significant difference in the partitioning rates of these metals in a cement kiln.[117] Even if there is a difference, this would not result in an arbitrary increase of allowable hazardous waste metals emissions. The thermal emission standards were calculated using thermal emissions data that are based on each source's compliance test. These tests were conducted at hazardous waste feed control levels that represented the upper bound of feed control levels these sources see on a day-to-day basis. To accomplish this, sources spiked metals into the hazardous waste prior to combusting the wastes. The amount of metals that were contained in the hazardous waste streams, after accounting for these spiked metals, far exceeded the metal levels that were contained in the raw material. Thus the differences in partitioning, if any, would likely be overshadowed by the fact that the majority of the metals were contained in the hazardous waste.

Notably, any partitioning bias that that may be present would also have been present during these compliance tests. As a result, this potential bias would be built into the emission standard and thus would not result in an arbitrary increase in allowable hazardous waste metals emissions because these sources will again demonstrate compliance under testing conditions similar to those used to generate the data used to calculate the MACT floors. We conclude that it is not necessary to provide additional prescriptive regulatory language that would require sources to demonstrate system removal efficiencies under testing conditions that exhibit a high ratio of hazardous waste metal content to raw material metal content because the regulations implicitly require sources to demonstrate hazardous waste metal feed control levels that represent the upper range of their allowable feed control levels.[118]

Comment: A commenter states that compliance with standards expressed in a thermal emissions format is problematic because the measurement of energy content of hazardous waste fuel blends is subject to significant variability due to the nature of the test. The commenter also claims that heating value measurements of waste streams that are mixtures of solids and liquids tend be biased high, which would inappropriately give these sources higher allowable metal emission limitation.

Response: There are standard ASTM procedures that reliably measure the energy content of the hazardous waste. Any parameter that is measured for compliance purposes is subject to method imprecision and variability. We do not believe that hazardous waste energy content measurements result in imprecision and variability above and beyond the measurement methods that are currently used to assure compliance with emission concentration-based standards.

The commenter did not provide evidence that supports the claim that energy content measurement and/or sampling methods consistently result in a positive bias. If a bias were consistently present for these types of wastes, then one would expect it to be also reflected in the measured data for which we based the emission standards, which would fully address the commenter's concern. Nonetheless, we note that all hazardous waste sampling and analysis procedures must be prescribed in each source's feedstream analysis plan, which can be reviewed by the permitting authority upon request. These feedstream analysis plans must ensure that sampling and analysis procedures are unbiased, precise, and that the results are representative of the feedstream. See § 63.1208(b)(8). More information on obtaining a representative samples can be found in EPA's SW-846 publication.[119] These procedures involve acquiring several sub-samples that provide integration over the breadth, depth and surface area of the waste container and obtaining replicate samples (see Ch. 13.3.1 of SW-846).

Comment: A commenter states that BTU measurements can be reported as either a higher heating value or a lower heating value, and suggests that EPA require sources to use the lower heating value calculation when determining allowable hazardous waste feed control levels. The commenter seems to imply that use of higher heating values will inappropriately result in higher allowable metal feed rates for fuel blends that contain aqueous waste.

Response: The BTU data in our database that we use to calculate the emission standards reflect higher heating values. It is standard practice in the incineration/combustion industry to report the gross heat of combustion (or Start Printed Page 59457higher heating value). We conclude that sources should use the higher heating value rather than the lower heating value for all compliance determinations because these are method-based emission standards. Fuel blends that contain aqueous wastes will not be inappropriately rewarded with higher allowable feed rates because any fuel mixture that contain aqueous mixtures will have lower reported heating values, irrespective of whether they are reported as higher heating values or lower heating values.[120]

E. Standards Can Be No Less Stringent Than the Interim Standards

Comment: Several commenters oppose EPA's position in the proposed rule that the replacement standards can be promulgated at a level no less stringent than the interim standards for incinerators, cement kilns, and lightweight aggregate kilns. In instances where the calculated replacement standard is less stringent than the interim standard, the commenters oppose EPA's position of “capping” the replacement standard at the level of the interim standard to prevent backsliding from those levels. Instead, commenters recommend that EPA calculate and finalize the existing and new source floor levels without regard to the interim standards. One commenter also notes that the interim standards are simply a placeholder without the necessary statutory basis to qualify as emission limitations for purposes of establishing MACT floors. Another commenter, however, supports EPA's position to prevent backsliding to levels less stringent than the interim standards.

Response: We maintain that the replacement standards can be no less stringent than existing standards, including the interim standards under §§ 63.1203-1205, for incinerators, cement kilns, and lightweight aggregate kilns. These standards were promulgated on February 13, 2002, and sources were required to comply with them no later than September 30, 2003, unless granted a one-year extension (see § 63.1206(a)). Thus, all hazardous waste combustors are currently complying with the interim standards. The comment that the standards lack some type of requisite statutory pedigree misses the central point of our interpretation of the statute: motivation for achieving a standard (be it regulatory compulsion, statutory requirement, or some other reason) is irrelevant in determining levels of MACT floors. National Lime v. EPA, 233 F. 3d at 640. What matters is the level of performance, not what motivated that level.

As a result, the replacement standards promulgated today ensure that sources will emit HAP at levels no higher than levels achieved under current regulations. We do this in this rule, when necessary, by either capping a calculated floor level by the interim standard (when both the calculated floor level and interim standard are expressed in the same format of the standard) or by adopting dual standards in cases where formats of the standard vary (so that comparison of stringency cannot be uniformly determined (as for cement kilns and lightweight aggregate kilns, as explained in the preceding section above and in the following response). In this case, the sources are subject to both the replacement and interim standards.

Comment: One commenter states that some proposed standards expressed in a thermal emissions format would allow some sources to emit semivolatile metals at levels higher than the interim standard. The commenter states that EPA reached incorrect conclusions when making relative stringency comparisons between standards expressed in a thermal emissions and mass concentrations format because, in part, EPA assumed an average F-factor (e.g., semivolatile metals for cement kilns).[121] In addition, the commenter notes that the actual relationship between standards expressed in terms of thermal emissions and mass concentrations is complex and depends on a number of factors. As a result, the commenter urges EPA to adopt dual standards (i.e., promulgate the MACT standard as both the standard expressed in a thermal emissions format and also the interim standard expressed in a mass concentration format) to prevent backsliding.

Response: Even though a source may operate in compliance with a standard expressed in a thermal emission format, a source may or may not also be in compliance with the corresponding mass concentration interim standard (e.g., the semi- and low volatile metal emission standards for cement and lightweight aggregate kilns of §§ 63.1204 and 63.1205, respectively). As reflected in the comment, making a judgment as to whether a replacement standard is more stringent than the interim standard for the HAP is not always a straight-forward calculation. As we discussed in the proposed rule [122] and echoed by the commenter, comparing standards in the thermal emissions format to those in a mass concentration format involves assumptions that vary on a site-specific basis and can vary over time, including the hazardous waste fuel replacement rate, contributions to emissions from nonhazardous waste inputs such as raw materials and nonhazardous waste fuels such as coal, how close to the standard a source elects to comply, the system removal efficiency demonstrated during testing, and the type and composition, including heating value, of fuels burned.

To ensure that sources operating under standards expressed in a thermal emissions format will not emit HAP metals at levels higher than currently achieved under the interim standards, we adopt a dual standard to prevent emissions increasing to levels higher than the interim standards. The dual standard structure includes both the standard expressed in a thermal emissions format and the interim standard, which is expressed in a mass concentration format. We apply this concept to several standards including semivolatile metals, low volatile metals, and mercury [123] for cement kilns and semivolatile metals and low volatile metals for lightweight aggregate kilns. This approach ensures that sources are not emitting HAP metals above the levels of the interim standards because we cannot reliably determine that emissions under a standard expressed in a thermal emissions format would not exceed the interim standard for all sources in the category. See §§ 63.1220(a)(2)-(a)(4), and (b)(2)-(b)(4) and 63.1221(a)(3)-(a)(4) and (b)(3)-(b)(4).

We evaluated the relative stringency of the standards expressed in the thermal emissions format compared to the interim standards for the entire source category in order to determine if the dual standard scheme could be avoided. We determined that we could not. For some HAP groups we found that many sources in the category would have the potential to exceed the interim Start Printed Page 59458standards for that HAP.[124] In this case, we considered simply “capping” the standard expressed in the thermal emission format by the interim standard (i.e., the promulgated standard would only be expressed in a mass concentration format). However, we conclude that this approach would not be appropriate because the standard expressed in a thermal emission format would likely be more stringent than the mass concentration for some sources, and the statute requires that MACT floors reflect this superior level of performance.

In other cases we found that the standards expressed in the thermal emissions format would not likely exceed the interim standards by the majority of sources operating under typical conditions.[125] While our analysis (based on information in our data base) shows in these cases that the emission standard expressed in a thermal emission format would not likely result in an exceedance of the interim standard, this conclusion may not be true because the assumptions may not be valid for a particular source or site-specific factors may change in future operations. For example, HAP metal emissions could increase over time due to increases in HAP contributions from raw materials or alternative raw materials. Given this potential, we adopt dual standards for the HAP metal standards in order to ensure that standards expressed in a thermal emissions format will not exceed emission levels achieved under the interim standards.[126]

Comment: Several commenters state that the interim standards do not reflect the average performance of the best sources, and so cannot be the basis for floor levels.

Response: In those few situations where we have established floor levels at the level of the interim standards, we have done so as the best means of estimating performance of the best performing sources. Based on the available data to us, the average of the best performing sources exceeds the level of the interim standards in a few instances. Under these circumstances, the binding regulatory limit becomes the best means available to us to estimate performance. See Mossville, 370 F. 3d at 1241-42 (accepting regulatory level as a floor standard where sources' measured performance is not a valid means of determining floor levels, and where such data contains results as high as those regulatory levels).

F. How Can EPA's Approach to Assessing Variability and its Ranking Methodologies Be Reasonable When They Result in Standards Higher Than the Interim Standards?

A commenter argued that EPA's floor methodologies, in particular its consideration of variability beyond that demonstrated in single test conditions, the SRE/feed and Air Pollution Control Device methodologies, must be arbitrary because in a few instances projected standards using these approaches were higher than the current interim standards, a level every source (not just the best performers) are achieving. Commenters also noted that one of the new source standards calculated under these approaches was higher than an existing source standard, another arbitrary result.

EPA believes that these seeming anomalies (which are infrequent) result from the database used to calculate performance and standards, rather than from the approaches to assessing variability or the two questioned floor methodologies. The data base is from test results which preceded EPA's adoption of the interim standards. Thus, the level of performance required by the later rule is not necessarily reflected in pre-rule test data. In confirmation, some of the standards computed using straight emission approaches also are higher than the interim standards. Other anomalies arise simply due to scarcity of data (floor levels for certain HAP emitted by lightweight aggregate kilns especially, where there are only nine sources total). In these situations there is a greater likelihood that one or more of the best performing sources will have relatively high emissions because we are required to use data from five sources to comprise the MACT pool whenever we have data from fewer than 30 sources, and a small amount of data can skew the result. See § 112(d)(3)(B).[127]

For example, many of the calculated new source chlorine floors were slightly higher than the calculated existing source standards because we assumed all sources with measured emissions below 20 ppmv were in fact emitting at 20 ppmv (see part four, section I.C). We generally are unable to differentiate a single best performing source among these best performers because many/all of the best performing sources emissions are adjusted to the same emission level. The calculated new source floor can be slightly higher than the existing source floor because the variability factor that is applied to the single best performing source is based on only one test condition (with three emission test runs). This results in a higher level of uncertainty relative to the existing source standard, which is based on a compilation of emissions data from several sources that have essentially the same projected emissions as a result of the method bias correction factor. The variability factor that is applied to the emissions of the single best performing source is therefore higher than the variability factor for the existing source floor because there are fewer degrees of freedom in the statistical analysis.[128] Likewise, many of the calculated solid fuel boiler new source standards were slightly higher than the calculated existing source standards because, as discussed above, there are fewer degrees of freedom when assessing the variability from a single best performing source. The solid fuel boiler “anomalies” also occur using a straight emissions methodology. See USEPA, “Technical Support Document for the HWC MACT Standards, Volume III: Selection of MACT Standards,” September, 2005, Section 19, for further discussion that summarizes and explains these so-called anomalies.

Start Printed Page 59459

IV. Use of Surrogates

A. Particulate Matter as Surrogate for Metal HAP

Comment: A commenter states that EPA's use of particulate matter as a surrogate for nonenumerated metals is unlawful and arbitrary and capricious because although particulate matter emissions may provide some indication of how good a source's end-of stack control of such metals is, it does not indicate what its actual metal emission levels are.[129] The commenter states that emissions of these metals can vary based on metal feed rate without having any appreciable effect on particulate matter emission levels. Thus a particulate matter standard does not necessarily ensure that metal emissions are reduced to the metal emission levels achieved by the relevant best performing sources. To support this assertion, the commenter states that EPA is on record saying “low particulate matter emissions do not necessarily guarantee low metal HAP emissions, especially in instances where the hazardous waste feeds are highly concentrated with metal HAP.” 69 FR at 21221.

Response: The final rule uses a particulate matter standard as a surrogate to control: (1) Emissions of nonenumerated metals that are attributable to all feedstreams (both hazardous waste and remaining inputs); and (2) all nonmercury metal HAP emissions (both enumerated and nonenumerated metal HAP) from the nonhazardous waste process feeds at cement kilns, lightweight aggregate kilns, and liquid fuel boilers (e.g., emissions attributable to coal and raw material at a cement kiln, and emissions attributable to fuel oil for liquid fuel boilers). Incinerators, liquid and solid fuel boilers may elect to comply with an alternative to the particulate matter standard that would limit emissions of all the semivolatile metal HAPs and low volatile metal HAPs. See § 63.1219(e).

The particulate matter standard is a necessary, effective, and appropriate surrogate to control nonmercury metal HAPs. The record demonstrates overwhelmingly that when a hazardous waste combustor emits particulate matter, it also emits nonmercury HAP metals as part of that particulate matter, and that when particulate matter is removed from emissions the nonmercury HAP metals are removed with it.[130] Nonmercury metal HAP emissions are therefore reduced whenever particulate matter emissions are reduced. The particulate matter standard thus is an effective and appropriate surrogate that assures sources are controlling these metal HAP with an appropriate back-end control technology. National Lime v. EPA, 233 F. 3d at 639. The nonenumerated metal HAP are no different than other semivolatile or low volatile metals in that they also will be effectively controlled with a back-end particulate matter air pollution control device.

We also considered the possibility of developing a standard for nonenumerated HAP metals instead of a PM standard (i.e., regulating these metals directly, rather than through use of a surrogate). We conclude for several reasons, however, that issuing emission standards for these nonenumerated metals in lieu of a particulate matter standard would not adequately control nonmercury metal HAPs to levels achieved by the relevant best performing sources.

We generally lack sufficient compliance test emissions data for the noneneumerated metals to assess the relevant best performing sources, because, as discussed below, most of these metals were not directly regulated pursuant to RCRA air emission standards.[131] Although we have more emissions data for these metals that are based on (so called) normal operations, we still lack sufficient emissions data to establish nonenumerated metal standards for all the source categories. Use of normal data may also be problematic because of the concern raised by the cement kiln and lightweight aggregate kiln stakeholders that our normal metals emissions data obtained from compliance tests are not representative of the range of actual emissions at their sources. Cement kiln and lightweight aggregate kiln stakeholders submitted long-term hazardous waste mercury feed control data that support their assertion. Although these stakeholders did not submit long-term normal hazardous waste feed control data for the nonenumerated metals, we can still see that use of the normal nonenumerated metal snapshot emissions in our database to determine MACT floors could raise similar concerns with respect to whether the normal data in fact represents average emissions at these sources, and their level of performance.

Use of particulate matter emissions data to assess the relevant best performers for nonenumerated metal HAP is therefore more appropriate for two reasons. Compliance test data better account for emissions variability and avoid the normal emissions bias discussed above. We also have much more particulate matter emissions data from more sources, which better allows us to evaluate the true range of emissions from all the sources within the source category and to assess and identify the relevant top performing 12 percent of the sources.

It would be inappropriate to assess total stack gas emissions of nonenumerated metals for cement kiln and lightweight aggregate kilns when determining the relevant best performers because these emissions would, in part, reflect the metal feed levels in these sources' nonhazardous waste process feedstreams. This is not appropriate because nonhazardous process feedstream control is not a feasible means of control. See part four, section III.B.1. A potential solution to this problem would be to identify the relevant best performers by assessing each source's hazardous waste thermal emissions for these nonenumerated metals (given that hazardous waste thermal emissions exclude by definition emissions attributable to inputs other than hazardous waste, i.e. raw materials and fossil fuels). This, however, would be problematic because, aside from the data limitation issues, the majority of the nonenumerated metals data reflect normal emissions which often do not contain the highest feed rates used by the source. As a result, we cannot assess performance on a thermal emissions basis because of the uncertainty associated with system removal efficiencies at such low metal feedrates. Furthermore, even if we could issue hazardous waste thermal emissions standards for these metals, a particulate matter emission standard would still be necessary to control nonmercury metal HAP emissions from the nonhazardous waste process feedstreams. Start Printed Page 59460

Emission standards for these nonenumerated metals could require sources to implement hazardous waste feed control (for these metals) to comply with the standard.[132] We are less assured that these sources were implementing hazardous waste feed control for these nonenumerated metals at the time they conducted the emissions tests (which serve as the basis for floor calculations) because most of these metals were never directly regulated pursuant to the RCRA emission standards.[133] This means that sources tended to optimize (or at least concentrate their efforts on) control of the metals that are regulated. Although these metals were being controlled with each source's back-end control device, sources may not have been controlling these metal feedrates because they probably were not subject to specific feedrate limitations (feed control of the enumerated metal HAP does not ensure feed control of these nonenumerated metal HAP). Furthermore, simultaneous feed control of all these metals, when combined with enumerated semivolatile and low volatile metals, may not be possible because the best performing sources for all these metals may collectively represent a hazardous waste feedstream that does not exist in practice (from a combined metal concentration perspective) because there likely would be different best performers for each of the metal HAP or metal HAP groups.[134] We thus conclude that back-end control as measured and assessed by each source's particulate matter emissions is the appropriate floor technology to assess when identifying the relevant best performers for nonenumerated HAP metals and estimating these sources' level of performance.

Comment: A commenter states that EPA's rationale for use of particulate matter as a surrogate for nonenumerated metals is flawed because EPA has provided no data in the proposal to justify its hypothesis that particulate matter is an appropriate surrogate for non-enumerated metal HAP. The commenter also states that the proposed emission standards for particulate matter for existing sources discriminate against boilers and process heaters that burn clean (i.e., little or very low concentrations of HAP metals) hazardous waste fuels. The commenter suggests that if there are sufficient data, EPA should consider developing an alternative emission standard for total HAP metals for new and existing liquid fuel boilers, as was done for the Subpart DDDDD National Emission Standards for Hazardous Air Pollutants for Industrial/Commercial/Institutional Boilers and Process Heaters.

Response: As previously discussed in this section, particulate matter reflects emissions of nonmercury metal HAPs because these compounds comprise a percentage of the particulate matter (provided these metals are fed into the combustion unit). The technologies that have been developed and implemented to control particulate matter also control nonmercury metal HAP. Since non-mercury metal HAP is a component of particulate matter, we can use particulate matter as a surrogate for these metals. Further justification for the use of particulate matter as a surrogate to control metal HAP is included in the technical support document.[135]

We conclude that we do not have enough nonenumerated metal emissions data to calculate alternative total metal emission floors for liquid fuel boilers. The most problematic of these metals are manganese and cobalt, where we have emission data from only three sources. We have much more compliance test particulate matter emissions data from liquid fuel boilers, and thus conclude that the particulate matter standard best reflects the emission levels achieved by the relevant best performers.

Similar to the above discussion, calculating an alternative total metal emissions floor raises questions regarding the method used to calculate such floors. Hazardous waste combustor metal emissions have traditionally been regulated in volatility groupings because the volatility of the metal affects the efficiency of back-end control (i.e., semivolatile metals are more difficult to control than low volatile metals because they volatilize in the combustor and then condense as small particulates prior to or in the emission control device). When identifying the best performing sources, we previously have, in general, only evaluated sources that have metal emissions information for every metal in the volatility grouping. This approach could prove to be problematic since it is not likely many sources will have emissions data for all the metals.

Although we could not calculate alternative total metal emission floor standards based on the available emissions data we have, we agree with the commenters' view that sources that burn hazardous waste fuels with low levels of nonenumerated metals should be allowed to comply with a metals standard rather than the particulate matter standard. We proposed an alternative to the particulate matter standard (see 69 FR at 21331) for incinerators, liquid, and solid fuel boilers that was a simplified version of the alternative particulate matter standard that is currently in effect for incinerators pursuant to the interim standards (see § 63.1206(b)(14)). We received no adverse comment and are promulgating this alternative as proposed. The alternative metal standards apply to both enumerated and nonenumerated metal HAP, excluding mercury. For purposes of these alternative requirements, each nonenumerated metal is classified as either a semivolatile or a low volatile metal and subsequently grouped with the associated semivolatile and low volatile enumerated metals. The semivolatile and low volatile metals standards under this alternative are the same as those that apply to other liquid fuel boilers, but the standard would apply to all metal HAP, not just those enumerated in the generic low volatile metal and semivolatile metal standards. See §§ § 63.1216(e), 63.1217(e) and 63.1219(e).

B. Carbon Monoxide/Hydrocarbons and DRE as Surrogates for Dioxin/Furan

Comment: One commenter states that the dioxin/furan floors for new and existing solid fuel boilers is unlawful and arbitrary and capricious. EPA established the floor for dioxin/furan for these sources as compliance with the carbon monoxide or hydrocarbon standard and the destruction and removal efficiency (DRE) standard. The Start Printed Page 59461commenter states that EPA has not shown that carbon monoxide or hydrocarbon emissions correlate to dioxin/furan emissions, and, accordingly, has not shown that the carbon monoxide or hydrocarbon standard, together with the DRE standard, are valid surrogates.

This commenter also states that it is inappropriate for EPA to use carbon monoxide or hydrocarbons and DRE as surrogates to establish dioxin/furan floors for liquid fuel boilers with wet or no air pollution control devices and for hydrochloric acid production furnaces. The commenter believes EPA inappropriately justifies these surrogates by claiming that a numerical dioxin/furan floor would not be replicable by the best sources or duplicable by the others. The commenter states that EPA has no discretion to avoid setting floors for a HAP just because it believes that HAP is not controlled with a technology. Rather, EPA must set floors reflecting the relevant best sources' actual performance. Such floors necessarily will be duplicable by the relevant best sources themselves. That they cannot be replicated by other sources is irrelevant according to the commenter.

In addition, the commenter states that EPA does not claim or demonstrate that the carbon monoxide and hydrocarbon floors for solid fuel boilers reflect the average emission levels achieved by the relevant best sources.

Finally, the commenter also notes that EPA appears to argue that its carbon monoxide or hydrocarbon standard and DRE standard could be viewed as work practice standards under section 112(h) which allows EPA to establish work practice standards in lieu of emission standards only if it is not be feasible to set the former. Because EPA has made no such demonstration, setting work practice standards to control dioxin/furan emissions from boilers would be unlawful according to the commenter.

Response: The commenter raises four issues: (1) Are the carbon monoxide/hydrocarbon standard and the DRE standard adequate surrogate floors to control dioxin/furan; (2) floors for existing sources must be established as the average emission limitation achieved by the best performing sources irrespective of whether the limitation is duplicable by the best performing sources or replicable by other sources; (3) EPA has not explained how the carbon monoxide and hydrocarbon floors reflect the average emission limitation achieved by the relevant best sources; and (4) EPA cannot establish work practice standards for dioxin/furan under section 112(h) because it has not demonstrated that setting an emission standard is infeasible under section 112(h)(1).

Carbon Monoxide and Hydrocarbons Are Adequate Surrogates to Control Dioxin/Furan when Other Controls Are Not Effective or Achievable. Carbon monoxide and hydrocarbons (coupled with the DRE standard) are the best available surrogates to control dioxin/furan emissions when a numerical floor would not be achievable and when other indirect controls, such as control of the gas temperature at the inlet of a dry particulate matter control device to 400F, are not applicable or effective.[136]

As we explained at proposal, operating under good combustion conditions to minimize emissions of organic compounds such as polychlorinated biphenyls, benzene, and phenol that can be precursors to dioxin/furan formation is an important requisite to control dioxin/furan emissions.[137] See 69 FR at 21274. Minimizing dioxin/furan precursors by operating under good combustion practices plays a part in controlling dioxin/furan emissions, and that role is substantially enhanced when there are no other dominant factors that relate to dioxin/furan formation and emission (e.g., operating a dry particulate matter control device at temperatures above 400F).

Carbon monoxide and hydrocarbons are widely accepted indicators of combustion conditions. The current RCRA regulations for boilers and hydrochloric acid production furnaces use emissions limits on carbon monoxide and hydrocarbons to control emissions of toxic organic compounds. See 56 FR 7150 (February 21, 1991) documenting the relationship between carbon monoxide, combustion efficiency, and emissions of organic compounds. In addition, carbon monoxide and hydrocarbons are used by many CAA standards for combustion sources to control emissions of organic HAP, including: MACT standards for hazardous waste burning incinerators, hazardous waste burning cement kilns, hazardous waste burning lightweight aggregate kilns, Portland cement plants, and industrial boilers; and section 129 standards for commercial and industrial waste incinerators, municipal waste combustors, and medical waste incinerators. Finally, hydrocarbon emissions are an indicator of organic hazardous air pollutants because hydrocarbons are a direct measure of organic compounds.

Commenters on our proposed MACT standards for hazardous waste incinerators, cement kilns, and lightweight aggregate kilns stated that EPA's own surrogate evaluation [138] did not demonstrate a relationship between carbon monoxide or hydrocarbons and organic HAP at the carbon monoxide and hydrocarbon levels evaluated. See 64 FR at 52847 (September 30, 1999). Several commenters on that proposed rule noted that this should not have been a surprise given that the carbon monoxide and hydrocarbon emissions data evaluated were generally from hazardous waste combustors operating under good combustion conditions (and thus, relatively low carbon monoxide and hydrocarbon levels). Under these conditions, emissions of HAP were generally low, which made the demonstration of a relationship more difficult. These commenters noted that there may be a correlation between carbon monoxide and hydrocarbons and organic HAP, but it would be evident primarily when actual carbon monoxide and hydrocarbon levels are higher than the regulatory levels. We agreed with those commenters, and concluded that carbon monoxide and hydrocarbon levels higher than those we established as emission standards for hazardous waste burning incinerators, cement kilns, and lightweight aggregate kilns are indicative of poor combustion conditions and the potential for increased emissions organic HAP. We continue to believe that carbon monoxide and hydrocarbons are adequate surrogates for organic HAP which may be precursors for dioxin/furan formation and note that the commenter did not explain why our technical analysis is problematic.

Emissions that Are Not Replicable or Duplicable Are Not Being “Achieved”. The commenter believes that floors must be established as the average emission limitation of the best performing sources irrespective of whether they are replicable by the best performing sources or duplicable by other sources. To the contrary, emission Start Printed Page 59462levels that are not replicable by the best performing sources are not being “achieved” by those sources and cannot be used to establish the floor.

For solid fuel boilers, we explained at proposal why dioxin/furan emissions are not replicable by the best performing sources (or duplicable by other sources): there is no dominant, controllable means that sources are using that can control dioxin/furan emissions to a particular level. See 69 FR at 21274-75. We explained that data and information lead us to conclude that rapid quench of post-combustion gas temperatures to below 400 °F—the control technique that is the basis for the MACT standards for dioxin/furan for hazardous waste burning incinerators, and cement and lightweight aggregate kilns—is not the dominant dioxin/furan control mechanism for coal-fired boilers. We believe that sulfur contributed by the coal fuel is a dominant control mechanism by inhibiting formation of dioxin/furan. Nonetheless, we do not know what minimum level of sulfur provides significant control. Moreover, sulfur in coal causes emissions of sulfur oxides, a criteria pollutant, and particulate sulfates. For this reason, as well as reasons stated at 69 FR 21275, we are not specifying a level of sulfur in coal for these sources as a means of dioxin/furan control.

The same rationale applies to liquid fuel boilers with no air pollution controls or wet air pollution control systems and to hydrochloric acid production furnaces—there is no dominant, controllable means that sources are using that can control dioxin/furan emissions to a particular emission level.[139] Thus, best performer dioxin/furan emissions are not replicable by the best performing sources (or duplicable by other sources). For these sources, the predominant dioxin/furan formation mechanism for other source categories—operating a fabric filter or electrostatic precipitator above 400F—is not a factor.

Given that these sources are not using controllable means to control dioxin/furan to a particular emission level, there is no assurance that the best performers can achieve in the future the emission level reported in the compliance test in our data base. Put another way, the test data do not reflect these sources' variability, and the variability is largely unquantifiable given the uncertainties regarding control mechanisms plus the environmental counter-productiveness of encouraging use of higher sulfur coal. Hence, that reported emission level is not being “achieved” for the purpose of establishing a floor.

