Regulation of Fuels and Fuel Additives: Standards for Reformulated and Conventional Gasoline

SUMMARY: Through the amended Clean Air Act of 1990, Congress mandated

that EPA promulgate new regulations requiring that gasoline sold in

certain areas be reformulated to reduce vehicle emissions of toxic and

ozone-forming compounds. This document finalizes the rules for the

certification and enforcement of reformulated gasoline and provisions

for unreformulated or conventional gasoline.

DATES: The regulations for the reformulated gasoline program are

effective on March 18, 1994. The incorporation by reference of certain

publications listed in the regulations is approved by the Director of

the Federal Register as of March 18, 1994. The information collection

requirements contained in 40 CFR part 80 have not been approved by the

Office of Management and Budget (OMB) and are not effective until OMB

has approved them. EPA will publish a document in the Federal Register

following OMB approval of the information collection requirements.

Retail sale of reformulated gasoline will begin on January 1, 1995,

as will the provisions for the ``simple model'' certification, the

anti-dumping program for conventional gasoline, and the associated

enforcement procedures. (For all ensuing sections of this document, the

program's beginning date of January 1, 1995 refers only to the retail

sale of reformulated gasoline.) Certification of reformulated gasoline

by the ``complex model'' and compliance with the Phase II performance

standards, will begin January 1, 1998 and January 1, 2000,

respectively.

ADDRESSES: Materials relevant to this FRM are contained in Public

Dockets A-92-01 and A-92-12, located at room M-1500, Waterside Mall

(ground floor), U.S. Environmental Protection Agency, 401 M Street SW.,

Washington, DC 20460. The docket may be inspected from 8 a.m. until 12

noon and from 1:30 p.m. until 3 p.m. Monday through Friday. A

reasonable fee may be charged by EPA for copying docket materials.

FOR FURTHER INFORMATION CONTACT:

Paul Machiele (reformulated gasoline requirements), U.S. EPA (RDSD-12),

Regulation Development and Support Division, 2565 Plymouth Road, Ann

Arbor, MI 48105, Telephone: (313) 668-4264.

George Lawrence (reformulated gasoline and anti-dumping enforcement

requirements), U.S. EPA (6406J), Field Operations and Support Division,

501 3rd Street, Washington, DC 20005, Telephone: (202) 233-9307.

SUPPLEMENTARY INFORMATION: Today's final rule is preceded by four

previous notices: an initial notice proposing standards for

reformulated and conventional gasoline (NPRM) published on July 9, 1991

(56 FR 31176), a supplemental notice (SNPRM) published on April 16,

1992 (57 FR 13416), an additional NPRM published on February 26, 1993

(58 FR 11722), and a notice of correction for Phase II standards

published on April 1, 1993 (58 FR 17175). Insofar as the rules

finalized today mirror the proposed standards, those previous documents

may be referred to.

Today's preamble explains the basis and purpose of the final rule,

focusing on issues that have been revised since the publication of the

correction notice for the Phase II performance standards (58 FR 17175).

Support documents, including the Regulatory Impact Analysis (RIA), are

available in Public Docket No. A-92-12.

To Request Copies of This Final Rule Contact: Delores Frank, U.S.

EPA (RDSD-12), Regulation Development and Support Division, 2565

Plymouth Road, Ann Arbor, MI 48105, Telephone: (313) 668-4295.

Copies of the preamble, the Final Regulatory Impact Analysis (RIA),

the Responses to Comments on Enforcement Provisions (RCEP), the complex

model, the simple model and the regulations for the reformulated

gasoline rulemaking are available on the OAQPS Technology Transfer

Network Bulletin Board System (TTNBBS). The TTNBBS can be accessed with

a dial-in phone line and a high-speed modem (PH# 919-541-5742). The

parity of your modem should be set to none, the data bits to 8, and the

stop bits to 1. Either a 1200, 2400, or 9600 baud modem should be used.

When first signing on, the user will be required to answer some basic

informational questions for registration purposes. After completing the

registration process, proceed through the following series of menus:

(M) OMS

(K) Rulemaking and Reporting

(3) Fuels

(9) Reformulated gasoline

A list of ZIP files will be shown, all of which are related to the

reformulated gasoline rulemaking process. The six documents mentioned

above will be in the form of a ZIP file and can be identified by the

following titles: ``PREAMBLE.ZIP'' (preamble); ``RIAFINAL.ZIP'' (RIA);

``ENFORCE.ZIP'' (RCEP); ``EPAFINAL.ZIP'' (complex model);

``MODFINAL.ZIP'' (simple model); ``REGFINAL.ZIP'' (regulations). To

download these files, type the instructions below and transfer

according to the appropriate software on your computer:

ownload, rotocol, xamine, ew, ist, or to exit: D filename.zip

You will be given a list of transfer protocols from which you must

choose one that matches with the terminal software on your own

computer. Then go into your own software and tell it to receive the

file using the same protocol. Programs and instructions for de-

archiving compressed files can be found via ystems Utilities from

the top menu, under rchivers/de-archivers.

I. Background

The purpose of the reformulated gasoline regulations is to improve

air quality by requiring that gasoline be reformulated to reduce motor

vehicle emissions of toxic and tropospheric ozone-forming compounds, as

prescribed by section 211(k)(1) of the Clean Air Act (CAA or the Act),

as amended. This section of the Act mandates that reformulated gasoline

be sold in the nine largest metropolitan areas with the most severe

summertime ozone levels and other ozone nonattainment areas that opt

into the program. It also prohibits conventional gasoline sold in the

rest of the country from becoming any more polluting than it was in

1990. This requirement ensures that refiners do not ``dump'' fuel

components that are restricted in reformulated gasoline and that cause

environmentally harmful emissions into conventional gasoline.

Section 211(k)(l) directs EPA to issue regulations that, beginning

in 1995, ``require the greatest reduction in emissions of ozone-forming

and toxic air pollutants (``toxics'') achievable through the

reformulation of conventional gasoline, taking into consideration the

cost of achieving such emission reductions, any non air-quality and

other air-quality related health and environmental impacts and energy

requirements.'' The Act mandates certain requirements for the

reformulated gasoline program. Section 211(k)(3) specifies that the

minimum requirement for reductions of volatile organic compounds (VOC)

and toxics for 1995 through 1999, or Phase I of the reformulated

gasoline program, must require the more stringent of either a formula

fuel or an emission reductions performance standard, measured on a mass

basis, equal to 15 percent of baseline emissions. Baseline emissions

are the emissions of 1990 model year vehicles operated on a specified

baseline gasoline. CAA compositional specifications for reformulated

gasoline include a 2.0 weight percent oxygen minimum and a 1.0 volume

percent benzene maximum.

For the year 2000 and beyond, the Act specifies that the VOC and

toxics performance standards must be no less than that of the formula

fuel or a 25 percent reduction from baseline emissions, whichever is

more stringent. EPA can adjust this standard upward or downward taking

into account such factors as feasibility and cost, but in no case can

it be less than 20 percent. These are known as the Phase II

reformulated gasoline performance standards. Taken together, sections

211(k)(1) and 211(k)(3) call for the Agency to set standards that

achieve the most stringent level of control, taking into account the

specified factors, but no less stringent than those described by

section 211(k)(3).

The reader may refer to the April 16, 1992 SNPRM (57 FR 13416) and

the February 26, 1993 NPRM (58 FR 11722) described in more detail

below), the February 1993 Draft Regulatory Impact Analysis (DRIA), the

Final Regulatory Impact Analysis (RIA), and Public Dockets A-91-02 and

A-92-12 for a thorough description of the goals and regulatory

development of the reformulated and anti-dumping programs and

discussions of a number of associated technical issues.

A. Regulatory Negotiation (Reg Neg)

Shortly after passage of the Clean Air Act Amendments of 1990, EPA

entered into a regulatory negotiation with interested parties to

develop specific proposals for implementing both the reformulated

gasoline and related anti-dumping programs. These parties included

representatives of the oil and automobile industries, vehicle owners,

state air pollution control officials, oxygenate suppliers, gasoline

retailers, environmental organizations, and citizens' groups. (See the

1991 NPRM for the members of the negotiating committee and a discussion

of the process for selecting them.)

In August 1991 the committee reached consensus on a program outline

and signed an ``Agreement in Principle'' describing that consensus. EPA

agreed to propose a two-step approach to reformulated gasoline. The

first step would take effect in 1995 and utilize a ``simple model'' to

certify that a gasoline meets applicable emission reduction standards.

The simple model allows certification based on a fuel's oxygen,

benzene, heavy metal and aromatics content and Reid Vapor Pressure

(RVP).

Under the second step, according to the regulatory negotiation

agreement, EPA would propose a ``complex model'' to supplant the simple

model for certifying compliance with these standards. Certification

under the complex model would take effect 4 years after it is

promulgated. EPA also agreed to propose the more stringent Phase II

emission performance standards.

B. July 9, 1991 NPRM (56 FR 31176)

The first NPRM for the reformulated gasoline program was published

prior to the conclusion on the regulatory negotiations. Normally, in a

negotiated rulemaking, such a reg-neg committee meets to develop a

proposed rule which will be acceptable to all parties. If consensus is

reached on a proposed rule, it is published as an NPRM. The committee

members and the entities they represent agree to support the proposal

and not to seek judicial review of the final rule if it has the same

substance and effect as the consensus proposal. In this case, EPA

published an NPRM while the advisory committee was still conducting

negotiations. The Agency believed that although consensus of the

members on an acceptable rule was possible, an NPRM was required at

that time in order to meet the statutory deadline.

The 1991 NPRM described the provisions of both a program to require

the sale of gasoline which reduces emissions of toxics and ozone-

forming volatile organic compounds (VOCs) in certain nonattainment

areas and a program to prohibit the gasoline sold in the rest of the

country from becoming more polluting. The 1991 notice described the

outline of the reformulated gasoline program as required by statutory

provisions and options that the regulatory negotiation committee

members were considering. Topics included in the 1991 proposal

consisted of the derivation of the emission standards, fuel

certification by modeling, opt-in provisions, credits, anti-dumping

requirements, and enforcement provisions for all aspects of the

reformulated gasoline program.

C. April 16, 1992 SNPRM (57 FR 13416)

As noted above, the Agency's SNPRM (57 FR 13416) reflected the

agreement reached in the regulatory negotiation that had been conducted

to develop reformulated gasoline regulations under section 211(k). The

Supplemental Notice of Proposed Rulemaking (SNPRM) described the

standards and enforcement scheme for both reformulated and conventional

gasoline. It also included specific proposals for the simple emission

model to be used in gasoline certification and enforcement.

D. February 26, 1993 NPRM (58 FR 11722)

In their comments on the SNPRM, the ethanol industry expressed

concern that the reformulated gasoline rulemaking, as proposed in the

SNPRM, effectively excluded ethanol from the reformulated gasoline

market. In an attempt to address their concern, the Agency proposed an

ethanol incentive program, at the direction of former President Bush,

intended to promote the use of ethanol (and other renewable oxygenates)

in reformulated gasoline. The objective of the proposed renewable

oxygenate program was to enhance the market share for renewable

oxygenates while, theoretically, maintaining the overall environmental

benefits of the reformulated gasoline simple model. This would be

accomplished by offsetting any increase in volatility that may result

from the inclusion of ethanol with volatility reductions that occur in

the rest of the RFG pool. This volatility balancing, however would not

take into account any increase in volatility in-use due to mixing of

ethanol and non-ethanol gasoline blends (commingling). The renewable

oxygenate program would not be required in class B areas (the South)

unless a state requested inclusion in the program. Thus, the NPRM (58

FR 11722) for reformulated gasoline proposed revisions to the simple

model, as well as to the associated anti-dumping, and enforcement

provisions. Also included in the NPRM were the proposed complex model

for certification of reformulated gasoline and the proposed Phase II

performance standards. The complex model is now scheduled to take

effect January 1, 1998. The complex model will provide a method of

certification based on the fuel characteristics such as oxygen,

benzene, aromatics, RVP, sulfur, olefins and the percent of fuel

evaporated at 200 and 300 degrees Fahrenheit (E200 and E300,

respectively). The NPRM also proposed Phase II standards for

reformulated gasoline which are to take effect in the year 2000, as

prescribed by section 211(k)(3) of the Clean Air Act (CAA). The

proposed VOC performance standard was 20-32 percent for class B and 26-

35 percent for class C. EPA proposed to set the toxic standard at 20 or

25 percent reduction since additional toxics control was not found to

be cost effective and, in most cases, these greater toxics reductions

were expected to occur through fuel reformulation for VOC control. The

NPRM also included proposed NOx performance standards of 0-16

percent in classes B and C. The proposed NOx standards greater

than zero were not required by the CAAA, but were proposed under the

authority of section 211(c)(1) in conjunction with the Phase II

reformulated gasoline standards of the Act since additional NOx

control was deemed beneficial and cost effective in reducing ambient

ozone levels.

E. Discussion of Major Comments and Issues

EPA received a number of comments on the first NPRM (56 FR 31176),

the SNPRM (57 FR 13416), and the latest NPRM (58 FR 11722) for

reformulated and conventional gasoline. Comments covered a wide range

of topics including regulatory procedure, certification standards,

modeling emissions by the simple and complex models, the role of

ethanol and other oxygenates in reformulated gasoline, vehicle testing,

the anti-dumping program, Phase II standards, cost-effectiveness, and a

number of enforcement-related issues. EPA has conducted an analysis of

the comments received and duly considered the significant issues.

Summaries of these comments and EPA's responses to them are contained

in the Final Regulatory Impact Analysis and the Summary and Analysis of

Comments which has been placed in the docket for this rulemaking

(Public Docket No. A-92-12). Since the publication of the NPRM, the

Agency has continued to develop the complex model. The first revisions

of the complex emissions model since 1993 NPRM publication for

reformulated gasoline have been provided to the public at a June 2,

1993 public workshop. EPA developed several complex model options in

July which was provided to the public. In October of 1993, a draft

version of the final complex model was released for public inspection

as well. All the iterations of the complex model since the publication

of the 1993 NPRM have been available to the public via a public

electronic bulletin board and in submittals to the EPA Air Docket,

Docket No. A-92-12.

All the various components of this rulemaking are being finalized

in today's notice. The additional time has allowed adequate public

review of the complex model and its implications for the reformulated

gasoline Phase II standards.

The remainder of this preamble is organized into the following

sections:

II. Treatment of Ethanol

III. Simple Model for Reformulated Gasoline Compliance

IV. Complex Model

V. Augmenting the Models Through Testing

VI. Phase II (Post-1999) Reformulated Gasoline Performance

Standards and NOx Standards for Reformulated Gasoline

VII. Enforcement

VIII. Anti-Dumping Requirements for Conventional Gasoline

IX. Anti-Dumping Compliance and Enforcement Requirements for

Conventional Gasoline

X. Provisions for Opt-In by Other Ozone Non-Attainment Areas

XI. Federal Preemption

XII. Environmental and Economic Impacts

XIII. Public Participation

XIV. Compliance With the Regulatory Flexibility Act

XV. Statutory Authority

XVI. Administrative Designation and Regulatory Analysis

XVII. Compliance With the Paperwork Reduction Act

XVIII. Notice Regarding Registration of Reformulated Gasolines

II. Treatment of Ethanol

A. Background

The April 16, 1992 proposal of the Simple Model and Phase I

standards was designed to be fuel and oxygenate neutral. Ethanol,

however, when added to gasoline in the amount needed to satisfy the

oxygen content requirement of the Act raises the Reid vapor pressure

(RVP) of the resulting blend by about 1 psi, making it more difficult

for ethanol blends to meet the mass VOC performance standards than

blends using other oxygenates. For ethanol to be blended with the RFG,

a blendstock gasoline with an RVP low enough to offset the increase

resulting from adding ethanol would have to be obtained.

