American Coke & Coal Chemicals Institute v. Environmental Protection Agency

Notice: This opinion is subject to formal revision before publication in the
Federal Reporter or U.S.App.D.C. Reports. Users are requested to notify the
Clerk of any formal errors in order that corrections may be made before the
bound volumes go to press.

United States Court of Appeals
FOR THE DISTRICT OF COLUMBIA CIRCUIT

Argued May 4, 2006 Decided July 11, 2006

No. 03-1039

AMERICAN COKE AND COAL CHEMICALS INSTITUTE,
PETITIONER

v.

ENVIRONMENTAL PROTECTION AGENCY,
RESPONDENT

Consolidated with
No. 05-1334

On Petitions for Review of an Order of the
Environmental Protection Agency

Fredric P. Andes argued the cause and filed the briefs for
petitioner.

Eric G. Hostetler, Attorney, U.S. Department of Justice,
argued the cause for respondent. On the briefs were John C.
Cruden, Deputy Assistant Attorney General, Brian H. Lynk,
 2

Attorney, and Richard T. Witt, Counsel, U.S. Environmental
Protection Agency.

Before: HENDERSON, ROGERS and GRIFFITH, Circuit
Judges.

Opinion for the Court filed by Circuit Judge ROGERS.

ROGERS, Circuit Judge: In 2002, the Environmental
Protection Agency promulgated a final rule under the Clean
Water Act (“CWA”), 33 U.S.C. § 1251 et seq. (2000), revising
certain nationwide limitations on water pollutant discharges
from sources in the cokemaking subcategory of the iron and
steel industry.1 The American Coke and Coal Chemicals
Institute (“Institute”) challenges four effluent limitations for
existing sources that apply to the recovery of by-products from
cokemaking. The Institute contends that the final limitations
and standards are not a logical outgrowth of the proposed rule
and violate its right to comment under both the CWA and the
Administrative Procedure Act (“APA”), 5 U.S.C. § 553(c)
(2000). The Institute also contends EPA was arbitrary and
capricious. Upon review of the record, we conclude that the
final limitations are a logical outgrowth of the proposed rule,
that EPA’s determinations of the limitations are based on a
reasonable and consistently explained methodology and
supported by the record, and that EPA reasonably determined
that the limitations are achievable. Accordingly, we deny the
petitions for review.

1
See Effluent Limitations Guidelines, Pretreatment
Standards, and New Source Performance Standards for the Iron and
Steel Manufacturing Point Source Category (“Final Rule”), 67 Fed.
Reg. 64,216, 64,220 (Oct. 17, 2002) (codified at 40 C.F.R. pt. 420).
 3

I.

The CWA declares a national goal of restoring and
maintaining the chemical, physical and biological integrity of
the nation’s waters, 33 U.S.C. § 1251, and prohibits all pollutant
discharges except where specifically authorized, id. § 1311(a).
Industries “directly” discharging “toxic” or “non-conventional”
pollutants into navigable waters must treat their wastewater so
as to comply with effluent limitations set with reference to the
capabilities of the “best available technology economically
achievable” to control those discharges (“BAT”). See id. §
1342(a), (b); id. § 1311(b)(2)(A). Industries “indirectly”
discharging such pollutants by discharging wastewater into
sewage systems connected to publicly owned treatment works
(“POTWs”), which treat the wastewater prior to its introduction
into public waterways, are required to reduce the level of
pollutants through treatment prior to such discharge (or
“pretreatment”) in order “to prevent the discharge of any
pollutant through [publically owned] treatment works.” Id. §
1317(b)(1) (italics added); see generally Weyerhaeuser Co. v.
Costle, 590 F.2d 1011, 1019-20 (D.C. Cir. 1978). EPA is
required to mandate pretreatment when a pollutant “interferes
with, passes through, or otherwise is incompatible with such
works.” 33 U.S.C. § 1317(b)(1). The pretreatment standards for
existing sources (“PSES”) are set with reference to the level of
pollution-control technology available to a given industry. See
id. Although the CWA does not specify which level of control
technology is to be the applicable reference in setting the PSES
standards, EPA has adopted the position that PSES standards are
to be set with reference to technologies analogous to BAT. See
Final Rule, 67 Fed. Reg. at 64,210.
 4

The manufacture of coke2 has been subject to regulation
under the CWA since 1982.3 In 1992, EPA entered into a
consent decree arising out of litigation based on a 1987
amendment to the CWA, 33 U.S.C. § 1314(m), which requires
that EPA issue biannual plans for the annual review and revision
of existing effluent guidelines and the promulgation of certain
new guidelines. See Final Rule, 67 Fed. Reg. at 64,219. EPA
agreed to undertake final action on eleven point-source
categories and on eight categories to be later designated by the
EPA, id. After completing a preliminary study, EPA selected
the iron and steel industry as the subject of a revised effluent
rule. Id.; see EPA Preliminary Study of the Iron and Steel
Category, No. 821-R-95-037 (1995). The most common
technique for manufacturing coke creates numerous by-
products, including the four toxic or non-conventional pollutants
at issue: benzo[a]pyrene, naphthalene, cyanide, and ammonia-
N.4