Finally, we note that beyond-the-floor controls such as activated carbon can control dioxin/furan to a particular emission level. If a source were to install activated carbon, it could achieve the level demonstrated in a compliance test, after adjusting the level to account for emissions variability to ensure the measurement was replicable. The commenter argues that such a result is mandatory under the straight emissions approach (the only way the commenter believes best performers can be determined). Doing so, however, would amount to a surreptitious beyond-the-floor standard (forcing adoption of a control technology not used by any existing source), without considering the beyond-the-floor factors set out in section 112(d)(2). In fact, we considered beyond-the-floor standards based on use of activated carbon for these sources—solid fuel boilers, liquid fuel boilers with wet or no emission control device, and hydrochloric acid production furnaces—but rejected them for reasons of cost. The cost-effectiveness ranged from $2.5 million to $4.9 million per gram TEQ of dioxin/furan removed. In contrast, the cost-effectiveness of the beyond-the-floor standard we promulgate for liquid fuel boilers equipped with dry emission control devices is $0.63 million per gram TEQ of dioxin/furan removed.[140]

Consequently, we are not promulgating a beyond-the-floor standard for dioxin/furan for these sources, and do not believe we should adopt such a standard under the guise of determining floor levels.

The Carbon Monoxide and Hydrocarbon Floors Are Appropriate MACT Floors. We explained at proposal why the carbon monoxide standard of 100 ppmv and the hydrocarbon standard of 10 ppmv are appropriate floors. See 69 FR at 21282. The floor level for carbon monoxide of 100 ppmv is a currently enforceable Federal standard. Although some sources are achieving carbon monoxide levels below 100 ppmv, it is not appropriate to establish a lower floor level because carbon monoxide is a conservative surrogate for organic HAP. Organic HAP emissions may or may not be substantial at carbon monoxide levels greater than 100 ppmv, and are extremely low when sources operate under the good combustion conditions required to achieve carbon monoxide levels in the range of zero to 100 ppmv.[141] (See also the discussion below regarding the progression of hydrocarbon oxidation to carbon dioxide and water). As such, lowering the carbon monoxide floor below 100 ppmv may not provide significant reductions in organic HAP emissions. Moreover, it would be inappropriate to establish the floor blindly using a mathematical approach—the average emissions for the best performing sources—because the best performing sources may not be able to replicate their emission levels (and other sources may not be able to duplicate those emission levels) using the exact types of good combustion practices they used during the compliance test documented in our data base. This is because there are myriad factors that affect combustion efficiency and, subsequently, carbon monoxide emissions. Extremely low carbon monoxide emissions cannot be assured by controlling only one or two operating parameters.

We proposed a floor level for hydrocarbons of 10 ppmv even though the currently enforceable standard for boilers and hydrochloric acid production furnaces is 20 ppmv because: (1) Although very few sources elect to comply with the RCRA standard for hydrocarbons rather than the standard for carbon monoxide, those that comply with the hydrocarbon standard have hydrocarbon levels well below 10 ppmv; and (2) reducing hydrocarbon emissions within the range of 20 ppmv to 10 ppmv may reduce emissions of organic HAP.

Although all sources are likely to be achieving hydrocarbon levels below 10 ppmv, it is not appropriate to establish a lower floor level because hydrocarbons are a surrogate for organic HAP. Although total hydrocarbons would be reduced at a floor level below 10 ppmv, we do not know whether Start Printed Page 59463organic HAP would be reduced substantially. As combustion conditions improve and hydrocarbon levels decrease, the larger and easier to combust compounds are oxidized to form smaller compounds that are, in turn, oxidized to form carbon monoxide and water. As combustion continues, carbon monoxide is then oxidized to form carbon dioxide and water. Because carbon monoxide is a difficult-to-destroy refractory compound (i.e., oxidation of carbon monoxide to carbon dioxide is the slowest and last step in the oxidation of hydrocarbons), it is a conservative surrogate for destruction of hydrocarbons, including organic HAP, as discussed above. As oxidation progresses and hydrocarbon levels decrease, the larger, heavier compounds are destroyed to form smaller, lighter compounds until ideally all hydrocarbons are oxidized to carbon monoxide (and then carbon dioxide) and water. Consequently, the relationship between total hydrocarbons and organic HAP becomes weaker as total hydrocarbon levels decrease to form compounds that are not organic HAP, such as methane and acetylene.[142]

Moreover, as discussed above for carbon monoxide, it would be inappropriate to establish the floor blindly using a mathematical approach—the average emissions for the best performing sources—because the best performing sources may not be able to replicate their emission levels (and other sources may not be able to duplicate those emission levels) using the exact types of good combustion practices they used during the compliance test documented in our data base. This is because there are myriad factors that affect combustion efficiency and, subsequently, hydrocarbon (and carbon monoxide) emissions. Extremely low hydrocarbon emissions cannot be assured by controlling only one or two operating parameters.

The Standards for CO and HC Are Not Work Practice Standards. The floor standards for CO or HC for boilers and hydrochloric acid production furnaces are quantified emission limits. The standards consequently are not work practice standards (even though they represent levels showing good combustion control). CAA section 302(k). EPA's reference to section 112(h)(1) at proposal (69 FR at 21275) was consequently erroneous.

C. Use of Carbon Monoxide and Total Hydrocarbons as Surrogate for Non-Dioxin Organic HAP 143

Comment: A commenter states that neither the total hydrocarbon nor carbon monoxide standard alone provides adequate surrogate control for organic HAP. Accordingly, EPA must include standards for both. Hazardous waste combustors could have total hydrocarbon levels below the standard during the carbon monoxide compliance tests, but higher total hydrocarbon levels at other times during normal operation because there are many variables that can affect total hydrocarbon emissions, and these will not all be represented during the carbon monoxide compliance test. The commenter states that EPA is on record stating that carbon monoxide limits alone may not by itself minimize organic emissions because products of incomplete combustion can result from small pockets within the combustion zone where adequate time, temperature, turbulence and oxygen have not been provided to completely oxidize these organics. The commenter also states that EPA is on record stating that total hydrocarbon levels can exceed good combustion condition levels when carbon monoxide levels are below 100 ppmv.

Response: The final rule requires compliance with destruction and removal efficiency and carbon monoxide or hydrocarbon standards as surrogates to control non-dioxin organic HAP emissions [144] from liquid fuel boilers, solid fuel boilers, and hydrochloric acid production furnaces. These are effective and reliable surrogates to control organic HAP. We conclude that simultaneous measurement of both total hydrocarbons and carbon monoxide with continuous emission monitors is not necessary because each serves as a reliable surrogate to control organic HAP emissions. The commenter has cited EPA preamble language that was included in the April 19, 1996 proposed rule for hazardous waste incinerators, cement kilns, and lightweight aggregate kilns. In that rule we proposed to require compliance with both the total hydrocarbon standard and the carbon monoxide standard. We requested comment on whether these requirements were redundant, and we later requested comment on whether we should allow sources to comply with either the carbon monoxide standard or the total hydrocarbon standard. We clarified, however, that allowing sources to comply with the carbon monoxide standard would be contingent on the source demonstrating compliance with the hydrocarbon standard during the compliance test. We believed this was necessary because we had limited data that showed a source could have total hydrocarbon levels exceeding 10 ppmv even though their carbon monoxide emission levels were below 100 ppmv. EPA subsequently promulgated this approach in the September 1999 Final Rule. 62 FR 52829.

Today's rule adopts the same approach for liquid and solid fuel boilers and hydrochloric acid production furnaces. We again conclude that it is not necessary to require sources to verify compliance with both of these standards on a continuous basis with two separate continuous emission monitors, given the redundancy of these measurement techniques. Total hydrocarbon emission measurements are a more direct indicator of organic HAP emissions than carbon monoxide. Hence, continuous compliance with this standard always assures that organic HAP are well controlled. Carbon monoxide is a conservative indicator of combustion efficiency because it is a product of incomplete combustion and because it is a refractory compound that is more thermally stable than hydrocarbons. The hydrocarbon products of incomplete combustion that are simultaneously formed during incomplete, or inefficient, combustion conditions can be subsequently oxidized later in the combustion process. In such instances carbon monoxide will likely still be prevalent in the exhaust gas even though the products of incomplete combustion were later oxidized. The conservative nature of carbon monoxide as an indicator of good combustion practices is supported by our data. At carbon monoxide levels less than 100 ppmv, our data indicates that there is no apparent relationship between carbon monoxide and hydrocarbons (other than that hydrocarbon levels are generally below 10 ppm when carbon monoxide levels are below 100 ppm). For example, a source with a carbon monoxide level of 1 ppm is no more likely to have lower Start Printed Page 59464measured hydrocarbons than a source achieving a carbon monoxide emission level of 100 ppm.[145]

We consider the few instances where the data showed total hydrocarbon levels above 10 ppmv while carbon monoxide levels are below 100 ppmv to be anomalies. Even so, we have accounted for this by requiring compliance with the hydrocarbon standard during the compliance test if a source elects to comply with the carbon monoxide standard. See §§ § 63.1216(a)(5)(i), 1217(a)(5)(i), and 1218(a)(5)(i).

We disagree with the commenter's assertion that the total hydrocarbon compliance demonstration during the compliance test is insufficient. Sources are required to establish numerous operating requirements based on operating levels that were demonstrated during the test, including minimum operating temperature, maximum feed rates, minimum combustion zone residence time, and operating requirements on the hazardous waste firing system that control liquid waste atomization efficiency. Sources must comply with these operating requirements on a continuous basis. Compliance with these requirements, in addition to the requirements to comply with the carbon monoxide and destruction and removal standards, adequately assure sources are controlling organic HAP emissions to MACT levels.

Comment: A commenter states that EPA's proposed use of surrogates for organic HAP do not ensure that each of the organic HAP (e.g., polychlorinated biphenyls and polyaromatic hydrocarbons) are reduced to the level of the HAP emitted by the relevant best performing sources. EPA has not shown the necessary correlation between either the total hydrocarbon or carbon monoxide standards and organic HAP, and neither is a reasonable surrogate according to the commenter.

Response: Carbon monoxide and total hydrocarbon monitoring are widely used and accepted indicators of combustion efficiency, and hence control organic HAP, which are destroyed by combustion.[146] Sources that are achieving carbon monoxide of emission levels of 100 ppm or a hydrocarbon emission levels of 10 ppm are known to be operating pursuant to good combustion practices. This is supported by an extensive data analysis we used to support identical standards for incinerators, cement kilns, and lightweight kilns which were promulgated in the September 1999 Final Rule. We are applying the same rationale to support these standards for boilers and hydrochloric acid production furnaces.

Today's rule requires continuous compliance with either a carbon monoxide and hydrocarbon standard, in combination with a destruction and removal efficiency standard, as surrogates to control organic HAP. We conclude that sources which comply with these standards are operating under efficient combustion conditions, assuring non-dioxin organic HAP are being oxidized, thus limiting emissions to levels reflecting MACT. Efficient combustion of hazardous waste minimizes emissions of organic HAP that are fed to the combustion chamber as well as emissions attributable to products of incomplete combustion that may form within the combustion chamber or post combustion. We are not capable of issuing emission standards for each organic HAP because of data limitations and because such emission standards may not be replicable by individual sources or duplicable by the other best performing sources because of the complex nature of combustion and post combustion formation of products of incomplete combustion.

V. Additional Issues Relating to Variability and Statistics

Many commenters raised issues relating to emissions variability and statistics other than those discussed above in Section III.A: (1) Variability dampening for data sets containing nondetects; (2) imputation of variability to address variability dampening for data sets containing nondetects; and (3) our analysis of variance procedures to identify subcategories. We present comments and responses on the remaining topics below.

A. Data Sets Containing Nondetects

Comment: One commenter states that EPA's approach of assuming measurements that are below detection limits are present at the detection limit dampens the variability of the data set. Thus, the variability of ranking parameters is understated when ranking sources to identify the best performers and emissions variability is understated when calculating the floor.

Response: We agree with the commenter. For the final rule, we use an approach to address nondetects whereby a value is assigned to each nondetect within its possible range such that the 99th percentile upper prediction limit for the data set (i.e., test condition runs for each source) is maximized. Although this approach maximizes the deviation among runs containing nondetect measurements, the test condition average is lower because we no longer assume the nondetect analyte is present at the level of detection. See response to comments discussion below for more information on this statistical approach to address variability of nondetects.

We use this measurement imputation approach to address variability of feedrate data sets containing nondetects for source ranking purposes and to address variability of emissions data sets containing nondetects when calculating floors. We do not apply the measurement implementation approach to system removal efficiency (SRE) data sets where feedrates or emissions contain nondetects, however. Statistical imputation of nondetect SREs is complicated given that SRE is derived from feedrate and emissions data, both of which could contain nondetect measurements.[147] Our inability to apply the imputation approach to SREs is not a major concern, however, because system removal efficiency is used as a source ranking criterion only (i.e., it is not used as the standard, except for hydrochloric acid production furnaces where there are no nondetect feedrate or emissions measurements), and there are few instances where system removal efficiencies are derived from nondetect feedrate or emissions data.

B. Using Statistical Imputation To Address Variability of Nondetect Values

On February 4, 2005, EPA distributed by email to major commenters on the proposed rule a direct request for comments on a limited number of issues that were raised by the public comments on the proposed rule. The nondetect measurement imputation approach discussed above was one of the issues for which we requested comment. We discuss below the major comments on the approach.

Comment: Most commenters state that they agree with either the concept or the approach in principle but cannot Start Printed Page 59465provide substantive comments. These commenters indicate they cannot provide substantive comments because they cannot determine the implications of using the approach given that we did not provide the resulting floor calculations. One commenter suggests that, before blindly applying this arbitrary estimate of a nondetect value, a reality check should be done to validate that this is reasonable by consulting what is published on the method variability, as well as by checking variability factors derived for other data in the database that are above the detection limit.

Another commenter voiced significant concerns with the approach. The commenter states that EPA contradicts its assumption at proposal that all data that are reported as nondetect are present at the detection limits by now admitting that the true value is between zero and the level of detection. The commenter concludes that EPA now proposes to retreat from its assumption that undetected pollutants are always present at the detection limits not because that assumption is false but because it does not generate sufficiently lenient floors. The commenter believes that this underscores that EPA's statistical analysis approach cannot possibly give an accurate picture of any source's actual emission levels. Accordingly, it cannot possibly satisfy EPA's obligation to ensure that its floors reflect the average emission levels achieved by the relevant best performing sources.

The commenter also states that EPA's imputation approach is independently flawed because it assumes—again inaccurately—that the value for a nondetect is always either the highest value or lowest value in the allowable range. In reality the undetected values will necessarily fall in a range between the highest and lowest, and thus yield less variability than EPA would assume.

Response: We agree in theory with the commenter who suggests that the results of the imputation approach should be checked to see if it overstates variability for nondetect data by comparing the results of the imputation approach with the actual variability for detected measurements in the data set. We considered comparing the relative standard deviation derived from the imputation approach for data sets with nondetects, to the relative standard deviation for the data set using a regression analysis. Under the regression analysis approach, we considered relating the relative standard deviation of detected data sets to the average measurement. We would determine this relationship for each standard for which we have nondetect data, and use the relationship to impute the standard deviation for a data set containing nondetects.[148]

We could not perform this analysis, however, because: (1) We have very few detected measurements for the data sets for several standards and could not establish the relationship between relative standard deviation and emission concentration for those data sets; and (2) moreover, for many data sets where detected measurements would have been adequate to establish the relationship, it would have been problematic statistically to extrapolate the relationship to the very low values assigned to the nondetect measurements (e.g., 100% of the detection limit; the value assigned by our statistical imputation approach).[149]

This commenter also suggests that we check the resultant standard deviation after imputation by consulting what is published on the method variability. The commenter did not explain, however, how method variability relates to the variability of nondetect data.

Moreover, we believe that the imputation approach is one approach we could have reasonably used to estimate variability of nondetect data. We first attempted to apply standard statistical techniques to address the nondetect issue. We investigated standard interval censoring techniques to calculate maximum likelihood estimates (MLE) of the average and standard deviation that provide the best fit for a normal distribution for the data containing nondetect values, taking into account that each nondetect data point can be anywhere within its allowable interval. These techniques are not applicable, however, to data sets where all data are nondetects, as is the case for many of our data sets. In that situation, we approximated the mean as the average of the midpoints of the nondetect intervals, and the standard deviation as one half of the possible range of the data.

After working with this MLE/Approximation approach for some time and iteratively developing complicated algorithms to address problems as they arose, we concluded that we needed a simpler approach that could be applied to all data sets. Accordingly, we developed the statistical imputation approach discussed in Section IV.A above.

For 22 separate floors, we compared the results of the approaches we considered for nondetects: (1) Nondetects present at the detection limit (i.e., full detection limit approach); (2) MLE; (3) MLE combined with an approximation approach (i.e., MLE/Approximation approach; and (4) statistical imputation.[150] The MLE approach was only applicable to 2 of the 22 floor data sets, and the numerical algorithm failed to converge on an answer for one of those. The MLE/Approximation approach sometimes results in floors that are unrealistically high (i.e., it calculated 5 of 22 floors that were higher than the statistical imputation approach, which always produces floors that are equal to or higher than assuming nondetects are present at the full detection limit), and sometimes fails to converge on an answer. Because of these limitations, we do not use either the MLE or MLE/Approximation approach.

We believe the statistical imputation approach is preferable to the full detection limit approach because it: (1) Accounts for variability of data sets containing nondetects; (2) can be applied to all data sets containing nondetects; and (3) results in reasonable floor levels. In most cases, floors calculated using statistical imputation are close to those calculated by the full detection limit approach. The statistical imputation approach can produce substantially higher floors than the full detection limit approach, however, when a relatively high nondetect is reported because of a high detection limit. Nonetheless, the statistical imputation approach calculated floors that were 30% higher than the full detection limit approach for only 2 of the 22 floors.

We reject the comment that our approach to handling nondetect data is a mere manipulation to raise the floor. The commenter observes that EPA appears to determine that its initial approach of assuming the worst-case for nondetect data—that the data are present at the detection limit—did not produce floors that were high enough, and consequently applies another manipulation—statistical imputation of nondetect measurements—that assumes the nondetect data are present at lower levels but nonetheless generates floors that are even higher than before. Although the commenter is correct Start Printed Page 59466about the outcome of our handling of nondetect data'the floors are generally higher after statistically imputing nondetect measurements than if nondetects are simply assumed to be present at the detection limit—our rationale for handling nondetects is sound. At proposal, we assumed that nondetects are present at the detection limit. We do not know (nor does anyone else) whether a nondetect value is actually present at 1% or 99% of the detection limit. We thought that assuming that all values were at the limit of detection would reasonably estimate the range of performance a source could experience for these nondetect measurements. This approach inherently maximizes the average emissions but minimizes emissions variability.

Commenters on the proposed rule state that assuming nondetects are present at the detection limit dampens emissions variability—a consideration necessary to ensure that a source's performance over time is estimated reasonably. Mossville, 370 F. 3d at 1242 (daily maximum variability must be accounted for in MACT standards [including floors] which must be achieved continuously). See also CMA, 870 F. 2d at 232 (EPA not even obligated to use data from plants that consistently reported nondetected values in calculating variability factors for best performing plants). We agree with these commenters, and are using the statistical imputation approach to address the concern. Relative to our proposed approach of assuming nondetect measurements are present at the detection limit, the statistical imputation approach reduces the average of the data set for a source while maximizing the deviation of the data set. These are competing and somewhat offsetting factors when calculating the floor for existing sources given that we use a modified 99th percentile upper prediction limit to calculate the floor—the floor is the average of the test condition averages for the best performers plus the pooled variance of their runs. See CMA, 870 F. 2d at 232 (upholding approach to variability for datasets with nondetect values where various conservative assumptions in methodology offset less conservative assumptions).

We further disagree with this commenter's view that the statistical imputation approach is independently flawed because it assumes that the value for a nondetect is always either the highest value or lowest value in the allowable range. The commenter states that, in reality, the undetected values will necessarily fall in a range between the highest and lowest, and thus yield less variability than EPA would assume. Although the commenter is correct that the true value of a nondetect measurement is likely to be in the range between the highest or lowest value possible rather than at either extreme, we do not know where the true value is within that range. To ensure that variability is adequately considered in establishing a floor, the statistical imputation approach, by design, maximizes the deviation by assuming the nondetect value is at one end of the range or the other, whichever results in a higher average for the data set.

C. Analysis of Variance Procedures To Assess Subcategorization

We use analysis of variance (ANOVA) to determine whether subcategories of sources have significantly different emissions. For two subsets of emissions, the variance of the data between the two subsets is compared to the variance within the subsets. The ratio of these two variances is called the F-statistic. The larger the F-statistic the more likely the underlying data distributions are different. To make a decision regarding the difference between the two subsets, we compare this calculated F-statistic to an F-value associated with a particular confidence level.

One commenter has raised several concerns with our use of the ANOVA procedure in the selection of incinerator subcategories.

Comment: The ANOVA procedure is based upon the assumption that the underlying distribution of both data sets has a normal shape. For incinerator emissions data this assumption is not valid. A log-probability plot shows that particulate emission data is better described by a lognormal distribution. Prior to conducting the ANOVA procedure, the data should be log-transformed.

Response: We use probability plots, Skewness Coefficients, and Correlation Coefficient/Shapiro-Wilks testing to evaluate whether it is more appropriate to analyze emissions data for ANOVA and floor calculations assuming the data represent a normal or lognormal distribution. We believe it is reasonable to assume the data represent a normal distribution for several reasons.

The purpose of the ANOVA subcategorization analysis is to determine if there is a significant difference in emission levels between potential subcategories to warrant establishing separate floors for the subcategories. Although in some cases it may appear that a data set in its entirety may be better represented by a lognormal distribution, the high emissions data causing the right-hand skew will be truncated when we identify the best performing sources—those with the lowest emissions—to calculate floors. This moves the appearance of a skewed distribution toward one that is more symmetric and thus, more representative of a normal distribution.

In addition, our analyses showed: (1) The probability plots do not suggest that either assumed distribution is significantly or consistently better; (2) the data set arithmetic averages tend to be in the neighborhood of the medians, indicating the data sets are not significantly skewed and more closely normal than lognormal; and (3) in some cases, neither assumed distribution could be statistically rejected.[151]

Comment: Some of the data sets used for comparison have very few members. This means that the within-group variance for a small data set would have to be very low for the two groups to be judged as separate.

Response: We agree, but note that as the sample sizes change, the critical values are also changing depending on the degrees of freedom.

Comment: Only emissions data were considered in the ANOVA tests. Feed rate and removal efficiency should have been considered as well.

Response: Differences between subcategories in feedrates or system removal efficiency are irrelevant if there is no significant difference in emissions between the subcategories. The purpose of considering subcategorization is to determine if there are design, operation, or maintenance differences between subcategories that could affect the type or concentration of HAP emissions and thus sources' ability to achieve the floor absent subcategorization. Consequently, it is appropriate to consider emissions only when evaluating subcategorization.

Comment: The confidence level used by EPA for the F-statistic in all cases was 95 percent. If the calculated F-statistic were equal to this 95 percent confidence value, it would mean that there is only a 5 percent chance that data for the two subsets were drawn from the same parent distribution. A less stringent (lower) confidence level would be more appropriate for this analysis.

The commenter evaluated particulate emissions for specialty incinerators (i.e., munitions, chemical weapons and mixed waste incinerators) and non-specialty incinerators (all others). The commenter log-transformed the data and Start Printed Page 59467determined that there was only a 30 percent chance that the two data sets could come from the same parent distribution. This result, together with the vastly different operating characteristics for the two types of incinerators, argues for their being treated as separate categories, according to the commenter.

Response: A confidence level of 95% assigns a probability of 0.95 of accepting the hypothesis when there is no difference between subcategories and hence a probability of 0.05 of rejecting a true hypothesis. This reduces the probability to 5% of rejecting a true hypothesis. A less stringent confidence level would increase the chances of rejecting a true hypothesis. The farther apart the averages of the two potential subcategories are, the more likely they are to be statistically different and the more likely you are to be wrong if you hypothesize that they are not different.

A 95% confidence level is most often used for ANOVA because it is generally believed that being wrong one time out of 20 is an acceptable risk for purposes of ANOVA. In addition, statisticians are comfortable with a 95% confidence level because, in a normal distribution, 95% of the data fall within 2 (actually 1.96) standard deviations of the mean.

Other confidence levels could be used for ANOVA—99% or 90%—if there is a good reason to deviate from the general default of 95%. A 99% confidence level is the second most commonly used confidence level and is generally used when it is very important that you be sure that you are right (i.e., where you can only accept the risk of being wrong 1 time out of 100) before you classify the populations (in this case subcategories) as different. Occasionally, but much less frequently, confidence levels of 90% or less are used. But, we note that these situations are so infrequent that some statistics books provide tables for the ANOVA F-statistic only at the 95% and 99% confidence levels.

For these reasons, we believe that the 95% confidence level is an appropriate level among those we could have reasonably selected.

VI. Emission Standards

A. Incinerators

Comment: A commenter states that EPA's subcategorization (and assignment of differing dioxin/furan standards as a result) between incinerators with wet or no air pollution control device and incinerators equipped with dry air pollution control devices or waste heat boilers is unlawful because incinerators equipped with a given type of pollution control equipment are not different “classes,” “types,” or “sizes” of source. The commenter implies that EPA justifies this subcategorization by stating that these sources have different emission characteristics, which is no less unlawful and arbitrary than subcategorizing based on the pollution control devices they use.

Response: We agree that it would not be appropriate to subcategorize source categories based on a given air pollution control technique. See 69 FR at 403 (Jan. 4, 2004). As stated at proposal, we do not subcategorize incinerators with respect to dioxin/furans based on the type of air pollution control device used. 69 FR at 21214. For example, with respect to dioxin/furans, it would not be appropriate subcategorize based on whether a source is using: (1) Good combustion practices; (2) a carbon bed; (3) an activated carbon injection system; or (4) temperature control at the inlet to its dry air pollution control device. These devices and practices are what control dioxin/furan emissions. Today's final rule does not subcategorize based on these control devices and practices. Instead, our subcategorization approach recognizes the potential of some emission control equipment to create pollutant emissions that subsequently must be addressed.[152]

Dioxin/furans are unique in that these pollutants are not typically present in the process inputs, but rather are formed in the combustor or in post combustion equipment. The primary cause of dioxin/furan emissions from incinerators not equipped with waste heat boilers is post combustion formation by surface-catalyzed reactions that occur within the dry air pollution system.[153] This is evidenced by the statistically significant higher dioxin furan emissions for incinerators with dry air pollution control systems compared to those without dry systems.

Incinerators with dry air pollution systems are designed to effectively control metal and particulate matter emissions through use of baghouses, electrostatic precipitators, etc. Incinerators that are designed in this manner have the potential for elevated dioxin/furan emissions because dry air pollution control systems provide locations where surface-catalyzed reactions can occur (e.g., on particles on fabric filter bags or electrostatic precipitator plates). Thus, for purposes of dioxin/furan formation and control, incinerators equipped with dry air pollution systems are in fact different “types” of incinerators because of their unique pollutant generation characteristics.

On the other hand, incinerators with wet air pollution control systems are generally designed to effectively reduce total chlorine emissions (with the use of wet scrubbers) and metals and particulate matter emissions. There generally is a tradeoff, however, in that these types of incinerators may not be as efficient in reducing particulate matter and metal emissions compared to incinerators that are equipped with baghouses and dry electrostatic precipitators. These types of incinerators generally do not have the potential to have elevated dioxin/furan emissions because they do not provide locations where surface catalyzed reactions can occur. For purposes of dioxin/furan emission formation and control, sources with wet air pollution control systems are thus likewise different types of incinerators.[154]

Subcategorizing dry air pollution systems and wet air pollution control systems for purposes of establishing a dioxin/furan standard is no different than subcategorizing incinerators equipped with waste heat boilers. The waste heat boiler is the origin of the dioxin/furan that is generated. These incinerators are designed to efficiently recover heat from the flue gas to produce useful energy. A result of this type of incinerator design, however, is that it also provides a location where surface catalyzed reactions can occur (i.e., the boiler tubes), potentially resulting in elevated dioxin/furan formation (and emissions if not properly controlled).

An alternative approach that does not subcategorize these sources, but rather identifies best performing sources as those sources with the lowest emissions irrespective of whether they have a wet Start Printed Page 59468or dry air pollution control device, would yield floors that would not be achievable unless all the sources, including the best performers, adopted beyond-the-floor technology. The calculated dioxin/furan floor for existing incinerators and liquid fuel boilers using such an approach would be 0.008 and 0.009 ng TEQ/dscm, respectively.[155] All of the best performing sources for these calculated floors had either wet air pollution systems or no air pollution control systems. The floor technology used by these sources is good combustion practices. As a result, these floor levels would not be replicable by these best performing sources nor duplicable by other sources through use of the same good combustion practices because of the uncertainties associated with dioxin/furan generation mechanisms and rates that can vary both within sources and across sources, potentially leading to significant variability in emission levels.[156] Sources equipped with wet or no air pollution systems would thus likely be required to install carbon systems to comply with these standards, a technology used by only four incinerators (none of which were best performers in the above discussed floor analysis). Such an outcome should be viewed as a beyond-the-floor technology and therefore assessed pursuant to the factors enumerated in section 112(d)(2). Furthermore, it is unclear, and perhaps doubtful, that these floors would be achievable by these sources even if they were to install beyond-the-floor controls such as activated carbon systems because no sources using activated carbon are currently achieving those floor levels. We therefore conclude that it is appropriate, and necessary, to subcategorize these types of incinerators for purposes of calculating dioxin/furan floor standards.