Ethanol industry representatives commented that obtaining such

blendstocks would be both difficult and expensive, because ``sub-RVP''

blendstocks would be more costly to refine and because blendstock

production would be controlled by petroleum refiners. Methyl tertiary

butyl ether (MTBE), an oxygenate which does not boost a fuel's RVP,

which is derived from methanol gas and the petroleum product

isobutylene and whose blends can readily be put through petroleum

pipelines, was thought to be the oxygenate of choice for most refiners.

Ethanol's representatives theorized that the oil industry would have a

desire to use MTBE over ethanol and, thus, little incentive to make the

sub-RVP blendstock necessary for ethanol blending. The ethanol industry

contended that a reformulated gasoline program which they argued would

effectively preclude ethanol was contrary to Congress' intent that

ethanol have a role in the program. They argued that the oxygen content

requirement of section 211(k)(2) was motivated in large part by a

desire to expand markets for ethanol. They noted the strong support

afforded the RFG legislative initiative by members of Congress from

agricultural states. They also cited statements in the legislative

history indicating some members' expectation that the RFG program would

provide an increasing market for ethanol.

Ethanol representatives contended that the benefits of ethanol use

justify its inclusion in the RFG program. Specifically, they explained

that ethanol is currently made in the United States from domestically-

grown grains, primarily corn, and thus represents an important domestic

and renewable source of energy. They further explained that to the

extent ethanol is used in place of imported petroleum products, it

promotes the nation's energy independence and improves its balance of

trade, and that ethanol use also strengthens the market for corn,

consequently reducing the need for price supports. Moreover, as a

biomass-based product, ethanol is potentially a renewable fuel to the

extent the energy derived exceeds any fossil fuel energy consumed in

producing the ethanol.

In view of ethanol's importance to the nation's energy security and

agricultural economy, ethanol representatives urged that the proposal

be revised to allow ethanol to effectively participate in the RFG

market. They suggested several possible revisions. For example, they

argued that the 1 psi waiver granted to certain ethanol blends by

section 211(h) of the CAA be applied to ethanol-blended RFG under

section 211(k). They reasoned that since Congress recognized in the

provision requiring nationwide reductions in fuel RVP that ethanol

required such a waiver, ethanol should receive a similar waiver if the

VOC performance standard for RFG sold in the smoggiest cities were

defined in terms of a required reduction in RVP.

If the section 211(h) waiver were not available to RFG ethanol

blends, the ethanol industry suggested that the VOC reduction

requirement take into account that specific VOCs from various

reformulated gasolines differ in their ozone formation potential. While

ethanol raises a fuel's volatility and thus its VOC emissions, they

argued that the resulting VOCs are less ozone-forming than those that

would otherwise occur. They urged that the 15 percent reduction

requirement should thus be interpreted to require a 15 percent

reduction in ozone-forming potential, not simply mass of ozone-forming

VOCs. Ethanol supporters suggested additional ways of encouraging or

even requiring ethanol use in RFG. The Governors Ethanol Coalition, for

instance, suggested that EPA require the RFG market to satisfy its

oxygenate requirements through a minimum percentage of domestically

produced renewable fuel.

Based on ethanol's importance to the nation's energy and

agricultural policy, President Bush on October 1, 1992 announced a plan

to allow ethanol to effectively compete in the RFG program, with the

expectation that, with barriers removed, ethanol use would grow. In

lieu of an RVP waiver, or inclusion of ozone reactivity this plan was

based upon provisions of section 211(k)(1) allowing the Administrator

to take into consideration cost, energy requirements, and other

specified factors in setting RFG performance standards. The most

significant part of this plan called for EPA to ``establish rules for

reformulated gasoline in all northern cities that will have the effect

of granting a one-pound waiver for the first 30 percent market share of

ethanol blends, while achieving environmental benefits comparable to

those provided for in EPA's proposed rule and regulatory negotiation.''

The environmental benefits of the proposed RFG program would be

maintained by offsetting any increase in volatility of RFG containing

ethanol with reductions in the volatility of the rest of the

reformulated gasoline pool. In response to the announcement by former

President Bush, EPA proposed on February 26, 1993 provisions to provide

an RVP (and VOC) incentive for the use in reformulated gasoline of

renewable oxygenates such as ethanol.

B. Concerns With the Proposal

At the time of the February 26, 1993 proposal, EPA had a number of

concerns with respect to its legality, energy benefits, and

environmental neutrality. Nevertheless, we proposed the provisions for

public comment in the hope that these concerns could be overcome based

on new data and information developed in-house or received through

public comment. Since the time of the proposal these concerns have been

enhanced. Additional data and information has been developed which

indicates that energy benefits would be unlikely to occur as a result

of the proposal. While the production of much of the ethanol in the

country produces on the margin more energy and uses less petroleum than

went into its production, a recent study by the Department of Energy

(refer to DOE's comments on the proposal) indicates that the margin

disappears when ethanol is mixed with gasoline. The energy loss and

additional petroleum consumption necessary to reduce the volatility of

the blend to offset the volatility increase caused by the ethanol

causes the energy balance and petroleum balance to go negative. Since

the potential energy benefits were the basis in the proposal for

providing the incentives for renewable oxygenates, the justification

for the proposal no longer exists.

Additional data and information has also been developed which

indicates that VOC emissions would increase significantly under the

proposal. As discussed in section I of the RIA, the commingling effect

of mixing ethanol blends with non-ethanol blends in consumer's fuel

tanks, the effect of ethanol on the distillation curve of the blend,

and unrestricted early use of the complex model combined result in

roughly a 6-7.5% increase in gasoline vehicle VOC emissions even though

there is no increase in the average RVP of in-use gasoline. As a

result, the proposal would have sacrificed 40 to 50 percent of the VOC

control that is required under section 211(k) for reformulated gasoline

in exchange for incentives for what is likely to have been only a

marginal increase in the market share of ethanol in reformulated

gasoline and no energy benefits or cost savings.

As discussed in section I of the RIA, ethanol is not excluded from

competing in the reformulated gasoline market under the provisions of

the April 16, 1992 SNPRM. As a result of the economic advantage of

ethanol over other oxygenates, ethanol should maintain a significant

market share under the reformulated gasoline program even without the

renewable oxygenate incentives proposed in the February 16, 1993

proposal. As a result, the actual ethanol market share increase as a

result of the renewable oxygenate provisions would be expected to be

far less than the maximum of 30% for which incentives were provided.

Given the relatively small increase in ethanol demand as a result of

the renewable oxygenate provisions in exchange for such a large loss in

the environmental control of the reformulated gasoline program, there

does not appear to be any justification for promulgating these

provisions.

Furthermore, comments were received from virtually all parties,

including ethanol industry representatives, that the proposal was

unworkable and would significantly increase the cost of the

reformulated gasoline program. While EPA maintains that the program

would have provided an economic incentive for the use of renewable

oxygenates in reformulated gasoline up to a 30% market share, EPA

acknowledges that the proposal would have intruded into the efficient

operation of the marketplace, impacting the cost of the reformulated

gasoline program. As a result, after taking into account the cost, non-

air quality and environmental impacts, and energy impacts, EPA has

found itself with no choice but to back away from the renewable

oxygenate provisions of the February 26, 1993 proposal.

C. Provisions for the Final Rule

In lieu of the renewable oxygenate proposal, EPA investigated a

number of options aimed at making the program more workable by reducing

the fuel tracking, recordkeeping, and enforcement burden associated

with the proposal. While such options tended to make the program more

workable from the standpoint of the refining and fuel distribution

processes, they also tended to either reduce the assurance that the

environmental benefits of the program would be achieved in all areas

covered by the RFG program, or to place additional restrictions on the

flexibility contained in the proposal for blending ethanol into

gasoline. Given this and the other concerns with the proposal (cost,

lack of energy benefits, significant environmental loss, etc.), EPA did

not believe these options to be appropriate or justifiable either under

the provisions of section 211(k) of the Act. The reader is referred to

the Final Regulatory Impact Analysis for a detailed discussion of the

renewable oxygenate program.

A number of commenters suggested alternative provisions (1.0 psi

RVP waiver for ethanol blends, inclusion of ozone reactivity in the

standard setting process, mandates for refiners to provide clear

gasoline blendstock for downstream blending with ethanol, etc.) to the

proposed renewable oxygenate program to allow ethanol to play a larger

role in the reformulated gasoline program. It was argued that without

such provisions ethanol would be excluded from the market entirely in

direct conflict with the intent of Congress in the CAA.

EPA, however, does not agree that ethanol is excluded from

competing in the reformulated gasoline marketplace under the provisions

of the April 16, 1992 proposal. In fact, as under the recently

implemented wintertime oxygenated fuels program, ethanol is expected to

significantly increase its market share under the reformulated gasoline

program, especially in Midwestern areas where ethanol enjoys State tax

incentives and relatively low distribution costs. In addition, not only

is ethanol expected to compete as an alcohol, but it also may compete

with methanol as an ether feedstock in the future. As a result, EPA

believes that the treatment of ethanol blends under the April 16, 1992

proposal is entirely consistent with the intent of Congress as

expressed in section 211(k) of the CAA.

The alternative provisions (1.0 psi RVP waiver for ethanol blends,

inclusion of ozone reactivity in the standard setting process, mandates

for refiners to provide clear gasoline blendstock for downstream

blending with ethanol, etc.) suggested by various commenters to further

enhance the competitiveness of ethanol in the reformulated gasoline

program are not appropriate. These provisions are both outside of EPA's

legal authority under the CAA, and indefensible from an environmental

and scientific standpoint. The 1.0 psi waiver for example, could easily

forfeit all VOC emission reductions otherwise achieved by the

reformulated gasoline program. A move away from the mass based

standards of the Act to reactivity based standards is not only

unsupportable on the basis of the available scientific information, but

even if EPA were able to do so, it would be unlikely to provide any

significant advantage for ethanol blends. As discussed in section I of

the RIA, the recent urban airshed modeling studies claiming that

ethanol blends with a 1.0 psi waiver do not increase ozone relative to

an MTBE blended reformulated gasoline are frought with invalid

assumptions and inconsistencies and are not applicable to the

reformulated gasoline situation. As a result, they provide no credible

scientific support for special provisions for ethanol in the context of

the reformulated gasoline program.

Given the lack of justification for the renewable oxygenate

provisions of the February 26, 1993 proposal, the options considered

for simplifying that proposal, and other alternative provisions

recommended by commenters, EPA is, thus, basing the oxygenate-related

provisions of the final rule on the provisions as proposed in the April

16, 1992 proposal. Despite this decision, EPA still believes ethanol

will be able to compete favorably in the reformulated gasoline market

either as a direct additive or as an ether feedstock as discussed

above. As such, EPA believes that the nationwide production of ethanol

will increase as a result of this rulemaking with corresponding

benefits to our Nation's agricultural sector. However, the increase may

not be as large as it otherwise would have been had an incentive

program been promulgated for ethanol. The reader is referred to section

I. of the RIA for additional description of the comments and

information which led up to this decision.

III. Simple Model for Reformulated Gasoline Compliance

In accordance with section 211(k) of the Clean Air Act, EPA

requires that in order for a gasoline to be certified as reformulated,

it must contain at least 2.0 weight percent oxygen, no more than 1.0

volume percent benzene, and no heavy metals (unless a waiver is

granted); result in no increase in NOX emissions; and achieve

required toxics and VOC emission reductions. The VOC, NOX, and

toxics emission requirements effective between January 1, 1995 and

December 31, 1997 and EPA's derivation of them are set forth below.

Two methods by which refiners can certify their fuel as meeting the

VOC, NOX, and toxics requirements of reformulated gasoline are

contained in this rulemaking. The first, by use of a ``Simple Model,''

is described in this section. A second method, the use of the ``Complex

Model'' is described in Section IV. Provisions for augmenting the

Complex Model through vehicle testing are described in Section V. For

reasons set forth in the April 16, 1992 SNPRM (57 FR 13417-13418) and

discussed Section V, vehicle testing is not an option as a separate,

stand-alone method of certification. First, models can better reflect

in-use emission effects since they can be based on the results of

multiple test programs. Second, individual test programs may be biased,

either intentionally or unintentionally. Third, fuel compositions tend

to vary due in part to factors beyond the control of fuel suppliers,

potentially requiring testing of each batch if a model is not used.

Finally, models make more efficient use of scarce and expensive

emissions effects data than is otherwise possible. For these reasons,

EPA believes that the modeling options promulgated by EPA are necessary

for the reformulated gasoline program to achieve its environmental

objectives and to minimize the costs of the program. Comments were

received suggesting that EPA allow certification based on testing as an

optional means of certification. However, for the same reasons

discussed above, EPA does not believe such an option would be

appropriate. EPA would have much less certainty that the results of the

test program were valid.

At the time of the simple model proposal, while a number of fuel

parameters were thought to impact emissions, data were sufficient for

only a few of these parameters (Reid vapor pressure, fuel oxygen,

benzene, and aromatics) to quantify their effect with reasonable

accuracy for use in an emissions model. For those additional parameters

which were thought to impact emissions in a directionally clear, but as

of yet unquantifiable manner (sulfur, T90, and olefins), EPA proposed

that they be capped at the refiner's 1990 average level to prevent

emission effects from changes in their levels from undercutting the

emission reductions achieved by the parameters contained in the simple

model. The effect of aromatics on VOC and NOX emissions was also

unclear, but instead of being capped, it was believed that the level of

aromatics would be controlled by the role aromatics plays in the

formation of air toxics emissions.

Data is now available to accurately quantify not only the effects

of RVP, oxygen, benzene, and aromatics on emissions, but also sulfur,

T90 (or E300), olefins, and T50 (or E200). The effects of these fuel

parameters are incorporated into the Complex Model described in Section

IV.

The Complex Model is the most accurate and complete model currently

available for use in the reformulated gasoline program. Absent any

other considerations, EPA would require use of the Complex Model for

purposes of certification. However, based on leadtime considerations,

EPA is allowing use of either the Simple or Complex Model during the

first three years of the reformulated gasoline program as proposed.

These lead time considerations were described in the April 1992

proposal (57 FR 13417-8). EPA is providing four years leadtime before

use of the Complex Model is mandatory to allow the regulated industry

adequate time to plan and design necessary refinery modifications,

obtain necessary permits and capital, complete construction, and

complete start-up and equipment shakedown. Furthermore, EPA has every

confidence that on average the refiners certifying their fuel using the

Simple Model will achieve the emission reductions that Congress

intended for the reformulated gasoline program.

Various comments were received criticizing the use of the Simple

Model for fuel certification, stating that it had limited flexibility,

discouraged innovation, penalized refiners producing cleaner than

average gasoline in 1990, and should be scrapped. Many of these

comments would appear to be resolved by the option available for early

use of the Complex Model. Therefore, in keeping with the need to

provide adequate lead time and the fact that compliance with the Simple

Model will produce the mandatory VOC and toxic emission reductions,

refiners will be permitted to use the simple model for certification

until December 31, 1997. Until this date, fuel suppliers will have the

option of using the complex model instead of the simple model to take

advantage of the effects of parameters contained in the complex model

but not contained in the simple model (as described in the following

paragraphs). The reader is referred to the April 16, 1992 SNPRM for

more discussion of these lead time provisions.