2
Coke is “cooked” coal. It is a fuel – used primarily in the
steel industry – created when the “impurities” of bituminous and other
varieties of coal have been driven off by baking raw coal in an airless
oven.
3
See Iron and Steel Manufacturing Point Source Category
Effluent Limitations Guidelines, Pretreatment Standards, and New
Source Performance Standards, 47 Fed. Reg. 23,258 (May 28, 1982).
4
Benzo[a]pyrene is an aromatic hydrocarbon that is
mutagenic and highly carcinogenic. Naphthalene, best known as the
primary ingredient of mothballs, is an aromatic crystalline
hydrocarbon that easily converts to gaseous matter at room
temperature without a vacuum (i.e., it “volatizes”) and is classified by
EPA a “toxic” pollutant. See 40 C.F.R. § 401.15. Cyanides are
chemical compounds containing a carbon atom triple-bonded to a
nitrogen atom, and certain cyanides are highly toxic (e.g., hydrogen
cyanide). Ammonia-N (or ammonia as nitrogen) is an inorganic,
 5

From 1997 to 1999, EPA conducted 67 visits to iron and
steel facilities in the United States and Canada in order to collect
information on each site’s manufacturing operations, wastewater
generation, and wastewater treatment systems.5 EPA in 1998
also solicited technical and economic information relevant to
promulgation of a revised rule from various participants in the
cokemaking industry through four surveys.6 On the basis of the
site visits and surveys, EPA selected 16 sites at which to
perform wastewater sampling in order to characterize the
effectiveness of the treatment processes. See Proposed
Development Document at § 3.3. During this period, EPA
conducted a variety of outreach efforts, in which the Institute
participated, id. at § 3.5, including five stakeholders’ meetings
between 1998 and 2000 at which EPA described its preliminary
position on the model technology options and data quality
protocols, and solicited further comment and relevant data. Id.7
EPA obtained additional information from secondary sources,
including trade journals, industry databases, and studies by

dissolved form of nitrogen. Depending on other variables, high levels
of ammonia-N are often toxic for aquatic life.
5
See EPA Development Document for the Proposed Effluent
Limitations Guidelines and Standards for the Iron and Steel
Manufacturing Point Source Category (“Proposed Development
Document”) at § 3.2, No. 821-B-00-011 (December 2000).
6
See id. at § 3.1; EPA Announcement of Information
Collection Activities: 1997 Iron and Steel Industry Survey, 62 Fed.
Reg. 54,453 (Oct. 20, 1997); see also OMB Responses to EPA
Information Collection Activities, 63 Fed. Reg. 74,023, 47,023 (Sept.
3, 1998) (approving EPA’s information collection request).
7
See generally EPA Effluent Guidelines Iron & Steelmaking
Rulemaking Process, http://www.epa.gov/ost/ironsteel (last visited
June 19, 2006).
 6

governmental and private entities. See id. at § 3.4-5. Because
the operation of POTWs would be relevant to any PSES
standards, EPA solicited information from the state and local
entities responsible for POTW treatment of the indirect
discharges from the industry. Id. at § 3.6. In general, any
information that EPA collected that was not subject to protection
from disclosure as confidential business information was made
part of the public record. See id. at §§ 3.2, 3.5-6.

EPA published proposed revisions to the effluent limitations
for the iron and steelmaking industry in December 2002,
including the limitations applicable to the “by-product recovery”
segment of the cokemaking subcategory.8 With regard to the
pollutants regulated through PSES standards, EPA stated that it
contemplated using its “traditional” pass-through analysis to
determine which pollutants would be subject to pretreatment
limitations pursuant to 33 U.S.C. § 1317(b)(1). See NOPR, 65

8
See Limitations Guidelines, Pretreatment Standards, and
New Source Performance Standards for the Iron and Steel
Manufacturing Point Source Category (“NOPR”), 65 Fed. Reg.
81,964, 81,978-79 (Dec. 27, 2000). The Proposed Development
Document described how EPA would derive its effluent limitations.
The NOPR and the development document identified: (1) the
pollutants that would be the subject of the limitations, how those
pollutants would be measured, id. at 82,016; Proposed Development
Document at §§ 4, 11; (2) the pollutants considered to “pass-through”
POTWs, thus requiring pretreatment, see 33 U.S.C. § 1317(b)(1), and
EPA’s pass-through analysis, Proposed Development Document at §§
11.9, 11.10; (3) the control technologies considered as the BAT or
BAT-equivalent models, id. at §§ 12.2.1, 14, the sources of
information on the operation of the model technologies, id. at §§ 3,
12.2, and the data-selection criteria, id. at § 12.1; and, (4) how EPA
would derive discharge limitations expressed in terms of an allowable
weight of discharged pollutant per ton of coke production, id. at § 12.
 7