B. Cement Kilns

1. Hg Standard

Comment: Several commenters recommend that EPA use a commenter-submitted dataset, which includes three years of data documenting day-to-day levels of mercury in hazardous waste fuels fired to all hazardous waste burning cement kilns, to identify a MACT floor for existing and new cement kilns. Several commenters state that existing cement kilns should have the option to comply with either of the following mercury standards: (1) A hazardous waste feed concentration limit, expressed in ppmw, based on an evaluation of the five best performing sources within the commenter-submitted dataset (documenting day-to-day levels of mercury in the hazardous waste over a three year period); or (2) a hazardous waste maximum theoretical emissions concentration (MTEC), expressed in units of μg/dscm, developed by projecting emissions of the best performing sources assuming mercury concentrations in the hazardous waste were at the source's 99th percentile level in the commenter-submitted dataset. To identify the best performing sources, the commenter suggests selecting the five sources with the lowest median mercury concentrations in the dataset. For existing sources, the commenters' evaluation yields a hazardous waste feed concentration limit of 3.3 ppmw and a stack concentration emission limit of 150 μg/dscm (rounded to two significant figures and considering mercury contributions only from the hazardous waste). For new cement kilns, the commenters recommend a mercury standard in the format of a hazardous waste feed concentration limit only, expressed in ppmw, based on the single source with the lowest 99th percentile level of mercury in hazardous waste. The commenters recommend a mercury standard of 1.9 ppmw for new sources.

Response: We agree with commenters that the commenter-submitted dataset documenting the day-to-day levels of mercury in hazardous waste fuels fired to all hazardous waste burning cement kilns is the best available data to identify floor levels for existing and new cement kilns. See discussion in Part Four, Section I.D. However, we disagree with the commenters' suggested format of the mercury standard for existing sources. Establishing the mercury standard as the commenters' suggest (i.e., 3.3 ppmw in the hazardous waste feed or 150 μg/dscm as a hazardous waste MTEC) fails to consider the interim mercury standards. As discussed in Part Four, Section III.E, there can be no backsliding from the levels of performance established in the interim standards. While not every source feeding hazardous waste with a maximum mercury concentration of 3.3 ppmw would exceed the interim standard, most sources using more than 50 percent hazardous waste as fuel (i.e., replacing at least half its fossil fuel with hazardous waste) would exceed the interim standard, emitting mercury higher than the levels allowed under §§ 63.1204(a)(2) and 63.1206(b)(15) of the interim standards.[157] The hazardous waste MTEC of 150 μg/dscm calculated by the commenters is also higher than the level currently allowed under § 63.1206(b)(15) of the interim standards. Since sources cannot backslide from the levels of the interim standards, if we were to accept the commenters' floor analysis results as presented (which we are not), then we would “cap” each calculated standard (i.e., 3.3 ppmw hazardous waste feed concentration and 150 μg/dscm in stack emissions) at the interim standard level. This would result in a mercury standard for existing sources of 3.3 ppmw hazardous waste feed and a hazardous waste feed MTEC of 120 μg/dscm or 120 μg/dscm as a stack gas concentration limit. We note this is similar to the mercury standard adopted today: a hazardous waste feed concentration limit of 3.0 ppmw and a hazardous waste feed MTEC of 120 μg/dscm or 120 μg/dscm as a stack gas concentration limit. For an explanation of why we derived a level of 3.0 ppmw from the data, see Section 7.5.3 of Volume III of the Technical Support Document.

The commenters' suggested new source mercury standard of 1.9 ppmw in the hazardous waste has the same deficiency. New sources with a hazardous waste fuel replacement rate of approximately 75% could emit mercury at levels higher than currently allowed under the interim standards. After capping the calculated standard at the interim standard level, we would identify the mercury standard for new sources as a hazardous waste concentration limit of 1.9 ppmw in the hazardous waste and a hazardous waste feed MTEC of 120 μg/dscm or 120 μg/dscm as a stack gas concentration limit. For reasons discussed in Section 7.5.3 of Volume III of the Technical Support Document, this is indeed the mercury standard we are promulgating for new cement kilns.

The commenters also suggest that the best performing sources should be identified as those with the lowest three-year median concentration of mercury in hazardous waste. Although this approach would be permissible, we conclude that it is more appropriate to identify the best performers (or single best performer for new sources) by Start Printed Page 59469selecting those with the lowest 99th percentile upper level mercury concentrations. (This is not a statistically determined upper prediction limit; there is sufficient data for an arithmetically calculated 99th percentile to reliably reflect sources' performance.) We believe that this approach best accounts for the variability experienced by best performing sources over time.

A detailed discussion of the MACT floor analysis for existing and new cement kilns is presented in Section 7.5.3 of Volume III of the Technical Support Document. In summary, the mercury standard for existing cement kilns is 3.0 ppmw in the hazardous waste feed and 120 μg/dscm as a hazardous waste maximum theoretical emission concentration feed limit or 120 μg/dscm as a stack gas concentration limit. For new sources the mercury standard is 1.9 ppmw in the hazardous waste feed and 120 μg/dscm as a hazardous waste maximum theoretical emission concentration feed limit or 120 μg/dscm as a stack gas concentration limit.[158]

Comment: Two commenters oppose EPA's proposed approach to base compliance with the mercury standard on averaged annual emissions. The commenters state an annual average would allow mercury emissions to exceed the interim standard because a source could burn high concentrations of mercury waste over a short period and still comply with an annual limit by burning low concentration wastes at other times. These commenters support the concept of a 12-hour rolling average feedrate limit (i.e., the current requirement under the interim standards) in conjunction with an emission standard no less stringent than the interim standard.

Response: We agree with these comments. Cement kilns must establish a 12-hour rolling average feedrate limit of mercury to comply with these standards. The mercury standards for cement kilns are “capped” at the interim standard level to prevent backsliding from the current level of performance. This is accomplished by expressing the standard as a limit on the mercury concentration in the hazardous waste (with the rolling average) and either an emission concentration limit or hazardous waste maximum theoretical emission concentration feed limit. See § 63.1209(l)(1)(iii).

2. Total Chlorine

Comment: One commenter states that the proposed MACT floor approach is inconsistent with the statutory definition of MACT because EPA's selection of a routinely achievable system removal efficiency (SRE) was arbitrary and not representative of the best performing sources. Instead, the commenter suggests EPA identify a MACT SRE based on the five sources with the best SREs and apply that SRE to the MACT chlorine feed level. Later, in supplemental comments, the same commenter suggests two alternative approaches to identify a floor level. One approach applies a ranking methodology based on emissions and chlorine feed, and the second suggested approach applies a triple ranking method based on emissions, feed, and chlorine SRE. Other commenters, however, supported EPA's proposed approach.

Response: We are adopting the same approach we proposed at 69 FR at 21259. As we explained, this is a variant of the SRE/Feed approach, the variant involving the degree of system removal efficiency achieved by the best performing sources. In summary, to determine the floor level we first identify the best performing sources according to their hazardous waste chlorine feedrate. The best performing sources are those that have the lowest maximum theoretical emissions concentration (MTEC), considering variability. We then apply an SRE of 90 percent (the specific point in contention) to the best performing sources' total MTEC (i.e., thus evaluating removal of total chlorine across the entire system, including chlorine contributions to emissions from all feedstreams such as raw materials and fossil fuels) to identify the MACT floor, which is expressed as a stack gas emissions concentration in parts per million by volume. This approach defines the MACT floor as an emission level that the best performing sources could achieve if the source limits the feedrate of chlorine in the hazardous waste to the MACT level (i.e., the level achieved by the average of the best performing five sources) while also achieving an SRE that accounts for the inherent variability in raw material alkalinity and (to a lesser degree) cement kiln dust recycle rates, and production requirements. 69 FR at 21259.

Under this approach, we are evaluating hazardous waste feed control as we do for other sources. One commenter objects to our determination that an SRE of 90 percent is representative of the best performing sources because we have not established a MACT SRE—the average SRE achieved by the best performing sources.

There is no doubt that the cement manufacturing process is capable of capturing significant quantities of chlorine when favorable conditions exist within the kiln system. Our usual approach of establishing an SRE by ranking the most efficient SREs taken from individual compliance tests, however, would result in a standard that would not be achievable because it may not be duplicable by the best performers or certainly would not be replicable by others, given that it is a function of various highly variable parameters, especially levels of alkali metals (e.g., sodium and potassium) and volatile compounds (e.g., chlorine and sulfur) in the raw materials. Alkalis and volatiles vary at a given best performer facility (in fact, at all facilities) as different strata are mined in the quarry, and across facilities due to different sources of raw materials. Raw material substitution is infeasible and counter to the objective of producing quality product (i.e., a product with low alkali content).

Cement kilns thus are not able to design or operate to achieve a specific SRE at the high (most efficient) end of the range of test conditions. This is demonstrated by our calculations of system removal efficiency data, which is essentially a collection of performance “snapshots.” See SRE data summarized in Table 1 at the end of this response; see also Mossville, 370 F. 3d at 1242 (maximum emission variability associated with raw material variability needs to be accounted for in MACT floor determination since the standard must be met at all times under all operating conditions). The performance data of the “apparent” best performers—upwards of 99 percent—identified by the commenter are simply a snapshot in the possible range of performance and are not replicable in the future due to factors which are uncontrollable by the source, as just explained. In confirmation, cement kilns achieving this level of removal in one test proved incapable of replicating their own result in other tests even though individual sources each have their own proprietary source of raw materials. See results in table for Giant (SC), Essroc (IN), Holcim (MO), Giant (PA), and LaFarge (KS) all Start Printed Page 59470of whom would violate a 99 + percent standard based on their own operating results.

Table 1.—Summary of System Removal Efficiency Data for Wet Process Cement Kilns 159

FacilityNumber Runs in Data BaseLow SRE Run (%)High SRE Run (%)Average SRE of All Runs (%)
LaFarge (OH)399.199.499.3
Giant (SC)2495.599.899.0
Essroc (IN)1397.399.998.7
Holcim (MO)696.499.998.4
LaFarge (KS)1295.799.398.1
Giant (PA)1787.799.497.1
Continental (MO)395.797.096.5
Ash Grove (AR)3785.198.895.1
Texas Industries (TX)688.897.093.6
Holcim (MS)976.599.290.0
159 See Section 3.6 of Volume II (Specific MACT Standards) of Comment Response Document, September 2005.

However, the data indicate that SRE is reasonably quantifiable to a point. Based on our data base of system removal efficiency information from 130 test conditions where total chlorine was evaluated, we conclude that a system removal efficiency of 90 percent is a reasonable estimate of MACT SRE.[160]

We also reject the commenter's three suggested alternative approaches to identify a MACT SRE to apply to the MACT feed level. The commenter's methods all suffer a common flaw: They fail to recognize and take into account the limitations of the total chlorine SRE data. For example, as just demonstrated, available data show that considering the SRE data associated with the most recent compliance test as a ranking factor will result in unachievable standards due to the varying effectiveness of chlorine capture (which impacts emissions) depending on the raw material mix characteristics. Considering only the most recent compliance test data as suggested yields results that are unachievable because the best performer's SRE data are likely biased high (e.g., sources that happen to test under favorable conditions are likely to be identified as best performers), which would not be replicable by even that source on a day-to-day basis.

3. Semivolatile and Low Volatile Metals

Comment: Commenters oppose EPA's proposed approach to treat each kiln as a separate and unique source in the SRE/Feed MACT floor analysis for cement kilns.[161] Commenters state that the approach is an improper way to perform a statistical analysis and reduces the variability in emissions that otherwise would be observed in a MACT pool of five unique sources. Variability is reduced because co-located kilns at the same plant share many of the factors that comprise front-end and back-end controls. As a result, the calculated MACT floors for SVMs and LVMs for cement kilns are too stringent. The commenters' recommended solution (in instances where co-located kilns are among the top five performers) is to use only the data from the best performing co-located kiln, exclude any lesser performing kilns at the plant site, and then include the next-best performing non-co-located kiln in the MACT pool. Implementing their recommendation, the commenters state that the MACT floor for SVMs increases from 4.0 × 10−4 to 7.4 × 10−4 lbs/MMBtu and the floor for LVMs increases from 1.4 × 10−5 to 1.8 × 10−5 lbs/MMBtu. Another commenter generally supports EPA's approach noting that the variability factor applied to the emissions data already accounts for variability.

Response: We consider sources that are not identical as unique sources and emissions data and information from unique sources are considered separate sources in the floor analyses. An example of an “identical” source in our data base is compliance test data from a similar on-site combustion unit used in place of a compliance test for another unit (i.e., emissions testing of an identical unit was not conducted). These sources and their associated data are called “data in lieu of” sources in our data based on the RCRA provisions under § 266.103(c)(3)(i). We acknowledge that co-located sources may in fact share certain similar operation features (e.g., use of raw material from the same quarry, use of the same coal and hazardous waste burn tank to fire the kilns); however, given that the co-located sources (except those designated as data in lieu of) are not designed identically, and given their hazardous waste feed control levels were not identical during testing, we conclude we must consider each source as a unique source in the floor analyses.[162]

Comment: Commenter states that EPA's proposed standards for new cement kilns are unachievable due to problems with its accounting for variability, in part because EPA did not consider geographic differences when assessing feed control levels. The concentrations of hazardous constituents in the waste in a particular region are likely to be different than in the waste from another geographical region due to types of industrial sectors located within each region. Sources cannot reasonably arrange for transportation of lower HAP wastes generated across the country and cannot treat the hazardous waste to remove or reduce HAP concentrations. The commenter cites several court decisions that support their assertions. Commenter believes that while this represents a problem for developing both the new and existing source floors, it is a greater predicament for the new Start Printed Page 59471source floor because this floor level is based on test data for only one source.

Response: We are not obligated to account for varying hazardous waste feed control levels occurring because of differing HAP generation rates in different locations (for commercial sources), or because different production process types generate higher or lower levels HAP concentration wastes. Hazardous waste feed control is a legitimate control technology. The commenter seems to suggest that we should subcategorize low feeding sources and high feeding sources based on their hazardous waste feed control level. This would inappropriately subcategorize sources based on differing levels of controls, which we do not do. See 69 FR at 403 (January 5, 2004). Nonetheless, as previously discussed, the SRE/Feed methodology lessens the impact of feed control variations across commercial units because it results in fewer situations where best performing back-end controlled sources (from a particulate matter emissions perspective) cannot achieve the semivolatile and low volatile metal design levels and floors.

For new source standards, the single best performing cement kiln sources for semivolatile metals and low volatile metals were not the lowest hazardous waste feed controlled source (both floors were based on sources with the fourth best, (i.e., lowest, hazardous waste feed control level). We therefore do not believe these sources are atypically low hazardous waste feeders relative to the other best performing sources in the existing source MACT pools.

C. Lightweight Aggregate Kilns

1. Mercury Standard

Comment: One commenter, an operator of lightweight aggregate kilns subject to this rule, recommends that EPA establish the mercury standard for lightweight aggregate kilns at a hazardous waste feed concentration limit of 3.3 ppmw for existing sources and 1.9 ppmw for new sources, which is the same standard suggested in public comments by a trade organization representing hazardous waste burning cement kilns. The commenter notes that these mercury limits are appropriate for lightweight aggregate kilns because the commenter's two lightweight aggregate manufacturing facilities participate in the same hazardous waste fuel market as the majority of cement kilns. Moreover, the commenter maintains that its parent company also owns and operates two cement kilns and that its lightweight aggregate kilns receive hazardous waste from many of the same generators that provide hazardous waste fuel to the cement kilns. Consequently, the commenter states that the cement industry's data set of actual mercury feed concentrations in the hazardous waste best represents the full range of hazardous waste fuel concentrations that exist in the waste fuel market (see also Part Four, Sections I.D and E).

Response: We disagree with the commenter. Although the cement industry's set of mercury feed concentration data in the hazardous waste may represent the full range of concentrations for the cement kiln source category, we cannot conclude the same for lightweight aggregate kilns because the commenter states that the mercury dataset are only applicable to its kilns.[163] Further, the commenter provides no specific information or data to support the conclusion that its suggested approach is justified for the other lightweight aggregate kiln facility.

We also disagree with the commenter as to the appropriateness of establishing the mercury standard in the format of a hazardous waste feed concentration (i.e., 3.3 ppmw for existing sources and 1.9 ppmw for new sources) for lightweight aggregate kilns. A hazardous waste feed concentration standard is improper for this source category because one lightweight aggregate kiln facility's sources (although not the commenter's) controls mercury emissions using wet scrubbing. Thus, a hazardous waste feed concentration standard would inappropriately limit the mercury concentration in hazardous waste for sources that use control equipment capable of capturing mercury. A source with control equipment should not be restricted to a hazardous waste feed concentration standard that is based on sources that can only control mercury emissions through limiting the amount of mercury in the hazardous waste.

In any case, as explained earlier in our discussion of cement kiln mercury standard, we believe that it is preferable to establish an emission standard to assure that the actual amount of mercury emitted by these sources is controlled by means of a numerical standard for stack emissions.

Comment: One commenter agrees that a source may not be able to achieve the mercury standard due to raw material contributions that might cause an exceedance of the emission standard in spite of a source using properly designed and operated MACT floor control technologies, including controlling the levels of metals in the hazardous waste. The commenter opposes the proposed alternative standard of 42 μg/dscm, which is expressed as a hazardous waste maximum theoretical emissions concentration. Instead, the commenter suggests that EPA maintain the alternative standard options of §§ 63.1206(b)(15) or 63.1206(b)(9).

Response: We agree with the commenter that the mercury standard should address the concern of raw material contributions causing an exceedance of the emission standard. We also agree that the proposed alternative standard of a hazardous waste maximum theoretical emissions concentration of 42 μg/dscm is an improper standard because the underlying data are unrepresentative. See discussion in Part Four, Section I.E. We note that the mercury standard promulgated today is 120 μg/dscm as a stack gas concentration limit or 120 μg/dscm as a hazardous waste maximum theoretical emission concentration feed limit. The alternative mercury standard sought by the commenter under § 63.1206(b)(15) is a limit of 120 μg/dscm as a hazardous waste maximum theoretical emission concentration, which is included in the mercury standard promulgated today. This should address the commenter's concern.

Comment: One commenter supports a mercury standard with short-term compliance limits (e.g., 12-hour rolling average feedrate limits) as opposed to the annual limit proposed.

Response: For reasons discussed in Part Four, Section I.E, we are using a different mercury dataset than at proposal. We solicited comment on a floor approach using these data in a notice [164] sent directly to certain commenters. We are adopting that approach today. The monitoring requirements of the mercury standard for lightweight aggregate kilns includes short-term averaging periods (i.e., not to exceed a 12-hour rolling average), as recommended by the commenter.

2. Total Chlorine Standard

Comment: One commenter supports excluding from the floor analysis all lightweight aggregate kiln sources that lack air pollution control devices for chlorine, such as scrubbing technology. The floor analysis should simply exclude sources without back-end controls according to the commenter. Start Printed Page 59472

Response: We disagree. For the final rule, we are using the SRE/Feed MACT floor approach which defines best performers as those sources with the best combined front-end hazardous waste feed control and back-end air pollution control efficiency. The commenter's suggestion would exclude emissions data from two of the three facilities in this source category even though valid emissions data from these sources are available (and therefore ordinarily to be used, see CKRC, 255 F. 3d at 867), and these sources achieved the best front-end hazardous waste feed control in the category. We note that the best feedrate controlled sources have hazardous waste thermal feed levels that are approximately one-fifth the level of the source's with back-end controls. These data describe the level of performance of sources in the category and must be evaluated in the MACT floor analysis. We also note that even if we were to implement the commenter's suggestion, the MACT floor results would not change for existing and new lightweight aggregate kilns because the total chlorine emissions data of the source with back-end air pollution controls (after considering variability) are higher than the standards promulgated today. Thus, the commenter's suggestion also would result in a standard that would be capped by the interim standard.

3. Beyond-the-Floor Standards

Comment: One commenter opposes EPA's proposed decision to promulgate a beyond-the-floor standard for dioxin/furans for existing and new lightweight aggregate kilns based on performance of activated carbon injection.

Response: For the final rule, we conclude that a beyond-the-floor standard for lightweight aggregate kilns is not warranted. The Clean Air Act requires us to consider costs and non-air quality impacts and energy requirements when considering more stringent requirements than the MACT floor. In the proposed rule, we estimated that the incremental annualized compliance costs for lightweight aggregate kilns to achieve the beyond-the-floor standard would be approximately $1.8 million and would provide an incremental reduction in dioxin/furan emissions of 1.9 grams TEQ per year (see 69 FR at 21262). At proposal we judged costs of approximately $950,000 per additional gram of dioxin/furan TEQ removed as justified, and, therefore, we proposed a beyond-the-floor standard. Since proposal, we made several changes to the dioxin/furan data base as the result of public comments. One implication of these changes is a lower national emissions estimate for dioxin/furans for lightweight aggregate kilns. We now estimate an incremental reduction in dioxin/furan emissions of 1.06 grams TEQ per year with costs ranging between $1.6 and $2.2 million per additional gram of dioxin/furan TEQ removed. Based on these costs and consideration of the non-air quality impacts and energy requirements (including more waste generated in the form of spent activated carbon, and more energy consumed), we conclude that a beyond-the-floor standard for existing and new lightweight aggregate kilns is no longer justified. For an explanation of the beyond-the-floor analysis, see Section 12.1.2 of Volume III of the Technical Support Document. We note that EPA also retains its authority under RCRA section 3005(c) (the so-called omnibus permitting authority) by which permit writers can adopt more stringent emission standards in RCRA permits if they determine that today's standards are not protective of human health and the environment.

D. Liquid Fuel Boilers

1. Mercury Standard Not Achievable When Burning Legacy Mixed Waste

Comment: One commenter states that the proposed liquid fuel boiler mercury standard is not achievable by a commercial boiler, DSSI (Diversified Scientific Services, Inc.) that burns mercury-bearing low level radioactive waste that is also a hazardous waste (so-called ‘mixed waste’) that was generated years ago (so-called, legacy waste). The waste is an organic liquid containing high concentrations of mercury. The boiler is equipped with a wet scrubber which provides good mercury control—93%, system removal efficiency according to the commenter.

The commenter states that the proposed liquid fuel boiler mercury standard is not achievable using feedrate control and/or additional back-end control. Waste minimization is not an option because the waste has already been generated. Further, available national treatment capacity for mercury-bearing, low-level radioactive organic hazardous waste is very limited. The only other hazardous waste combustion facility authorized to treat such waste is the Department of Energy incinerator at Oak Ridge, Tennessee. Waste treatment volumes at that facility are restricted by the mercury feed rate limitation for the incinerator. In addition, the feedrate of the waste cannot be practicably reduced because of the large back-log of waste that must be treated.

The commenter suggests that their boiler be subject to the incinerator mercury standard because the mixed waste has far higher concentrations of mercury than wastes burned by other boilers and, as a consequence, the boiler is more incinerator-like with respect to the feedrate of mercury.

Response. We agree with the commenter's suggestion. The final rule subjects this commercial liquid fuel boiler to the mercury standard for incinerators. We are classifying this source as a separate type of source for purposes of the mercury standard, because the type of mercury-containing waste it processes is dramatically different from that processed by other liquid fuel boilers, effectively making this a different type of source for purposes of a mercury standard [165] . The source thus feeds mercury at concentrations exceeding that of any boiler but at concentrations within the range processed by hazardous waste incinerators. The maximum test condition average MTEC [166] for mercury for the remaining liquid fuel boilers is 20 μg/dscm. All the liquid fuel boiler mercury data represent “normal” data, i.e., data that were not spiked. (The lack of spiked data in the liquid fuel boiler data base, in and of itself, indicates that these sources do not process mercury-bearing waste and do not need the operational flexibility gained by spiking to account for occasional higher concentration mercury wastes.) DSSI's 2002 mercury test condition average MTEC was spiked to 3500 μg/dscm. In other words, DSSI needs the operational flexibility to feed 175 times more mercury than any other liquid fuel boiler. Incinerators, on the other hand, had mercury MTECs that ranged to 110,000 μg/dscm in 2002. In fact, DSSI's mercury feed rate is the eighth highest of the 40 incinerators, including DSSI, for which we have 2002 mercury feed rate data. DSSI's process feed is thus within the upper range of mercury feed found at incinerators.

We believe it is well within the broad discretion accorded us in section 112(d)(1) to subcategorize among “types” and “classes” of sources within a category. See also Weyerhaeuser v. Costle, 590 F. 2d at 254, n. 70 (D.C. Cir. 1978) (similar raw waste characteristics justify common classification) and Chemical Manufacturers Ass'n v. EPA, 870 F. 2d 177, 253-54 and n. 340 (5th Start Printed Page 59473Cir. 1989) (same). We note that this boiler will be subject to the liquid fuel boiler standards for all HAP other than mercury (the only HAP where the issue of appropriate classification arises).

Not surprisingly, given the disparity in waste concentration levels, the DSSI boiler, even though equipped with back end control comparable to best performing commercial incinerators, achieves mercury emission levels less than an order of magnitude higher than the other hazardous waste-burning liquid fuel boilers, few of which use back end control that is effective for mercury.[167] This emission disparity likewise indicates that DSSI is treating a different type of waste than other liquid fuel boilers.

The nature of the mercury-bearing waste further confirms that it is of a different type than that processed by other hazardous waste burning liquid fuel boilers. The waste is a remediation waste, a type of waste burned routinely by commercial hazardous waste incinerators but almost never by a liquid fuel boiler.

Moreover, the waste is a legacy, mixed waste generated decades ago in support of the United States' strategic nuclear arsenal. It is not amenable to the types of control all other liquid fuel boilers use to reduce mercury emissions—some type of feed control or other minimization technique. We investigated whether any waste minimization options are feasible for this waste, and find that they are not. Normally, waste minimization is accomplished by one of three means: eliminating the use of mercury in the process to prevent it from being in the waste; pretreating the waste before burning to remove the mercury; or sending it to another facility better suited to handle the waste. Changing the production process to eliminate or reduce the mercury content of the waste is not an option because this waste has already been generated. Pretreatment is already practiced to the maximum extent feasible by settling out and separating the heavier mercury from the liquid components after thermal desorbtion. The remaining organic liquid that is burned by the mixed waste boiler contains concentrations of mercury (in organo-mercury and other organic soluble forms) that are orders of magnitude higher than burned by other liquid fuel boilers. Much of the waste cannot be feasibly pretreated to remove mercury because this legacy, mixed waste comes from many highly diverse sources. It is not practical or feasible to investigate how to remove the mercury from wastes of such varied and unique origins.

Only one other facility could potentially treat this mixed waste, DOE's incinerator at Oak Ridge, Tennessee, whose permit allows the incinerator to manage mixed waste. However, waste treatment volumes for mercury-bearing wastes at that facility are restricted by the mercury feed rate limitation in the incinerator's permit. The DOE incinerator alone cannot assure national capacity for mercury-bearing, low-level radioactive organic hazardous waste. In addition, the back-end emission controls of the mixed waste boiler are superior to those used by most incinerators, including the Oak Ridge incinerator. This boiler uses a highly effective wet scrubbing system—the principal MACT floor back-end control for mercury used by incinerators—that achieves over 93% system removal efficiency. This is superior control compared to most incinerators, including the one at Oak Ridge which achieves 75 to 85% removal.[168]

Thus, this mixed waste boiler is reasonably classified a different type of source with respect to mercury waste than other hazardous waste-burning liquid fuel boilers, based on the nature of the waste burned and confirmed by the source's mercury emissions. We note that, although the final rule subjects only the DSSI mixed waste boiler to the incinerator mercury standard, we would conclude that any other liquid fuel boiler with the same fact pattern (i.e., that met the same criteria as the DSSI boiler as discussed above) should also be subject to the incinerator mercury standard rather than the liquid fuel boiler mercury standard.

Comment. One commenter states that EPA's standards for all sources must reflect the actual emission levels achieved by the relevant best sources. If EPA wishes to subject the boiler source and incinerators to the same emission standards, however, it is entirely within the Agency's power to do so.

Response. We agree. There is no functional difference between this boiler and incinerators with respect to mercury feed rate and the type of waste processed (incinerators often treat remediation wastes). Therefore, the most relevant sources for the purposes of clarification in this case are incinerators, not liquid fuel boilers.

Accordingly, we have classified DSSI as an incinerator for purposes of a mercury standard (i.e., made it subject to the mercury standard for incinerators), and have included the DSSI mercury data with the incinerator data when assessing mercury standards for incinerators.

Comment. In something of a contradiction, the same commenter argues that the mixed waste boiler source (DSSI) does not claim that it cannot meet the relevant mercury standard for liquid fuel boilers, but only that it cannot do so “using either feedrate control or MACT floor back end emission control.” Floors must reflect the emission levels that the relevant best sources actually achieve, not what is achievable through the use of a chosen emission control technology. It is flatly unlawful—and essentially contemptuous of court—for EPA even to entertain the source's argument that the source should be subject to a less stringent emission standard based on the levels they believe would be achievable through the use of one chosen control technology.

The commenter also states that the source acknowledges that it could achieve a better emission level, and apparently meet the relevant standards, by using activated carbon. Their argument that doing so would generate large quantities of spent radioactive carbon does not support its attempt to avoid Clean Air Act requirements; the alternative to the source accumulating large quantities of radioactive carbon is releasing large quantities of radioactive and toxic pollution into the environment.

Response. DSSI cannot meet the liquid boiler mercury standard because it burns a unique waste that resembles wastes processed by hazardous waste incinerators (in terms of mercury concentration and provenance) and is unlike any mercury-containing waste burned by the remaining liquid fuel boilers. See the earlier discussion showing that DSSI needs the operational flexibility to feed 175 times more mercury than any other liquid fuel boiler, but that DSSI's process feed is within the upper range of mercury feed found at incinerators.

We agree that DSSI is processing different types of mercury-bearing wastes than those combusted by all other liquid fuel boilers. We believe that establishing a different mercury standard for DSSI is warranted, as it would for any source with demonstrably unique, unalterable feedstock which is Start Printed Page 59474more difficult to treat than that processed by other sources otherwise in the same category.