A. Simple VOC Emissions Model

The simple model for VOC emissions is comprised of fuel

specifications for RVP and oxygen. Fuels sold at retail outlets must

have an RVP during the high ozone season (June 1 through September 15)

of no more than 7.2 psi in VOC control region 1 (the southern areas

typically covered by ASTM class B during the summer) and 8.1 psi in VOC

control region 2 (the northern areas typically covered by ASTM class C

during the summer).1 The differences in climate between these two

types of areas requires a corresponding difference in gasoline

volatility to achieve the same emissions effect. The period of June 1

through September 15 was chosen for the high ozone season because most

of the ozone violations occur during this period. (See 56 FR 24242 for

a discussion of the determination of this period.)

---------------------------------------------------------------------------

\1\Lower RVP limits apply for fuels that comply under averaging.

RVP controls also apply from May 1 to May 31 for facilities upstream

of retail outlets. These issues are discussed elsewhere in this

proposal.

---------------------------------------------------------------------------

Section 211(k)(3) of the Act requires that at a minimum

reformulated gasoline comply with the more stringent of either a 15%

reduction in VOC emissions or a formula fuel described in that section,

whichever is greater. EPA has determined that the formula fuel would

achieve less than a 15% reduction in VOC. As such, the minimum VOC

emission reduction required by the Act is 15%. As discussed in section

IV, EPA believes that the VOC emission reduction in VOC control region

2 from a fuel with an RVP of 8.1 psi and 2.0 weight percent oxygen will

be sufficient to achieve the minimum 15% VOC emission reduction

relative to the Clean Air Act baseline gasoline (which has an RVP of

8.7 psi). In VOC control region 1, an 8.1 psi RVP fuel with 2.0 percent

oxygen (which would meet the minimum 15% reduction requirement relative

to the CAA baseline fuel) would actually have greater emissions than a

fuel meeting EPA's Phase II RVP control standards for VOC control

region 1 (maximum RVP of 7.8 psi). EPA believes that when Congress

designated cities for inclusion in the reformulated gasoline program

that it intended the program to provide emissions reductions in

addition to those provided by the Phase II RVP requirements. If EPA

merely required reformulated gasoline in VOC control region 1 to meet

the RVP requirement for VOC control region 2, then no reduction in VOC

emissions would accrue under the first phase of the reformulated

gasoline program beyond those mandated by Phase II RVP standards. EPA

projects that relative to Phase II RVP control levels, a fuel with 7.2

psi RVP and 2.0 weight percent oxygen would provide VOC emission

reductions in VOC control region 1 similar to those obtained in VOC

control region 2.

While requiring reformulated gasoline sold in VOC control region 1

to have an RVP of no more than 7.2 psi goes beyond the minimum

requirement stated in section 211(k)(3), section 211(k)(1) authorizes

EPA to require emission reductions in VOC control region 1 of this

magnitude because they are achievable considering costs, other air

quality and non-air quality impacts, and the energy implications of

such a requirement.

Similarly, EPA believes that additional VOC reductions are

obtainable if refiners are allowed to meet the RVP and oxygen standards

through averaging. If refiners wish to take advantage of averaging, EPA

thus will require their average RVP for both VOC control regions 1 and

2 to be reduced by 0.1 psi to 7.1 and 8.0 psi, respectively, and the

average oxygen concentration to be increased to 2.1 weight percent

oxygen. For additional discussion of the rationale for the more

stringent standard in VOC control region 1 and the increase in

stringency of the averaging standards, the reader is referred to the

April 16, 1992 SNPRM.

B. Simple NOx Emissions Model

The Clean Air Act requires that there be no NOX emissions

increase from reformulated fuels. Based on data available during the

regulatory negotiations and at the time of the April 16, 1992 proposal,

it appeared that fuel oxygen content and the type of oxygenate used may

have an impact on NOX emissions while no other simple model

parameter appeared to have such an impact. Due to the statutory

requirement for oxygenate use, and the lack of any other parameters in

the simple model by which refiners could offset any NOX increase,

EPA needed to place restrictions on the amount of oxygen that could be

added to the fuel in order to prevent NOX emission increases. EPA

proposed on the basis of the data then available that MTBE blends

containing up to 2.7 weight percent (wt%) oxygen and other blends

containing up to 2.1 wt% oxygen would be presumed to result in no

NOX increase. Greater oxygenate concentrations could not be

permitted due to the risk of NOX emission increases.

When additional data became available, however, there did not

appear to be any significant difference between the NOX emission

effects of oxygen from different oxygenates. Furthermore, it appeared

that reducing the concentration of a number of additional fuel

parameters (aromatics, olefins, sulfur, etc) could reduce NOX

emissions. Since these fuel parameters all tend to be reduced to

varying degrees when oxygenates are added to gasoline, EPA proposed in

its February 26, 1993 proposal that all oxygenates be assumed to result

in no NOX emission increase under the simple model up to 2.7 wt%

oxygen.

Under the final Complex Model discussed in Section IV, oxygen has

been found to result in no NOX increase, in fact, it results in a

very slight decrease. However, the other changes that occur to the fuel

when oxygenates are added both increase and decrease NOX emissions

(increases in E200 increase NOX emissions while reductions in

sulfur, olefins, aromatics, and increases in E300 reduce NOX

emissions). Typically the effect of these other fuel changes will be to

further reduce NOX emissions. However, there is no control placed

on E200 levels under the simple model, and the levels of sulfur,

olefins, an E300 are only constrained to the refiner's 1990 baseline

levels (aromatics is controlled indirectly to some degree by the toxics

requirement). As a result, there is no assurance under the simple model

that oxygenate addition will not increase NOX emissions. The more

oxygenate added, the greater the increase in E200, and the greater the

possibility for a NOX increase. For this reason EPA believes it is

still appropriate to cap the maximum oxygen content under the Simple

Model at 2.7 wt%. Any higher oxygen concentrations will require use of

the complex model.

However, for a number of reasons, EPA believes it is appropriate

for any oxygenate up to 3.5 weight percent oxygen to be presumed to

result in no NOX emission increase under the simple model during

those months without ozone violations (e.g., winter months) unless a

state requests that oxygenate levels be limited to the 2.7 wt% oxygen

level applicable during those months with ozone violations. First,

although there are a number of concerns associated with NOX

emissions, the main concern of focus in this rulemaking is ozone which

is for the most part a summertime problem. Second, while there is no

assurance that individual batches of gasoline containing more than 2.7

wt% oxygen will not increase NOX emissions, the increase, if any,

would be small (i.e., likely less than 1 percent). Third, on average

across all fuel produced by all refiners in an area, a NOX

reduction may still occur. Fourth, there are benefits to the use of

oxygenates during the winter months (lower CO and air toxics emissions)

that may be more important to individual states than the certainty that

no one batch of fuel increases NOX emissions relative to the 1990

baseline.

A state may make a request for the 2.7 wt% oxygen limit to apply

during the non-ozone season when it believes that the use of higher

oxygenate levels would interfere with attainment or maintenance of

another ambient air quality standard (other than ozone) or another air

quality problem. This proposal parallels the Regulatory Negotiation

Agreement of August 16, 1991 and EPA's letter to the Renewable Fuels

Association dated August 14, 1991.

C. Simple Toxics Emissions Model

Under section 211(k)(3), EPA must at a minimum require the more

stringent of either a specified formula fuel or a 15 percent reduction

in toxics emissions from that of baseline gasoline. All five of the

toxic air pollutants that section 211(k)(10) of the Act specifies for

control through reformulated gasoline (benzene, 1,3-butadiene,

polycyclic organic matter (POM), formaldehyde, and acetaldehyde) also

fall under the category of VOCs. Exhaust emissions include unburned

benzene and benzene formed from other aromatics during the combustion

process. Benzene, an aromatic compound, is a natural component of

gasoline and, as such, is present in evaporative, running loss and

refueling emissions (nonexhaust emissions). However, nonexhaust VOC and

benzene emissions data are only available in sufficient quantities

under high ozone test conditions. Therefore, nonexhaust benzene

emissions are not considered outside of the high ozone season. The four

other toxic air pollutants subject to control by reformulated gasoline

are not present in gasoline and hence are solely products of

combustion.

The equations that represent the simple model for air toxics

emissions are shown in section 80.42 of the regulations. The derivation

and referenced work is given in the regulatory impact analysis.

Only minor changes were made to the proposed simple toxics model.

One change excluded ethane from the exhaust VOC baseline emissions as

discussed below in Section III.D.3. The weight fractions of the various

toxics as a function of VOC have also been adjusted accordingly,

resulting in no net change in predicted toxics performance for a

particular fuel. At the request of commenters, EPA has also included

the oxygenates tertiary amyl methal ether (TAME) and ethyl tertiary

amyl ether (ETAE) as well as provisions for other oxygenates and mixed

oxygenates. Due to their similar chemical makeup, methyl ethers (such

as TAME) and ethyl ethers (such as ETAE) are to be modeled using the

same equations as for MTBE and as for ETBE, respectively. Higher

alcohols will be modeled using the same equations as for ethanol.

Higher ethers will be modeled as ETBE for all air toxics, since ETBE

was the highest ether for which toxics data were available.

D. Baseline Determination

Where the performance standard is more stringent than the formula,

the Act requires EPA to promulgate standards for the performance of

reformulated gasoline that are relative to emission levels from

baseline vehicles using baseline fuel. In order to determine whether

fuels meet the performance requirements of reformulated gasoline under

the simple model, EPA must therefore establish the baseline to which

the emission performance of reformulated fuels are to be compared. The

following discussion describes how EPA derived the emission baselines.

1. Control Periods

Before the emission baselines can be determined, the time frame

over which fuel performance will be evaluated must be identified.

Section 211(k) of the Act requires control of VOC emissions during the

``high ozone season.'' For the purposes of this rulemaking, the high

ozone season is defined to be June 1 through September 15. This period

covers the vast majority of days during which the national ambient air

quality standard for ozone is exceeded nationwide and is consistent

with the period covered by EPA's gasoline volatility control

requirements. All gasoline at service stations must thus comply with

the reformulated gasoline requirements during this period. Also in

keeping with the gasoline volatility control rulemaking the ``VOC

control Period'' for compliance with the reformulated gasoline

provisions upstream from the service station (necessary to ensure

complying fuel is available at the service stations during the high

ozone season) is May 1 through September 15.

2. Baseline Gasoline

The fuels to be used in determining baseline emissions are

unchanged from the February 26, 1993 proposal and are shown below.

Table III-1.--Baseline Fuel Compositions

------------------------------------------------------------------------

Summer Winter

------------------------------------------------------------------------

Sulfur, ppm....................................... 339 338

Benzene, volume percent........................... 1.53 1.64

RVP, psi.......................................... 8.7 11.5

Octane, R+M/2..................................... 87.3 88.2

T10, degrees F.................................... 128 112

T50, degrees F.................................... 218 200

T90, degrees F.................................... 330 333

Aromatics, volume percent......................... 32.0 26.4

Olefins, volume percent........................... 9.2 11.9

Saturates, volume percent......................... 58.8 61.7

------------------------------------------------------------------------

3. Definition of Ozone-Forming VOC

The Act requires reductions in emissions of ozone-forming VOCs.

This interpretation is consistent with the focus of Section 211(k) on

the areas with the most extreme ozone pollution problem. EPA proposed

in April 16, 1992 that methane would be excluded from the definition of

VOC on the basis of its low reactivity in keeping with past EPA

actions, but included all other VOCs including ethane. EPA further

proposed, however, that should the Agency modify the definition of VOC,

we might do so for the reformulated gasoline rulemaking as well. As

discussed in the February 26, 1993 proposal, EPA has also modified the

definition of VOC to exclude ethane in a separate Agency rulemaking (57

FR 3941). As a result, the performance of fuels meeting the VOC

emission requirements under the simple model are expressed on a non-

methane, non-ethane basis. This change resulted in slight changes to

the simple model equations previously proposed, but the overall results

of the simple model are essentially unaffected.

4. Simple Model Baseline

The following table shows the baseline emissions under the simple

model which result from the assumptions discussed above. Since the

MOBILE model does not estimate toxics emissions, however, separate data

and information was necessary to determine their baseline emissions.

The toxics baseline was developed in essentially the same manner as

that proposed in the April 16, 1992 proposal. An explanation of this

derivation can be found in Section II of the RIA.

Table III-2.--Simple Model Baseline Emissions

------------------------------------------------------------------------

Summer

-------------------------- Winter

Region 1 Region 2

------------------------------------------------------------------------

Exhaust VOCs (g/mi).............. 0.444 0.444 0.656

Non-Exhaust VOC (g/mi)........... .856 .766 0

Total VOCs (g/mi)................ 1.30 1.21 0.656

Exhaust Benzene (mg/mi).......... 30.1 30.1 40.9

Evaporative Benzene.............. 4.3 3.8 0.0

Running Loss Benzene............. 4.9 4.5 0.0

Refueling Benzene................ 0.4 0.4 0.0

1,3-Butadiene.................... 2.5 2.5 3.6

Formaldehyde..................... 5.6 5.6 5.6

Acetaldehyde..................... 4.0 4.0 4.0

POMs............................. 1.4 1.4 1.4

--------------------------------------

Total TAPs (mg/mi)........... 53.2 52.1 55.5

------------------------------------------------------------------------

E. Phase I Performance Standards Under the Simple Model

Section 211(k)(3) directs EPA to require, at minimum, that Phase I

reformulated gasoline comply with the more stringent of two alternative

VOC and toxics emission requirements--either a performance standard of

a 15 percent reduction from baseline levels on a mass basis, or

compositional requirements specified as a formula in Section

211(k)(3)(A). The formula effectively defines a set of maximum or

minimum fuel parameter specifications. In evaluating which requirement

is more stringent, EPA is to consider VOC and toxics separately.

The stringency of the formula is best evaluated by determining the

emissions performance of the fuels that would be certifiable if EPA

were to impose the requirements of Section 211(k)(3)(A). A gasoline

would meet these requirements if it (1) had no more than 1.0 volume

percent benzene, (2) had no more than 25 volume percent aromatics, (3)

had no less than 2.0 weight percent oxygen, and (4) met the

requirements for detergent additives and lead content. The formula does

not specify or limit any additional gasoline properties, and therefore

a wide variety of fuels with very different properties would qualify as

complying with the formula. For example, the formula specifies the

weight percent oxygen but does not specify the type of oxygenate. If

EPA were to impose the requirements of Section 211(k)(3)(A), then any

approved oxygenate could be used to meet the formula's oxygen

requirement, as long as it was blended to achieve the required weight

percent oxygen. The same would be true of sulfur levels, distillation

characteristics, olefin levels, RVP levels, and so on. As long as the

formula's requirements were met, the fuel would be certifiable if EPA

were to base its certification requirements on Section 211(k)(3)(A).

To evaluate the emissions performance of the various fuels that

would comply with the formula requirements, EPA used the Phase I

complex model. Given the Phase I baseline emission levels, EPA

considers the complex model to be the most appropriate means of

evaluating emissions performance since it incorporates the Agency's

most recent, complete, and accurate knowledge of the effects of fuel

properties on VOC and toxics emissions. Since many of the fuel

parameters that are not specified for the formula affect VOC and toxics

emissions, the various possible formula fuels exhibit a wide variety of

emission performance levels as these unspecified parameters vary.

According to the Complex Model, requirements based on many possible

formula fuels would be less stringent than requirements based on the 15

percent minimum reduction requirements of Section (211)(k)(3)(B). In

addition, the lack of specificity of the formula fuel would make

establishment of an equivalent emissions performance standard

impossible, since one or more possible formula fuels would fail to meet

any specific standard.