Fed. Reg. at 82,011.9 EPA planned to compensate for gaps in the
data, due to the impossibility of measuring certain pollutants
found in very low concentration levels, by making certain
assumptions about the value of “non-detect” datapoints. See
NOPR, 65 Fed. Reg. at 82,012. EPA planned to use only a
subset of the data on POTW performance in the 1982 POTW
Study to estimate the effectiveness of the “well-operated
POTW.” As relevant, EPA would estimate the percentages of
pollutant removed by POTW treatment based upon data from
POTWs that treated influent containing pollutants at a
concentration of ten times the minimum level that could be
detected. Id. Given that the treatment technologies generally
reduced the concentration of pollutants by at least 90%, id., the
“ten times minimum level” criteria was designed to ensure that
EPA’s estimation of the removal percentages would be
“meaningful” because it would be based upon the comparison of
two measured concentration readings (before and after
treatment). Id. Where there remained datapoints indicating that
a pollutant’s concentration was below a detectable level, EPA
substituted the minimum detectable level of concentration. Id.
In electing these data-editing criteria, EPA understood that both
might lead to biases in favor of finding that POTW treatment

9
Under this methodology, a pollutant is deemed to pass
through a POTW when the median percentage of the pollutant
removed by well-operated POTWs is less than the median percentage
removed by direct dischargers using BAT technologies to remove the
same pollutant, see NOPR, 65 Fed. Reg. at 82,011. The NOPR
identified a study of the effectiveness of POTWs dating from 1982 and
covering the survey years 1977 though 1980, see EPA Fate of Priority
Pollutants in Publicly Owned Treatment Works (“POTW Study”), No.
440/1-82/303 (Sept. 1982), as EPA’s “primary source” of data on the
operation of benchmark “well-operated POTWs.” See NOPR, 65 Fed.
Reg. at 82,011; see also Proposed Development Document at § 3.6.
 8

was either more or less effective than it actually was.10

EPA evaluated four treatment technologies as possible BAT
models for the direct discharge standards and two technological
options as the models for the PSES standard. Id. at 82,016-17.
In the NOPR, EPA identified one of the four direct discharge
treatment technologies — referred to a BAT-3 — as the model
for its proposed direct discharge standard. Id. at 82,016. EPA
“co-propos[ed]” two model treatment technologies — PSES-1
and PSES-3 — as the basis for the PSES standards. Id. at 82,
017. The Proposed Development Document identified, as the
second of EPA’s general data-selection “criteria,” that data
would be accepted as representative of the technological model
only if the “facility had . . . demonstrate[d] good operation of the
treatment component, as indicated by pollutant removals across
the treatment system and treatment system effluent quality,”
with “good operation” being determined based on such things as
site visits, survey responses, and direct measurement of the
reduction of pollutants. Proposed Development Document at §
12.1. The Proposed Development Document also stated that
“EPA intends to further review and possibly revise the data
selection methodology.” Id.

10
The “ten times minimum level” criteria potentially could
lead to an over-estimation of the amount of pollutant removed by
POTWs, because treatment technologies tend to be more effective
when they treat higher concentrations of pollutant. NOPR, 65 Fed.
Reg. at 82,013. On the other hand, the choice to assign the minimum
detectable concentration level to non-detect datapoints could lead to
an under-estimation of the effectiveness of the treatment. Id. at
82,012.
 9

After an extended comment period,11 EPA promulgated the
final rule in October 2002.12 Benzo(a)pyrene and cyanide were
subjected to direct discharge limitations. Naphthalene, cyanide
and ammonia-N were subject to indirect discharge limitations.13
EPA’s pass-through analysis for naphthalene remained
substantially unchanged.14 The procedures used to derive the
direct- and indirect-discharge effluent limitations also did not
change.15 However, the Final Rule departed from the NOPR in

11
EPA extended the initial 60-day comment period (until
April 2001) and held additional meetings with stakeholders, including
the Institute, which in February 2001 made presentations on the
economic effects of the proposed effluent limitations. 66 Fed. Reg.
17,842; see Final Development Document at § 3.6. The Institute also
submitted written comments on the proposed rule, many of which
EPA directly addressed in the EPA Comment Response Document at
20-20 through 20-45, 20-47 though 20-50, 20-61 through 20-62 (EPA
Contract No. 68-C6-0044, Work Assignment 5-01 (April 30, 2002)).
EPA also considered late-received comments. See id. at 26-1.