How DSSI chooses to comply with the incinerator mercury standard (for example, whether it must use some other type of emissions control technology) is not germane to this decision. We note that today's mercury standard for incinerators will force this source to lower its mercury emissions, since it is unlikely that it can meet today's 120 μg/dscm standard at all times without some changes in operations.

Comment. The source argues that waste minimization is not feasible for legacy mixed waste that has already been generated. It is not possible to travel back in time and unmake mixed legacy waste that already has been created. That obvious fact, however, lends no support to their argument that it should be allowed to burn mixed legacy waste with less stringent emission standards, according to one commenter.

Response. As discussed above, the mercury standard for liquid fuel boilers is not achievable for this source because it is a different type and class of boiler, based on the type of mercury-containing hazardous waste it processes. Because this boiler has mercury feed rates that resemble those of incinerators—not liquid fuel boilers—and waste minimization is not possible, subjecting the boiler to the mercury incinerator standard is a reasonable means of sub-categorization pursuant to the discretionary authority provided us by section 112(d)(1) of the Clean Air Act.

Comment. The commenter states that it is entirely possible to dispose of mixed legacy waste without burning it. Specifically, currently available technologies such as chemical oxidation and precipitation can be used to treat mixed legacy waste without burning it—and without releasing mercury into the air. Therefore, mixed legacy waste should not be burned at all; it should be disposed of safely through the application of one of these more advanced technologies.

Response. First, these wastes must be treated before they can be land disposed. RCRA sections 3004(d), (g)(5), and (m). They also must meet a standard of 0.025 mg/l measured by the Toxicity Characteristic Leaching Procedure before land disposal is permissible. 40 CFR 268.40 (standard for “all other nonwastewaters that exhibit the characteristic of toxicity for mercury”).[169] EPA's technical judgment is that it would be very difficult to meet this standard by any means other than combustion. Moreover, as an organic liquid, the waste is readily amenable to treatment by combustion. In addition, combustion is a legal form of treatment for the waste. EPA did not propose to change or otherwise reconsider these treatment standards in this rulemaking, and is not doing so here. We note, however, that 40 CFR 268.42 and 268.44 provide means by which generators and treatment facilities can petition the Agency to seek different treatment standards from those specified by rule, and set out requirements for evaluating such petitions.

We note further that, because this waste is radioactive, exceptional precautions need to be taken in its handling. The nonthermal treatment alternatives mentioned by the commenter ignore the potential for radiation exposure if nonthermal treatment is used. Concerns (some of which are mentioned in DSSI's comment) include: Nonthermal treatment would (or could) increase worker exposure; desire to reduce handling of radioactive materials in general; need to avoid contaminating equipment that subsequently requires decontamination or handling as radioactive material; minimizing the generation of additional radioactive waste residues; reducing the amount of analysis of radioactive materials, which causes potential exposure, generation of radioactive wastes and equipment; wastes are varied and often of small volumes, which makes it difficult to develop routine procedures. Nonthermal treatment alternatives are also not currently available to DOE to manage the diversity and volume of DOE mixed waste. It is thus our belief that the commenter has not fully explored the implications of its position, especially with regard to radiation exposure.

If the commenter wishes to pursue this issue, EPA believes the appropriate context is through the Land Disposal Restriction mechanisms described above.

Comment. The commenter states that the source argues that feedrate control is not “practical.” There appears to be no record evidence indicating what would make feedrate control impractical and why any such obstacle could not be overcome.

Response. Feedrate control to the extent necessary to achieve the liquid fuel boiler standards is not practical for reasons just discussed. This source is one of two available sources that is authorized to treat mixed waste, and the other source is not likely to have the ability to burn mercury-bearing organic waste in the future due to permit limitations and size constraints.

Comment. The commenter states that mixed legacy waste should not be burned at all. If there are truly no other facilities that are currently permitted to dispose of mixed legacy waste, such waste should be stored until a facility that can treat such waste safely—e.g., through chemical oxidation—can be permitted.

Response. The commenter's suggestion is beyond the scope of today's rulemaking. The suggestion is also illegal, since RCRA prohibits the storage of hazardous waste for extended periods. See RCRA section 3004(j); and Edison Electric Inst. v. EPA, 996 F. 2d 326, 335-37 (DC Cir. 1993) (illegal under RCRA section 3004(j) to store hazardous waste pending development of a treatment technology). EPA also notes that it retains authority under RCRA section 3005(c) (the so-called omnibus permitting authority) by which permit writers can adopt more stringent emission standards in RCRA permits if they determine that today's standards are not protective of human health and the environment.

2. Different Mercury, Semivolatile Metals, Chromium, and Total Chlorine Standards for Liquid Fuel Boilers Depending on the Heating Value of the Hazardous Waste Burned

Comment. Several commenters state that liquid fuel boilers should have an alternative concentration-based standard in addition to the thermal emission-based standard. Liquid fuel boilers are typically “captive” units that burn waste fuels generated from on-site or nearby manufacturing operations, rather than accepting wastes from a wide variety of other sources. Because they have captive fuel sources, operators generally do not have fuel blending capabilities. Liquid fuel boilers “burn what they have,” and as such have very limited operational flexibility. EPA should not penalize boilers that have the same mass concentrations of metals or chlorine in their waste compared to other boilers, but which wastes have a lower heating value than wastes burned by other boilers. (The “penalty” is that emissions limits that are normalized by the heating value of the hazardous waste require that less volume of lower heating value waste can be burned compared to higher heating value fuel.) This problem is made worse by the limited data base for liquid fuel boilers, Start Printed Page 59475the lack of historical data to verify that these standards are achievable over time, and having most or all of the measured emissions below detection limits. In addition, most of the mercury and semivolatile metal data EPA has in the data base were obtained during normal operations and while the source demonstrated compliance with RCRA's chromium standard—the other metals data were available only because stack method Method 29 reports data for all RCRA metals, even ones that are not at issue for the compliance test. (Sources generally elected to comply with the BIF Tier I metals emissions levels, but Tier III for chromium. Thus, the Method 29 test for chromium will give emissions results for all the metals—even those not subjected to stack testing—not just chromium.)

Response. As explained earlier in Part Four, Section V.A., EPA has selected normalizing parameters that best fit the input to the combustion device. A thermal normalizing parameter (i.e., expressing the standards in terms of amount of HAP contributed by hazardous waste per thermal content of hazardous waste) is appropriate where hazardous waste is being used in energy-recovery devices as a fuel, since the waste serves as a type of fuel. Using a thermal normalizing parameter in such instances avoids the necessity of subcategorizing based on unit size.

The commenters raise the other side of the same issue. As the commenters point out, some liquid fuel boilers burn lower Btu hazardous waste because that is the waste available to them, and those with waste that has a low heating value are, in their words, “penalized,” compared to those with a high(-er) heating value. Also, since these are not commercial combustion units, they normally lack the opportunity to blend wastes of different heating values to result in as-fired high heating value fuels. If boiler standards are normalized by hazardous waste heating value, sources with lower heating value waste must either reduce the mass concentration of HAP or increase the waste fuel heating value (or increase the system removal efficiency) compared to sources with wastes having the same mass concentration of HAP but higher heating value.

Moreover, the thermal normalizing parameter is not well suited for a hazardous waste that is not burned entirely for its fuel value. In cases where the lower heating value waste is burned, the boiler is serving—at least in part—as a treatment device for the lower heating value hazardous waste. When this occurs, the better normalizing parameter is the unit's gas flow (a different means of accounting for sources of different size), where the standard is expressed as amount of HAP per volume of gas flow (the same normalizing parameter used for most of the other standards promulgated in today's final rule.)

The commenters requested that liquid fuel boilers be able to select the applicable standard (i.e., to choose between normalizing parameters) and further requested that we assess the performance of these units (for the purpose of establishing concentration-based MACT floor levels) by using the same MACT pool of best performing sources expressed on a thermal emissions basis.

Neither of these suggestions is appropriate. Choice of normalizing parameter is not a matter of election, but rather reflects an objective determination of what parameter is reasonably related to the activity conducted by the source. Moreover, the commenter's suggestion to use thermal emissions to measure best performance for a concentration-based standard does not make sense. It arbitrarily assumes that the best performers with respect to low and high heating value wastes are identical.

Instead, we have established two subcategories among the liquid fuel boilers: those burning high and those burning low heating value hazardous waste. The normalizing parameter for sources burning lower energy hazardous waste is that used for the other hazardous waste treatment devices, gas flow rate, so that the standard is expressed as concentration of HAP per volume of gas flow (a concentration-based form of the standard.) The normalizing parameter for sources burning higher energy content hazardous waste is the thermal parameter used for energy recovery devices, such as cement kilns and lightweight aggregate kilns. For the purposes of calculating MACT floors, the best performers are then drawn from those liquid fuel boilers burning lower energy hazardous waste for the lower heating value subcategory, and from those liquid fuel boilers burning higher energy hazardous waste for the higher heating value subcategory [170] . (See Section 23.2 of Volume III of the Technical Support Document for more information.)

Moreover, liquid fuel boilers are not irrevocably placed in one or the other of these subcategories. Rather, the source is subject to the standard for one or the other of these subcategories based on the as-fired heating value of the hazardous waste it burns at a given time. Thus, when the source is burning for energy recovery, then the thermal emissions-based standard would apply. When the source is burning at least in part for thermal destruction, then the concentration based standard would apply. This approach is similar to how we have addressed the issue of normalization in other rules where single sources switch back and forth among inputs which are sufficiently different to warrant separate classification.[171]

We next considered what an appropriate as-fired heating value would be for each liquid fuel boiler subcategory. Although we have used 5000 Btu/lb (the heating value of lowest grade fuels such as scrap wood) in past RCRA actions as a presumptive measure of when hazardous waste is burned for destruction (see, e.g. 48 FR 11159 (March 16, 1983)), we do not think that measure is appropriate here. We used the 5,000 Btu/lb level to delineate burning for destruction from burning for energy recovery at a time when that determination meant the difference between regulation and nonregulation. See 50 FR 49166-167 (Nov. 29, 1985). This is a different issue from choosing the most reasonable normalizing parameter for regulated units (i.e., units which will be subject to a standard in either case).

Instead, we are adopting a value of 10,000 Btu/lb as the threshold for subcategorization. This is approximately the heating value of commercial liquid fossil fuels. 63 FR 33782, 33788 (June 19, 1998) It is also typical of current hazardous waste burned for energy recovery. Id. Moreover, EPA has used this value in its comparable fuel specification as a means of differentiating fuels from waste. See id. and Table 1 to 40 CFR section 261.38, showing that EPA normalizes all Start Printed Page 59476constituent concentrations to a 10,000 Btu/lb level in its specification for differentiating fuels from wastes.

We next examined the waste fuel being burned at cement kilns and lightweight aggregate kilns, which burn hazardous waste fuels to drive the process chemistry to produce products[172] , to cross-check whether 10,000 Btu/lb is a reasonable demarcation value for subcategorizing. 10,000 Btu/lb is the minimum heating value found in burn tank and test report data we have for cement kilns and lightweight aggregate kilns [173] . We believe the cement kiln and light weight aggregate kiln data confirm that this is an appropriate cutpoint, since these sources are energy recovery devices that blend hazardous wastes into a consistent, high heating value fuel for energy recovery in their manufacturing process.

We then separated the liquid fuel boiler emissions data we had into two groups, sources burning hazardous waste fuel with less than 10,000 Btu/lb and all other liquid fuel boilers, and performed separate MACT floor analyses. (See Sections 13.4, 13.6, 13.7, 13.8, and 22 of Volume III of the Technical Support Document.) We calculated concentration-based MACT standards for these sources from their respective mercury, semivolatile metals, chromium, and total chlorine data.

Liquid fuel boilers will need to determine which of the two subcategories the source belongs in at any point in time. Thus, you must determine the as-fired heating value of each batch of hazardous waste fired so that you know the heating value of the hazardous waste fired at all times.[174] If the as-fired heating value of hazardous wastes varies above and below the cutpoint (i.e., 10,000 Btu/lb) at times, you are subject to the thermal emissions standards when the heating value is not less than 10,000 Btu/lb and the mass concentration standards when the heating value is less than 10,000 Btu/lb. To avoid the administrative burden of frequently switching applicable operating requirements between the subcategories, you may elect to comply with the more stringent operating requirements that ensure compliance with the standards for both subcategories.

Comment: EPA's attempt to give actual performance two different meanings within a single floor approach is unlawful, unexplained, internally inconsistent, and arbitrary. If EPA believes that mass-based emissions constitute sources' actual performance, the best performing sources must be those with the best mass based emissions—not thermal emissions.

Response: As just explained, we agree with this comment, and have developed MACT floors independently for the two subcategories of liquid fuel boilers. Thus, we have defined two separate MACT pools based on the thermal input of the waste fuel and derived two separate and consistent MACT standards for sources when they burn solely for energy recovery, and when they do not.

We also note that a source cannot “pick and choose” the less stringent of the two standards and comply with those. The source must be in compliance with the set of standards that apply.

3. Alternative Particulate Matter Standard for Liquid Fuel Boilers

Comment: A commenter requested that EPA establish standards that allow boilers the option to comply with either a concentration-based particulate matter standard or thermal emissions-based particulate matter standard.

Response: We determined that it is appropriate to express the particulate matter emission standard as a concentration-based standard consistently across source categories and not to give boilers the option to comply with a thermal emissions-based particulate matter standard. As discussed in Part Four, Section III.D as well as the preceding section, metal and chlorine concentration-based emission standards can be biased against sources that process more hazardous waste (from an energy demand perspective), in part because the SRE/Feed methodology assesses feed control of each source when identifying the best performing sources; the ranking procedure thus favors sources with lower percentage hazardous waste firing rates (keeping all other assessment factors equal). The thermal emission standard format eliminates this firing rate bias, which amounts to a limitation on the amount of raw material (hazardous waste fuel to an energy recovery device) that may be processed, when identifying best performing sources.

The methodology we use to identify best performing sources for particulate matter emissions is not affected by the firing rate bias in the manner that metal and chlorine emissions are. This is primarily because we define best performing sources as those with the best back-end air pollution control technology; feed control is not assessed (specifically ash feed control) for raw materials, fossil fuel, or unenumerated HAP metal in the hazardous waste. The hazardous waste firing rate bias is therefore not present when we identify the best performing particulate matter sources because a source's hazardous waste firing rate is not a direct factor in the ranking procedure.

We also note that four of the nine best performing liquid fuel boilers for particulate matter are equipped with fabric filters. Particulate matter emissions from sources equipped with fabric filters are not significantly affected by ash inlet loading. This is not true for metals and chlorine, given metal and chlorine emissions from fabric filters tend to increase at increased feed rates. See Volume III of the Technical Support Document, Sections 5.3 and 7.4. We conclude that the hazardous waste firing rate issue is not a concern for these sources given their particulate matter emissions would not be significantly affected by increased hazardous waste firing rates.

4. Long-term, Annual Averaging Is Impermissible

Comment: Standards expressed as long-term limits are legally impermissible because those levels, by definition, would sometimes be greater than the average emission levels achieved by the best performing sources. Compliance also must be measured on a continuous basis, under section 302(k) of the Act. Thus, floor levels (and standards) for mercury expressed as long-term limits are illegal.

Response: The commenter maintains that the statutory command in section 112(d)(3)(A) to base floor standards for existing sources on “the average emission limitation achieved by the best performing 12 percent of * * * existing sources” precludes establishing standards expressed as long term averages because certain daily values could be higher. We do not accept this position. The statute does not state what type of “average” performance EPA must assess. Long term, i.e., annual, averaging of performance is quite evidently a type of average, and so is permissible under the statutory text. Moreover, it is reasonable to establish Start Printed Page 59477standards on this basis (the standards being the average of the best performing sources, expressed as a long-term average), where sufficient data exist. Indeed, since the principal health concern posed by the emitted HAP is from chronic exposure (i.e. cumulative exposure over time), long-term standards (which reduce the long-term distribution of emitted HAP) arguably would be preferable in addressing the chief risks posed by these sources' emissions.

We establish standards with long-term averaging limits whenever we use normal data to estimate long-term performance. We do this in the few instances where there are insufficient data (whether normal data or compliance test data) to estimate each source's short term emission levels (e.g., mercury and semivolatile metal standards for liquid fuel boilers).[175] One or two snapshot data based on normal operations are not likely to reflect a source's short-term operating levels in part because feed control levels can vary over time.[176] See Mossville, 370 F. 3d at 1242 (varying feed rates lead to different emission levels, and this variability must be encompassed within the floor standard because the standard must be met at all times). As a result, snapshot normal emissions, when averaged together, better reflect a source's long term average emissions. An emission standard based on normal data that is averaged together, but expressed as a short-term limit, would not be achievable by the best performing sources because it would not adequately account for their emissions variability. See National Wildlife Federation v. EPA, 286 F. 3d at 572-73 (“[c]ontinuous operation at or near the daily maximum would in fact result in discharges that exceed the long-term average. Likewise, setting monthly limitations at the 99th percentile would not insure that the long-term average is met”). Long-term limits better account for this variability because such limits allow sources to average their varying feed control levels over time while still assuring average emissions over this period are below the levels demonstrated by the best performing sources.

Indeed, under the commenter's approach where no averaging of intra-source data would be allowed, sources would not be in compliance with the standards during the performance tests themselves. The tests consist of the average of three data runs, so half of the emissions-weighted data points would be impermissibly higher than the average during the test used to derive today's emission standards.

EPA also does not see that section 302(f) of the Act, cited by the commenter, supports its position. That provision indicates that the emission standards EPA establishes must limit the quantity, rate, or concentration of air pollutants on a continuous basis. A standard expressed as a long-term average does so by constraining the overall distribution of emissions to meet a long-term average. Also, long term limits result in emission standards that are lower than those that otherwise would be implemented on a short-term basis. The short-term limit would have to reflect the best performing sources' short term emissions variability (i.e., the maximum amount of variability a source could experience during a single test period). National Wildlife Federation, 286 F. 3d at 571-73.

Comment: Other commenters argued the opposite point, that ERA has no data to show that an annual average is achievable, and EPA should establish a longer averaging period.

Response: We believe that all sources can achieve the mercury and semivolatile metals standards for liquid fuel boilers on an annual basis using some combination of MACT controls, i.e., feed control, back end control, or some combination of both. We agree that we have a small data set for these standards, but also believe that it is intuitive that a liquid fuel boiler can meet these standards on an annual basis, because one year is sufficiently more than any seasonal (i.e., several month long) production of certain items that may not be represented by the tests we have.

This informs us that an average of less than a year may not be achievable. It does not inform us that averaging of more than a year is required, since variations that occur with a year are averaged together. An annual average is sufficient for a source to determine whether an individual waste stream impacts negatively on the compliance of the liquid fuel boiler and take measures to address the issue.

5. Gas Fuel Boilers

Comment: How can a boiler burning only gaseous waste also be burning hazardous waste? Uncontained gases are not considered hazardous waste under RCRA. Why are boilers that burn only gasses part of the liquid fuel boiler subcategory?

Response: We agree with the commenter that boilers that burn gasses are unlikely to burn hazardous wastes. However, gas fuel hazardous waste boilers have existed in the past,[177] and we believe we need to define a MACT standard for them. Therefore, we included gas fuel boilers in the liquid fuel boiler subcategory for reasons cited in the proposed rule. See 69 FR at 21216.

E. General

1. Alternative to the Particulate Matter Standards

Comment: Commenters state that some incinerators are currently complying with the alternative to the particulate matter standard provision pursuant to the interim standards. See § 63.1206(b)(14). The eligibility and operating requirements for the alternative to the particulate matter standard in the Interim Standards are different than the proposed alternative to the particulate matter standard in the replacement rule. Specifically, the proposed alternative to the particulate matter standard would no longer require sources to demonstrate a 90% system removal efficiency or a minimum hazardous waste metal feed control level to be eligible for the alternative. Commenters request that EPA clarify in the final rule that the proposed alternative to the particulate matter standard supersedes the requirements in the Interim Standards.

Response: We are finalizing the alternative to the particulate matter standard for incinerators as proposed, with the exception that the alternative metal emission limitations have been revised as a result of database changes since proposal. See § 1219(e) and part three, section II.A. We considered superseding the interim standard alternative to the particulate matter standard requirements (63.1206(b)(14)) immediately (upon promulgation) by replacing it with the revised alternative Start Printed Page 59478standard provisions finalized in today's rule. Although the eligibility requirements for the alternative to the particulate matter standard finalized today are less stringent than the interim standard requirements, the metal emission limitations that are also required by the alternative finalized today are by definition equivalent to or more stringent than the metal limitations in the interim standard alternative. We therefore cannot completely supersede the interim standard provisions immediately (upon promulgation) because sources have three years to comply with more stringent standards. We are instead revising the interim standard provisions of § 63.1206(b)(14) to only reflect the revised alternative standard eligibility criteria (specifically, we have removed the requirements to achieve a given system removal efficiency and hazardous waste metal HAP feed control level).[178] These eligibility criteria revisions become effective immediately with respect to the interim standards because they are less stringent than the current requirements. Sources should modify existing Notifications of Compliance and permit requirements as necessary prior to implementing these revised procedures.

Comment: One commenter is opposed to the alternative to the particulate matter standard because it ignores the health effects/benefits that are attributable to particulate matter.

Response: Particulate matter is not defined as a hazardous air pollutant pursuant the NESHAP program. See CAA 112(b)(1). We control particulate matter as a surrogate for metal HAP. See part four, section IV.A. As a result, a particulate matter standard is not necessary in instances where metal HAP emission standards can alternatively and effectively control the nonmercury metal HAP that is intended be controlled with the surrogate particulate matter standard. The alternative to the particulate matter standard in the final rule accomplishes this. We acknowledge that particulate matter emission reductions result in health benefits. That in itself does not give EPA the authority under § 112(d)(2) to directly regulate particulate matter, however.

2. Assessing Risk as Part of Consideration of Nonair Environmental Impacts

Comment: Commenter states that EPA has inappropriately failed to consider emissions of persistent bioaccumulative pollutants in its beyond-the-floor analysis despite EPA's acknowledgment that these HAPs have non-air quality health and environmental impacts.

Response: EPA has taken the consistent position that considerations of risk from air emissions have no place when setting MACT standards, but rather are to be considered as part of the residual risk determination and standard-setting process made under section 112 (f) of the statute. EPA thus interprets the requirement in section 112 (d) (2) that we consider “non-air quality health and environmental impacts” as applying to the by-product outputs from utilization of the pollution control technology, such as additional amount of waste generated, and water discharged.[179] EPA's interpretation was upheld as reasonable in Sierra Club v. EPA, 353 F. 3d 976, 990 (D.C. Cir. 2004) (Roberts, J.).

VII. Health-Based Compliance Alternative for Total Chlorine

A. Authority for Health-Based Compliance Alternatives

Comment: One commenter states there is no established health threshold for either HCl or chlorine.

Response: Although EPA has not developed a formal evaluation of the potential for HCl or chlorine carcinogenicity (e.g., for IRIS), the evaluation by the International Agency for Research on Cancer stated that there was inadequate evidence for carcinogenicity in humans or experimental animals and thus concluded that HCl and chlorine are not classifiable as to their carcinogenicity to humans (Group 3 in their categorization method). Therefore, for the purposes of this rule, we have evaluated HCl and chlorine only with regard to non-cancer effects. In the absence of specific scientific evidence to the contrary, it has been our policy to classify non-carcinogenic effects as threshold effects. RfC development is the default approach for threshold (or nonlinear) effects.

Comment: One commenter states that the proposal is an inappropriate forum for bringing forward such a significant change in the way that MACT standards are established under Section 112(d) of the Clean Air Act. A precedent-setting change of the magnitude that EPA has raised should be discussed openly and carefully with all affected parties, rather than being buried in several individual proposed standards.

Response: Including health-based compliance alternatives for hazardous waste combustors does not mean that EPA will automatically provide such alternatives for other source categories. Rather, as has been the case throughout the MACT rule development process, EPA will undertake in each individual rule to determine whether it is appropriate to exercise its discretion to use its authority under CAA section 112(d)(4) in developing applicable emission standards. Stakeholders for those affected rules will have ample opportunity to comment on the Agency's proposals.

Comment: One commenter states that the proposed approach is contrary to the intent of the CAA which explicitly calls for a general reduction in HAP emissions from all major sources nationwide through the establishment of MACT standards based on technology, rather than risk, as a first step.

Response: For pollutants for which a health threshold has been established, CAA section 112(d)(4) allows the Administrator to consider such threshold level, with an ample margin of safety, to establish emission standards.

Comment: One commenter states that the proposed approach would take the national air toxics program back to the time-consuming NESHAP process that existed prior to the Clean Air Act Amendments of 1990.

Response: We disagree that allowing a health-based compliance alternative in the final rule will alter the MACT program or affect the schedule for promulgation of the remaining MACT standards. Today's rule is the last MACT rule to be promulgated, and the health-based compliance alternative did not delay promulgation of the rule.

Comment: A commenter is concerned that the proposal would remove the benefit of the “level-playing field” that would result from the proper implementation of technology-based MACT standards.

Response: Providing health-based compliance alternatives in the final rule for sources that can meet them will assure the application of a uniform set of requirements across the nation. The final rule and its criteria for demonstrating eligibility for the health-based compliance alternatives apply uniformly to all hazardous waste combustors except hydrochloric acid Start Printed Page 59479production furnaces. The final rule establishes two baseline levels of emission reduction for total chlorine, one based on a traditional MACT analysis and the other based on EPA's evaluation of the health threat posed by emissions of HCl and chlorine. All hazardous waste combustor facilities must meet one of these baseline levels, and all facilities have the same opportunity to demonstrate that they can meet the alternative health-based emission standards. We also note that additional uniformity is provided by limiting the health-based compliance alternatives for incinerators, cement kilns, and lightweight aggregate kilns to the emission levels allowed by the Interim Standards.

Comment: Several commenters state that site-specific emission limits are inappropriate under section 112(d)(4) because they are not emission standards. One commenter asserts that the Agency's position that the limits are based on uniform procedures is flawed because the process allows “any scientifically-accepted, peer-reviewed risk assessment methodology for your site-specific compliance demonstration.” This is not a “uniform” procedure, according to the commenter. There are a host of variables that influence the results of an accepted methodology. The commenter reasons that, without some standardization of those variables, there is no uniform or standard analysis. Each permitting authority could establish its view of appropriate variables; there would be no national consistency.

Several other commenters assert that EPA has the authority to establish an exposure-based emission limit for total chlorine. One commenter notes that one issue that often arises when considering risk-based standards is whether EPA has authority under section 112 to establish an exposure-based emission limit. The commenter states that the concern seems to be that some stakeholders construe the Act's statutory provisions as requiring uniform emission limitations at all facilities, rather than emissions that are measured at places away from the source and that vary from facility to facility. The commenter does not see any legal impediment to establishing exposure-based limits.

The commenter notes that, first, under section 112, EPA has authority to establish “emission standards.” Emission standards are defined to be a requirement established by the State or the Administrator which limits the quantity, rate or concentration of emissions of air pollutants on a continuous basis * * * to assure continuous emission reduction, and any design, equipment, work practice or operational standard promulgated under this chapter. EPA's alternate risk-based emission standard will limit the quantity, rate or concentration of the emissions. The commenter states that there is no requirement in the definition that specifies where the emission standard is to be measured, nor is there such a requirement anywhere in the statute.

Second, the commenter notes that EPA's proposed exposure-based limit will result in facilities establishing operating parameter limitations, or OPLs. These OPLs qualify as emission limitations because they are “operational standards” being promulgated under section 112, according to the commenter. They will be measured at the facility, not at the point of exposure. Finally, the commenter reasons that the limitations EPA is establishing are uniform. They uniformly protect the individual most exposed to emission levels no higher than a hazard index of 1.0. Consequently, the commenter believes that there is nothing in the statute that prevents the Agency from promulgating exposure-based emission standards.

Response: We agree with the commenters who believe the Agency has the authority to establish health-based compliance alternatives under a national exposure standard. In particular, we agree with the commenter that the health-based compliance alternatives are national standards since they provide a uniform and national measure of risk control, and also that the health-based compliance alternatives are “emission standards” because they limit the quantity, rate or concentration of total chlorine emissions.

Section 112(d)(4) authorizes EPA to bypass the mandate in section 112(d)(3) in appropriate circumstances. Those circumstances are present for hazardous waste combustors other than hydrochloric acid production furnaces. Section 112(d)(4) provides EPA with authority, at its discretion, to develop health-based compliance alternatives for HAP “for which a health threshold has been established,” provided that the standard reflects the health threshold “with an ample margin of safety.”

Both the plain language of section 112(d)(4) and the legislative history indicate that EPA has the discretion under section 112(d)(4) to develop health-based compliance alternatives for some source categories emitting threshold pollutants, and that those standards may be less stringent than the corresponding MACT standard (including floor standards) would be.[180] EPA's use of such standards is not limited to situations where every source in the category or subcategory can comply with them. As with technology-based standards, a particular source's ability to comply with a health-based standard will depend on its individual circumstances, as will what it must do to achieve compliance.

In developing health-based compliance alternatives under section 112(d)(4), EPA seeks to ensure that the concentration of the particular HAP to which an individual exposed at the upper end of the exposure distribution is exposed does not exceed the health threshold. The upper end of the exposure distribution is calculated using the “high end exposure estimate,” defined as “a plausible estimate of individual exposure for those persons at the upper end of the exposure distribution, conceptually above the 90th percentile, but not higher than the individual in the population who has the highest exposure” (EPA Exposure Assessment Guidelines, 57 FR 22888, May 29, 1992). Assuring protection to persons at the upper end of the exposure distribution is consistent with the “ample margin of safety” requirement in section 112(d)(4).