In past proposals, EPA has evaluated the formula fuel by assigning

levels for unspecified parameters at their level in baseline gasoline,

as defined in section 211(k)(9)(B) of the Act. However, such an

interpretation would not eliminate the problems described above, since

the oxygenate type would remain unspecified. Hence the requirements of

a formula could be met by a range of fuels, each based on different

oxygenates, even if unspecified parameters were to be set to baseline

levels, and this range of fuels would exhibit a range of emission

performance levels. While the Complex Model attributes identical

effects to oxygen in different chemical forms for most pollutants, it

incorporates emission effects that depend on the type of oxygenate used

for nonexhaust benzene, acetaldehyde, and formaldehyde emissions. EPA

therefore ran the complex model for several fuels, varying the type of

oxygenate and holding other parameters not specified by the formula at

statutory baseline levels.

The VOC emission reductions from baseline levels for all such

formula fuels were less than 15 percent. EPA therefore based the VOC

emission requirements for Phase I reformulated gasoline on the 15

percent reduction minimum performance standard, since this standard is

more stringent than the requirements of the formula.

For toxics performance, EPA separately evaluated the emissions

performance of fuels that met the formula requirements and contained

statutory baseline levels of unspecified fuel properties for VOC

control regions 1 and 2, since nonexhaust benzene emissions would

differ in these two regions. EPA also evaluated such fuels with

different oxygenate types. The results are shown in Table II-3. These

results include both summer and winter effects, weighted based on the

share of vehicle miles traveled in each season.

Table II-3.--Phase I Toxics Emissions Performance of Formula Fuels

------------------------------------------------------------------------

Percent reduction from

CAAB levels

Oxygenate type -------------------------

VOC control VOC control

region 1 region 2

------------------------------------------------------------------------

ETBE.......................................... 11.82 11.65

Ethanol....................................... 13.16 13.01

MTBE.......................................... 16.33 16.15

TAME.......................................... 16.81 16.67

------------------------------------------------------------------------

The results indicate that whether a formula fuel (with unspecified

fuel parameters at statutory baseline levels) meets the 15% minimum

performance requirement of section 211(k)(3)(B) depends on the type of

oxygenate used. If EPA were to impose the formula requirements of

section 211(k)(3)(A), the results presented in Table II-3 indicate that

not all gasolines which could be certified as reformulated would

achieve at least a 15 percent reduction in toxics mass emissions, even

if unspecified fuel properties were set at statutory baseline levels.

If EPA were to require a 15 percent emissions reduction in accordance

with section 211(k)(3)(B), however, all fuels would achieve this

minimum level of reductions. EPA therefore believes that the formula

requirements of section 211(k)(3)(A) are not as stringent as the

performance standard set forth in Section 211(k)(3)(B).

The minimum performance standard for Phase II is even more

stringent than the Phase I standards. EPA has therefore determined that

the performance standard is more stringent than the formula for both

VOCs and toxics, for both Phase I and Phase II. EPA must therefore set

its Phase I requirements for both VOCs and toxics to be no less

stringent than the 15 percent emission reduction performance standard

required by section 211(k)(3)(B). EPA has considered whether it should

require greater reductions in toxics mass emissions than that required

by the 15 percent minimum performance standard. However, the Agency has

concluded that more stringent toxics requirements are not cost-

effective, as is discussed more fully in Section VI. Hence EPA has set

the Phase I toxic emission performance standard at the minimum 15

percent reduction from baseline levels required by the Act. Compliance

with this standard must be demonstrated using the appropriate emission

models throughout Phase I.

Under the authority of section 211(k)(1), EPA believes that the

greater flexibility and reduced cost afforded to gasoline refiners and

importers by an averaging program allow EPA to require a greater

reduction in toxics emissions than is required under section 211(k)(3).

As discussed in Section VII, the Agency believes it appropriate, when

the air toxics standard is met on average, that it be 1.5 percentage

points more stringent than standards met on a per-gallon basis. EPA

estimates that the approximate 1.5 percentage point margin will be

sufficient to recoup any compliance margin refiners would have

otherwise had to maintain to ensure achievement of the toxics

requirements in the absence of an averaging program. In sum, the

tighter averaged standard should have the potential to increase the

environmental benefits of the reformulated gasoline program while not

increasing the cost of obtaining those benefits. As a result, the air

toxics performance standard when met on an annual average basis is set

at a 16.5% reduction from baseline levels.

F. Applicability (1995-7)

The Simple Model described in this section is effective beginning

January 1, 1995 with the beginning of the reformulated gasoline program

as a means by which fuel producers can certify that their fuel meets

the requirements for reformulated gasoline. The Complex Model described

in Section IV will not be required to be used for fuel certification

until January 1, 1998.

Until January 1, 1998, refiners who produce reformulated gasoline

will have a choice of certifying their gasoline by using either the

Simple Model or the Complex Model. EPA proposed three options for

establishing the performance standards under early, optional use of the

Complex Model. Under one option, if a refiner opts to utilize the

Complex Model before January 1, 1998 the reformulated gasoline can have

no worse VOC, NOX, or toxic emissions performance than would be

predicted by the Complex Model for a Simple-Model fuel (minimum 2.0

percent oxygen, maximum 1.0 percent benzene, and maximum RVP of 8.1 psi

in Class C areas and 7.2 psi in Class B areas) having that refiner's

average 1990 levels of sulfur, olefins, and T90 (E300). The second

option was a variation of the first, in that refiners producing

gasoline for use in only the southern reformulated gasoline areas (VOC

control region 1) could measure their fuel performance against the CAA

baseline gasoline as an alternative to their own 1990 refinery

baseline. The third option, proposed by EPA in February 1993, would

extend the second option to all reformulated gasoline areas.

The rationales for these options are discussed in detail in EPA's

proposals. Many of the comments were also received prior to the

proposals, and as such were addressed there. As a result, the reader is

referred back to the proposals for additional discussion. After

considering the comments, EPA has decided to promulgate the first

option. First, under this option each refiner will have to achieve the

same reductions, whether they use the simple model or the complex

model. The option to use either model increases refiner flexibility,

but will not change the emissions reductions required for a refiner

prior to mandatory use of the complex model in 1998. EPA believes that

the reductions required under the simple model are achievable

considering all relevant factors and will continue to be so under the

optional use of the complex model. In fact, the additional flexibility

of using the complex model would in some cases make them even more

reasonable.

Second, the other two options create an incentive for early use of

the complex model by those refiners who would then have a less

stringent performance standard than under the simple model. This would

produce on average an increase in overall emissions for reformulated

gasoline compared to average emissions if only the simple model was

allowed. Refiners with individual baselines for sulfur, T90 and olefins

that are lower than the CAA baseline would, under the second and third

options, get credit for emission benefits for these parameters, and

could use this to justify a less stringent RVP control than required

under the simple model. There would be no parallel disincentive to

early use of the complex model for refiners with higher baselines which

would result in an increase in their required reductions. This

imbalance in the expected early use of the complex model could easily

lead to an average 1-2 percentage point reduction in the average

emission performance of reformulated gasoline from 1995-7 as discussed

in section I of the RIA. Based on this negative environmental impact,

and the reasonableness of the complex model performance standard under

the first option, EPA has decided to promulgate the first option

described above for early use of the complex model.

G. Enforcement of the Early Use Option

Additional controls over reformulated gasoline certified using the

``early-use'' complex model are necessary for the operation of the

downstream enforcement mechanisms of VOC and NOX emissions

performance minimums, and covered area gasoline quality surveys. These

restrictions are necessary because under the restricted early-use

approach being promulgated, VOC, toxics, and NOX percentage

reductions are calculated from a baseline fuel using the refiner's 1990

baseline levels of sulfur, T-90, and olefins. As a result, the

reformulated gasolines produced by different refiners (or in some

cases, at different refineries) under this option will likely each meet

different percentage reduction standards for VOC, toxics, and NOX.

Therefore, the performance of a fungible mixture of complex model

gasolines produced by different refiners at different refineries could

not be predicted, nor could be evaluated.\2\

---------------------------------------------------------------------------

\2\Beginning in 1998, certification of reformulated gasoline

using the simple model will no longer be an option, and all

reformulated gasoline will be certified using the complex model.

Also beginning in 1998, all refiners and importers will calculate

emissions performance reductions from Clean Air Act average

gasoline; individual refiner baselines will not be relevant to

reformulated gasoline. As a result, the difficulties with downstream

enforcement and surveys will be resolved.

---------------------------------------------------------------------------

In order for the per-gallon minimums for VOC and NOX emissions

performance to be monitored by downstream regulated parties and

enforced by EPA, the baseline for a given gasoline sample must be

known. Without knowledge of the baseline, it is not possible to

determine whether the fuel complies with the per-gallon minimums, since

it will be different for each refinery. Similarly, in order for the

gasoline quality surveys to function under early use of the complex

model, the baseline from which to determine the emission performance

for VOC, toxics, and NOX must be known. Without knowledge of the

baseline, it is not possible to determine whether the complex model

fuels in an area on average meet the per-gallon standards.

EPA received comments from two industry groups representing the

refining industry on this issue. Both commenters stated that EPA should

require that ``early-use'' complex model gasolines subject to different

baselines be segregated through the gasoline distribution system. EPA

is adopting this suggested approach as the best (and perhaps only)

means of accommodating both the restricted early-use option and

downstream enforcement of per-gallon minimums and gasoline quality

surveys.

Under this approach, gasoline sampled at any point in the

distribution system would have known values for VOC, toxics, and

NOX emissions performance that meet the per-gallon and minimum

standards. Today's rule requires that these values must be included in

the product transfer documents for ``early-use'' complex model

gasoline, to inform downstream parties and EPA of the relevant per-

gallon and minimum values.

Today's rule prohibits the commingling throughout the distribution

system, including at retail outlets, of ``early-use'' complex model

gasoline that is subject to different baselines. One commenter stated

that the segregation of this gasoline should be through the terminal

level only. EPA disagrees with this comment because segregation through

the retail level also is necessary in order for gasoline quality

surveys to function. Survey samples are taken at retail outlets, and

the survey requires that the relevant per-gallon values for VOC,

toxics, and NOX emissions performance must be known for each

sample.

EPA realizes that restrictions on commingling of ``early-use''

complex model gasolines constitutes a significant constraint on the use

of this option, because most gasoline used in the United States is

transported as a fungible commodity. As a result, EPA anticipates that

before 1998 the complex model will be used only in limited situations.

This might occur where a refiner has a gasoline transportation system

that is dedicated from the refinery through the retail level, or where

the cost advantages of using the complex model are sufficiently large

to offset the difficulties of segregation. In spite of these

constraints, EPA sees no alternative to requiring segregation controls

over ``early-use'' complex model gasoline.

IV. Complex Model

The complex model described in this section has undergone

significant changes since it was first proposed in the February 1993

NPRM. These changes have been made in response to three key factors:

EPA's improved understanding of the relationship between fuel

characteristics and emissions, EPA's use of more appropriate data

analysis methods, and comments received in response to the February

NPRM, a public workshop held on May 25, 1993, and EPA's July 14, 1993

docket submission that described a number of alternative complex

models. The key elements in the complex model being promulgated today

are discussed in this section. This discussion also addresses the major

substantive comments received by EPA regarding the complex model. A

more detailed description of the model and its derivation, including a

detailed summary and analysis of comments, can be found in Section IV

of the RIA.

Baseline Emissions

As discussed in Section III, EPA is using a July 11, 1991 version

of MOBILE4.1 to estimate baseline emissions from light-duty vehicles

for the simple model, assuming a basic inspection and maintenance

program. This baseline was developed in the regulatory negotiation and

was at the time the best estimate of the in-use emission performance of

1990 vehicles from which to ensure that the minimum performance

standards required by section 211(k) of the Clean Air Act would be

achieved.

Since that time the Agency has developed a new version of the

MOBILE model, MOBILE5a, for use by the states in demonstrating

compliance with the national ambient air quality standard for ozone. As

proposed in the February 26, 1993 proposal, EPA will use MOBILE5a in

conjunction with an enhanced I/M program to establish the emission

baseline for Phase II of the reformulated gasoline program beginning in

the year 2000. EPA, however, has decided to retain the MOBILE4.1 and

basic I/M baseline assumption for the simple model during Phase I of

the RFG program. Switching to a MOBILE5a baseline for Phase I would

have required reformulated fuels to meet a slightly more stringent RVP

standard to maintain the minimum VOC emissions performance required by

the Act. The majority of the VOC emission reductions achieved by RFG

are from nonexhaust emissions; under MOBILE5a, nonexhaust VOC emission

reductions are less effective in reducing overall VOC emissions than

are exhaust VOC reductions, while the opposite is true under MOBILE4.1.

Thus, in order to provide refiners with sufficient leadtime to complete

the investments needed to meet the requirements of the program, the

baseline for the Simple Model is determined using MOBILE4.1.

When replacement of the Simple Model with the Complex Model is

required in 1998, the issue again arises as to whether a more stringent

standard should be required by shifting to use of MOBILE5a in

determining the baseline. MOBILE5a clearly provides a more recent

estimate of the mobile source VOC inventory than does MOBILE4.1.

However, many of the changes made in MOBILE5a were intended to

significantly increase the accuracy of the exhaust emission estimates

while similar changes which would have increased the accuracy of the

nonexhaust VOC emission estimate were not incorporated for various

reasons, including the limited time available to revise the MOBILE

model. As a result, the proportional contribution of exhaust and

nonexhaust VOC emissions to the in-use VOC inventory may not be any

more accurate in MOBILE5a than in MOBILE4.1 even though MOBILE5a

provides a more accurate assessment of the total contribution of mobile

sources to the entire VOC inventory by virtue of its greater accuracy

in estimating exhaust VOC emissions. Since it is the relative

proportions of exhaust and nonexhaust VOC emissions and not the overall

magnitude of the mobile source VOC inventory which determines how

difficult it will be for refiners to meet the overall VOC standard in

1998, it is unclear whether MOBILE5a would be more appropriate to use

in 1998 than MOBILE4.1.

A simple model fuel evaluated using the complex model achieves more

than the minimum 15% requirement of the Act using the MOBILE4.1

baseline exhaust/nonexhaust ratio but less than the 15% requirement

using the MOBILE5a baseline exhaust/nonexhaust ratio. Given the

uncertainty in the actual in-use exhaust/nonexhaust ratio during this

interim period, it is difficult to know whether or not the 15% actually

would be achieved in-use by a fuel meeting the requirements of the

Simple Model. Using MOBILE4.1 to determine the baseline in 1998 would

introduce some risk that the 15% minimum performance requirement of the

Act would not be met in-use by a fuel meeting the requirements of the

Simple Model. However, this risk is relatively small in magnitude (less

than three percentage points of emission reduction are at stake) and

duration (the risk exists for only two years). On the other hand, using

MOBILE5a to determine the 1998 baseline would result in some risk that

refiners would be required to incur greater costs to achieve a more

stringent standard than the minimum required by the Act. This greater

stringency would have the effect of creating a third interim phase to

the RFG program.

Given the uncertainty in determining whether a MOBILE4.1-based

performance standard or a MOBILE5a-based standard more accurately

reflects the in-use conditions in 1998, the potential disruption to

refinery operations (even if only for a small increase in the

stringency of the fuel reformulation requirements), the fact that a

more stringent standard in 1998 was not discussed or envisioned as part

of the regulatory negotiation process, and the fact that any risk to

the environment is small and of short duration, EPA does not believe it

to be appropriate to base the Phase I complex model standards on

MOBILE5a and require refiners to meet a more stringent performance

standard in 1998. As a result, EPA will retain MOBILE4.1 with basic I/M

as the basis for the Phase I performance standards under the Complex

Model in 1998.

In summary, EPA has retained the VOC and NOX baselines

proposed in the SNPRM, including the relevant I/M assumptions, for use

with the complex model prior to 2000. The onset of the Phase II

performance standards in 2000 will increase the overall stringency of

the standards, and a new baseline based on MOBILE5A will not, by

itself, be the cause of new investment by refiners. By this time,

enhanced I/M programs should be fully operational in nearly all

reformulated gasoline areas. Therefore, baseline VOC and NOX

emission levels to be used with the complex model in Phase II are based

on MOBILE5A's estimate of emissions from light-duty vehicles and trucks

with enhanced I/M.

Baseline estimates of toxics emissions are not available directly

from the MOBILE models. The nonexhaust toxics model bases its estimates

of nonexhaust toxics on the RVP and benzene levels of the fuel. Since

both of these levels are specified for Clean Air Act baseline (CAAB)

gasoline, EPA has used the nonexhaust toxics model to determine the

baseline nonexhaust toxics emission level. The exhaust toxics baseline

has been estimated by multiplying the exhaust toxics emission level

predicted by the complex model for CAAB gasoline by the ratio of

baseline exhaust VOC emissions to the average exhaust VOC emission

measurement in the complex model database. Since the five regulated

exhaust toxic pollutants are all classified as VOCs, this adjustment

sets the baseline exhaust toxics level equal to the exhaust toxics

levels that would have been observed if the vehicles represented by the

complex model database had VOC emission levels representative of in-use

vehicles when tested on CAAB gasoline. No comments were received

opposing this approach, which is discussed in more detail in Section

III of the RIA.

In evaluating the performance of simple model fuels, EPA has

focused its attention on the average refiner. The need to compensate

for differences between individual refinery baselines and the Clean Air

Act baseline when the use of the complex model becomes mandatory has

been communicated in past proposals, workshops, and the discussions

associated with the Agreement in Principle. Hence refiners have been

given adequate notice that if their baseline fuel produces higher

emissions than CAAB fuel, then they must offset such emissions when the

use of the complex model becomes mandatory in 1998. The four years

before use of the complex model becomes mandatory is adequate leadtime

for refiners. Refiners undertaking investments to comply with the

simple model requirements have been made aware of these requirements,

and this transition process was inherent in the regulatory negotiation

agreement and in prior proposals. EPA recognizes that the precise

emissions impact of individual refiner baselines could not be

determined with confidence until the Complex Model was promulgated.

However, refiners were aware of at least one course of action that

would satisfy the requirements of the program under the complex model,

namely to alter their baseline fuel to match the Clean Air Act baseline

prior to meeting the simple model requirements.

Baseline emissions of VOC, NOx, and toxics are given in Table

IV-1 for Phase I and in Table IV-2 for Phase II. Summer and winter

baselines are shown for both phases, with summer baseline emissions for

VOC Control Regions 1 and 2 shown separately. The toxics emission

baseline shown in Table IV-1 is applicable only during 1998 and 1999

and for those refiners choosing to use the complex model prior to 1998;

the baselines shown in Table IV-2 are applicable in 2000 and beyond.

Table IV-1.--Phase I Baseline Emissions, Milligrams/Mile

------------------------------------------------------------------------

Summer

Pollutant --------------------------------------

Region 1 Region 2 Winter

------------------------------------------------------------------------

Running loss VOC................. 430.77 390.42 0.00

Hot soak VOC..................... 264.61 229.96 0.00

Diurnal VOC...................... 125.09 108.71 0.00

Refueling VOC.................... 40.01 40.01 0.00

------------------------------------------------------------------------

Nonexhaust VOC................... 860.48 769.10 0.00

Exhaust VOC...................... 446.00 446.00 660.00

Total VOC........................ 1306.48 1215.10 660.00

NOx.............................. 660.00 660.00 750.00

Running loss benzene............. 4.92 4.46 0.00

Hot soak benzene................. 3.02 2.63 0.00

Diurnal benzene.................. 1.30 1.13 0.00

Refueling benzene................ 0.42 0.42 0.00

------------------------------------------------------------------------

Nonexhaust toxics................ 9.66 8.63 0.00

Exhaust benzene.................. 26.10 26.10 37.57

Acetaldehyde..................... 2.19 2.19 3.57

Formaldehyde..................... 4.85 4.85 7.73

1,3-butadiene.................... 4.31 4.31 7.27

POM.............................. 1.50 1.50 2.21

------------------------------------------------------------------------

Exhaust toxics................... 38.95 38.95 58.36

Total toxics................. 48.61 47.58 58.36

------------------------------------------------------------------------

Table IV-1.--Phase II Baseline Emissions, Milligrams/Mile

------------------------------------------------------------------------

Summer

Pollutant --------------------------------------

Region 1 Region 2 Winter

------------------------------------------------------------------------

Running loss VOC................. 328.53 294.15 0.00

Hot soak VOC..................... 84.11 80.97 0.00

Diurnal VOC...................... 93.34 63.62 0.00

Refueling VOC.................... 53.33 53.33 0.00

------------------------------------------------------------------------

Nonexhaust VOC................... 559.31 492.07 0.00

Exhaust VOC...................... 907.00 907.00 1341.00

Total VOC.................... 1306.48 1215.10 1341.00

NOX.............................. 1340.00 1340.00 1540.00

Running loss benzene............. 3.75 3.36 0.00

Hot soak benzene................. 0.96 0.93 0.00

Diurnal benzene.................. 0.97 0.66 0.00

Refueling benzene................ 0.56 0.56 0.00

------------------------------------------------------------------------

Nonexhaust toxics................ 6.24 5.51 0.00

Exhaust benzene.................. 53.54 53.54 77.62

Acetaldehyde..................... 4.44 4.44 7.25

Formaldehyde..................... 9.70 9.70 15.34

1,3-butadiene.................... 9.38 9.38 15.84

POM.............................. 3.04 3.04 4.50

------------------------------------------------------------------------

Exhaust toxics................... 80.10 80.10 120.55

Total toxics................. 86.34 85.61 120.55

------------------------------------------------------------------------

Exhaust Emissions Model

1. Data Sources

The relationship between fuel properties and exhaust emissions is

complex and the theory behind such relationships continues to be

developed. As a result, EPA has asked industry, state regulatory

agencies, and other organizations with relevant test data to make their

data available to the Agency to ensure that this rule is based on as

much relevant information as possible. The complex model described in

the following section is based on data generated from a number of

exhaust emissions testing programs. These programs, their design

intent, and their limitations are discussed in Section IV.A of the RIA.

Data from these programs were excluded from EPA's analysis if the data

were not based on a valid FTP measurement cycle, if the vehicle in

question did not employ 1990-equivalent emission control technology, if

the vehicles did not exhibit stable, repeatable emissions performance,

or if the data were clearly inconsistent with the bulk of the data

available to EPA (based on statistical considerations). In addition,

data from programs that did not measure nonmethane hydrocarbon

emissions were not used to develop EPA's exhaust VOC complex model. The

Agency believes its analysis considered all valid, and relevant data on

the exhaust emissions effect of fuel modifications when used in 1990

model year and equivalent vehicles that was available at the time the

model was developed.

2. Analysis Method

Exhaust emissions are affected by both vehicle and fuel

characteristics. Since the test programs described above generally

involved different vehicles, different fuels, and in some cases

different test procedures, the analysis required to determine the

relationship between fuel properties and emissions is complex. However,

EPA believes that the methods used to develop the complex model

considers and addresses these complexities appropriately. EPA utilized

statistical analysis techniques to isolate the effects of fuel

modifications on exhaust emissions of VOC, NOX, and toxics from

other factors affecting exhaust emissions.

At a series of six public workshops held over the past two years,

the Agency presented its views on data sources, analysis methods, and

preliminary emissions models for public review and comment. The Agency

also requested other organizations to share their data, analysis

expertise, and emissions models at these workshops. The methods used to

develop the model promulgated today appropriately incorporate the

comments and suggestions regarding the analysis process received at the

workshops, as well as other comments and suggestions received from

industry, state and federal government authorities, and other

interested parties during the course of this rulemaking. Information

regarding the workshops, public comments and suggestions, and EPA's

analysis methods can be found in Docket A-92-12. The approach chosen by

EPA to analyze the available data is summarized below and is discussed

more fully in Section IV.A of the RIA.

Since the vehicle and the fuel both affect exhaust emissions, EPA's

analysis separated exhaust emissions into fuel components and vehicle

components. In all test programs analyzed by EPA, the single most

significant determinant of the level of emissions from a given vehicle

on a given fuel was the vehicle itself. Fuel properties exert a much

smaller influence on exhaust emissions than do vehicle characteristics

such as emission control system technology, vehicle mileage, catalyst

efficiency, oxygen sensor efficiency, engine size, engine design,

vehicle size, fuel efficiency, vehicle maintenance, etc. To identify

the effects of fuel property modifications on emissions, EPA found it

necessary to identify the effect of each vehicle on emissions and

separate this effect from the fuel effects. For vehicles used in more

than one test program, EPA found it necessary to determine the vehicle

effect separately for each test program since vehicle effects were

observed to change between studies.

The fuel components of exhaust emissions were separated into two

main categories. The first category consisted of the effects of

individual fuel parameters. For example, the effect of sulfur on

NOX emissions was best modeled by a relationship containing a

linear sulfur term (of the form c1S, where c1 is a constant

and S is the sulfur level) and a second-order sulfur term (of the form

c2S2, where c2 is a constant). The second category of

fuel terms consisted of interactive effects between two fuel

parameters. For example, EPA's analysis found that the effect of

aromatics on hydrocarbon emissions is related to the E300 level of the

fuel. This effect cannot be represented as an aromatics or E300 effect

alone but must be represented as an interactive term of the form

c3AE, where c3 is a constant, A is the aromatics level, and E

is the E300 level.

In the February 1993 proposal, EPA indicated that it planned to

make several changes to the method used to develop the complex model.

As discussed in that proposal and in the RIA, fuels can be

characterized in terms of a number of different sets of fuel

parameters. EPA used the results of individual fuel studies and its

public workshops to select the set of fuel parameters used to model

exhaust emissions in its February 1993 proposal. At that time, the

Agency indicated that it might alter its choice of parameters to

represent gasoline distillation characteristics from a temperature

basis (using T50 and T90) to a percent evaporated basis (using E200 and

E300, the percentage of the fuel's volume that evaporates when heated

to 200 deg.F and 300 deg.F, respectively). For reasons outlined in the

February 1993 NPRM and section IV.A of the RIA, EPA has chosen to make

this change and has converted its exhaust emission models to a percent

evaporated basis since the NPRM was issued, removing the T50 and T90

terms from its models in the process. The Auto/Oil Heavy Hydrocarbon

and EPA Phase II Reformulated Gasoline Test Program studies have been

added to the complex model database. Finally, EPA has changed the

confidence level required to permit terms to remain in the model to 90

percent, in keeping with the approach used in developing the simple

model. The Agency was not able to determine the influence of the type

of aromatic compounds in fuels, specifically heavy aromatics, on

exhaust emissions, and hence such terms have not been included in the

complex model at this time.

Because vehicles can have different emission control systems, the

Agency anticipated that fuel modifications would have different

emission effects on different types of cars. To account for these

differences, EPA's February 1993 proposal divided vehicles into two

``emitter classes'' (normal and higher emitters) based on their exhaust

emission levels. EPA then subdivided vehicles in each emitter class

into ``technology groups'' based on the emission control technology

with which each vehicle was equipped. However, as discussed in the

NPRM, EPA was concerned that technology group distinctions among higher

emitters might not be appropriate, since such vehicles' high level of

emissions indicated that their emission control systems were not

functioning properly. In addition, the limited quantity of data for

higher emitters made it difficult to identify genuine differences in

emissions response between higher emitters of different technology

groups. Many commenters expressed similar concerns. Hence the model

promulgated today does not divide higher emitters into technology group

categories but retains such distinctions when analyzing normal

emitters. In response to numerous comments, EPA attempted to reduce the

number of normal emitter technology groups. However, as discussed in

section IV.A of the RIA, EPA was unable to identify an appropriate

basis for consolidation. EPA considers its retention of emitter class

and technology group distinctions to be justified by the presence of

statistically significant fuel effects specific to individual emitter

classes and technology groups in today's complex model.

At the same time, EPA recognized the validity of comments received

from a number of sources that (1) many emission effects were likely to

be consistent across multiple technology groups or across emitter

classes, and (2) insufficient data were available to model many

potential terms, particularly interactive terms. The approach used by

EPA to construct the complex model proposed in February 1993 did not

incorporate these legitimate concerns. To do so, EPA has utilized a

modified version of the ``unified'' approach advocated by API and other

commenters (as described in the RIA) to develop today's complex model.

This modeling approach, the statistical criteria used by EPA in

conjunction with this approach, and the techniques used to simplify the

models are discussed in detail in section IV.A of the RIA and are

summarized below.

First, interactive terms were permitted to enter the models only

when sufficient data were available. The model proposed in the February

1993 NPRM permitted all interactive terms to enter the models,

regardless of whether sufficient data were available to estimate such

an effect, and it did not apply statistical criteria to evaluate

whether terms added to the model introduced more risk of inaccuracy in

the model than they removed.

Second, preliminary models for higher emitting vehicles were

constructed based solely on data from such vehicles. Only those terms

that satisfied EPA's statistical criteria (discussed at length in the

RIA) were retained. These criteria included measures to balance

overfitting (introducing too many terms to explain the observed data)

and underfitting (not including terms necessary to explain the observed

data). The NPRM model did not include measures to prevent overfitting.

Third, the entire database was analyzed using the unified approach.

The effects of each term on emissions was divided into two parts: an

average effect across all vehicles, and a series of adjustment terms

for each technology group and for higher emitters. Only those terms

that satisfied EPA's statistical criteria were retained, with two

exceptions. Higher emitter adjustment terms were retained regardless of

statistical significance since they had been found to be statistically

significant when examining the higher emitter data separately. EPA was

concerned that failure to do so might cause genuine higher emitter

effects to be ``washed out'' by the greater number of data for normal

emitters. In addition, some overall terms were retained for hierarchy

reasons despite low statistical significance. For example, a linear

term for a given fuel parameter (e.g., E300) might not be significant

while a squared term for the same parameter (e.g., E300\2\) might be

significant. Since the mathematical form of the squared terms includes

the corresponding linear effects, the linear term would be retained

regardless of significance to preserve the model's hierarchical

structure. The importance of hierarchy was emphasized by a number of

workshop participants and commenters, as discussed in the RIA. The NPRM

model included separate terms for each technology group and emitter

class and hence did not include terms to represent the average effect

of a fuel parameter across all vehicles. The NPRM model also did not

incorporate hierarchy considerations.

Fourth, outlying and overly influential data were dropped from the

database and the model was re-estimated based on the remaining data.

Outlying data consist of observations that differ from the average

observed effect by so large a margin that they are more likely to

represent observational error, reporting error, or other measurement

artifacts than genuine phenomena. Outlying data can obscure genuine

emissions effects. Influential data consist of observations that by

themselves materially affect the resulting model, i.e., the model would

differ materially if they were excluded. In a database the size of the

Complex Model database, individual data points should not have such

unusually large effects. Excluding outlying and influential

observations is standard statistical practice. The NPRM model did not

exclude either type of observation.

Fifth, terms were deleted from the resulting model to avoid

overfitting and collinearity problems. Overfitting occurs when so many

terms are included in a regression model that the expected error due to

the erroneous inclusion of a term exceeds the expected error due to not

including the term. Collinearity problems occur when the fuel

parameters included in the model are correlated with one another in the

fuels tested. For example, the addition of oxygenate to gasoline causes

E200 to increase. The oxygenate-containing fuels in the complex model

database tend to have higher E200 values than fuels without oxygenate.