We agree with the view of several commenters that section 112(d)(4) is appropriate for establishing health-based compliance alternatives for total chlorine for hazardous waste combustors other than hydrochloric acid production furnaces. Therefore, we have established such compliance alternatives for affected sources in those categories. Affected sources which believe that they can demonstrate compliance with the health-based compliance alternatives may choose to comply with those compliance alternatives in lieu of the otherwise applicable MACT-based standard.

Comment: One commenter states that the risk assessments would not provide an ample margin of safety because background exposures are not taken into account. There is no accounting for other chlorine compounds from other sources at the facility, or from other neighboring facilities. The commenter believes that there is no evidence in the section 112(f) residual risk assessments produced thus far that emissions from collocated sources will actually be pursued by EPA. The commenter also notes that the Urban Air Toxics program cannot be relied upon to address ambient background. This program, Start Printed Page 59480required under section 112(k), was to be completed by 1999. However, the strategy has not been finalized and the small amount of activity in this area is focused on voluntary emission reductions rather than federal requirements. Finally, the commenter notes that control of criteria pollutants via State Implementation Plans to achieve compliance with the NAAQS is problematic. For particulate matter (PM) and ozone, new NAAQS were set in 1997 and seven years later the nonattainment designations are still being determined. The designation process will be followed by a 3 year period to prepare State Implementation Plans and several more years to carry out those plans. In the meantime, there will be high levels of PM and ozone in the air near many hazardous waste combustors in New Jersey which will exacerbate exposures to chlorine and hydrogen chloride.

Response: Total chlorine missions from collocated hazardous waste combustors must be considered in establishing health-based compliance alternatives under § 63.1215. Ambient levels of HCl or chlorine attributable to other on-site sources, as well as off-site sources, are not considered, however. As we indicated in the Residual Risk Report to Congress and in the recent residual risk rule for Coke Ovens, the Agency intends to consider facility-wide HAP emissions as part of the ample margin of safety determination for CAA section 112(f) residual risk actions. 70 FR at 19996-998 (April 15, 2005); see also, 54 FR at 38059 (Sept. 14, 1989) (benzene NESHAP).

Comment: Several commenters state that acute exposure guideline levels (AEGLs) are once-in-a-lifetime exposure levels. They assert that, because short term exposures at a Hazard Index greater than 1.0 may occur more than once in a lifetime, using AEGLs for the purpose of setting risk-based short-term limits for HCl and chlorine does not provide an “ample margin of safety.”

Response: To assess acute exposure, we proposed to use acute exposure guideline levels for 1-hour exposures (AEGL-1) as health thresholds. We have investigated commenters' concerns, however, and conclude that AEGLs are not likely to be protective of human health because individuals may be subject to multiple acute exposures at a Hazard Index greater than 1.0 from hazardous waste combustors. Consequently, we use acute Reference Exposure Levels (aRELs) rather than acute exposure guideline levels (AEGLs) as acute exposure thresholds for the final rule. See also Part Two, Section IX.D above. Acute RELs are health thresholds below which there would be no adverse health effects while AEGL-1 values are health thresholds below which there may be mild adverse effects.

Acute exposures are relevant (in addition to chronic exposures) and the acute exposure hazard index of 1.0 could be exceeded multiple times over an individual's lifetime. Although we concluded at proposal that the chronic exposure Hazard Index would always be higher than the acute exposure Hazard Index, and thus would be the basis for the total chlorine emission rate limit, this conclusion relates to acute versus chronic exposure to a constant, maximum average emission rate of total chlorine from a hazardous waste combustor. See 69 FR at 21300. We explained that acute exposure must nonetheless be considered when establishing operating requirements to ensure that short-term emissions do not result in an acute exposure Hazard Index of greater than 1.0. This is because total chlorine and chloride feedrates to a hazardous waste combustor (e.g., commercial incinerator) can vary substantially over time. Although a source may remain in compliance with a feedrate limit with a long-term averaging period (e.g., 12-hour, monthly, or annual) based on the chronic Hazard Index, the source could feed chlorine during short periods of time that substantially exceed the long-term feedrate limit. This could result potentially in emissions that exceed the one-hour (i.e., acute exposure) Hazard Index. Consequently, we discussed at proposal the need to establish both short-term and long-term total chlorine and chloride feedrate limits to ensure that neither the chronic exposure nor the acute exposure Hazard Index exceeds 1.0.[181]

We conclude that 1-hour Reference Exposure Levels (aRELs) are a more appropriate health threshold metric than AEGL-1 values for hazardous waste combustors given that the acute Hazard Index limit of 1.0 may be exceeded multiple times over an individual's lifetime, albeit resulting from uncontrollable factors. The California Office of Health Hazard Assessment has developed acute health threshold levels that are intended to be protective for greater than once in a lifetime exposures. The acute exposure levels are called acute Reference Exposure Levels and are available at http://www.oehha.ca.gov/​air/​acute_​rels/​acuterel.html.

The 1-hour REL values for hydrogen chloride and chlorine are 2.1 mg/m3 and 0.21 mg/m3, respectively. The AEGL-1 values for hydrogen chloride and chlorine are 2.7 mg/m3 and 1.4 mg/m3, respectively. Although there is little difference between the 1-hour REL and AEGL-1 values for hydrogen chloride, the 1-hour REL for chlorine is substantially lower than the AEGL-1 value.

In summary, we believe that aRELs are a more appropriate health threshold metric than AEGL-1 values for establishing health-based compliance alternatives for hazardous waste combustors because aRELs are “no adverse effect” threshold levels that are intended to be protective for multiple exposures.

Comment: One commenter states that the health-based compliance alternative is unlawful because the proposal does not address ecological risks that may result from uncontrolled HAP emissions, including risks posed to those areas where few people currently live, but sensitive habitats exist.

Response: An ecological assessment is normally required under CAA section 112(d)(4) to assess the presence or absence of “adverse environmental effects” as that term is defined in CAA section 112(a)(7). To identify potential multimedia and/or environmental concerns, EPA has identified HAP with significant potential to persist in the environment and to bioaccumulate. This list does not include hydrogen chloride or chlorine.

We also note that health-based total chlorine emission limits for incinerators, cement kilns, and lightweight aggregate kilns cannot be higher than the current Interim Standards. See § 63.1215(b)(7). Thus, the ecological risk from total chlorine emissions from these sources will not be increased under the health-based limits.

In addition, we note that only 2 of 12 solid fuel boilers have total chlorine emissions higher than 180 ppmv, and only 1 liquid fuel boiler has emissions higher than 170 ppmv. Thus, boilers generally have low total chlorine emissions which would minimize ecological risk.

Consequently, we do not believe that emissions of hydrogen chloride or chlorine from hazardous waste boilers will pose a significant risk to the environment, and facilities attempting to comply with the health-based Start Printed Page 59481alternatives for these HAP are not required to perform an ecological assessment.

B. Implementation of the Health-Based Standards

Comment: Several commenters are concerned that the health-based compliance alternative will place an intensive resource demand on state and local agencies to review and approve facilities' eligibility demonstrations, and State and local agencies may not have adequate expertise to review and approve the demonstrations. One commenter states that permitting authorities do not have the expertise to review eligibility demonstrations that are based on procedures other than those included in EPA's Reference Library, as would be allowed. The commenter also states that, if the health-based compliance alternative is promulgated, EPA should establish one standard method for the analyses so there is consistency nationwide. If EPA offers more than one method, EPA should do all of the risk assessment reviews, instead of passing the responsibility, without clear direction, to the permitting authorities, according to the commenter.

Response: The health-based compliance alternatives for total chlorine that EPA has adopted in the final rule should not impose significant resource burdens on states. The required compliance demonstration methodology is structured in such a way as to avoid the need for states to have significant expertise in risk assessment methodology. We have considered the commenters' concerns in developing the criteria defining eligibility for these compliance alternatives, and the approach that is included in the final rule provides clear, flexible requirements and enforceable compliance parameters. The final rule provides two ways that a facility may demonstrate eligibility for complying with the health-based compliance alternatives. First, look-up tables allow facilities to determine, using a limited number of site-specific input parameters, whether emissions from their sources might cause the Hazard Index limit to be exceeded. Second, if a facility cannot demonstrate eligibility using a look-up table, a modeling approach can be followed. The final rule presents the criteria for performing this modeling.

Only a portion of hazardous waste combustors will submit eligibility demonstrations for the health-based compliance alternatives. Of these sources, several should be able to demonstrate eligibility based on simple analyses—using the look-up tables. However, some facilities will require more detailed modeling. The criteria for demonstrating eligibility for the compliance alternatives are clearly defined in the final rule. Moreover, under authority of RCRA section 3005(c)(3), multi-pathway risk assessments will typically have already been completed for many hazardous waste combustors to document that emissions of toxic compounds, including total chlorine, do not pose a hazard to human health and the environment. Thus, state permitting officials have already reviewed and approved detailed modeling studies for many hazardous waste combustors. The results of these studies could be applied to the eligibility demonstration required by this final rule.

Because these requirements are clearly defined, and because any standards or requirements created under CAA section 112 are considered applicable requirements under 40 CFR part 70, the compliance alternatives would be incorporated into title V programs, and states would not have to overhaul existing permitting programs.

Finally, with respect to the burden associated with ongoing assurance that facilities that opt to do so continue to comply with the health-based compliance alternatives, the burden to states will be minimal. In accordance with the provisions of title V of the CAA and part 70 of 40 CFR (collectively “title V”), the owner or operator of any affected source opting to comply with the health-based compliance alternatives is required to certify compliance with those standards every five years on the anniversary of the comprehensive performance test. In addition, if the facility has reason to know of changes over which the facility does not have control, and these changes could decrease the allowable HCl-equivalent emission rate limit, the facility must submit a revised eligibility demonstration. Further, before changing key parameters that may impact an affected source's ability to continue to meet the health-based emission standards, the source is required to evaluate its ability to continue to comply with the health-based compliance alternatives and submit documentation to the permitting authority supporting continued eligibility for the compliance alternative. Thus, compliance requirements are largely self-implementing and the burden on states will be minimal.

Comment: One commenter suggests that the look-up tables would have more utility if EPA developed tables for each source category to ensure the HCl-equivalent emission rate limits reflected stack parameters representative of each source category. Similarly, another commenter notes that a look-up table designed to be applicable to all hazardous waste combustors is very conservative and will have limited utility. This commenter does not suggest that EPA develop look-up tables for each class of hazardous waste combustors, however. Rather, the commenter suggests that since look-up tables have already been developed for industrial boilers that do not burn hazardous waste [182] hazardous waste combustors should be allowed to use those look-up tables instead of the look-up tables proposed for hazardous waste combustors.

Response: We noted at proposal that the emission rates provided in the look-up table for hazardous waste combustors are more stringent than those promulgated for solid fuel industrial boilers that do not burn hazardous waste. This is because the key parameters used by the SCREEN3 atmospheric dispersion model (i.e., stack diameter, stack exit gas velocity, and stack exit gas temperature) to predict the normalized air concentrations that EPA used to establish HCl-equivalent emission rates for solid fuel industrial boilers that do not burn hazardous waste are substantially different for hazardous waste combustors. Thus, the maximum HCl-equivalent emission rates for hazardous waste combustors would generally be lower than those EPA established for solid fuel industrial boilers that do not burn hazardous waste.

Nonetheless, we agree with the commenter's concerns that the look-up tables would have more utility if they better reflected the range of stack properties representative of hazardous waste combustors. Accordingly, we examined the stack parameters for all hazardous waste-burning sources in our data base (except for hydrochloric acid production furnaces that are not eligible for the health-based emission standards). After analyzing the relationships among the various stack parameters (i.e., stack height, stack diameter, stack gas exhaust volume, and exit temperature), we concluded that the look-up table should be modified to treat both stack diameter and stack height as independent variables rather than relying on stack height alone.

We developed separate tables for short-term (i.e., 1-hour) HCl-equivalent Start Printed Page 59482emissions limits to protect against acute health effects and long-term (i.e., annual) emission limits to protect against chronic effects from exposures to chlorine and hydrogen chloride. As discussed above, we used the acute Reference Exposure Level (aREL) developed by Cal-EPA as the benchmark for acute health effects. We used EPA's Reference Concentrations (RfC) as the benchmark for chronic health effects from exposures occurring over a lifetime.

Emission limits in the look-up table are expressed in terms of HCl-toxicity equivalent emission rates (lbs/hr). To convert your total chlorine emission rate (lb/hr) to an HCl-equivalent emission rate, you must adjust your chlorine emission rate by a multiplicative factor representing the ratio of the HCl health risk benchmark to the chlorine health risk benchmark. For 1-hour average HCl-equivalent emission rates, the ratio is the ratio of the aREL for HCl (2100 micrograms per cubic meter) to the aREL for chlorine (210 micrograms per cubic meter), or a factor of 10.[183] For annual average emissions, the ratio is the ratio of the RfC for HCl (20 micrograms per cubic meter) to the RfC of chlorine (0.2 micrograms per cubic meter), or a factor of 100. See § 63.1215(b).

We used the SCREEN3 air dispersion model to develop the emission limits in the look-up tables. SCREEN3 is a screening model that estimates air concentrations under a wide variety of meteorological conditions in order to identify the meteorological conditions under which the highest ambient air concentrations are likely to occur and what the magnitude of the ambient air concentrations are likely to be. The SCREEN3 model implements the procedures in EPA's “Screening Procedures for Estimating the Air Quality Impact of Stationary Sources, Revised” (EPA-454/R-92-019, U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards, Research Triangle Park, NC, October 1992). Included are options for estimating ambient air concentrations in simple elevated terrain and complex terrain. Simple elevated terrain refers to terrain elevations below stack top. We did not use the complex terrain option in the development of the look-up tables because of the site-specific nature of plume impacts in areas of complex terrain. Therefore, the look-up tables cannot be used in areas of complex terrain (which we define generally as terrain that rises above stack top). Sources located in complex terrain (i.e., as a practical matter, sources other than those that are located in flat or simple elevated terrain as discussed below and thus cannot use the look-up tables) must use site-specific modeling procedures to establish HCl-equivalent emission rates.

We looked at two generic terrain scenarios for purposes of the look-up table. In one we assumed the terrain rises at a rate of 5 meters for every 100 meter run (i.e., a slope of 5 percent) and that terrain is “chopped off” above stack top (following the convention for such analyses in simple elevated terrain). In the other we assumed flat terrain. As can be seen from the tables in § 63.1215, the emission limits with flat terrain are significantly higher than those with simple elevated terrain. To reasonably ensure that the emission limits are not substantially over-stated (e.g., by a factor of 2), the simple elevated terrain table must be used whenever terrain rises to an elevation of one half (1/2) the stack height within a distance of 50 stack heights.

For both the simple elevated terrain and flat terrain scenarios, we performed model runs for urban and rural dispersion conditions, with and without building downwash. We selected the highest (ambient air concentration) values at each distance from among the four runs for each of the terrain scenarios.

As can be seen from the tables in § 63.1215, the HCl-equivalent emission rate limits range from 0.13 pounds per hour on an annual average (for a 0.3 meter diameter stack that is 5 meters tall that lies within 30 meters of the property boundary) to 340 pounds per hour (for a 4.0 meter diameter stack that is 100 meters tall that lies 5000 meters from the property boundary) when located in simple elevated terrain. In flat terrain, the range is from 0.37 to 1100 pounds per hour on an annual average. This contrasts with the look-up table at proposal, where the comparable range was from 0.0612 pounds per hour (for a 5 meter stack height at a distance of 30 meters) to a maximum of 18 pounds per hour (for stack heights of 50 meters or greater, at distances of 500 meters or greater).

If you have more than one hazardous waste combustor on site, the sum of the ratios for all combustors of the HCl-equivalent emission rate to the HCl-equivalent emission rate limit cannot exceed 1.0. See § 63.1215 (c)(3)(v). This will ensure that the Hazard Index of 1.0 is not exceeded considering emissions from all on-site combustors.

Comment: Several commenters state that facilities should be allowed to establish an averaging period for the total chlorine and chloride feedrate limit that is shorter than an annual rolling average. Commenters are referring to the feedrate limit to ensure compliance with the annual average HCl-equivalent emission rate limit. Commenters are concerned with the data handling issues that could arise from calculating, recording, and reporting an annual rolling average feedrate level that is updated hourly, and note that a shorter averaging period would make the limit more stringent.

Response: We agree with commenters, and conclude, moreover, that a 12-hour averaging period rather than an annual averaging period will be imposed on the vast majority of sources as a practical matter. This is because sources must establish a limit on the feedrate of total chlorine and chloride to ensure compliance with the semivolatile metals emission standards. See § 63.1209(n). The feedrate limit for total chlorine and chloride is established under § 63.1209(n) as the average of the hourly rolling averages for each test run, and the averaging period is 12 hours. Thus, the averaging period for the feedrate limit for semivolatile metals—12-hour rolling average updated hourly—trumps the annual rolling average averaging period that would otherwise apply here.[184]

Sources may also demonstrate compliance with the semivolatile metals standard by assuming all semivolatile metals in feedstreams are emitted. See § 63.1207(m)(2). Sources that do not have emission control equipment, such as most liquid fuel boilers, are particularly likely to use this approach. Under this approach, there is no concern regarding increased volatility of metals as chlorine feedrates increase, and such sources are not subject to a feedrate limit for chlorine for compliance assurance with the semivolatile metal standard. These sources may establish an averaging period for the feedrate of total chlorine and chloride for compliance with the health-based compliance alternative for total chlorine of not to exceed one year.[185]

Start Printed Page 59483

Comment: Several commenters offered suggestions on whether a short-term feedrate limit was needed for total chlorine and chloride (i.e., chlorine) as EPA suggested, and if EPA continues to consider it necessary, how the limit should be established.

One commenter states that it is not necessary to set short-term limits for chlorine feedrates. If EPA concludes that short-term limits are necessary, however, the commenter recommended these options: (1) Cap the feedrate at a level that is extrapolated up to the feedrate associated with Interim Standard for incinerators; (2) if the facility uses the site-specific option to set emission limits, the dispersion models can easily be used to set a 1-hour (or longer) limit; and (3) if the facility uses the look up table (which at proposal provided only annual average HCl-equivalent emission rate limits), a short-term limit can be set based on a multiplier of the annual limit'10 times the annual limit as recommended by documents in EPA's Air Toxics Risk Assessment Reference Library.

Another commenter states that, if EPA were to promulgate a short-term feedrate limit, the EPA-endorsed factor of 0.08 employed to translate maximum hourly concentrations to annual concentrations could be used to identify the maximum hourly feedrate limit.

Finally, another commenter states that extrapolation of the chlorine feedrate (from the level during the comprehensive performance test when the source documents compliance with the annual average HCl-equivalent emission rate limit) should be allowed to 100% of the 1-hour average HCl-equivalent emission rate limit because numerous safety factors have already been included in the health risk threshold values, look-up tables, and modeling demonstration.

Response: At proposal, we explained that sources would establish an annual average feedrate limit on chlorine as the feedrate level during the comprehensive performance test demonstrating compliance with the annual average HCl-equivalent emission rate limit.[186] Only long-term exposures—maximum annual average exposures—need be considered when confirming that the chlorine feedrate during the comprehensive performance test (i.e., average of the hourly rolling averages for each run) is acceptable because the annual exposure Hazard Index limit (i.e., not to exceed 1.0) would always be exceeded before the 1-hour Hazard Index limit (i.e., not to exceed 1.0). Thus, the feedrate limit associated with annual exposures would always be more stringent than the feedrate limit associated with 1-hour exposures. See 69 FR at 21299.

We further explained at proposal, however, the need to establish a short-term feedrate limit for chlorine to ensure that the 1-hour HCl-equivalent emission rate did not exceed the 1-hour average HCl-equivalent emission rate limit due to variability in the chlorine feedrate during the annual averaging period for the feedrate limit. We requested comment on approaches to establish this 1-hour chlorine feedrate limit, including extrapolating feedrates to 100% of the 1-hour average HCl-equivalent emission rate limit. See 69 FR at 21304.

In the final rule we have corrected and refined these procedures. The final rule requires you to establish a long-term chlorine feedrate limit to maintain compliance with the annual average HCl-equivalent emission rate limit as either: (1) The chlorine feedrate during the comprehensive performance test if you demonstrate compliance with the semivolatile metals emission standard during the test (see § 63.1209(o)); or (2) if you comply with the semivolatile metals emission standard under § 63.1207(m)(2) by assuming all metals in the feed to the combustor are emitted, the annual average HCl-equivalent emission rate limit divided by [1 − system removal efficiency] where you demonstrate the system removal efficiency during the comprehensive performance test. See discussion in Part Two, Section IX.H, of this preamble. If you establish the chlorine feedrate limit based on the feedrate during the performance test to demonstrate compliance with the semivolatile metals emission standard, the averaging period for the feedrate limit is a 12-hour rolling average. If you establish the chlorine feedrate limit based on the system removal efficiency during the performance test, the averaging period is up to an annual rolling average.

The final rule also requires you to establish an hourly rolling average chlorine feedrate limit if you determine under § 63.1215(d)(3) that the 1-hour average HCl-equivalent emission rate limit may be exceeded. That feedrate limit is established as the 1-hour HCl-equivalent emission rate limit divided by [1 − system removal efficiency].

Under § 63.1215(d)(3), you must establish an hourly rolling average chlorine feedrate limit unless you determine considering specified criteria that your chlorine feedrates will not increase over the averaging period for the long-term chlorine feedrate limit (i.e., 12-hour rolling average or (up to) annual rolling average) to a level that may result in an exceedance of the 1-hour average HCl-equivalent emission rate limit. The criteria that you must consider are: (1) The ratio of the 1-hour average HCl-equivalent emission rate based on the total chlorine emission rate you select for each combustor to the 1-hour average HCl-equivalent emission rate limit for the combustor; and (2) the potential for the source to vary chlorine feedrates substantially over the averaging period for the long-term chlorine feedrate limit.

For example, if a source's primary chlorine-bearing feedstreams have a relatively constant chlorine concentration over the averaging period for the chlorine feedrate limit to ensure compliance with the annual average HCl-equivalent emission rate limit (e.g., generally 12-hours), as may be the case for commercial sources feeding from large burn tanks or on-site sources where chlorine levels in wastes are fairly constant, you may conclude that there is little probability that 1-hour feedrates would vary substantially over the averaging period. Thus, a 1-hour rolling average chlorine feedrate limit may not be warranted. Even if chlorine feedrates could vary substantially over the long-term feedrate averaging period, however, an hourly rolling average feedrate limit still may not be warranted if the source's 1-hour average HCl-equivalent emission rate is well below the 1-hour HCl-equivalent emission rate limit. See Part Two, Section IX.H, of this preamble for a discussion of the relationship between emission rates, emission rate limits, and feedrate limits.

We disagree with the commenter who states that short-term chlorine feedrate limits are not necessary. The 1-hour average HCl-equivalent emission rate limit could potentially be exceeded for sources with highly variable chlorine feedrates and where the 1-hour HCl-equivalent emission rate is relatively high compared to the 1-hour HCl-equivalent emission rate limit. The 1-hour average HCl-equivalent emission rate limit could be exceeded even though the source remains in compliance with the annual average HCl-equivalent emission rate limit (and, Start Printed Page 59484moreover, the 12-hour rolling average or (up to) annual rolling average chlorine feedrate limit).

We agree with commenters that suggest that the hourly rolling average chlorine feedrate limit should be extrapolated from performance test feedrates up to 100% of the 1-hour average HCl-equivalent emission rate limit. The final rule requires you to establish the hourly rolling average feedrate limit (if a limit is required under § 63.1215(d)(3)) as the 1-hour HCl-equivalent emission rate limit divided by [1 − system removal efficiency]. Establishing the hourly rolling average feedrate in this manner ensures that the 1-hour HCl-equivalent emission rate limit is not exceeded, and thus that the aREL-based Hazard Index of 1.0 is not exceeded.

We also agree in principle with commenters that suggest that the hourly rolling average feedrate limit be based on the 1-hour average HCl-equivalent emission rate limit which is based on emissions modeling. These commenters suggested that we use a multiplier of 10 or 12.5 (i.e., 1/0.08) to project 1-hour average HCl-equivalent emission rate limits from the annual average HCl-equivalent emission rate limits. Rather than use these approaches to project 1-hour average emissions from annual average emissions, however, we use emissions modeling to develop look-up tables for both 1-hour average HCl-equivalent emission rate limits and annual average HCl-equivalent emission rate limits. For sources that use site-specific risk assessment to demonstrate eligibility, they will use the same models to estimate 1-hour average maximum ambient concentrations. Thus, the final rule uses modeling to establish directly 1-hour average HCl-equivalent emission rate limits rather than approximating those limits from annual average HCl-equivalent emission rate limits as commenters suggest. In summary, the final rule requires you to establish the 1-hour average HCl-equivalent emission rate limit by either using Tables 3 or 4 in § 63.1215 to look-up the limit, or conducting a site-specific risk analysis. Under the site-specific risk analysis option, the 1-hour average HCl-equivalent emission rate limit would be the highest emission rate that the risk assessment estimates would result in an aREL-based Hazard Index not exceeding 1.0 at any off-site receptor location.

We do not agree that the short-term feedrate limit should be capped at the level corresponding to the Interim Standards for incinerators, cement kilns, and lightweight aggregate kilns. The final rule caps the total chlorine emission rate and the annual average HCl-equivalent emission rate limit at the level equivalent to the Interim Standard for total chlorine. Thus, the long-term chlorine feedrate limit (12-hour rolling average or (up to) an annual rolling average) is capped at the level corresponding to the Interim Standards for incinerators, cement kilns, and lightweight aggregate kilns. The hourly rolling average feedrate limit to maintain compliance with the 1-hour average HCl-equivalent emission rate limit, however, can exceed the numerical value of the long-term chlorine feedrate limit because the 1-hour average HCl-equivalent emission rate limit is substantially higher than the annual average HCl-equivalent emission rate limit. Thus, capping at the interim standard level is inappropriate unless the interim standard were somehow re-expressed as a 1-hour limit.

Comment: Many commenters state that requiring prior approval of the eligibility demonstration would be unworkable. Commenters are concerned that the permitting authority may not approve the demonstration prior to the compliance date even though the source has submitted complete and accurate information and has responded to any requests for additional information in good faith. Commenters are also concerned that the permitting authority may disapprove the demonstration too late for the source to take other measures to comply with the total chlorine MACT standard. Once commenter recommends the following alternative approach: (1) If the regulatory agency does not act on a risk demonstration within the 6-month period, it is conditionally deemed approved; and (2) if a risk demonstration is disapproved, the source would have to comply with the MACT emission standards no later than three years after notice of disapproval and, in the interim, sources would comply with current emission limits for total chlorine.

Another commenter suggests that, if the permitting authority has neither approved nor disapproved the eligibility demonstration by the compliance date, the source may begin complying on the compliance date with the alternative health-based limits specified in the eligibility demonstration.

Finally, another commenter states that facilities should be granted a three-year extension of the compliance date if the Agency denies a good-faith eligibility demonstration. The commenter is concerned that sources will not have time to install additional controls or take other measures after a denial is issued but prior to the compliance date.

Response: We agree with commenters that requiring prior approval of the eligibility demonstration may be unworkable for the reasons commenters suggest. We also agree with commenters that sources who make a good-faith eligibility demonstration but whose demonstration is denied by the permitting authority may need additional time to install controls or take other measures to comply with the MACT emission standards.

Accordingly, the final rule does not require prior approval of the eligibility demonstration for existing sources. If your permitting authority has not approved your eligibility demonstration by the compliance date, and has not issued a notice of intent to disapprove your demonstration, you may nonetheless begin complying, on the compliance date, with the HCl-equivalent emission rate limits and associated chlorine feedrate limits you present in your eligibility demonstration.

In addition, the final rule states that the permitting authority should notify you of approval or intent to disapprove your eligibility demonstration within 6 months after receipt of the original demonstration, and within 3 months after receipt of any supplemental information that you submit. A notice of intent to disapprove your eligibility demonstration, whether before or after the compliance date, will identify incomplete or inaccurate information or noncompliance with prescribed procedures and specify how much time you will have to submit additional information or comply with the total chlorine MACT standards. The permitting authority may extend the compliance date of the total chlorine MACT standards to allow you to make changes to the design or operation of the combustor or related systems as quickly as practicable to enable you to achieve compliance with the total chlorine MACT standards.

Comment: One commenter states that proposed § 63.1215(f)(1)(A) should have required sources to conduct a new comprehensive performance test only if there are changes that would decrease the HCl-equivalent emission rate limit below the HCl-equivalent emission rate demonstrated during the comprehensive performance test. Similarly, the commenter suggests that a retest should not be required if a change increases the HCl-equivalent emission rate limit but the source elects to maintain the current feedrate limit.

Another commenter states that the Agency should clarify that if there are any changes that are not controlled by the facility owner/operator, and the Start Printed Page 59485facility is required to change its design or operation to lower chlorine emissions to address the changes, the facility may request up to three years to make such changes.

Response: We generally agree with the commenters and have revised the rule as follows: (1) A new comprehensive performance test is required to reestablish the system removal efficiency for total chlorine only if you change the design, operation, or maintenance of the source in a manner that may decrease the system removal efficiency (e.g., the emission control system is modified in a manner than may decrease total chlorine removal efficiency); and (2) if you use the site-specific risk analysis option for your eligibility demonstration and changes beyond your control (e.g., off-site receptors newly residing or congregating at locations exposed to higher ambient levels than originally estimated) dictate a lower HCl-equivalent emission rate limit and you must make changes to the design, operation, or maintenance of the combustor or related systems to comply with the lower limit, you may request that the permitting authority grant you additional time to make those changes as quickly as practicable.