In a sense, one can predict the E200 value of a fuel by knowing its

oxygen content. Hence these two parameters would be considered to be

highly collinear. Since regression models are developed under the

assumption that terms are not collinear, the presence of strong

collinearities can introduce error into the regression. Today's complex

model takes both collinearity and overfitting into account by using a

standard statistical criterion called Mallow's Cp criterion to

remove terms which introduce large overfitting and collinearity

problems. This approach resulted in a simpler, more reasonable, and

statistically more sound model than had been proposed in the February

1993 NPRM. It should be noted that high emitter terms forced into the

model earlier in the process could be dropped at this stage of the

analysis. Measures were taken to limit collinearity problems in the

NPRM model, but overfitting concerns and the Cp criterion were not

addressed.

Sixth, the contribution of each remaining term to the model's

explanatory power was estimated, and those terms whose contribution

summed to less than one percent were deleted (i.e., the retained terms

accounted for 99 percent of the explanatory power of the model) to

simplify the form of the model without materially reducing its ability

to predict the emissions impact of fuel modifications. This step was

not taken during development of the NPRM model.

Finally, the resulting models for each technology group within the

set of normal emitting vehicles were consolidated into a single

equation using a random balance approximation. The details of that

approximation are given in Section IV.A of the RIA. This step was not

taken during development of the NPRM model.

The results of EPA's modeling efforts confirms the importance of

technology group and emitter class distinctions, as can be seen by

examining the differences in the exhaust emission equations for

specific normal emitter technology groups or for normal and higher

emitter class categories (as discussed in greater detail in the RIA).

Efforts to reduce the number of technology group categories for normal

emitters were not successful. Efforts to subdivide higher emitters by

their emission characteristics such as exhaust hydrocarbon to NOX

ratio did not improve the quality of EPA's higher emitter model.

However, as discussed above, EPA found it unnecessary to separate

higher emitters by technology group. This modification reflects EPA's

belief, supported by preliminary field information, that one or more

emission control components on higher emitters tend to be

malfunctioning, which renders a classification scheme based on vehicle

equipment questionable.

3. Exhaust Model

As was discussed in the April 1992 and February 1993 proposals, the

weight assigned to each technology group or emitter class for modeling

purposes was set equal to its contribution to in-use emissions for each

pollutant. The weight assigned to each emitter class was set equal to

its projected contribution to in-use emissions. The weighting factor

assigned to normal emitters was then broken down further by technology

group, again according to their projected contribution to in-use

emissions. These estimates and projections are essentially unchanged

from the February 1993 proposal, although minor changes have been made

to reflect more complete information about the fraction of 1990 sales

accounted for by each technology group. The rationale for, derivation

of, and renormalization of the weighting factors themselves are

discussed in more detail in the RIA.

Various commenters indicated that they considered EPA's previously

proposed models were too complex. In response, the Agency has modified

its analysis method in several ways. The resulting method, described in

Section IV.B.2, results in exhaust emission models containing two

equations for each pollutant instead of as many as sixteen separate

equations, as was the case for the model proposed in February 1993.

Each equation also has far fewer terms than the February 1993

equations. However, EPA does not believe that today's less complicated

complex model is less accurate than the complex models presented at

public workshops or in the February proposal. This belief is based on

the models' comparable explanatory power (as reflected in their similar

R\2\) and the superior accuracy of today's model in accounting for the

emission effects seen in the vehicle testing programs that comprise the

complex model database. Today's VOC and NOX models are based on

the most accurate of the three sets of models included in EPA's July

14, 1993 docket submittal, while also taking into account relevant

comments regarding specific aspects of the models. Today's toxics

models are a further simplification of the models included in the July

1993 docket submittal in response to comments received by EPA on its

docket submittal. These points are discussed more fully in Section IV.A

of the RIA.

The specific equations that comprise the complex model can be found

in section 80.45 of the regulations for this rule. Their derivation is

discussed in detail in Section IV.A of the RIA. The range of parameter

values for which these equations are valid is discussed in Section D

and in Section IV.D of the RIA. As discussed in Section V, refiners are

required to submit data to augment the model if they wish to certify

fuels with properties that fall outside this range as reformulated

gasolines.

C. Nonexhaust Model

Nonexhaust emissions are less strongly affected by vehicle design

and are influenced by fewer fuel characteristics than are exhaust

emissions. In addition, the theoretical principles involved in

nonexhaust emissions (which include evaporative, running loss, and

refueling emissions) are better understood, and nonexhaust emission

control technologies are more consistent across vehicles, than are

exhaust emissions and emission control technologies. Since the

relationship between fuel properties and nonexhaust emissions is less

complex and better understood than for exhaust emissions, there was

much less need for EPA to generate additional data to evaluate

nonexhaust emissions than was the case for exhaust emissions. EPA was

able to base its nonexhaust VOC emission model on data generated from

EPA's ongoing nonexhaust emissions testing program that has been used

to develop EPA's MOBILE emission inventory models, specifically the

MOBILE4.1 and MOBILE5.0A models. EPA believes this data to be

sufficient to model the relationship between fuel properties and

nonexhaust VOC emissions for the purposes of this rule. Additional

information about MOBIL4.1 and MOBILE5.0A can be found in Dockets A-91-

02 and A-92-12.

EPA is in the process of developing an enhanced model of nonexhaust

VOC emissions, based on a more complete set of theoretical principles

and additional test data, that is expected to be more accurate and more

widely applicable to oxygenated fuels than the MOBILE models. A

preliminary version of this model was discussed at a public workshop

held on August 25, 1992, and materials related to this model have been

placed in the docket for this rulemaking. At this time, however, this

enhanced nonexhaust VOC emissions model is not complete and hence is

not incorporated in today's complex model.

The nonexhaust VOC model in today's complex model is based on

correlations between RVP and nonexhaust VOC emissions derived from the

July 11, 1991 version of MOBILE4.1 for Phase I of the reformulated

gasoline program (1995-1999) and from MOBILE5A for Phase II (2000 and

beyond). This approach is consistent with the definition of baseline

emissions set forth in Section IV.A and is based on the same

considerations outlined in that section.

To develop the correlations shown below, the MOBILE models were

used with temperatures of 69 to 94 degrees Fahrenheit for Class B areas

and 72 to 92 degrees Fahrenheit for Class C areas. As discussed in

Section IV.A, a basic inspection and maintenance program was assumed

for Phase I while an enhanced I/M program was assumed for Phase II. In

addition, the presence of Stage II evaporative emissions recovery

systems with an overall vapor recovery efficiency of 86 percent was

assumed (as discussed in the SNPRM and NPRM). EPA is in the process of

promulgating requirements for onboard refueling emission controls which

may be more effective at controlling refueling emissions than Stage II

vapor recovery systems. However, these requirements did not apply to

1990 model year vehicles and hence cannot be incorporated into the

model for certification purposes. In addition, EPA has chosen not to

incorporate the effects of onboard refueling controls in its evaluation

of the effects of reformulated fuels on emissions from the entire in-

use vehicle fleet, which includes vehicles from a number of different

model years. This decision was made for several reasons. First,

requirements for onboard refueling controls have not yet been

finalized, making evaluation of their impact on in-use emissions

difficult. Second, onboard refueling controls are not expected to be

required on all new vehicles until 2000 and are not expected to be

present on the bulk of in-use vehicles for several years after that

time. Third, while onboard controls are expected to be more efficient

at controlling refueling emissions than Stage II controls, the

difference is not expected to be large in areas affected by the

reformulated gasoline program and will affect only a small portion of

total nonexhaust VOC emissions. Since EPA's analysis of the additional

benefits of onboard vapor recovery controls is not yet available, and

since such benefits are expected to be small relative to overall

emissions, EPA has chosen to retain its assumptions regarding Stage II

vapor recovery in forecasting the effects of fuel modifications on

nonexhaust VOC emissions from the in-use vehicle fleet.

The only toxic air pollutant covered by the reformulated gasoline

program that is found in nonexhaust emissions is benzene, which is a

natural component of gasoline. The other four toxic air pollutants

listed in section 211(k) are solely products of fuel combustion and

hence are not found nonexhaust emissions. As discussed in the SNPRM,

the Agency's correlation between fuel benzene content and summer non-

exhaust benzene emissions is based on results from General Motors'

proprietary model of tank vapors, as confirmed independently by EPA-

generated data using a number of fuels. Both the derivation and

verification of the non-exhaust benzene emissions model are discussed

more fully in the RIA. The nonexhaust benzene emission model also

depends on the RVP of the fuel, as is the case for the nonexhaust VOC

emission model. The derivation of the nonexhaust benzene and VOC models

is discussed more fully in the RIA.

D. Range/Extrapolation

Like all regression models, the complex model is not valid for all

possible input values. The range of fuel parameter values over which

the complex model accurately predicts vehicle emissions is given in

Table IV-3. These ranges are based on the range of data used to develop

the models and on comments received by the Agency on this issue. The

limits proposed in the February 1993 were, in some cases, narrower than

the range of data used to develop the complex model. In addition, the

limits proposed in the NPRM would have prevented a number of very low

emitting fuels from being certified using the model.

Table IV-3.--Parameter Ranges for Which the Complex Model Can Be Used

------------------------------------------------------------------------

Valid range for:

---------------------------

Fuel Parameter Reformulated Conventional

fuel fuel

------------------------------------------------------------------------

Aromatics, vol %............................ 0-50 0-55

E200, %..................................... 30-70 30-70

E300, %..................................... 70-100 70-100

Olefins, vol %.............................. 0-25 0-30

Oxygen, vol %............................... 0-3.7 0-3.7

RVP, psi.................................... 6.4-10 6.4-11

Sulfur, ppm................................. 0-500 0-1000

Benzene, vol %.............................. 0-2.0 0-4.9

------------------------------------------------------------------------

EPA has received a number of comments requesting alterations in the

model's range. After considering these comments and re-evaluating the

data on which the complex model is based, EPA has modified the range

limits. In some cases, EPA has chosen to extrapolate the complex model

slightly beyond the range for which data were available in order to

allow additional fuels, both conventional and reformulated, to be

evaluated using the model without recourse to expensive and time-

consuming vehicle testing. These extrapolations are limited to those

parameters whose effects appear to be well-characterized by the complex

model. A detailed discussion of the limits of the available data, EPA's

rationale for extending the valid range of the model for some

parameters, and the extrapolation method used to extend the model can

be found in Section IV.D of the RIA.

E. Winter

While the VOC performance standard for reformulated fuels applies

only in the summer, the toxics and no-NOx-increase requirements

apply year-round. EPA therefore recognized the need to model the

exhaust toxics and NOx emissions performance of reformulated

gasolines during the winter months as well as during the high ozone

season. Modeling winter emissions performance, however, presented a

number of difficulties. First, the data sources described earlier

provided data on emissions performance only under summer conditions and

for gasolines with RVP levels typical of summer gasolines. Second, the

RVP levels of fuels included in the complex model database ranged from

7 to 10 psi, while winter fuels tend to have RVP levels in the 11.5 psi

range and are not restricted by other regulations. Hence the complex

model cannot be used directly for fuels with typical winter RVP levels.

RVP's impact on canister loading and subsequent purging is thought

to be the primary cause of its effects on exhaust emissions. Since data

do not exist on the effects of winter fuels on canister loading under

winter conditions, the Agency is not able at this time to model the

effects of winter RVP levels on exhaust emissions. To avoid making

unsound or speculative predictions, EPA proposed and is now

promulgating a requirement that for purposes of evaluating emissions

effects using the complex model, the RVP of winter fuels be set at the

summer statutory baseline RVP value. In effect, this requirement builds

into the model the assumption that the RVP level of winter gasolines

has no effect on NOx or exhaust toxics emissions. As a result,

refiners will not be required to alter the RVP levels of winter

gasolines. Refiners will receive neither benefit nor penalty for

changing the RVP of their winter gasolines. To evaluate winter fuels

using the complex model, an RVP value equal to that of summer baseline

gasoline (8.7 psi) must be used instead of the fuel's actual RVP. Doing

so effectively removes the contribution of RVP to winter exhaust

emissions.

When sufficient data is developed on the emissions impact of winter

RVP levels under winter ambient conditions, EPA will be able to revise

the complex model accordingly. Until then, EPA believes it is more

appropriate to assume that RVP levels have winter exhaust emission

effects than to speculate about the magnitude of such impacts.

In its prior proposals, EPA had proposed that winter nonexhaust

emissions, including winter nonexhaust benzene emissions, be considered

zero. EPA received a number of comments requesting that both baseline

emissions and the nonexhaust toxics model include winter nonexhaust

benzene emissions. This request was based on the belief that the year-

round benzene limits would result in reduced nonexhaust benzene

emissions in the winter months. EPA has evaluated this claim, taking

into account temperature ranges and the effects of inspection and

maintenance programs on such emissions. EPA acknowledges the validity

of this claim, since winter nonexhaust emissions, including nonexhaust

benzene emissions, are likely to be nonzero under all winter

temperature ranges. In the past, the lack of sufficient data on

nonexhaust emissions under winter temperature conditions has prevented

EPA from developing reliable, accurate models of winter nonexhaust

emissions. The commenters provided a limited quantity of data on winter

nonexhaust emissions to support their claim. However, the data

submitted in support of this claim were based on measurements of

nonexhaust emissions from vehicles with very low nonexhaust emissions.

EPA's analysis indicates that these vehicles are not representative of

in-use vehicles. In addition, the chemical composition of the measured

nonexhaust emissions were characteristic of resting losses (losses that

occur due to permeation through fuel system components) rather than of

diurnal, hot soak, or running loss emissions. Resting losses are not

included in EPA's baseline emission estimates, so EPA does not consider

it appropriate to include resting losses in its nonexhaust emission

models. Finally, no data were submitted on nonexhaust benzene emissions

from fail vehicles under winter conditions. Since nonexhaust benzene

emissions from such vehicles will comprise a significant portion of

winter nonexhaust benzene emissions, EPA is concerned that a model

based on the submitted data would not provide accurate estimates of

such emissions. Given the theoretical merits of the claim, however, EPA

will consider including a model of winter benzene nonexhaust emissions

in the complex model in the future when sufficient data become

available.

F. Fungibility

EPA has long recognized the importance of maintaining a fungible

fuel system, in which complying gasolines can be mixed freely without

resulting in mixtures that do not themselves comply with regulatory

requirements. Fungibility is essential to smooth, cost-effective

operation of fuel distribution systems such as pipelines. The Agency

has received numerous comments on the need to maintain fungibility. At

the same time, the Agency considers it essential that gasolines

certified as reformulated meet all required emission performance levels

in the field. In cases where the effects of a given fuel parameter on

emissions are non-linear, it is possible for two complying fuels to

produce a non-complying fuel when mixed.

The complex model contains a number of nonlinear terms, which

introduces the possibility that gasolines which comply with this rule's

requirements in isolation would not comply if mixed with other

complying fuels. EPA has been concerned with this possibility and has

undertaken extensive analyses to determine its likelihood and to

develop methods to cope with its occurrence. EPA's analyses, which have

utilized methods that have been supported by a number of organizations,

indicate that the complex model promulgated in today's rule will not

create fungibility problems despite its inclusion of nonlinear terms.

This analysis is explained in greater detail in Section IV.F of the

RIA.

G. Future Model Revisions

The complex model promulgated in this rulemaking reflects EPA's

best understanding of the relationship between fuel characteristics and

vehicle emissions. However, EPA expects future research to clarify this

relationship. EPA also recognizes that changes in in-use vehicle

emission control programs (e.g., I/M programs) will continue to occur

and that these changes may alter the relationship between fuel

characteristics and in-use emissions. In addition, the Agency is

concerned that augmentations to the model through vehicle testing

(Section V) may, over time, accumulate to the point that a revised

complex model, incorporating the current complex model database and all

relevant information gathered since then, would be beneficial. As

discussed in Section V, EPA plans to issue revised complex models when

the Agency deems that sufficient new information is available to

warrant such action. Model revisions will be developed through a formal

rulemaking process.

H. Complex Model Performance of Simple Model Fuels

Fuels qualifying as reformulated under the simple model must meet

specified benzene, oxygen, and RVP requirements while also satisfying

the toxics performance standard. The RVP requirement differs between

VOC control regions, and the requirements and standards also vary

depending on whether compliance is being achieved on a per-gallon or

averaging basis. In addition, levels of other fuel parameters are only

specified under the simple model in terms of deviations from each

refiner's baseline fuel. Evaluating the performance of simple model

fuels under the complex model is difficult since fuel properties can

vary widely.

However, it is possible to evaluate a set of fuels that are

representative of expected, typical simple model fuels. EPA expects

most refiners to pursue compliance on average (for all or part of their

product slate) in order to maximize flexibility in day-to-day refinery

operations and recoup compliance margins. Given present and projected

conditions, EPA also expects that MTBE and ethanol will be the most

commonly used oxygenates during Phase I of the reformulated gasoline

program. The fuels specified in Tables IV-4 and IV-5 below include

fuels designed to meet the requirements of the simple model in both VOC

control regions and using both oxygenates. The level of olefins,

sulfur, E200, and E300 have been set to Clean Air Act baseline levels,

while the level of aromatics has been set at the level necessary to

comply with the toxics requirements of the simple model. Aromatics

levels were assumed to be the same for summer and winter fuels.

Table IV-4.--Typical Simple Model Fuels Using MTBE

[Under Averaging]

------------------------------------------------------------------------

Fuel

-------------------------------------------------------

1 2 3 4

------------------------------------------------------------------------

Fuel

Description:

Season........ Summer...... Summer...... Winter...... Winter

VOC Control 1........... 2........... 1........... 2

Region.

Fuel

Parameter:.

RVP, psi...... 7.1......... 8.0......... N/A......... N/A

Oxygen, wt%... 2.1......... 2.1......... 2.1......... 2.1

Benzene, vol%. 0.95........ 0.95........ 0.95........ 0.95

Aromatics, 27.5........ 26.3........ 27.5........ 26.3

vol%.

Olefins, vol%. 9.2......... 9.2......... 11.9........ 11.9

E200, %....... 41.......... 41.......... 50.......... 50

E300, %....... 83.......... 83.......... 83.......... 83

Sulfur, ppm... 339......... 339......... 338......... 338

------------------------------------------------------------------------

Table IV-5.--Typical Simple Model Fuels Using Ethanol

[Under Averaging]

------------------------------------------------------------------------

Fuel

-------------------------------------------------------

5 6 7 8

------------------------------------------------------------------------

Fuel

Description:

Season........ Summer...... Summer...... Winter...... Winter

VOC Control 1........... 2........... 1........... 2

Region.

Fuel

Parameter:.

RVP, psi...... 7.1......... 8.0......... N/A......... N/A

Oxygen, wt%... 2.1......... 2.1......... 2.1......... 2.1

Benzene, vol%. 0.95........ 0.95........ 0.95........ 0.95

Aromatics, 25.5........ 24.3........ 25.5........ 24.3

vol%.

Olefins, vol%. 9.2......... 9.2......... 11.9........ 11.9

E200, %....... 41.......... 41.......... 50.......... 41

E300, %....... 83.......... 83.......... 83.......... 83

Sulfur, ppm... 339......... 339......... 338......... 338

------------------------------------------------------------------------

The performance of these fuels according to the complex model

(using the MOBILE4.1 baseline as previously discussed) is summarized in

Table IV-6.

Table IV-6.--Performance of Typical Simple Model Fuels Under the Phase I Complex Model

[Under Averaging]\1\

----------------------------------------------------------------------------------------------------------------

Emission reduction versus CAAB fuel (percent)

----------------------------------------------------------------

Fuel Nonexhaust

Exhaust VOC VOC Total VOC NO X Toxics

----------------------------------------------------------------------------------------------------------------

1.............................................. 7.92 51.42 36.57 1.46 27.33

2.............................................. 5.35 23.93 17.11 1.28 24.57

3.............................................. 0.33 N/A 0.33 -0.21 12.83

4.............................................. 0.80 0.00 0.80 0.04 13.87

5.............................................. 8.64 51.42 36.82 1.90 25.70

6.............................................. 6.09 23.93 17.38 1.76 22.56

7.............................................. 3.55 N/A 3.56 0.58 11.52

8.............................................. 4.01 N/A 4.01 0.88 12.48

----------------------------------------------------------------------------------------------------------------

\1\Performance of summer fuels (#s 1, 2, 5, 6) given relative to that of Clean Air Act summer baseline fuel.

Performance of winter fuels (#s 3, 4, 7, 8) given relative to that of the winter baseline fuel defined in

Section III.

I. Phase I Performance Standards Under the Complex Model

All fuels produced during Phase I of the reformulated gasoline

program must meet the VOC, toxics, and NOX requirements of the

Act. Fuels certified using the complex model in Phase I must show

either no increase in NOX emissions from baseline levels on a per-

gallon basis as discussed in the February 1993 proposal or a 1.5%

reduction from baseline levels on average as discussed in Section VII.

In addition, as discussed in Section III.E., such fuels must result in

either a 15% reduction in total toxics emissions from baseline levels

on a per-gallon basis or a 16.5% reduction in total toxics emissions

from baseline levels on average.

With regard to the VOC standards, EPA considers fuels produced to

meet the provisions of the simple model to be producible. Thus, as

discussed in the February 1993 proposal, EPA believes it feasible to

base the Phase I standards for VOC emissions on the performance of

fuels that meet the Simple Model requirements, provided that this

performance is more stringent than minimum performance required by the

Act. EPA considers the fuels whose VOC performances were evaluated in

Section IV.H to be representative of Simple Model fuels. Under the

reformulated gasoline program, VOC emissions are controlled only during

the high ozone season. For this reason, the VOC performance standard

has been determined by the performance of the Phase I summer fuels

presented in Section IV.H. Since these fuels achieve emissons

reductions that equal or exceed the minimum requirements set forth in

the Act, the VOC performance standard during Phase I for fuels

certified under the complex model has been based on the performance of

these fuels. Setting the VOC performance standards in 1998-1999 equal

to this VOC performance level, which EPA believes to be a reasonable

estimate of the average performance of fuels produced in 1995-1997,

preserves the integrity of the two-phase program specified by Congress

and is consistent with the Agreement in Principle signed in 1991.

The summer VOC performance of ``typical'' high ozone season simple

model reformulated gasolines according to the complex model is

presented in Table IV-6. In VOC Control Region 1, the simple model fuel

reduces VOC emissions by 36.6 percent for the MTBE-containing fuel

(Fuel 1) and 36.8 percent for the ethanol-containing fuel (Fuel 5).

Since the 1998 performance requirements in VOC Control Region 1 are to

be based on the performance of typical simple model fuels, and since

Fuels 1 and 5 both satisfy the simple model requirements and are

considered by EPA to be representative of typical simple model fuels,

EPA has set its 1998 performance standards in VOC Control Region 1 so

as to permit both of these fuels to meet the 1998 performance

standards. In addition, EPA considers Fuel 1 to be more representative

of typical simple model fuels in VOC Control Region 1 since MTBE does

not boost fuel RVP levels to the extent that ethanol does. As was

discussed in the April 1992 and February 1993 proposals, EPA believes

that per-gallon performance standard should be set 1.5 percentage

points below the averaging performance standard. Hence high ozone

season fuels certified using the complex model during Phase I of the

reformulated gasoline program must provide a VOC emission reduction

from baseline levels of 36.6 percent when complying on average and 35.1

percent when complying on a per-gallon basis. Similarly, high ozone

season fuels certified using the complex model during Phase I in VOC

Control Region 2 must provide a VOC emission reduction from baseline

levels of 17.1 percent when complying on average and 15.6 percent when

complying on a per-gallon basis. These standards are summarized in

Table IV-7 for both VOC control regions, under averaging and per-gallon

compliance. Note that a negative performance standard signifies a

reduction from baseline emission levels.

Table IV-7.--Reformulated Gasoline Performance Standards Relative to Clean Air Act Baseline Gasoline for 1998-

1999

[Percent]

----------------------------------------------------------------------------------------------------------------

VOC control region 1 VOC control region 2

Emission ---------------------------------------------------------------

Average Per gallon Average Per gallon

----------------------------------------------------------------------------------------------------------------

VOC............................................. -36.6 -35.1 -17.1 -15.6

Toxics.......................................... -16.5 -15.0 -16.5 -15.0

NOX............................................. -1.5 0.0 -1.5 0.0

----------------------------------------------------------------------------------------------------------------

In summary, the per-gallon and averaging VOC performance standards

under the complex model during Phase I is set by the performance of the

corresponding simple model fuel when evaluated using the complex model.

The toxics performance standard is set at the statutory requirement of

a 15 percent reduction from baseline levels for per-gallon compliance

and a 16.5 percent reduction for compliance on average. Similarly, the

NOX performance standard under the complex model during Phase I

must satisfy the no NOx increase requirement on a per-gallon

basis, or meet a 1.5% reduction for compliance on average.

V. Augmenting the Models Through Testing

During the regulatory negotiation process, vehicle testing and

emission modeling procedures for certifying that a gasoline complies

with the NOX, toxics, and VOC requirements were discussed.

Emission models such as the simple model described in Section III and

the complex model described in Section IV offer several advantages over

testing to determine emission effects. First, models can better reflect

in-use emission effects since they can be based on the results of

multiple test programs. Second, individual test programs may be

intentionally or unintentionally biased due to vehicle selection, test

design, and analysis methods. Third, fuel compositions tend to vary due

in part to factors beyond the control of fuel suppliers such as

variations in crude oil compositions and the inherent variability of

refining processes. As a result, without one or more modeling options,

each batch of fuel would have to be tested to ascertain its emission

performance. Such levels of testing are neither desirable (because of

the potential for intentional or unintentional bias in vehicle test

programs) nor practical (because of the time and expense involved in

vehicle testing). Fourth, models make more efficient use of scarce and

expensive emission effects data than is possible otherwise. For these

reasons, EPA believes that the modeling options outlined above are

necessary for the reformulated gasoline program to achieve its

environmental objectives and to minimize the costs of the program.

These emission models, however, reflect currently-available

information and hence do not allow refiners to take advantage of

emission benefits derived from new fuel additives or changes in fuel

parameters not contained in the models. To allow for fuel technology

development and innovation, the Agency also believes that testing has a

role in certification as a means of supplementing the models. This

section contains a detailed discussion of the provisions EPA is

promulgating regarding the conditions under which testing is permitted,

the manner in which test results can be used to supplement the models,

and the minimum requirements for vehicle testing programs. As was first

outlined in the February 1993 NPRM, the vehicle testing process

described in this section has undergone significant changes since it

was first proposed in the April 1992 SNPRM. These changes have been

made in response to changes in EPA's approach to modeling the

relationship between fuel properties and emissions, as described in

Section IV, and comments received in response to the April 1992 and

February 1993 proposals. The following discussion addresses the major

substantive comments received by EPA regarding certification of fuels

by vehicle testing. A detailed summary and analysis of comments can be

found in Section IV.G of the RIA.

A. Applicability of Testing

Vehicle testing is the primary way that the effects of various

gasoline formulations on motor vehicle emissions can be determined. As

described above, data from vehicle testing programs forms the bulk of

the basis for the simple and complex models.

EPA believes that fuel certification through single test programs

is inherently less reliable than certification through a testing-based

model. The simple and complex models developed by EPA are based on a

far greater amount of testing than would be available from any single

test program. These models incorporate and balance the varying and

conflicting results of numerous test programs. The statistical

variation associated with an individual test program may cause a fuel

to show emission effects during testing that would not occur in-use.

Therefore, EPA proposes that testing only be permitted to augment the

models for fuel effects that are not covered in the models.

B. Augmenting the Simple Model

Due to the belief that fuels certified by vehicle testing should be

evaluated in conjunction with the most complete emission model

available to more accurately determine the emission benefits of the

fuels being tested, EPA proposed that vehicle testing be permitted to

augment the simple model only for the effect of oxygenates on NOx

emissions beyond the simple model's oxygen caps. All other testing was

to have been performed to augment the complex model. Based on data

collected since the time of the proposal on the effect of oxygenates on

NOx, EPA no longer believes it appropriate to augment the simple

model even in the limited manner described above. Considerably more

data are available in the complex model database regarding the effect

of oxygenates on NOx emissions than would be provided by any

individual test program. Therefore, testing can only be performed to

augment the complex model. Fuels with oxygen concentrations in excess

of 2.7 weight percent must be certified using the complex model.

C. Augmenting the Complex Model

EPA believes that the objective of testing under the complex model

should be to evaluate the emission effects of fuels whose emission

effects cannot be adequately represented by the model. Such fuels would

include fuels claiming emission effects from parameters not included in

the complex model and fuels containing complex model parameters at

levels beyond the range covered by the model. Without this constraint,

it may be possible for a fuel producer to use the statistical variation

associated with testing to claim emission effects through testing which

would not be demonstrated in-use, when tested to a greater degree, or

when modeled. For example, a fuel that would fail to meet the VOC

requirement by a small margin when evaluated under the complex model

could be tested and shown to meet the VOC requirement due to the

testing error associated with any vehicle testing program. In addition,

allowing testing of existing modeled parameters essentially would make

the complex model, and the associated emission performance standards, a

fluid target. Fuel producers would lose the certainty associated with a

fixed model and the confidence that their capital investments will be

useful for a fixed amount of time. Therefore, vehicle testing can be

used only to determine the emission effects of parameters not

adequately represented by the complex model. The emission effects of

the fuel parameter in question will be determined by combining the

emission effects determined through vehicle testing with the emission

effects predicted by the complex model. Furthermore, each testing

program can be used to identify the effects of only one new fuel

parameter, unless the changes in other fuel parameters are a natural

and inherent consequence of the primary fuel modification. Without this

constraint, EPA believes that accurate determination of the effects of

specific fuel parameters would be more difficult due to the inherent

variability in testing programs and the increased opportunities for

gaming.

In addition, fuel suppliers opting to augment the complex model

through vehicle testing must examine the extent to which emissions are

affected when fuels certified with the augmented complex model are

mixed with other fuels. The Agency is concerned with two potential

problems when different fuels are combined. First, the emission effects

of a parameter, as determined from vehicle testing, may not behave

linearly as fuels with one level of the parameter are mixed with fuels

with different levels of the same parameter. The degree to which this

process occurs is referred to in this notice as the parameter's

dilution effect. Dilution effects are evident in the complex model

proposed in February 1993 and in the model being promulgated today.

Second, the emission effects of various fuel parameters may be affected

by the level of other fuel parameters. The degree to which this process

occurs is referred to in this notice as an interactive effect. If such

effects are present (as in the complex model proposed in February 1993

and in the complex model being promulgated today), actual emission

performance of the fuel mixture in-use could be worse than emission

performance predicted from the complex model augmented by vehicle

testing results. Therefore, the testing process must be structured so

as to identify dilution and interactive effects.

D. Advance Approval of Test Programs

Given the number of factors involved in designing a test program,

the potential for inappropriate design is high. EPA wishes to avoid

submittal of petitions based on test data from poorly designed programs

in order to assure that the time and money invested in such programs is

well-spent and to assure that all augmentations to the model are based

on accurate data from well-designed test programs. Hence EPA will

require petitioners to obtain advance approval from the Agency for

their proposed vehicle testing programs. EPA will consider petitions to

augment the model only if based on the results of approved testing

programs. Furthermore, EPA retains the discretion to evaluate other

data when evaluating petitions to augment the complex model and when

determining the nature, extent, and limitations of the augmentation.