Comment: Several commenters state that the proposed approach for calculating chlorine emissions to address the potential bias using Method 26/26A attributable to high bromine or sulfur levels in feedstreams is not statistically valid. They indicate that the approach could lead to collection of total chlorine, hydrogen chloride and chlorine data that are contradictory and difficult to apply in a compliance situation. One commenter suggests that using Method 26/26A results for sources with bromine and sulfur dioxide, while recognizing that there is bias in the sampling method, will result in a valid compliance approach.

Response: We agree with commenters that the proposed approach to avoid the bias when feedstreams contain high levels of bromine or sulfur (bromine/chlorine ratio in feedstreams of greater than 5 percent, or sulfur/chlorine ratio in feedstreams of greater than 50 percent) during the comprehensive performance test may be problematic. The proposed approach would have required you to use Method 320/321 or ASTM D 6735-01 for hydrogen chloride measurements, to use Method 26/26A for total chlorine (i.e., hydrogen chloride and chlorine combined) measurements, and to calculate chlorine levels by difference. The potential problem is that chlorine emission levels are generally a very small portion of total chlorine measurements, and variability in the hydrogen chloride or total chlorine measurements due to method imprecision or other factors could result in inaccurate estimations of chlorine emission levels.

We do not agree, however, that using Method 26/26A for chlorine measurements for combustors feeding high levels of bromine or sulfur is acceptable-the chlorine measurement may be biased low. Chlorine emission levels must be determined as accurately as possible given that the long-term health threshold for chlorine is 100 times the threshold for HCl, and the short-term health threshold for chlorine is 10 times the threshold for HCl (i.e., using current RfCs and aRELs). To ensure that a conservative estimate of the chlorine emission rate is used to establish the alternative health-based emission limits and to address commenters' concerns, the final rule requires that you determine chlorine emissions to be the higher of: (1) The chlorine value measured by Method 26/26A, or an equivalent method; or (2) the chlorine value calculated by difference between the combined hydrogen chloride and chlorine levels measured by Method 26/26A, or an equivalent method, and the hydrogen chloride measurement from EPA Method 320/321 or ASTM D 6735-01, or an equivalent method.

Comment: Several commenters state the procedures for calculating HCl-equivalent emission rates cannot merely reference an outside source, such as a Web site, unless that reference specifies that the contents of the source are as of a date certain. To specify use of health threshold values that can change over time provides inadequate opportunity for notice and comment on the regulation.

Response: We believe that the best available sources of health effects information should be used for risk or hazard determinations. To assist us in identifying the most scientifically appropriate toxicity values for our analyses and decisions, the Web site to be used for RfCs identifies pertinent toxicity values using a default hierarchy of sources, with EPA's Integrated Risk Information System (IRIS) being the preferred source. The IRIS process contains internal and external peer review steps and IRIS toxicity values represent EPA consensus values. When adequate toxicity information is not available in IRIS, however, we consult other sources in a default hierarchy that recognizes the desirability of these qualities in ensuring that we have consistent and scientifically sound assessments. Furthermore, where the IRIS assessment substantially lags the current scientific knowledge, we have committed to consider alternative credible and readily available assessments (e.g., the acute Relative Exposure Levels established by the California Office of Health Hazard Assessment). For our use, these alternatives need to be grounded in publicly available, peer-reviewed information. We agree with the commenter that the issue of changing toxicity values is a general challenge in setting health-based regulations. However, we are committed to establishing such regulations that reflect current scientific understanding, to the extent feasible.

C. National Health-Based Standards for Cement Kilns

Comment: One commenter states that our suggestion at proposal that it would be appropriate to establish a single national emission rate type standard applicable to all cement kilns based on the worst-case scenario cement kiln is unduly burdensome as it discounts the benefits of improved dispersion realized by facilities that have invested in taller stacks that minimize downwash effects. The commenter recommends a dual limit for cement kilns such that the HCl equivalent emission rate is limited to both: (1) A 130 ppmv total chlorine emission standard (the Interim Standard) coupled with a chlorine feedrate limit based on a 12-hour rolling average; and (2) a Hazard Index of 1.0.

Response: We have decided not to include a separate national standard for cement kilns in the final rule for several reasons: (1) We have no assurance that the Cl2/HCl volumetric ratio exhibited during the most recent compliance test, and that was the basis for the commenter documenting in a study [187] that the Hazard Index of 1.0[188] was not exceeded, is representative of ratios in the past or future; (2) the commenter's recommended emission standard for cement kilns—130 ppmv total chlorine emission limit and a Hazard Index of 1.0—is equivalent to the requirements under § 63.1215 applicable to other hazardous waste combustors to establish site-specific emission limits; (3) the MACT standard for total chlorine for cement kilns is 120 ppmv such that the health-based standard that the commenter recommends—130 ppmv, Start Printed Page 59486the Interim Standard—would provide little compliance relief; and (4) even though the final rule does not provide a separate national health-based standard for cement kilns, cement kilns may apply for the health-based compliance alternatives applicable to other hazardous waste combustors.

Prior to publication of the proposed rule, the commenter submitted results of site-specific risk assessments for all cement kiln facilities showing that both the long-term and short term Hazard Index of 1.0 would not be exceeded at any facility assuming: (1) Sources emit total chlorine at the Interim Standard level of 130 ppmv; and (2) total chlorine emissions are apportioned between HCl and chlorine according to the apportionment exhibited during the most recent compliance test.

At proposal, we requested comment on how to ensure that the 130 ppmv concentration-based standard would ensure that total chlorine emission rates (lb/hr) would not increase to levels that may exceed the Hazard Index limit of 1.0 given that: (1) The partitioning ratio between HCl and chlorine could change over time such that a larger fraction of total chlorine could be emitted as chlorine, which has a much lower health risk threshold; and (2) the mass emission rate of total chlorine could increase. See 69 FR at 21306.

The commenter has addressed the concern about the mass emission rate of total chlorine potentially increasing by suggesting that the health-based standard include a limit on the feedrate of total chlorine and chloride at the level used in their risk assessment supporting a separate national standard for cement kilns. The commenter has also addressed the concern about the HCl and chlorine apportionment ratio changing over time by suggesting that the standard also include a requirement that the Hazard Index of 1.0 not be exceeded. We agree that sources need to account for variability in the chlorine to HCl ratio (see § 63.1215(b)(6)) and that periodic checks to ensure that the Hazard Index of 1.0 is not exceeded are needed. We believe the best way to ensure that the health-based compliance alternatives for total chlorine for cement kilns are protective with an ample margin of safety is through the procedures of § 63.1215 where site-specific emission rate limits are established rather than under a separate national standard for cement kilns.

VIII. Implementation and Compliance

A. Compliance Assurance Issues for both Fabric Filters and Electrostatic Precipitators (and Ionizing Wet Scrubbers)

1. Implementation Issues

Comment: Several commenters state that design and performance specifications and explicit detailed test procedures to determine conformance with the specifications are needed so that manufacturers can certify that their bag leak detection systems and particulate matter detection systems meet applicable criteria. Absent design and performance specifications and test procedures, commenters assert that the “manufacturer's certification” cannot ensure the performance capabilities of the devices.

Response: In general, we believe adherence to manufacturer's written specifications and recommendations is an appropriate approach to reasonably ensure performance of a bag leak detection system or particulate matter detection system, and we have retained that provision in the final rule. We agree, however, that there may be cases where other procedures are more appropriate than the manufacturer's recommendations to ensure performance of a bag leak detection system or particulate matter detection system. Consequently, the rule allows you to request approval for alternative monitoring procedures under § 63.1209(g)(1).[189] We note that you may use references other than EPA's Guidance Document, “Fabric Filter Bag Leak Detection Guidance,” September 1997 to identify appropriate performance specifications for the bag leak detection system or particulate matter detection system, including: PS-11 for PM CEMS; PS-1 for opacity monitors; and CPS-001 for opacity monitoring below 10% opacity. You may use these references to support your request for additions to, or deviations from, manufacturer's specifications.

Comment: One commenter states that bag leak detection systems and particulate matter detection systems should have a detection limit of 1.0 mg/acm to ensure peak performance is maintained rather than explicitly allowing sources to request approval for a detection limit on a site-specific basis as the rule currently allows. Several other commenters state that the bag leak detection system or particulate matter detection system need not have a detection limit as low as 1.0 mg/acm to detect increases in normal emissions. One commenter believes that bag leak detection systems installed on cement kilns should be allowed to have a detection limit of 10 mg/acm because: (1) A detection limit requirement of 10 mg/acm is more than sufficient to protect the particulate matter emission limit and to detect increases in particulate matter concentration given that the current particulate matter emission limit for existing kilns is 63 mg/dscm; (2) a detection limit requirement of 10 mg/acm is consistent with the requirement for bag leak detection systems in Subpart LLL, Part 63, for cement plants that choose to install bag leak detection systems on finish mills and raw mills, for bag leak detection systems and particulate matter detection systems installed on lime kilns under Subpart AAAAAA, and for industrial boilers under Subpart DDDDD; (3) a 10 mg/acm detection limit is achievable using state-of-the-art transmissometers (the actual instrument used in a continuous opacity monitoring system (COMS) at cement plants having kiln stack diameters of 2-3 meters, or greater; and (4) it is unclear if any bag leak detection system device can actually be demonstrated to achieve a 1.0 mg/acm detection limit except by extrapolation from tests conducted at higher dust loadings and theoretical arguments based on signal-to-noise ratios or other parameters. This commenter also recommends that EPA establish a 10 mg/am3 detection limit for all cement kilns rather than provide for site-specific determinations because allowing site-specific determinations is likely to create confusion in the selection of monitoring devices and further complicate the manufacturer's certification of performance requirements.

Response: The current requirement for the bag leak detection system sensitivity/detection limit applicable to incinerators and lightweight aggregate kilns is 1.0 mg/acm unless you demonstrate under § 63.1209(g)(1) that a lower sensitivity (i.e., higher detection limit) would detect bag leaks. We proposed to apply the bag leak detection system requirements to all hazardous waste combustors equipped with fabric filters and promulgate that requirement today. Although we also requested comment whether detection limits higher than 1.0 mg/acm should be allowed, none of the comments has convinced us to alter our view that the rule should allow higher detection limits on a site-specific basis. Similarly, Start Printed Page 59487we believe that the same detection limit requirement should apply to particulate matter detection systems that you may elect to use for compliance monitoring for your electrostatic precipitator or ionizing wet scrubber in lieu of site-specific operating parameter limits.

Both bag leak detection systems and particulate matter detection systems must be able to detect particulate emission in the range of normal concentrations. For example, to establish the alarm level for the bag leak detection system, you must first adjust detector gain/sensitivity and response time based on normal operations. Although the alarm level for particulate matter detection systems will be established based on operations during the comprehensive performance test or higher (see discussion below), the detector must be responsive within the range of normal operations for you to effectively minimize exceedances of the alarm level.

The range of normal emission concentrations will generally be well below both the particulate matter standard and emissions during the comprehensive performance test. Consequently, we disagree with commenters that believe the detection limit need only be within the range of emissions at the particulate matter emission standard. On the other hand, normal emissions may be well above 1.0 mg/acm such that a higher detection limit (e.g., 10 mg/acm) may be appropriate on a site-specific basis.

We also disagree with the comment that bag leak detection systems (or particulate matter detection systems) may not be able actually to achieve a 1.0 mg/acm detection limit. EPA is aware of bag leak detection system instruments certified to meet levels of 0.2 mg/m[3] and particulate matter detection systems can readily achieve detection limits well below 1.0 mg/acm.[190]

Comment: One commenter states that a continuous opacity monitoring system (COMS) that can achieve a detection level of 10 mg/acm or less can be used to monitor electrostatic precipitator performance. The commenter believes that allowing a COMS for compliance under Subpart EEE is also appropriate because cement kilns will be operating under the requirements of Subpart LLL (for cement kilns that do not burn hazardous waste) at times, which requires compliance with an opacity standard using a COMS.

Response: You may use a COMS (i.e., transmissometer) that meets the detection limit requirement as discussed above (i.e., 1.0 mg/acm or a higher detection limit that you document under an alternative monitoring petition under § 63.1209(g)(1) would routinely detect particulate matter loadings during normal operations) as the detector for your bag leak detection system or particulate matter detection system.

2. Compliance Issues

Comment: One commenter states that, if the bag leak detection system or particulate matter detection system exceeds the alarm level or an operating parameter limit (OPL) is exceeded, the automatic waste feed cutoff (AWFCO) system must be initiated. Allowing a source to exceed the alarm level for 5% of the time in a 6-month period does not ensure continuous compliance.

Response: Although the AWFCO system must be initiated if an OPL is exceeded, we believe that allowing exceedances of the bag leak detection system or particulate matter detection system alarm level up to 5% of the time in a 6-month period is reasonable. Requiring initiation of the AWFCO for an exceedance of an OPL is reasonable because sources generally can control directly the parameter that is limited. Examples are the feedrate of metals or chlorine, or pressure drop across a wet scrubber. Bag leak detection systems and particulate matter detection systems, however, measure mass emissions of particulate matter, a parameter that is affected by many interrelated factors and that is not directly controllable. We believe that the 5 percent alarm rate is a reasonable allowance for sources due to difficult-to-control variations in particulate matter emissions. More important, although the bag leak detection system and particulate matter detection system measure mass emissions of particulate matter, the detector response is not correlated to particulate matter emission concentrations to the extent necessary for compliance monitoring.[191] Thus, triggering the alarm level is not evidence that the particulate matter emission standard has been exceeded.

The purpose of a BLDS or PMDS is to alert the operator that the PM control device is not functioning properly and that corrective measures must be undertaken. We believe that using a BLDS or PMDS for compliance assurance better minimizes emissions of PM (and metal HAP) than use of operating parameter limits (which are linked to the automatic waste feed cutoff system). APCD operating parameters often have an uncertain relationship to PM emissions while the BLDS and PMDS provide real-time information on actual PM mass emission levels.[192]

Comment: One commenter states that requiring a notification if the bag leak detection system or particulate matter detection system set point is exceeded more than 5% of the time in a 6-month period is not cost-effective. Sources using bag leak detection systems have not linked exceedances to the data logging system and would incur an expense to do so.

Response: We continue to believe that limiting the aggregate duration of exceedances in a 6-month period is a reasonable approach to gage the effectiveness of the operation and maintenance procedures for the combustor. We note that recent MACT standards for several other source categories use this approach, including standards for industrial boilers and process heaters and standards for lime kilns.

Comment: One commenter states that EPA did not present a rationale for requiring a notification within 5 working days if the bag leak detection system or particulate matter detection system set point is exceeded more than 5% of the time during a 6-month period. The commenter notes that this notice is not required under the Subpart DDDDD boiler and process heater MACT. The commenter also notes that the source is required to take corrective measures under both the operation and maintenance plan and bag leak detection systems and particulate matter detection systems requirements. The commenter believes that requiring a report to the permitting authority is duplicative, unnecessary, and increases the burden on regulated facilities without providing additional protection to human health or the environment.

Response: If a source exceeds the alarm set point more than 5% of the time in a 6-month period, it is an indication that the operation and maintenance plan may need to be revised. Requiring the source to report the excess exceedances to the permitting Start Printed Page 59488authority serves as a notification that the authority may need to review the operation and maintenance plan with the source to determine if changes are warranted.

We agree with the commenter, however, that it is not necessary to require that the report be submitted within five working days of the end of the 6-month block period. Consequently, the final rule requires you to submit the report within 30 days of the end of the 6-month block period. Allowing 30 days to submit the report rather than 5 days as proposed is reasonable. We are concerned that 5 days may not be enough time to complete the report given that several exceedances toward the end of the 6-month block period may cause you to exceed the 5% time limit and that there may be many individual exceedances that need to be included in the report. We acknowledge that it may take some time to prepare the report given that you must describe the causes of each exceedance and the revisions to the operation and maintenance plan you have made to mitigate the exceedances.

Comment: One commenter notes that there is no guidance on how to calculate when the set-point has been exceeded more than 5 percent of the operating time within a 6 month period. The commenter notes that the MACT for industrial boilers and process heaters provides minimal instruction on how this is to be done, but it is not specific enough to enable facilities to ensure that they are in compliance with this requirement.

Response: For the final rule, we have adopted the procedures specified in the industrial boiler and process heater MACT for calculating the duration of exceedances of the set point. Those procedures are as follows:

1. You must keep records of the date, time, and duration of each alarm, the time corrective action was initiated and completed, and a brief description of the cause of the alarm and the corrective action taken.

2. You must record the percent of the operating time during each 6-month period that the alarm sounds.

3. In calculating the operating time percentage, if inspection of the fabric filter, electrostatic precipitator, or ionizing wet scrubber demonstrates that no corrective action is required, no alarm time is counted.

4. If corrective action is required, each alarm shall be counted as a minimum of 1 hour.

Although the commenter indicates that these procedures are not specific enough to ensure that sources are in compliance with the requirements, the commenter did not indicate the deficiencies or suggest additional requirements. If you need additional guidance on compliance with this provision, you should contact the permitting authority.

Comment: One commenter supports the approach of listing the shutting down of the combustor as a potential—but not mandatory—corrective measure in response to exceeding an alarm set point. Several commenters suggest, however, that EPA should specify that corrective measures could include shutting off the hazardous waste feed rather than shutting down the combustor. Other commenters state that it is inappropriate to imply that shutting down the combustor must be part of a corrective measures program for responding to exceedance of a set point. These commenters believe that the requirement to take corrective action upon the alarm is sufficiently protective. The facility should determine if shutting down the combustor is a necessary response to avoid noncompliance with a standard.

Response: You must operate and maintain the fabric filter, electrostatic precipitator, or ionizing wet scrubber to ensure continuous compliance with the particulate matter, semivolatile metals, and low volatile metals emission standards. Your response to exceeding the alarm set point should depend on whether you may be close to exceeding an operating parameter limit (e.g., ash feedrate limit for an incinerator or liquid fuel boiler equipped with an electrostatic precipitator) or an emission standard. If so, corrective measures should include, as commenters suggest, cutting off the hazardous waste feed. Corrective measures could also include, however, shutting down the combustor as the ultimate immediate corrective measure if an emission standard may otherwise be exceeded. Consequently, the final rule continues to require you to alleviate the cause of the alarm by taking the necessary corrective measure(s) which may include shutting down the combustor. This provision does not imply that shutting down the combustor is the default corrective measure. Rather, it implies that the ultimate immediate response, absent other effective corrective measures, would be to shut down the combustor.

Comment: One commenter states that periods of time when the combustor is operating but the bag leak detection system or particulate matter detection system is malfunctioning should not be considered exceedances of the set-point.

Response: If the bag leak detection system or particulate matter detection system is malfunctioning, the source cannot determine whether it is operating within the alarm set point. Accordingly, it is reasonable to consider periods when the bag leak detection system or particulate matter detection system is malfunctioning as exceedances of the set point.

B. Compliance Assurance Issues for Fabric Filters

Comment: One commenter states that establishing the set point for the bag leak detection system at twice the detector response achieved during bag cleaning as recommended by EPA guidance would not be sensitive enough to detect gradual degradation of the fabric filter, nor would it be low enough to require the operator of the source to take corrective measures that would ensure effective operation of the baghouse over time.

Response: The commenter expresses the same concern that EPA raised at proposal. See 69 FR at 21347. We have concluded, however, that it may be problematic to establish an alarm set point for fabric filters based on operations during the comprehensive performance test. This is because, as noted in earlier responses and at 69 FR at 21233, it is much more difficult to “detune” a fabric filter than an electrostatic precipitator to maximize emissions during the performance test.[193] Consequently, emissions from fabric filters that have not been detuned during the performance test may not be representative of the range of normal emissions caused by factors such as bag aging. Baghouse performance degrades over time as bags age. In addition, establishing the alarm set point based on operations during the performance test for baghouses that have not been detuned would establish more stringent compliance requirements on sources that perform the best—the lower the emissions, the lower the alarm set point. This would unfairly penalize the best performing sources.

For these reasons, the final rule requires you to establish the alarm set-point for bag house detection systems using principles provided in USEPA, “Fabric Filter Bag Leak Detection Guidance,” (EPA-454/R-98-015, September 1997).

Comment: One commenter states that the bag leak detection system requirement should not apply to the coal mill baghouse for cement kilns with indirect-fired coal mill systems where a fraction of kiln gas is taken Start Printed Page 59489from the preheater and routed to the coal mill and subsequently to a baghouse before entering the stack. The commenter notes that the PM in this gas is nearly exclusively coal dust, and the baghouse is substantially smaller than the baghouse for the kiln.

Response: We believe that a bag leak detection system is a reasonable approach to monitor emissions for the coal mill baghouse to ensure compliance with the particulate matter (and semivolatile and low volatile metals) emission standards. These systems are inexpensive to install and operate. Annualized costs are approximately $24,000.[194] Although the commenter did not suggest an alternative monitoring approach, and we are not aware of a less expensive and effective approach, we note that sources may petition the permitting authority under § 63.1209(g)(1) to request an alternative monitoring approach.

C. Compliance Issues for Electrostatic Precipitators and Ionizing Wet Scrubbers

Comment: Several commenters believe that a particulate matter detection system may not be necessary for monitoring of electrostatic precipitators and ionizing wet scrubbers. Commenters state that site-specific operating parameter limits linked to the automatic waste feed cutoff system can effectively monitor and ensure the performance of electrostatic precipitators and ionizing wet scrubbers. Particulate matter detection systems on cement kilns would have to operate in a high moisture stack environment (all kilns burning hazardous waste that are equipped with electrostatic precipitators are wet process kilns), with the potential for condensation and/or water droplet interference. Commenters state that when water droplets are present, many of these devices are not applicable.

Response: The final rule provides sources equipped with electrostatic precipitators or ionizing wet scrubbers the alternative of using a particulate matter detection system or establishing site-specific operating parameter limits for compliance assurance. If a particulate matter detection system is used, corrective measures must be taken if the alarm set point is exceeded. If the source elects to establish site-specific operating parameter limits, the limits must be linked to the automatic waste feed cutoff system.

In response to commenters' concern that high moisture stack gas may be problematic for particulate matter detection systems, we note that extractive light-scattering detectors and beta gauge detectors can effectively operate in high moisture environments. We acknowledge, however, that the cost of these extractive detector systems is substantially higher than transmissometers or in situ light-scattering detectors.

Comment: One commenter states that EPA must set minimum total power requirements for both ionizing wet scrubbers and electrostatic precipitators because allowing permit officials to establish compliance operating parameters on a site-specific basis frustrates the intention of the CAA by obviating the requirements for federal standards. The commenter asserts that a minimum total power requirement is monitorable, recordable, and reportable, three requirements that are necessary for these facilities to come into, and remain in compliance with, their Title V operating permits.

Other commenters state that electrostatic devices are not easily characterized by operating parameters in a “one-size-fits-all” fashion. The significant operating parameters for electrostatic devices are secondary voltage, secondary current, and secondary power (the product of the first two items). The relationship between these parameters and performance of the unit differ between applications and unit types. For example, inlet field power can increase as unit performance appears to decrease. In this case, an operating parameter other than secondary power by field would be more appropriate. The commenter notes that, in its various proposals over the years, EPA has discussed a number of approaches to establish operating parameter limits for electrostatic devices, including: Minimum total secondary power; minimum secondary power by field; pattern of increasing power from inlet to outlet field; and minimum secondary power of the last 1/3 of fields (or the last field). Commenters have also proposed: minimum specific power (secondary power divided by flue gas flow rate); minimum secondary voltage and/or secondary current; and total secondary voltage and/or secondary current. The commenter concludes that it is not surprising that there is so little agreement on the right approach, because different units and applications respond differently. EPA's proposal to let facilities and local regulators determine the best approach is far wiser than regulating from a distance.

Response: We agree with the commenters that state that it is not practicable to establish operating parameter limits that would effectively ensure performance of all electrostatic devices. Accordingly, the final rule continues to allow sources to establish site-specific operating parameter limits for these devices.

We disagree with the commenter's assertion that site-specific operating parameter limits obviate the requirements for federal standards. The site-specific operating parameter limits merely reflect the truism that no two sources are identical and so what each needs to do to comply with the uniform standards may differ. The final rule provides consistent, federally-enforceable emission standards. Necessary flexibility in compliance assurance for those emission standards does not undermine the uniformity of those standards. In addition, we disagree with the commenter's concern that without a minimum power limit, there will be no monitorable, recordable, and reportable Title V permit limits for electrostatic devices. To the contrary, site-specific operating parameter limits can and will be monitored, recorded, reported, and linked to the automatic waste feed cut-off system. And, if a source elects to use a particulate matter detection system in lieu of establishing site-specific operating parameter limits, the detector response will be monitored, recorded, reported, and linked to requirements to take corrective measures if the alarm set point is exceeded.

Comment: One commenter asserts that the use of electrostatic precipitator total power input data (sum of the product of kilovolts times milliamps for each electrostatic precipitator field) is one acceptable approach as a site-specific parameter to monitor electrostatic precipitator performance. Limits on power input for each field (or particular fields) are not warranted.

Response: A limit on total power input to a multifield electrostatic device is generally not an acceptable operating parameter for compliance assurance. We have documented that when total power input was held constant for a four-field electrostatic precipitator while the power input to the fourth field was decreased, emissions of particulate matter doubled from 0.06 gr/dscf to 0.12 gr/dscf. See 66 FR at 35143 (July 3, 2001). Thus, if the total power input during the comprehensive performance test were used as the operating parameter limit, particulate matter emissions could exceed the emission Start Printed Page 59490standard because of changes in other parameters that were not limited even though total power input did not exceed the parametric limit.

Notwithstanding our concern that a limit on total power input to a multifield electrostatic device is generally not an effective operating parameter for compliance assurance, this does not preclude you from documenting to the permitting authority that total power input is an effective compliance assurance parameter for your source. See § 63.1209(m)(1)(iv).

Comment: Several commenters suggest that the rule should offer various approaches to establish an achievable particulate matter detection system alarm level on a site-specific basis in lieu of the approach we proposed (i.e., average detector response during the comprehensive performance test): (1) Use the 2 times the maximum peak height or 3 times the baseline concepts developed in EPA's bag leak detection guidance documents; (2) allow spiking to set the alarm set point given that PS 11 allows for spiking as a way to calibrate PM CEMs; (3) establish the limit as the 99th percentile upper prediction limit of the average response during each performance test run instead of the average of the test run averages; (4) allow upward extrapolation from the average of the test run averages to some percentage of the particulate matter emissions standard (fraction could be variable depending upon how close to the standard the facility is during the compliance test); or (5) set the alarm point at the maximum test run.

Response: We agree with several of the commenters' suggestions: explicitly allowing spiking (and emission control device detuning) during the comprehensive performance test to maximize controllable operating parameters to simulate the full range of normal operations; and upward extrapolation of the detector response. See discussion below.

The final rule is consistent with commenters' suggestion to establish the alarm level for particulate matter detection systems on fabric filters based on the concepts in the Agency's guidance document on bag leak detection systems. Commenters made this suggestion in response to our request for comments on requiring particulate matter detection systems on fabric filters and establishing the alarm level based on the detector response during the comprehensive performance test. See 69 FR at 21347. The final rule requires bag leak detection systems on all fabric filters and suggests that you establish the alarm level using concepts in the bag leak detection system guidance.[195]

Neither the suggestion to establish the alarm level at the 99th percentile upper prediction limit (UPL99) based on the average response during the comprehensive performance test runs nor the suggestion to establish the alarm level at the maximum test run response would control PM emissions at the level achieved during the performance test or provide some assurance that the PM standard was not being exceeded, unless the detector response is correlated to PM concentrations. For example, if the detector response does not relate linearly to PM concentration (or if the response changes w/changes in particulate characteristics), the UPL99 detector response could relate to a much higher (e.g., 99.9th percentile) PM concentration. In addition, even if the detector response were correlated to PM concentration, there is no assurance that the correlation would be consistent over the range of the average detector response during the performance test to the UPL99 detector response. Note that under PS-11 for PM CEMS, even after complying with rigorous procedures to correlate the detector response to PM concentrations, the detector response may be extrapolated only to 125% of the highest PM concentration used for the correlation. Thus, the final rule does not use these approaches to establish the alarm level.

If you elect to use a particulate matter detection system in lieu of site-specific operating parameters for your electrostatic precipitator or ionizing wet scrubber, you must establish the alarm level using either of two approaches. See Appendix C of USEPA, “Technical Support Document for HWC MACT Standards, Volume IV: Compliance with the HWV MACT Standards,” September 2005. Under either approach, you may maximize controllable operating parameters during the comprehensive performance test to simulate the full range of normal operations (e.g., by spiking the ash feedrate and/or detuning the electrostatic device).[196]

You may establish the alarm set-point as the average detector response of the test condition averages during the comprehensive performance test.

Alternatively, you may establish the alarm set point by extrapolating the detector response. Under the extrapolation approach, you must approximate the correlation between the detector response and particulate matter emission concentrations during an initial correlation test. You may extrapolate the detector response achieved during the comprehensive performance test (i.e., average of the test condition averages) to the higher of: (1) A response that corresponds to 50% of the particulate matter emission standard; or (2) a response that correlates to 125% of the highest particulate matter concentration used to develop the correlation.