This data may include the existing complex model database, additional

vehicle testing programs, and other augmentation applications.

Petitioners are required to include the following information when

submitting a test program plan for approval: the fuel parameter to be

evaluated for emission effects; the number and description of vehicles

to be used in the test, including model year, model name, VIN number,

mileage, emission performance, technology type, and vehicle

manufacturer; the methods used to procure and prepare the vehicles for

testing; the fuels to be used in the testing program, characterized as

defined in Section V.I.5; the pollutants and emission categories to be

evaluated; the methods and precautions to be used to ensure that the

effects of the parameter in question are independent of the effects of

other parameters already included in the complex model; a description

of the quality assurance procedures to be used during the test program,

and the identity and location of the organization performing the

testing. EPA anticipates and encourages petitioners to submit the

information listed above in stages beginning with the most general and

ending with the most specific in order to streamline the approval

process and eliminate wasted effort. EPA will work with petitioners to

remedy unsatisfactory aspects of their proposed testing program.

These provisions provide the Agency with greater assurance that

petitioners would not selectively report test results to the Agency

that support their petitions. Petitioners would still be able to

``game'' the testing process by pre-screening vehicles to obtain a test

fleet with the desired sensitivity to the proposed parameter. However,

such a test fleet would have to be re-tested as part of the formal test

program and hence would be subject to the variability inherent in

vehicle testing, which would tend to reduce the gaming benefits from

pre-screening. EPA believes that the risks and costs associated with

re-testing will tend to dissuade petitioners from attempting to

manipulate the testing process in this manner.

EPA further requires that the results of all approved testing

programs be submitted to the Agency, even if the parameter in question

proves not to provide an emission benefit. The Agency believes this

requirement is necessary to ensure that all available data is at the

Agency's disposal when evaluating proposed augmentations to the complex

model and when updating the model itself. EPA does not intend to use

this provision to limit legitimate, innovative test programs. Rather,

EPA is only interested in preventing the creation of artificial fuel

parameters that claim to be the source of emission effects which are in

reality only normal statistical variability.

An example may help clarify the problems that can arise if testing

is permitted for such artificial parameters. The level of C10+

aromatics (aromatics whose molecules contain ten or more carbon atoms)

influences a fuel's E200, E300, and total aromatics levels. A testing

program to identify the effects of C10+ aromatics may indicate that an

emission effect from such compounds exists when the effect is actually

due to differences in the fuels' E200, E300, and total aromatics levels

or to the inherent statistical variability associated with vehicle

testing. A petition for approval of a test program to identify the

effects of C10+ aromatics would be required to identify specific

measures to be taken to isolate the emission effects of C10+ aromatics

from those of E200, E300 and total aromatics, all three of which are

included in the complex model. In this example, EPA might require that

certain test fuels contain identical levels of E200, E300, and total

aromatics; that more rigorous statistical tests be used to identify

genuine C10+ aromatics effects beyond those already incorporated in the

complex model for E200, E300, and total aromatics; that the fuels used

in the test program meet more detailed compositional criteria to ensure

their representativeness; or that additional vehicles and/or fuels be

tested. This provision helps assure that the effects observed in

vehicle testing programs are genuine and will occur in-use.

E. Exclusive Rights to Augmentation

EPA's April 1992 and February 1993 proposals discussed the

advantages and disadvantages of providing a system of exclusive rights

to model augmentations. EPA has given this matter further

consideration, including consideration of comments regarding exclusive

rights. The Agency has concluded that the reasons given in its April

1992 proposal for not providing a system of exclusive rights are still

valid. Hence the regulations governing augmentation of the complex

model through vehicle regulation being promulgated today do not provide

for exclusive rights to augmentations. Each augmentation will be

available to any refiner desiring to utilize it, and no restrictions

are provided under this rulemaking for exclusive rights, other than

those granted under other legal code (e.g., patent law). The Agency

does not believe adequate authority exists to promulgate exclusive

rights provisions under this rulemaking. Furthermore, as discussed in

the April 16, 1992 proposal, there are a number of reasons from

economic, administrative, and air quality perspectives that make open

use of model augmentations a desirable public policy.

To allow interested parties to review and comment on a model

augmentation, EPA will publish a description of the augmentation and

its supporting data and information for public comment prior to

approving an augmentation for use. In keeping with the provision of the

Act, EPA will take into account any comments received, and act upon any

request received for fuel certification through model augmentation

within 180 days of such a request being completed.

F. Duration of Augmentation

In its April 1992 proposal, EPA proposed that augmentations would

remain in effect until the next subsequent complex model update was

issued. EPA further proposed that if an augmentation had been valid for

three or fewer years upon implementation of the subsequent update to

the complex model, then refiners were permitted to continue using the

augmentation in conjunction with the previous complex model for an

additional length of time, subject to certain restrictions. EPA has

received a number of comments on this proposal. Today's rule includes a

set of limitations on the duration of the augmentation that incorporate

some elements of these comments. These limitations are described below.

The Agency is concerned that fuel suppliers not be allowed to claim

emission effects in perpetuity based on the testing program described

in this section due to the smaller degree of statistical confidence in

such effects compared to those included in an updated complex model.

The Agency also recognizes the need for fuel suppliers to recoup

investments made to reformulate gasoline, including investments to

utilize the emission effects identified through vehicle testing.

Therefore, petitioners will be permitted to use emission effects

determined through vehicle testing only for a limited period of time.

In general, this period of time extends until an updated version of the

complex model takes effect. Updates to the complex model will be issued

by EPA through a formal rulemaking process at such time that the Agency

determines that sufficient additional data has become available to

warrant issuing such an update. Since some augmentations may be in

place for a relatively short period of time before the model is

updated, the Agency may not be able to adequately assess the

augmentation. However, if a proposed update to the complex model is

issued within three years of the time at which the augmentation takes

effect, then fuel suppliers may be permitted to continue using the

augmentation to determine the emission effects of reformulated

gasolines. Specifically, if the Agency does not formally accept,

reject, or modify the augmentation in question for inclusion in the

updated complex model, then the augmentation will remain available

until the next update to the model takes effect. If the Agency reviews

the augmentation and either excludes the augmentation entirely or

includes the augmentation in a modified form, then the augmentation

will remain available for use in its original form, in conjunction with

the complex model for which the augmentation was issued, to those fuel

producers who can demonstrate to the Administrator's satisfaction that

they have begun producing fuels that are certified using the

augmentation. In such cases, the augmentation may continue to be used

for five years from the date the augmentation took effect or for three

years of fuel production, whichever is shorter.

For the reasons discussed above, augmentations to the model for the

effects of a given parameter over a particular range are permitted only

once. Regardless of whether the emission effects of a parameter are

included in an updated model, the augmentation can neither be used nor

renewed (even with data from a second identical test program) once the

maximum time period for use of a model augmented with the effects of

that parameter has expired. Further testing is permitted, however, to

provide EPA with the additional data needed to include the effect in a

future update to the model.

G. Limits on the Range of an Augmentation

Fuel suppliers will be permitted to claim the emission effects of

augmentations only to the extent that the test program measured the

effects of the fuel parameter in question over the range in question.

If the parameter is included in the complex model, then the

augmentation will be valid for fuels containing levels of the parameter

between the level tested in the test program and the nearest limit of

the complex model (as described in Section IV). If the parameter is not

included in the complex model, then the augmentation will be valid for

fuels containing levels of the parameter between the candidate and

baseline levels (i.e., the levels found in Addition Fuels 1 and 3 in

Table V.1). This provision is intended to be consistent with the limits

on the application of the simple and complex models as expressed in

Sections III and IV.

H. EPA Approval, Confirmatory Testing, and Fees

In the process of reviewing a model augmentation, EPA must confirm

the accuracy of the test results. To this end, EPA intends to monitor

the petitioner's test program. The Agency also reserves the right to

perform confirmatory testing to assure the validity of the test results

and the emission performance of the reformulated fuel before allowing

augmentation of the model. EPA further reserves the right to collect

fees any lawful of an amount sufficient to recoup all costs associated

with such confirmatory testing. EPA anticipates that if any

confirmatory testing is performed that it will be of a limited nature

and focused only on those aspects of the test program which are

unexpected or contrary to prior test programs and engineering

knowledge. Since EPA has not proposed methods to be used to calculate

and collect such fees, these provisions will be handled through a

subsequent rulemaking.

I. Test Requirements

1. Winter Testing

To be certified as reformulated, a gasoline must meet the air

toxics and NOX emission requirements year-round; the oxygen,

benzene, and heavy metal content requirements year-round, and the VOC

emission requirements in the high ozone season. As discussed in Section

IV of this notice and Sections III and IV of the RIA, the Agency does

not have sufficient data to model winter exhaust emissions. While

differences between the effects of fuel parameters under summer and

winter conditions beyond those discussed in Section IV may exist, the

Agency does not have any evidence to date to suggest that they are

significant. Therefore, EPA will apply the exhaust models developed for

summer emissions to winter fuels as well for purposes of determining

their air toxics and NOX emissions. The Agency is concerned that

allowing winter testing for some fuel parameters while modeling the

effects of other parameters based on summer emission data creates the

possibility of ``gaming'' the testing process. Fuel suppliers could use

the summer model to determine the effects of parameters that would

behave unfavorably under winter conditions and use winter testing to

determine the effects of parameters that would behave favorably under

winter conditions. This possibility may result in fuels being certified

for winter use (through a combination of winter testing and summer

modeling) that result in smaller emission reductions in-use than are

intended by the Act or than would occur by using the summer model.

Therefore, EPA is at this time requiring that all testing be performed

under summer ambient conditions. As the Agency gathers additional data

in the future with which to revise the model, EPA will consider whether

sufficient winter test data exists to permit the development of winter

NOX and air toxics models. If such models can be developed, the

Agency will consider whether to allow winter testing.

2. Pollutants to be Measured

To the extent testing is performed to augment the complex model, it

must be performed to determine the emission effects on all the

pollutants covered by the reformulated gasoline certification

requirements, including toxics (carbon monoxide and carbon dioxide

emissions must also be measured to permit validation of test results).

Failure to have such a requirement might result in important emission

effects being overlooked and could allow fuel producers to ``game'' the

certification requirements by permitting them to utilize the modeling

option for one pollutant and the test results for another pollutant

when it would be advantageous. The resulting certified reformulated

gasolines may not meet all of the applicable emission reduction

requirements in-use. For example, the model augmented by test results

may indicate that a fuel meets the VOC requirement but fails the toxics

requirement, while the model alone may indicate that the fuel meets the

toxics requirement but fails the VOC requirement. Allowing the

petitioner to claim the toxics emission effects predicted by the model

while claiming VOC benefits determined through testing would ignore

fuel effects on toxics that may not be addressed by the model.

Testing costs would be significantly reduced if only VOC and

NOX emissions were measured by testing, and toxics emissions were

allowed to be modeled. However, since the testing option can only be

used when the candidate fuel's parameters fall outside of the range of

the model, EPA believes that adequate information seldom would be

available to allow toxics emissions from such fuels to be modeled

adequately if adequate information on VOC and NOX emissions were

not available. If a fuel parameter is expected to affect VOC or

NOX and is not covered by the model, toxics emissions may very

well be affected and should be measured.

It should be noted, however, measurement of toxics emissions for

the fuels used to determine interactive effects (discussed below in

section IV.I.4.) need not be performed. During development of the

complex model, EPA found that interactive effects for air toxics are

either statistically insignificant, impossible to discern given the

accuracy and extent of available data, or too small to contribute

substantially to the model's explanatory and predictive power. The

complex model being promulgated today contains no interactive terms for

air toxics emissions for these reasons, and hence EPA considers it

unnecessary to require testing for interactive effects on air toxics.

Specifically, toxics emissions need not be measured when testing

additional Extension Fuels to determine interactive effects or when

testing Addition Fuels 4, 5, 6, and 7, as described in Section V.I.5.

However, EPA reserves the right to require that toxics be measured

during vehicle testing programs when evidence exists that adverse

interactive effects may exist for toxics. In particular, EPA reserves

the right to require testing for interactive toxics effects if future

revisions to the complex model include such effects.

To better optimize the test program for the particular fuel

parameter being evaluated, the Administrator may approve a request to

waive certain pollutant measurement requirements contained in this

section. Any such waiver would have to be obtained in advance of

vehicle testing. A request for such a waiver must include an adequate

justification for the requested change, including the rationale for the

request and supporting data and information. Such a request must

justify the reason that measurement of certain pollutants clearly is

not necessary, and identify those pollutants for which additional

testing may be warranted. For example, a petition might note that

reducing the concentration of a specific high molecular weight aromatic

decreased VOC emissions even though the overall concentration of

similar aromatics remained unchanged. The petitioner may be able to

justify a reduced need for toxics measurement based on the results of

other studies which show that toxics are proportional to total

aromatics rather than to individual aromatics species. In exchange,

additional testing may be justified for VOC emissions to enable a

greater degree of statistical confidence in the test results. As a

result, the fuel supplier may be able to present EPA with sufficient

justification to warrant increased testing for VOC emissions and

decreased testing for toxics emissions.

3. Exhaust and Nonexhaust Testing

VOC and air toxics emissions occur in both exhaust and nonexhaust

emissions. However, EPA believes that the relationship between fuel

characteristics and nonexhaust emissions is known with greater

certainty and precision than the relationship between fuel

characteristics and exhaust emissions. Nonexhaust emissions are a much

simpler phenomenon to model than exhaust emissions. Nonexhaust

emissions are driven primarily by well-understood principles of

physical chemistry and are modified by devices such as charcoal

canisters that are relatively easily modeled. Exhaust emissions, by

contrast, involve combustion and catalysis reactions that are not as

well understood theoretically and are much more difficult to model. In

addition, exhaust emissions are estimated directly from the Federal

Test Procedure (FTP) utilizing the Urban Dynamometer Driving Schedule,

while nonexhaust emissions are estimated from both FTP and non-FTP test

cycles in a complex process. Finally, data on nonexhaust emissions is

much more extensive and internally consistent than data for exhaust

emissions. For these reasons, EPA is restricting testing to augment the

model to exhaust emission testing. Vehicle testing of nonexhaust

emissions will not be accepted by EPA as the basis for augmentations to

the nonexhaust emission model promulgated in today's rulemaking.

EPA reserves the right to revise the nonexhaust emission model in

the future to reflect new data acquired by the Agency, with such

revisions taking effect after the start of Phase II of the program. In

particular, either a new MOBILE model or ongoing research aimed at

modeling nonexhaust emissions as a function of true vapor pressure over

a range of temperatures may provide the basis for a revised nonexhaust

model. The nonexhaust complex model being promulgated today relies on

the Reid vapor pressure (RVP) to characterize fuels' nonexhaust

emission characteristics. However, RVP is measured at a fixed fuel

temperature (100 deg.F), while nonexhaust emissions occur over a wide

range of fuel temperatures (80 deg.F to 130 deg.F). Since different

oxygenates alter the relationship between RVP and true vapor pressure

at a given temperature to different extents, EPA believes that a model

based on true vapor pressure would be more accurate for fuels

containing oxygenates than a model based solely on RVP.

By permitting nonexhaust emissions from a given fuel to be

estimated only from models and exhaust emissions to be estimated based

in part on vehicle testing, EPA believes that the accuracy of fuel

emission estimates will be enhanced. EPA also believes that this

restriction will focus testing resources on those emission effects

which the model predicts with the least degree of certainty (i.e.,

ex

This text is long and has been trimmed here. Open the source document for the complete record.