To establish an approximate correlation of the detector response to particulate matter emission concentrations, you should use as guidance Performance Specification-11 for PM CEMS (40 CFR Part 60, Appendix B), except that you need only conduct only 5 runs to establish the initial correlation rather than a minimum of 15 runs required by PS-11. In addition, the final rule requires you to conduct an annual Relative Response Audit (RRA) for quality assurance as required by Procedure 2—Quality Assurance Requirements for Particulate Matter Continuous Emission Monitoring Systems at Stationary Sources, Appendix F, Part 60.[197] The RRA is required on only a 3-year interval, however, after you pass two sequential annual RRAs.

The rule requires only minimal correlation testing because the particulate matter detection system is used for compliance assurance only—as an indicator for reasonable assurance that an emission standard is not exceeded. The particulate matter detection system is not used for compliance monitoring—as an indicator of continuous compliance with an Start Printed Page 59491emission standard. Because particulate matter detection system correlation testing and quality assurance is much less rigorous than the requirements of PS-11 for a PM CEMS, the particulate matter detection system response cannot be used as credible evidence of exceedance of the emission standard.

D. Fugitive Emissions

Comment: A commenter does not support EPA's proposed approach to allow alternative techniques that can be demonstrated to prevent fugitive emissions without the use of instantaneous pressure limits given that the CAA requires continuous compliance with the standards and given positive pressure events can result in fugitive emissions, irrespective of facility design.

Response: Rotary kilns can be designed to prevent fugitive emissions during positive pressure events. As stated in the February 14, 2002 final rule, and subsequently in the April 20, 2004 proposed rule, there are state-of-the-art rotary kiln seal designs (such as those with shrouded and pressurized seals) which are capable of handling positive pressures without fugitive releases. See 67 FR at 6973 and 69 FR at 21340. We have included documentation of such kiln designs in the docket.[198] Instantaneous combustion zone pressure limits thus may not be necessary to assure continuous compliance with these fugitive emission control requirements. Our approach to allow alternative techniques that have been demonstrated to prevent fugitive emissions is therefore reasonable and appropriate. We note that these alternative techniques must be reviewed and approved by the appropriate delegated regulatory official.[199]

Comment: A commenter disagrees with EPA's clarification that fugitive emission control requirements apply only to fugitives attributable to the hazardous waste, given that the CAA does not distinguish between HAP emissions that come from hazardous waste streams and other HAP emissions.

Response: The fugitive emission control requirements in today's final rule originated from the RCRA hazardous waste combustion fugitive emission control requirements for incinerators and boilers and industrial furnaces.[200] The primary focus of these RCRA requirements is to ensure hazardous waste treatment operations are conducted in a manner protective of human health and the environment.[201] It is therefore appropriate to clarify that the intent of this requirement is to control fugitive emission releases from the combustion of hazardous waste.

Furthermore, MACT requirements for source categories that do not combust hazardous waste (e.g., industrial boilers, Portland cement kilns, and commercial and industrial solid waste incinerators) do not have combustion chamber fugitive emission control requirements for the non-hazardous waste inputs or outputs (e.g., clinker product for cement kilns or coal and natural gas fuels for industrial boilers). We have previously taken the position that emissions not affected by the combustion of hazardous waste (e.g., clinker coolers, raw material handling operations, etc.) are regulated pursuant to the applicable nonazardous waste MACT rules.[202] ,[203] We conclude the clarification that the fugitive emission control requirements applies only to fugitive emissions that result from the combustion of hazardous waste is appropriate because it regulates emissions attributable to nonhazardous waste streams to the same level of stringency that otherwise would apply if the source did not combust hazardous waste.[204]

Comment: A commenter states that the instantaneous monitoring requirements are inappropriate because (1) EPA has not demonstrated that the average of the top 12% of boilers are capable of operating with no instantaneous deviations from the negative pressure requirements; and (2) these requirements, though not standards themselves, effectively increase the stringency of the standard itself beyond what even the best available technology can achieve.

Response: As previously discussed, the fugitive emission control requirements included in today's rule originated from the RCRA hazardous waste combustion chamber fugitive emission control requirements. These provisions allow sources to control fugitive emissions by “maintaining the combustion zone pressure lower than atmospheric pressure, or an alternative means of control equivalent to maintenance of combustion zone pressure lower than atmospheric pressure.” All sources that must comply with the provisions of this rule are, or were, required to control fugitive emissions from the combustion unit pursuant to RCRA.

The monitoring requirements in today's rule do not increase the stringency of the standard beyond what the best available technology can achieve. Although we do not have data that confirm negative pressure is being maintained on an instantaneous basis (as we define it)[205] for at least 12 percent of the boilers, we believe this is current practice and readily achievable by most sources.[206] These requirements have been in force for many years, and there is no basis for stating that they are unachievable (EPA is not aware of industrywide noncompliance with these provisions, the necessary premise of the comment). First, maintaining negative pressure is the option that most boilers elect to implement to demonstrate compliance with the RCRA fugitive emission control requirements. Second, negative pressure is readily achieved on an instantaneous basis in boilers through use of induced draft fans. Third, the requirements we are finalizing today for boilers are identical to the fugitive emission control requirements that hazardous waste incinerators, cement kilns, and lightweight aggregate kilns are currently complying with pursuant to the EEE interim standard regulations. See § 63.1206(c)(5). Most of these sources maintain negative combustion chamber pressure through use of induced draft fans, providing further evidence that continuously maintaining combustion Start Printed Page 59492zone pressure lower than ambient pressure is readily achievable by well designed and operated boilers.[207]

We note that use of instantaneous pressure monitoring is not a requirement. A source can elect to implement any of the four compliance options to control combustion system leaks as well as request to use alternative monitoring approaches. See §§ 63.1206(c)(5) and 63.1209(g). The instantaneous pressure monitoring option offers sources a method that satisfies the intent of the rule that can be applied at numerous sources. The inclusion of this requirement in today's rule is thus an attempt to simplify the review process for both regulators and affected sources; the absence of prescriptive compliance options in this case may likely result in time-consuming site-specific negotiations that would prolong the review and approval of comprehensive performance test workplans.

Comment: A commenter believes that requiring an instantaneous waste-feed cutoff when these pressure excursions occur is short-sighted and will result in greater HAP emissions. The commenter recommends EPA instead allow the use of reasonable pressure averaging periods in lieu of instantaneous pressure requirements.

Response: As discussed in the February 14, 2002 Final Amendments Rule, automatic waste feed cutoffs are appropriate non-compliance deterrents, and are necessary whenever an operating limit is exceeded. See 67 FR at 6973. Pressure excursions that result in combustion system leaks (and subsequently lead to automatic waste feed cutoffs) should be prevented by maintaining negative pressure in the combustion zone. We agree that needless triggering of automatic waste feed cutoffs due to short term pressure fluctuations that do not result in combustion system leaks would provide less environmental protection, not more. Today's rule offers three alternative options that do not require the use of instantaneous pressure monitoring to control combustion system leaks. See § 63.1206(c)(5). The use of averaging periods in these alternatives is not prohibited. Sources that elect to use an alternative compliance option must demonstrate that the alternative method is equivalent to maintaining combustion zone pressure lower than ambient pressure or, that the alternative approach prevents fugitive emissions.

E. Notification of Intent To Comply and Compliance Progress Report

1. Notice of Intent To Comply

In the NPRM, we proposed to re-institute the Notification of Intent to Comply (NIC) because we felt that it offered many benefits in the early stages of MACT compliance. As discussed in the 1998 “fast track” rule (63 FR 33782) and in the proposal, the NIC serves several purposes: as a planning and communication tool in the early implementation stages, to compensate for lost public participation opportunities when using the RCRA streamlined permit modification procedure to make upgrades for MACT compliance, and as a means to share information and provide public participation opportunities that would be lost when new units are not required to comply with certain RCRA permit requirements and performance standards. Please refer to the proposal at 69 FR 21313-21316 for additional discussion of the regulatory history, purpose, and implementation of the NIC provisions.

Overall, most commenters supported our decision to finalize NIC provisions. However, they also feel that the NIC should only be required for sources that have not completed a NIC previously (i.e., Phase 2 sources or Phase 1 sources that did not meet the previous NIC deadline) and for sources that need to make upgrades to comply with the final standards (i.e., either Phase 1 or Phase 2). They suggest that if sources do not need to make upgrades, then they should not be required to complete the NIC process, if they had done so previously. To require a second NIC would only add to the administrative burden and is not in line with Agency efforts to reduce reporting burdens. We agree that if Phase 1 sources do not need to make upgrades to comply with the Replacement Standards and if they completed the NIC process before, then it is not necessary to do so again.

In addition to the comment discussed above, a few commenters proposed that for sources who must still comply with the NIC because they wish to make upgrades, that the NIC public notice be combined with the Title V re-opening or renewal public notice. They point out that sources with existing Title V permits will have their permits re-opened or renewed to incorporate the new applicable requirements (i.e., Phase 1 Replacement or even Phase 2 Standards) shortly after the NIC public notice and meeting are to occur. Title V permit re-openings and renewals require: public notice, a minimum of 30 days for comment, and an opportunity to request a hearing.

While we do agree that the Title V re-opening and renewal requirements provide adequate information to the public and an opportunity for the public to comment and request a hearing, we are concerned that the timing requirements for the NIC may not correspond with the timing requirements for title V permit reopenings, revisions, and renewals. The public review of the draft NIC and subsequent public meeting are scheduled to occur 9 and 10 months, respectively, after the rule's effective date. On the other hand, Title V permits for major sources that have a remaining permit term of greater than 3 years from the rule's promulgation date will need to be re-opened, but this re-opening may not occur until 18 months beyond the promulgation date of the rule. Also, Title V permits that have a remaining permit term of less than 3 years from the rule's promulgation date will need to be renewed, but the timing of the renewal is contingent upon the individual permit term, not the timing requirements for public review of the draft NIC and public meeting. Thus, we do not believe there is ample opportunity to combine the requirements of the NIC and Title V process for the vast majority of sources.[208] Also, those sources that need to make upgrades to comply with the final standards and that need to modify any applicable conditions in their RCRA permit will not be able to request the streamlined modification procedure (see 40 CFR 270.42(j)) until they meet the NIC requirements. So the earlier they comply with the NIC requirements, the earlier they can begin upgrading their combustion units.

Another commenter suggested a change to the regulations at § 63.1210(c)(1) to account for sources that will cease burning hazardous waste prior to or on the compliance date. The regulations, as proposed, require sources to hold an informal public meeting to discuss anticipated activities described in the draft NIC even if they plan to cease burning hazardous waste. The commenter also suggested a similar change to § 63.1210(b)(2) that requires the draft NIC be made available for public review no later than 30 days Start Printed Page 59493prior to the public meeting. We agree with the commenter that it does not make sense to require sources that intend to cease burning hazardous waste to submit a NIC that discusses anticipated activities that will allow them to achieve compliance with the standards. We also agree that it is not necessary for those sources to hold an informal public meeting, since there are no MACT compliance activities to discuss. However, we believe that the public should be provided notice of the draft NIC so that they are aware of the source's intentions to cease burning and the steps (and key dates) the source will undertake to stop hazardous waste combustion activities.

With regard to Phase 2 sources, we had proposed that all Phase 2 sources comply with the same NIC requirements as the Phase 1 sources. Commenters did not express opinions in favor or against the NIC for Phase 2 sources. We believe that the NIC is beneficial in several respects. As mentioned previously, it serves as a planning and communication tool in the early implementation stages, it compensates for lost public participation opportunities when using the RCRA streamlined permit modification procedure to make upgrades for MACT compliance, and it is a tool to share information and provide public participation opportunities that would be lost when new units are not required to comply with certain RCRA permit requirements and performance standards. Ultimately, it creates more public confidence in the permitting process and so promotes a more stable regulatory environment.

For today's rule, we are finalizing our decision to re-institute the NIC provisions for Phase 1 and Phase 2 sources. We are including a few minor changes and clarifications to improve the proposed regulatory language based on commenters' suggestions. Section 63.1210(b) is revised so that Phase 1 sources that previously complied with the NIC requirements, and that do not need to make upgrades to comply with the Replacement Standards, are not required to comply with the NIC again. Sections 63.1210(b)(1)(iv) and (b)(2) have been revised and (c)(5) has been added so that sources that intend to cease burning hazardous waste prior to or on the compliance date are only required to prepare a (draft) NIC, make a draft of the NIC available for public review no later than 9 months after the effective date of the rule, and submit a final NIC to the Administrator no later than one year following the effective date of the rule. Last, we have revised language in § 63.1210(b) based upon a commenter's concerns that the term you “will” implies that sources are required meet their “estimated” dates for achieving key activities. We have removed “will” and replaced it with “anticipate” to more accurately represent the objective of the NIC, which is for sources to communicate their plans for complying with the standards in two years.

2. Compliance Progress Report

In the proposal, we explained why we thought a compliance progress report would be beneficial. In short, we believed it would help regulatory agencies determine whether Phase 1 and Phase 2 sources were making sufficient headway in their efforts to meet the compliance date. The progress report would be due to the regulatory agency at the midway point of the 3 year compliance period and would serve to update the information the source provided in its NIC. However, because we do not have any experience to draw upon regarding the value of the progress report, we requested comment on whether or not it should be required.

In response to our request for comment, all commenters were opposed to the progress report. They cited several reasons, with the most consistent one being that the progress report serves no useful purpose and imposes unnecessary additional burdens on sources. As we discussed above, sources and regulatory agencies will be focusing on the NIC as well as initial Title V applications, re-openings, revisions, and renewals during this three year compliance period. We agree with the commenter who noted that there is already significant interaction between sources and regulatory authorities during this period. Furthermore, we learned through implementation of the Interim Standards that some regulatory agencies found it difficult to manage the notices, applications, requests, and test plans that were due prior to the compliance date. Therefore, we have decided not to finalize any compliance progress report requirements for today's rule.

F. Startup, Shutdown, and Malfunction Plan

Comment: One commenter states that an exceedance of a standard or operating requirement during a malfunction should be a violation not only because source owners and operators need an incentive to minimize exceedances caused by malfunctions, but also because an exemption for malfunction periods would violate the plain language of the CAA. The commenter notes that an emission standard is defined by 42 U.S.C. § 7602(k) as a standard that “limits the quantity, rate, or concentration of emissions of air pollutants on a continuous basis, including any requirement relating to the operation of maintenance of a source to assure continuous emission reduction, and any design, equipment, work practice or operational standard * * *.” The commenter concludes that a standard that contains a malfunction exemption does not apply “on a continuous basis” as required by the statute. Likewise, the commenter concludes that an exemption for startup and shutdown periods would also violate this unambiguous statutory language.

The commenter also notes that, although some courts have held that a technology-based standard must provide some kind of an exemption for unavoidable technology failures, the rationale for such an exemption is that the underlying standard is based on the performance of a particular control technology that cannot be expected to function properly all of the time. The commenter believes that neither the rationale nor the exemption apply to section 112(d) standards, which are not based on the performance of any particular technology but instead must reflect the “maximum degree of reduction” that can be achieved, irrespective of the measures used by a source to achieve that reduction. CAA § 112(d)(2).

The commenter states that, even assuming for the sake of argument that EPA has authority to depart from the statutory language and carve out a startup, shutdown, and malfunction exemption, any such exemption must be narrowly drafted to apply only where a source demonstrates that a violation was unavoidable. See, e.g., Marathon Oil, 564 F.2d at 1272-73. As EPA recognizes, emission exceedances that occur during SSM events are frequently avoidable. See 69 FR at 21339/3 (noting that “proper operation and maintenance of equipment” helps avoid exceedances during startup, shutdown, and malfunction events), 69 FR at 21339/2 (describing the industry view that “some” exceedances that occur due to malfunctions are unavoidable). Thus, the commenter concludes that, even if a Marathon Oil-type exemption applies to a § 112(d) standard, it would be unlawful and arbitrary for EPA to exempt sources from liability for all emission exceedances occurring during startup, shutdown, and malfunction events. Rather, such an exemption could only apply where a source demonstrates that a given exceedance was unavoidable. Start Printed Page 59494

Many other commenters state that it would be illegal to require compliance with the emission standards and operating requirements during startup, shutdown, and malfunction events. The commenters note that EPA and the courts have long recognized that technology fails at times, despite a source's best efforts to maintain compliance. For this reason, the courts have recognized that technology-based standards such as EPA's § 112(d)(2) MACT standards must account for such unavoidable technology failures if the standards are to be truly “achievable.” Thus, the standards must excuse noncompliance with the actual emission standards during startup, shutdown, and malfunction events.

These commenters also note that EPA took the position in the September 1999 final MACT rule for hazardous waste combustors that exceedance of an operating requirement during startup, shutdown, or malfunction events was a violation if hazardous waste remained in the combustion chamber. The commenters note that industry groups challenged the rule, and while the D.C. Circuit did not reach this issue because it vacated the emission standards, it pointed out that “industry petitioners may be correct that EPA should have exempted HWCs from regulatory limits during periods of startup, shutdown, and malfunction, permitting sources to return to compliance by following the steps of a startup, shutdown, and malfunction plan filed with the Agency.” CKRC v. EPA, 255 F.3d 855, 872 (2001). Commenters conclude that, after reading this language, EPA officials wisely decided that hazardous waste combustors should not be required to meet the MACT emission standards and operating limits during startup, shutdown, and malfunction events.

Response: We agree with commenters who state that sources must be exempt from technology-based emission standards and operating limits during startup, shutdown, and malfunction events. Technology is imperfect and can malfunction for reasons that are not reasonably preventable. The regulations must provide relief for such situations. We believe that existing case law supports this position. See, e.g., Chemical Mfr's Ass'n v. EPA, 870 F. 2d at 228-230 (daily maximum limitations established at 99th percentile reasonable because rules also provide for upset defense for unavoidable exceedances); Marathon Oil v. EPA, 541 F. 2d at 1272-73 (acknowledged by commenter). As commenters noted, the D.C. Circuit also intimated in CKRC that some type of exception from compliance with standards during startup, shutdown and malfunction periods was required.

We do not agree with the commenter who contends that the § 112(d) MACT standards are not technology-based standards because they are not based on the performance of any particular technology but instead must reflect the “maximum degree of reduction” that can be achieved, irrespective of the measures used by a source to achieve that reduction. On the contrary, the standards must reflect the average performance of the best performing sources, which performance is achieved using technical controls—air pollution control devices, and for some pollutants, hazardous waste feedrate control. Those controls can fail for reasons that are not reasonably preventable. We note further that the situation was the same in the Clean Water Act cases which the commenter seeks to distinguish. Like section 112(d) standards, Clean Water Act standards are technology-based (reflecting Best Practicable Technology or Best Available Technology, see CWA sections 304 (b) and 301 (b)) and do not require use of any particular type of technology. See also Mossville, 370 F. 3d at 1242 (EPA must account for foreseeable variability in establishing MACT floor standards).

We agree with the commenter who states that any exemption from the emission standards and operating requirements during malfunctions must apply only where a source demonstrates that a violation was unavoidable. We note that the term malfunction is defined in § 63.2 as “any sudden, infrequent, and not reasonably preventable failure of air pollution control and monitoring equipment, process equipment, or a process to operate in a normal or usual manner which causes, or has the potential to cause, the emission limitations in an applicable standard to be exceeded. Failures that are caused in part by poor maintenance or careless operation are not malfunctions.” We believe this definition largely addresses the commenter's concern.

We acknowledge, however, that emissions can increase during malfunctions and potentially exceed the standards and agree that exceedances must be minimized. Accordingly, the final rule (and the current rule for incinerators, cement kilns, and lightweight aggregate kilns) requires that sources maintain compliance with the automatic hazardous waste feed cutoff system during malfunctions and notify the permitting authority if they have 10 or more exceedances of an emission standard or operating limit during a 6-month block period when hazardous waste is in the combustion chamber. See § 63.1206(c)(2)(v). This will alert the permitting authority that the source's operation and maintenance plan may not be adequate to maintain compliance with the emission standards and that the authority may need to direct the source to revise the plan under § 63.6(e)(3)(vi). Finally, we note that sources must report all excess emissions semiannually under § 63.10(e)(3) if an emission standard or operating limit is exceeded, including during malfunctions.

Comment: One commenter states that any exemption for emission exceedances during startup, shutdown, or malfunction events would violate the RCRA mandate for standards necessary “to protect human health and the environment.” 42 U.S.C. 6924(a). The commenter reasons that, because EPA's RCRA standards are health-based rather than technology-based, no unavoidability defense is available. Given that EPA concludes that the hazardous waste combustor MACT rule satisfies both its CAA and RCRA mandates, the emission standards and operating requirements cannot be waived during startup, shutdown, and malfunction events.

Response: We agree that the RCRA mandate to ensure protection of human health and the environment applies at all times, including during startup, shutdown, and malfunction events. Accordingly, the existing MACT requirements for incinerators, cement kilns, and lightweight aggregate kilns give sources the option of continuing to comply with RCRA permit requirements to control emission during these events, or to comply with special MACT requirements that are designed to be proactive and reactive and intended to be equivalent to the incentive to minimize emissions during these events provided by the RCRA requirements. See existing § 63.1206(c)(2)(ii). The special MACT requirements require sources to include proactive measures in the startup, shutdown, and malfunction plan to minimize the frequency and severity of malfunctions and to submit the startup, shutdown, and malfunction plan to the permitting authority for review and approval. We proposed to require boilers and hydrochloric acid production furnaces to comply with those same provisions providing for equivalence between the two sets of requirements, and promulgate those provisions today.

Comment: One commenter states that the rule should clarify the definitions of startup, shutdown, and malfunctions to preclude sources from improperly Start Printed Page 59495classifying as unavoidable exceedances those exceedances that could have been avoided had the source implemented an appropriate operation and maintenance plan. Many other commenters state that the current definitions in § 63.2 clearly define these terms.

Response: We believe the definitions of startup, shutdown, and malfunction are clearly defined in § 63.2, and combined with the startup, shutdown, and malfunction plan requirements, will preclude sources from improperly classifying as malfunctions events that could have been reasonably prevented by following appropriate procedures in the operation and maintenance plan. As discussed above, the definition of malfunction clearly states that failures that are caused in part by poor maintenance or careless operation are not malfunctions.

Comment: One commenter states that all stack bypasses, automatic waste feed cutoffs, and excursions from the operating parameter limits should be considered malfunctions.

Response: All failures resulting in stack bypasses, automatic waste feed cutoff, and excursions from the operating parameter limits are not malfunctions. As discussed above, failures caused in part by poor maintenance or careless operation are not malfunctions.

Comment: One commenter states that the rule should require sources to expand the startup, shutdown, and malfunction plan to address specific proactive measures that the source has considered and is taking to minimize the frequency and severity of malfunctions. Many other commenters believe that it is not necessary to expand the scope of the startup, shutdown, and malfunction plan beyond that required under § 63.6(e)(3) for other MACT source categories.

Response: We do not believe that it is necessary to expand the scope of the startup, shutdown, and malfunction plan generically for all hazardous waste combustors to address specific proactive measures that the source has considered and is taking to minimize the frequency and severity of malfunctions. Imposing additional requirements in particular situations is appropriate, however. For example, as discussed above, this expanded plan is required for sources that elect to meet the RCRA mandate using provisions of the startup, shutdown, and malfunction plan. See § 63.1206(c)(2)(ii). In addition, the plan with expanded scope may be appropriate for sources that have demonstrated an inability to minimize malfunctions. Consequently, the permitting authority should consider expanding the scope of the startup, shutdown, and malfunction plan on a site-specific basis under authority of § 63.6(e)(3)(vii) if the source has excessive exceedances during malfunctions. See § 63.1206(c)(2)(v)(A)(3) defining excessive exceedances during malfunctions and requiring reporting of the exceedances in the excess emissions report required under § 63.10(e)(3).

Comment: Two commenters state that all startup, shutdown, and malfunction plans should be submitted for review and approval by the delegated authority and made available for a 60-day public review period. Review and approval of the plans is needed in light of EPA's acknowledgment that most excess emissions would occur during startup, shutdown, and malfunctions. One of these commenters also believes that the regulations should provide for the public review period to be extended as necessary to accommodate a thorough public review. The reviewing authority should be required to provide a written response to public comments explaining any decision to reject a public comment suggesting ways for a facility to limit emissions during startup, shutdown, and malfunction events.

Many other commenters have concerns with requiring review and approval of startup, shutdown, and malfunction plans, except as required under § 63.1206(c)(2)(ii) for sources that elect to meet the RCRA mandate using provisions of the startup, shutdown, and malfunction plan as discussed above.

Response: Commenters express the same views here that they expressed under the rulemaking the Agency recently completed to revise the startup, shutdown, and malfunction plan requirements of the General Provisions applicable to all MACT source categories. See 68 FR at 32589-93 (May 30, 2003).

EPA concluded in that final rule that the Administrator may at any time request in writing that the owner or operator submit a copy of any startup, shutdown, and malfunction plan (or a portion thereof). Upon receipt of such a request, the owner or operator must promptly submit a copy of the requested plan (or a portion thereof) to the Administrator. In addition, the Administrator must request that the owner or operator submit a particular startup, shutdown, or malfunction plan (or a portion thereof) whenever a member of the public submits a specific and reasonable request to examine or to receive a copy of that plan or portion of a plan.

These provisions to provide the Administrator and the public with access to startup, shutdown, and malfunction plans, coupled with the provisions of § 63.6(e)(3)(vii) under which the Administrator must require the source to make changes to a deficient plan, should ensure that startup, shutdown, and malfunction plans are complete and accurate. We note that under § 63.6(e)(3)(vii) the Administrator must require the source to revise the plan if the plan: (1) does not address a startup, shutdown, or malfunction event that has occurred; (2) fails to operate the source (including associated air pollution control and monitoring equipment) during a startup, shutdown, or malfunction event in a manner consistent with the general duty to minimize emissions; (3) does not provide adequate procedures for correcting malfunctioning process and/or air pollution control and monitoring equipment as quickly as practicable; or (4) includes an event that does not meet the definition of startup, shutdown, or malfunction listed in § 63.2.

The commenter advocating that all hazardous waste combustors should be required to submit their startup, shutdown, and malfunction plans for review and approval did not explain why the concerns the Agency expressed in the General Provisions rulemaking (see 68 FR at 32589-93) are not valid for hazardous waste combustors. Accordingly, we do not believe it is appropriate to deviate from the General Provisions to require that all hazardous waste combustors submit their startup, shutdown, and malfunction plans for review.

G. Public Notice of Test Plans

1. What Are the Revised Public Notice Requirements for Test Plans?

Prior to the proposal, it was brought to our attention that the Agency did not provide any direction in the 1999 final rule regarding how and when sources should notify the public, what the notification should include, or where and for how long performance test plans should be made available. Consequently, we proposed to add clarifying language to the § 63.1207(e)(2) public notification requirement for approved performance test and CMS performance evaluation test plans because we believe that providing opportunities for timely and adequate public notice is necessary to fully inform nearby communities of a source's plans to initiate important waste management activities. The proposed clarifications are based upon the RCRA Expanded Public Participation Rule (60 FR 63417, December 11, 1995) requirements for Start Printed Page 59496public notification of an impending trial burn test. As a result, we did not feel that the clarifications imposed any new or additional requirements upon sources that will conduct a MACT comprehensive performance test or confirmatory performance test.

Commenters generally supported the clarifications to the public notice.[209] However, they suggested a change to the proposed requirement to provide notice of test plan approval no later than 60 days prior to conducting the test. The basis for suggesting a change is that many sources had not received approval of their test plans 60 days prior to the deadline for initiating their test under the Interim Standards. Moreover, several sources did not receive approval until well after the deadline for initiating the test. The problem created for these sources is that the required 60 day notification of the approved test plan effectively determines when the source will be able to begin its test. In other words, its test would need to be postponed until the approved test plan had been noticed for 60 days. Thus, commenters provided several possible alternatives.

One alternative that would avoid causing delays to testing is to require the public notice when the source submits its test plan. Although this fulfills the notification requirement, this alternative has a shortfall: The notice would occur at least one year (barring any extensions) in advance of the test and given this long period of time, the test plan is likely to be modified prior to approval. A second alternative is to provide notice of the test plan 60 days before the test as before, but regardless of approval status. This alternative is improved over the first, but still faces the same problem of potentially not offering the public an opportunity to view a final approved plan. A third alternative is to issue notice of the test plan as soon as it is approved. With this alternative, the public will have the most up-to-date information; however, it may not be until a few days prior to commencement of the test. Ideally, the second and third alternatives could be combined to provide the best possible chance of providing the public with an approved test plan in a reasonable period of time prior to the test. On the other hand, that would potentially require the facility to issue two notices if the test plan is not approved 60 days prior to the test. We do not believe this would be reasonable given that sources will be focused on activities associated with the impending test.

In consideration of practicality, we believe that the second alternative provides an adequate solution. As we mentioned, the drawback is that the public may not have the opportunity to view an approved test plan. However, we believe it is more important that the public be aware of a source's plans (i.e., how and when) for conducting the performance test.[210] This way, if they have questions, there will be 60 days in which they may contact the regulatory authority or the source before the test is scheduled to begin. This alternative will also eliminate the conflict associated with the confirmatory performance test. The regulations at § 63.1207(e)(1)(ii) specify that a source must submit to the regulatory authority its notice of intent to conduct a confirmatory performance test and the applicable test plans at least 60 calendar days prior to the date the test is to begin. Since we are no longer requiring that the test plans be approved before issuing public notice, sources would then provide notice of their confirmatory performance test plan to the public at the same time they submit their notice of intent and test plans to the regulatory authority. Therefore, we are requiring that sources issue the public notice of test plans 60 days in advance of commencing the performance test, whether their test plans have been approved or not. The regulations at § 63.1207(e)(2) have been revised accordingly.

One last concern related to the public notice of approved test plans involves sources that choose to conduct a performance test without an approved test plan (e.g., both time extensions provided by §§ 63.7(h) and 63.1207(e)(3) have expired or due to other circumstances, the source has elected to begin the test without approval). Because we did not believe any sources would choose or need to do so, we did not propose any guidance or regulations specific to issuing notice to the public of their test plans. Nevertheless, a few commenters raised this possibility indirectly in their discussion of the problematic 60 day notice of approved test plan requirement. The revised proposal addresses this concern by no longer requiring that test plans be approved before issuing public notice. Thus, sources that choose to begin their test without an approved plan will have complied with the requirement to issue public notice. Irrespective of the public notice requirements for noticing test plans, we expect that sources will notify their regulatory authority of their decision to proceed with their test in the absence of plan approval.

2. What Are the Revised Public Notice Requirements for the Petition To Waive a Performance Test?

In the Final Amendments Rule (67 FR 6968, February 14, 2002), the Agency did not provide direction regarding how, when, where, and what should be included in the public notice for a petition for time extension if the Administrator fails to approve or deny test plans.[211] In the proposal, we believed it important to provide clarification regarding when the notice must be issued and what it should contain. Thus, we proposed to revise paragraph § 63.1207(e)(3)(iv).

We received only one comment in response to the proposed requirements. The commenter did not express any concern over the requirements themselves, but rather suggested a change to terminology used. The commenter feels that the terms “to waive a performance test” or “waiver” as used in § 63.1207(e)(3)(iv) could be confusing to readers when we are actually referring to a time extension for commencing the test. Although we agree the terminology could be confusing, 40 CFR 63.1207(e)(3) clearly uses the term “waiver” in the context of an extension of time to conduct the performance test at a later date, implying that the deadline can be waived in this specific situation. The use of the term waiver is derived from the General Provisions requirements for requesting a waiver of performance tests (§ 63.7(h)). Thus, § 63.7(h)(3) provides the basis by which sources may petition, in the form of a waiver, for a time extension under § 63.1207(e)(3). In consideration of the above and that the existing regulations of § 63.1207(e)(3)(i)-(iii) consistently use the term waiver, we do not feel that a change to § 63.1207(e)(3)(iv) is warranted.

H. Using Method 23 Instead of Method 0023A

Comment. Most commenters support our proposal to allow the use of Method 23 instead of Method 0023A if a source includes this request in the comprehensive test plan to the permitting authority. Some commenters believe that Method 23 should be Start Printed Page 59497allowed in all cases without prior approval or on a source category basis.

Response. We proposed to allow sources to use Method 23 for dioxin and furan testing instead of SW-846 Method 0023A in situations where the enhanced procedures found in Method 0023A would not increase measurement accuracy. We proposed this change in the July 3, 2001, proposed rule, and again in the April 20, 2004, proposal. See 66 FR at 35137 and 69 FR at 21342.

The final rule promulgates this change as proposed. See § 63.1208(b)(1)(i). You may use Method 23 in lieu of Method 0023A after justifying use of Method 23 as part of your performance test plan that must be reviewed and approved the delegated permitting authority. You may be approved to use Method 23 considering factors including whether previous Method 0023A analyses document that dioxin/furan are not detected, are detected at low levels in the front half of Method 0023A, or are detected at levels well below the emission standard, and the design and operation of the combustor has not changed in a manner that could increase dioxin/furan emissions. We note that coal-fired boilers and combustors equipped with activated carbon injection systems may not be able to support use of Method 23, however, because these sources' stack gas is likely to contain carbonaceous particulate. Thus, these sources are likely to benefit the most from using Method 0023A.

The final rule does not automatically allow use of Method 23 for particular source categories because we cannot assess whether all sources in a category meet the conditions for use of Method 23—generally that quality assurance may not be improved—such as those listed above. These determinations can only be made on a site specific basis by the permitting authority most familiar with the particular source.

Comment: Commenters do not believe that an additional petition process (i.e., under § 63.1209(g)(1)) is necessary before allowing use of Method 23. Instead, EPA should require that the use of Method 23 should be submitted with the test plan to the regulatory agency for approval.

Response: We agree that a separate petition is unnecessary. Sources should include a justification to use Method 23 in the performance test plan that is submitted for review and approval. This will allow the permitting authority to determine whether use of Method 23 is appropriate for the source.

Comment: Two commenters state that “the justification of the use of Method 23 will not be by the existing system of a petition to EPA, but will be included as a part of the performance test plan that is submitted to the delegated regulatory authority for review and approval. This means that the expertise, training, and decision-making will not be consistent across the country. This is especially a problem because of the severe resource, training and staff reductions among the delegated regulatory authorities across the country and from region to region. The decision to allow or disallow use of Method 23 should come specifically, for each case, from EPA consideration of the submitted justification, based on the knowledge and expertise of trained and experienced EPA staff. This is important for uniformly applying the testing requirements all across the country.”

Response: We disagree, and we believe the responses to comments in today's rule make clear when Method 23 is an acceptable substitute for Method 0023A. If the source has carbon in the flue gas, as is the case with coal-fired boilers, boilers with carbon injection, and other sources likely to have a substantial amount of carbonaceous particulate matter in the flue gas, Method 0023A will generally be preferable because it includes procedures to account for dioxin and furan bound to carbonaceous particulate matter found in the probe and filter. In other situations, Method 23 will generally give the same results at a lower cost.

I. Extrapolating Feedrate Limits for Compliance With the Liquid Fuel Boiler Mercury and Semivolatile Metal Standards

Comment: One commenter questions whether allowing sources to extrapolate metal feedrates downward from the levels achieved during the comprehensive performance test to establish a metal feedrate limit will ensure compliance with the emission standards.

Response: The mercury and semivolatile metals standards for liquid fuel boilers are annual average emission limits where compliance is established by a rolling average mercury feedrate limit with an averaging period not to exceed an annual rolling average (updated hourly).[212] We use this approach because the emissions data used to establish the standards are more representative of normal emissions than compliance test emissions.[213]

As we explained at proposal, to ensure compliance with the mercury and semivolatile metal emission standards for liquid fuel boilers, you must document during the comprehensive performance test a system removal efficiency for the metals and back-calculate from the emission standard a maximum metal feedrate limit that must not be exceeded on an (not to exceed) annual rolling average. See 69 FR at 21311-12. If your source is not equipped with an emission control system (such as activated carbon to control mercury) for the metals in question, however, you must assume zero system removal efficiency. This is because, although a source that is not equipped with an emission control system may be able to document a positive system removal efficiency in a single test, that removal efficiency is not likely to be reproducible. Rather, it is likely to be an artifact of the calculation of emissions and feeds rather than a removal efficiency that can reliably be repeated.

To ensure that you can calculate a valid, reproducible system removal efficiency for sources equipped with a control system that effectively controls the metal in question, you may need to spike metals in the feed during the comprehensive performance test at levels that may result in emissions that are higher than the standard. This is appropriate because compliance with an emission standard derived from normal emissions data is based on compliance with an (not to exceed) annual average feedrate limit calculated as prescribed here, rather than compliance with the emission standard during the comprehensive performance test.[214]

The commenter is concerned that downward extrapolation from the levels achieved during the comprehensive performance test to establish a metal feedrate limit may not ensure Start Printed Page 59498compliance with the standard because system removal efficiency may be lower at lower feedrates.

This is a valid concern, and we have investigated it since proposal. We conclude that downward extrapolation of feedrates for the purpose of complying with the mercury and semivolatile metals emission standards for liquid fuel boilers will ensure compliance with the emission standards under the conditions discussed below.

We investigated the theoretical relationship between stack gas emissions and feedrate considering vapor phase metal equilibrium, the chlorine, mercury, and semivolatile metal feedrates for liquid fuel boilers in our data base, and the mercury and semivolatile emission standards for liquid fuel boilers.[215] We considered sources equipped with dry particulate matter controls and sources equipped with wet particulate matter controls.

Sources Equipped with Dry Controls. For sources equipped with dry controls other than activated carbon, mercury is not controlled. Thus, you must assume zero system removal efficiency. Consequently, if you are in the low Btu subcategory and comply with the mercury standard expressed as a mass concentration (μg/dscm), the mercury feedrate limit expressed as an MTEC (maximum theoretical emission concentration, μg/dscm) is equivalent to the emission standard.[216] If you are in the high Btu subcategory and comply with the mercury standard expressed as a hazardous waste thermal emission concentration (lb/MM Btu), the mercury feedrate limit expressed as a hazardous waste thermal feed concentration (lb/MM Btu) is also equivalent to the emission standard.

For semivolatile metals, the theoretical relationship between emissions and feedrate indicates that downward extrapolation introduces only a trivial error'0.17% at an emission rate 100 times the standard irrespective of the level of chlorine present. Id. Nonetheless, to ensure the error is minimal and to be practicable, you should limit semivolatile emissions during the comprehensive performance test to five times the emission standard.

Sources Equipped with Wet Scrubbers. For sources equipped with wet scrubbers, we conclude that the approach we use for semivolatile metals for dry scrubbers will also be appropriate to extrapolate a semivolatile metal feedrate limit for wet scrubbers. To ensure that downward extrapolation of the feedrate limit is conservative and to be practicable, you should limit semivolatile metal emissions during the comprehensive performance test to five times the emission standard.

For mercury, ensuring control with wet systems is more complicated because the level of chlorine present affects the formation of mercuric chloride which is soluble in water and easily controlled by wet scrubbers. Elemental mercury has very low solubility in scrubber water and is not controlled. The worst-case situation for conversion of elemental mercury to soluble mercuric chloride would be when the chlorine MTEC is lowest and the mercury MTEC is highest. We conclude that downward extrapolation of mercury feedrates is conservative for feedstreams that contain virtually no chlorine, e.g., below an MTEC of 100 μg/dscm. In addition, we conclude that downward extrapolation is appropriate [217] for boilers feeding chlorinated feedstreams provided that during the performance test: (1) Scrubber blowdown has been minimized and the scrubber water has reached steady-state levels of mercury prior to the test (e.g., by spiking the scrubber water); (2) scrubber water pH is minimized (i.e., you establish a minimum pH operating limit based on the performance test as though you were establishing a compliance parameter for the total chlorine emission standard); and (3) temperature of the scrubber water is maximized (i.e., you establish a maximum scrubber water temperature limit).

J. Temporary Compliance With Alternative, Otherwise Applicable MACT Standards

Comment: One commenter requests clarification on the requirements applicable to a source that switches to an alternative mode of operation when hazardous waste is no longer in the combustion chamber under the provisions of § 63.1206(b)(1)(ii). The commenter suggests that § 63.1206(b)(1)(ii) can imply that the complete compliance strategy needs to be switched over to the alternative section 112 or 129 requirements, even though compliance with the Subpart EEE requirements for monitoring, notification, reporting, and recordkeeping remains environmentally protective under Subpart EEE. For example, the commenter notes that § 63.1206(b)(1)(ii) could be incorrectly interpreted to require a source to comply with illogical requirements when the source temporarily switches to alternative, otherwise applicable standards, including standards testing and opacity monitoring under the alternative section 112 or 129 requirements. The commenter states that this interpretation makes little sense because a source that temporarily changes its mode of operation will continue to do testing under Subpart EEE, Part 63, or, in the case of opacity, the alternative section 112 requirements for cement kilns would necessarily require duplicate systems and compliance with redundant limits because a source may already be using a bag leak detection system or a particulate matter detection system. The commenter suggests only requiring sources to comply with the otherwise applicable emission standards under the alternative section 112 or 129 requirements while still operating under the various associated compliance requirements of Subpart EEE, part 63.

Response: The commenter requests clarification of § 63.1206(b)(1)(ii), which states that if a source is not feeding hazardous waste to the combustor and the hazardous waste residence time has expired (i.e., the hazardous waste feed to the combustor has been cut off for a period of time not less than the hazardous waste residence time), then the source may elect to comply temporarily with alternative, otherwise applicable standards promulgated under the authority of sections 112 and 129 of the Clean Air Act.[218] As we have explained in previous notices,[219] sources that elect to invoke § 63.1206(b)(1)(ii) to become temporarily exempt from the emission standards and operating requirements of Subpart EEE, Part 63, remain an affected source under Subpart EEE (and only Subpart EEE) until the source is no longer an affected source by meeting the requirements specified in Table 1 of § 63.1200. Of course, a source can elect not to use the alternative requirements for compliance during periods when Start Printed Page 59499they are not feeding hazardous waste, but, if so, the source must comply with all of the operating and monitoring requirements and emission standards of Subpart EEE at all times.[220] To implement § 63.1206(b)(1)(ii) a source defines the period of compliance with the otherwise applicable sections 112 and 129 requirements as an alternative mode of operation under § 63.1209(q). In order to be exempt from the emission standards and operating requirements of Subpart EEE, a source documents in the operating record that they are complying with the otherwise applicable Section 112 and 129 requirements specified under § 63.1209(q).

The commenter recommends that the complete compliance strategy need not be switched over to the alternative section 112 and 129 requirements when temporarily switching to the alternative standards. In general, we disagree. The intent of § 63.1206(b)(1)(ii) is to ensure that a source is complying with all requirements of sections 112 and 129 as an alternative mode of operation in lieu of the requirements under Subpart EEE. In the 1999 final rule we stated that the source must comply with all otherwise applicable standards under the authority of sections 112 and 129. Specifically, the source must comply with all of the applicable notification requirements of the alternative regulation, comply with all of the monitoring, recordkeeping, and testing requirements of the alternative regulation, modify the Notice of Compliance (or Documentation of Compliance) to include the alternative mode(s) of operation, and note in the operating record the beginning and end of each period when complying with the alternative regulation. See 64 FR at 52904. A source that elects to comply with otherwise applicable standards under § 63.1206(b)(1)(ii) must specify all requirements of those standards, not only the emission standards applicable under the sections 112 and 129 standards, but also the associated monitoring and compliance requirements and notification, reporting, and recordkeeping requirements in the operating record under § 63.1209(q).

The commenter suggests that a source should be able to comply with the otherwise applicable emission standards, while continuing to operate under the associated compliance requirements for the HAP under Subpart EEE. An example would be a cement kiln source complying with the dioxin and furan monitoring requirements under § 63.1209(k) of Subpart EEE for the dioxin and furan standards under § 63.1343(d) under Subpart LLL. We did not determine, when promulgating the provisions of §§ 63.1206(b)(1)(ii) and 63.1209(q)(1), that the monitoring provisions under Subpart EEE are equivalent to the associated monitoring requirements under the otherwise applicable 112 and 129 standards, or indeed, whether they are even well-matched. Such a determination would require notice and opportunity for comment, which we have not provided. However, this should not be interpreted to mean that a similar determination could not be made on a site-specific basis given that the MACT general provisions allow a source to request alternative monitoring procedures under § 63.8(f)(4). Certainly, a source can apply under this provision that the compliance requirements under Subpart EEE satisfy the associated monitoring requirements under the otherwise applicable 112 and 129 standards.

We also disagree with the commenter that emissions testing under the alternative standards of sections 112 and 129 is an example of an illogical requirement under § 63.1206(b)(1)(ii). Performance testing generally is required to demonstrate compliance with the emission standards and to establish limits on specified operating parameters to ensure compliance is maintained. In order to take advantage of the alternative under § 63.1206(b)(1)(ii), a source needs to show that compliance with and establish operating parameter limits for the otherwise applicable standards of sections 112 and 129. Thus, testing in order to establish operating parameter limits will be necessary. However, this does not mean that a separate performance test with the alternative sections 112 or 129 standards is necessarily required. We note that a source can make use of the performance test waiver provision under § 63.7(h) of the general provisions to request that the performance test under the alternative sections 112 and 129 standards be waived because the source is meeting the relevant standard(s) on a continuous basis by continuing to comply with Subpart EEE for the relevant HAP. This approach may be practicable for sources that can demonstrate that their level of performance during testing under Subpart EEE, including the associated operating and monitoring limits, will undoubtedly ensure continuous compliance with the emissions standards and the associated operating limits of alternative sections 112 and 129 standards.

Finally, the commenter notes that Subpart LLL (the alternative section 112 standards for cement kilns) includes opacity monitoring while Subpart EEE may not. The commenter states that this unnecessarily would require duplicate systems and compliance with redundant limits because of the bag leak detection and particulate matter detection system requirements under Subpart EEE. We respond that Subpart LLL specifies opacity as a standard (see § 63.1343(b)(2)), and, therefore, cement kilns subject to Subpart EEE must comply with the opacity standard when electing to comply temporarily with the requirements of Subpart LLL. We note that the opacity standard under Subpart EEE does not apply to cement kilns that are equipped with a bag leak detection system under § 63.1206(c)(8) and to sources using a particulate matter detection system under § 63.1206(c)(9). However, a cement kiln may use an opacity monitor that meets the detection limit requirements as the detector for a bag leak detection system or particulate matter detection system. See Part Four, Section VIII.A-C of the preamble.

K. Periodic DRE Testing and Limits on Minimum Combustion Chamber Temperature for Cement Kilns

Comment: Several commenters oppose the need for cement kilns that burn at locations other than the normal flame zone to demonstrate compliance with the destruction and removal efficiency (DRE) standard during each comprehensive performance test. These commenters recommend that EPA remove the requirement of § 63.1206(b)(7)(ii) for cement kilns citing that existing rule provisions (i.e., the requirements under § 63.1206(b)(5) pertaining to changes that may adversely affect compliance) are sufficient to require additional DRE testing after changes are made that may adversely affect combustion efficiency. Commenters question EPA's position that cement kilns that burn hazardous waste at locations other than the normal flame zone demonstrate a variability in DRE sufficient to justify the expense of re-testing for DRE with each performance test. Commenters point to EPA's data base that includes DRE results from over 30 tests with nearly 250 runs showing consistent DRE results, including sources burning hazardous waste at locations other than the normal flame zone, being achieved by cement kilns. The commenters note several burdens associated with DRE Start Printed Page 59500testing that do not result in improved environmental benefit including the purchase of expensive exotic virgin chemicals for performance testing, the risks to workers and contractors associated with the handling of these chemicals, and increasing the length of operation at stressful kiln operating conditions necessary to conduct DRE testing at minimum combustion chamber temperatures. Alternatively, commenters recommend that EPA revise the DRE requirements such that periodic testing is no longer required for cement kilns (that burn at locations other than the normal flame zone) after they have successfully achieved the DRE standard over multiple testing cycles (e.g., two or three) under similar testing regimes. That is, the source should only be required to demonstrate compliance with the DRE standard a maximum of two or three times until the source (that burns at locations other than the normal flame zone) modifies the system in a manner that could affect the ability of it to achieve the DRE standard.

Response: We are revising the requirements of § 63.1206(b)(7)(ii) such that cement kilns that feed hazardous waste at locations other than the normal flame zone need only demonstrate compliance with the DRE standard during three consecutive comprehensive performance tests provided that the source has successfully demonstrated compliance with the DRE standard in each test and that the design, operation, and maintenance features of each of the three tests are similar. These revisions do not affect sources that burn hazardous waste only in the normal flame zone.[221]

Prior to today's change, we required sources that feed hazardous waste in locations other than the flame zone to perform periodic DRE testing every 5 years to ensure that the DRE standard continues to be achieved over the life of the unit. See § 63.1206(b)(7)(ii). We justified this requirement because of concerns that sources that feed hazardous waste at locations other than the flame zone have a greater potential of varying DRE performance due to their hazardous waste firing practices. As we stated in the 1999 rule, we were concerned that the DRE may vary over time due to the design and operation of the hazardous waste firing system, and that those variations may not be identical or limited through operating limits set during a single DRE test (similar to what we concluded for sources that burn hazardous waste only in the normal flame zone). See 64 FR at 52850.

Commenters now question the need for subsequent DRE testing at cement kilns that feed hazardous waste at locations other than the normal flame zone once a cement kiln demonstrates compliance with the MACT DRE standard. The regulatory requirement for the destruction and removal efficiency standard has proved to be an effective method to determine appropriate process controls necessary for the combustion of hazardous waste. We are not convinced that only one DRE test is sufficient to ensure that a cement kiln that burns hazardous waste at locations other than the normal flame zone will continue to meet the DRE standard because temperatures are lower and gas residence times are shorter at the other firing locations. This is especially true given the industry trend to convert to the more thermally efficient preheater/precalciner kiln manufacturing process.[222] Precalciner kilns use a secondary firing system (i.e., flash furnace) at the base of the preheater tower to calcine the raw material feed outside the rotary kiln. This results in two separate combustion processes that must be controlled “ one in the kiln and the other in the flash furnace. The gas temperature necessary for calcining the limestone raw material in the flash furnace is lower than the temperature required making the clinker product. We conclude, therefore, that it is necessary, in spite of the concerns raised by commenters, to retain periodic DRE testing to ensure continued compliance with the DRE standard necessary for the control of nondioxin/furan organic HAP.

We also acknowledge, however, the concerns raised by the commenters. Our DRE data base of operating cement kilns includes results from approximately 25 DRE tests and nearly 200 runs.[223] All data show compliance with the DRE standard. Of these, approximately one-quarter of the data are from cement kilns that burned hazardous waste at locations other than the normal flame zone (e.g., injecting waste at midkiln in a wet process kiln), but we do not have DRE results from every operating cement kiln. Considering available DRE data and the concerns of the commenters, we believe that DRE testing during three consecutive comprehensive performance tests is sufficient to provide needed certainty about DRE performance while reducing the overall costs and toxic chemical handling concerns to the regulated source. Thus, we are revising the requirements of § 63.1206(b)(7)(ii) such that cement kilns that feed hazardous waste at locations other than the normal flame zone need only demonstrate compliance with the DRE standard during three consecutive comprehensive performance tests provided that the source has successfully demonstrated compliance with the DRE standard in each test and that the design, operation, and maintenance features of each of the three tests are similar. If a facility wishes to operate under new operating parameter limits that could be expected to affect the ability to meet the DRE standard, then the source would need to conduct another DRE test. Once the facility has conducted another three DRE tests under the new operating limits, then subsequent DRE testing would not be required. Accordingly, we are revising the requirements of § 63.1206(b)(7)(ii).

Comment: Several commenters support EPA's proposal to delete the requirement to establish an operating limit on the minimum combustion chamber temperature for dioxin/furans under § 63.1209(k)(1) for cement kilns. These commenters point to the high temperatures of approximately 2500°F required to make the clinker product. These high temperatures are fixed by the reaction kinetics and thermodynamics occurring in the burning zone and cannot be reduced below minimum values at the whim of the operator and still make a marketable product. In addition to deleting the minimum combustion chamber temperature limit for dioxin/furans, commenters also recommend, for similar reasons, that EPA delete the minimum combustion chamber temperature requirement under § 63.1209(j)(1) associated with the destruction and removal efficiency standand. Commenters note that demonstrating the minimum temperature requires operating under stressful operating conditions that can Start Printed Page 59501lead to upset conditions and potentially damage the integrity of the manufacturing equipment. Other commenters oppose, however, deletion of the minimum combustion chamber temperature limit for cement kilns. These commenters state that all combustion sources, including cement kilns, must meet a minimum combustion chamber temperature limit to control dioxin/furans and organic HAP emissions given that some cement kilns feed hazardous waste at locations other than the high temperature clinker-forming zone of the kiln.

Response: We are deleting as proposed the requirement to establish a minimum combustion chamber temperature limit for dioxin/furan under § 63.1209(k)(2) for cement kilns. See 69 FR at 21343. However, we retain the requirement for cement kilns to establish and comply with a minimum combustion chamber temperature limit for the destruction and removal efficiency standard under § 63.1209(j)(1).[224]

As discussed in the 1999 rule, nondioxin/furan organic hazardous air pollutants are controlled by the DRE standard and the carbon monoxide and hydrocarbon standards. See 64 FR at 52848-52852. This standard was not reopened in the present rulemaking. We note, however, that the DRE standard determines appropriate process controls necessary for the combustion of hazardous waste. Establishing and monitoring a minimum temperature of the combustion chamber is a principal factor in ensuring combustion efficiency and destruction of toxic organic compounds. As discussed in the previous response, we believe this is especially true given the industry trend to convert to the more thermally efficient preheater/precalciner kiln manufacturing process, which use two separate combustion processes. We conclude that it is necessary, in spite of the concerns raised by commenters, to retain the minimum combustion chamber temperature limit as related to the DRE standard to ensure that combustion efficiency within the entire kiln system is maintained for the control of nondioxin/furan organic HAP.

However, we acknowledge the difficulties that cement kiln operators face in establishing a minimum combustion chamber temperature limit, including the stressful operating conditions necessary to establish the limit. As we stated at proposal, our data indicate that limiting the gas temperature at the inlet to the particulate matter control device is a critical parameter in controlling dioxin/furan emissions in cement kilns. See 69 FR at 21344. Therefore, we believe that an operating limit on the minimum combustion chamber temperature is less important to ensure compliance with the dioxin/furan standard than to ensure compliance with the DRE standard. Thus, we remove the requirement to establish a minimum combustion chamber temperature limit for dioxin/furan under § 63.1209(k)(2) for cement kilns. This change does not affect the other operating parameter limits under § 63.1209(k) that must be established for dioxin/furans, including a limit on the gas temperature at the inlet to the particulate matter control device.

Comment: One commenter supports the use of previous minimum combustion zone temperature data, regardless of the test age, in lieu of conducting new, stressful DRE testing. That is, if a cement kiln is required to conduct future DRE tests, then the source should not have to re-establish a minimum combustion chamber temperature limit during the new test. Rather, the source should have the option to submit minimum combustion chamber temperature results in lieu of re-establishing the limit.

Response: We reject the commenter's suggestion for reasons discussed above. We believe that it is necessary to retain the link between the minimum combustion chamber temperature limit and the DRE test itself, which will ensure that the combustion efficiency of the entire system will be maintained for the control of nondioxin/furan organic HAP.

Comment: One commenter supports deletion of the minimum combustion chamber temperature requirement for dioxin/furan under § 63.1209(k)(2) for lightweight aggregate kilns.

Response: We reject the commenter's suggestion. Our data base of dioxin/furan emissions data shows substantial variability in test results at each source.[225] This may indicate that factors other than limiting kiln exit gas temperatures may be influencing significantly dioxin/furan formation in lightweight aggregate kilns. As such, we conclude that removing the minimum combustion chamber temperature limit would not be appropriate at this time due to the uncertain nature of dioxin/furan formation in lightweight aggregate kilns. Thus, we are retaining the requirement to establish a minimum combustion chamber temperature limit for dioxin/furans under § 63.1209(k)(2) and § 63.1209(j)(1) for lightweight aggregate kilns.

L. One Time Dioxin and Furan Test for Sources Not Subject to a Numerical Limit for Dioxin and Furan

Comment. Commenters support the one-time dioxin/furan test for sources not subject to a numerical dioxin and furan standard. Commenters agree that previous testing should be allowed to document the one time test.

Response. The final rule requires sources that are not subject to a standard with numerical dioxin and furan levels [226] to conduct a one-time dioxin and furan test as part of their initial comprehensive performance testing: lightweight aggregate kilns that elect to control the gas temperature at the kiln exit rather than comply with a dioxin/furan standard of 0.20 ng TEQ/dscm, solid fuel boilers, liquid fuel boilers with wet or no air pollution control systems, and HCl production furnaces. We will use these data as part of the process of addressing residual risk under CAA section 112(f) and evaluating future MACT standards under section 112(d)(6). The results may also be used as part of the RCRA omnibus permitting process.

Comment. EPA proposed that source not subject to a numerical dioxin and furan limit conduct a dioxin and furan test under worst-case conditions. Commenters state that operating under worst-case conditions is inconsistent with the CAA Section 112(f) process, which is to consider actual (i.e., normal) emissions. Commenters suggest that we require the tests be conducted under normal to above normal conditions.

Response. Section 112 (f) standards evaluate allowable emission levels, although actual emissions levels may also be considered. See 70 FR at 19998-Start Printed Page 5950219999 (April 15, 2005). Although we agree with the commenter that, in general, emissions in the range of normal to maximum are considered for section 112(f) determinations, we believe that dioxin/furan testing to provide information of use in section 112(f) residual risk determinations should be conducted under conditions where controllable operating conditions are maximized to reflect the full range of expected variability of those parameters which can be controlled. This is because dioxin/furan emissions may relate exponentially with the operating conditions that affect formation. We believe that dioxin/furan emissions relate exponentially with gas temperature at the inlet to an ESP or fabric filter,[227] and are concerned that emissions may also relate exponentially with the operating parameters (discussed below) that affect emissions from sources subject to the one-time dioxin/furan emissions test. Emissions testing under operating conditions that are in the range of “normal to above normal” may be exponentially lower than emissions under operating conditions reflecting maximum daily variability of the source. Since testing under normal operating conditions makes no effort to assess operating variability, emissions during such testing would fail to reflect expected daily maximum operating variability and so would not represent time-weighted average emissions and would under-represent health risk from chronic exposure.

Although we acknowledge that sources will not exhibit maximum operating variability each day of operation, we believe that it is important to assess the upper range of emissions that these sources may emit to properly evaluate under section 112(f) whether the MACT standards for dioxin/furan for these sources (i.e., absent a numerical emission standard) protect public health with an ample margin of safety.[228]

In addition, we note that emissions reflecting daily maximum variability would be most useful for section 112(d)(6) determinations in the future because they would represent the full range of emissions variability that results from controllable operating conditions.

For these reasons, the final rule requires sources to test under feed and operating conditions that are most likely to reflect maximized expected daily variability of dioxin/f