12
See supra note 1; EPA Development Document for Final
Effluent Limitations Guidelines and Standards for the Iron and Steel
Manufacturing Point Source Category (“Final Development
Document”), No. 821-R-02-004 (April 2002). Various other
accompanying documents explain specific decisions, such as the
exclusion of particular data-points.

13
See Final Rule, 67 Fed. Reg. at 64,233.
14
Compare Final Development Document at § 12.2.2-3, with
Proposed Development Document at 11.9-10; id. at Table 10-2.
15
Compare generally Final Development Document at § 14,
with Proposed Development Document at § 12. These procedures may
be divided into two principal stages. First, EPA derived limits on the
 10

permissible daily and monthly concentrations of specific pollutants in
effluent wastewater from data from facilities with well-designed and
operated model treatment technologies. Second, EPA converted these
limits into limitations expressed in terms of the maximum weight of
pollutant allowed per ton of coke production. Greatly simplified, EPA
derived the maximum allowed concentration by first selecting the
sampling episodes from facilities operating either BAT-1 or PSES-1
technologies and then calculating an average performance level for
each episode (the “episode-specific” long term average or “LTA”)
using either the arithmetic average or modified delta-lognormal
distribution. Then, EPA adopted the median of the episode-specific
LTAs as the “option-specific LTA” for that treatment technology.
Compare Final Development Document at § 14.8.1, with Proposed
Development Document at § 12.7.1. In order to allow for the
predictable variation in the amount of pollutant removed by the
technological option or model, the option-specific LTAs would be
multiplied by both a daily and a monthly “variability factor.” Thus,
EPA would derive two limits expressed in terms of maximum
concentration: a maximum daily limit, which was calculated as the
product of the option-specific LTA and the daily variability factor; and
a maximum monthly average limit, which was calculated as the
product of the option-specific LTA and the monthly variability factor.
In order to derive the two variability factors, EPA first calculated daily
and monthly episode-specific variability using the modified delta-
lognormal distribution and then adopted the mean of the daily and
monthly episode-specific variability factors as the daily and monthly
option-specific variability factors. Compare Final Development
Document at § 14.8.2, with Proposed Development Document at
12.7.4. Where the variability factors could not be calculated for
specific pollutants, EPA borrowed the variability factors from a
different pollutant treated by the same technological option. This was
the case with benzo(a)pyrene. Compare Final Development
Document at § 14.8.3, with Proposed Development Document at §
12.7.5. To calculate the concentration-based limits for
benzo(a)pyrene, EPA borrowed the variability factors calculated for
naphthalene treated under BAT-1, a decision it justified because
benzo(a)pyrene and naphthalene were “chemically similar” and BAT-
 11

four ways that are relevant to this appeal:

(1) EPA rejected BAT-3 as the technological model for
the direct-discharge limitations, because it was “not
economically achievable” by industry, in favor of the less
costly (and less effective) BAT-1. See Final Rule, 67 Fed.
Reg. at 64,233.

(2) EPA adopted PSES-1 as the technological model
for the PSES standards, the less costly and less effective of
the two co-proposed options.16 Id. at 64,234. EPA

1 effectiveness in treating these chemicals would exhibit comparable
variability. Final Development Document at § 14.8.3.

Having calculated the maximum daily and the monthly
average concentration limits for each pollutant, EPA then converted
these limits into daily and monthly average limits based on the weight
of pollutant allowed per ton of coke produced. This stage involved
multiplying the concentration-based limits by a “production
normalized flow rate,” by which EPA sought to capture the amount of
water typically discharged per ton of coke production, and certain
conversion factors, which would handle the translation between units.
See Final Development Document at §§ 13.2 & 14.9.1. By
multiplying the concentration limits by an amount of water
discharged, EPA could make the jump from limits expressed in terms
of average maximum concentrations to limits expressed in terms of
pounds of allowable pollutant per ton of production. See generally
Nat’l Wildlife Fed’n v. EPA, 286 F.3d 554, 571-73 (D.C. Cir. 2002)
(per curiam).
16
BAT-1 was the technological model of the previous direct
discharge effluent limitations. See NOPR, 65 Fed. Reg. at 81,986. As
relevant, BAT-1 involves two chief water treatment processes: “free
and fixed ammonia stripping” and “biological treatment with
clarification.” See Final Development Document at § 9.1.1. PSES-1
is equivalent to BAT-1 but dispenses with the biological treatment
 12

predicted that setting the pretreatment limit based upon
PSES-3 would result in plant closures and job losses. Id.
It concluded that PSES-1 represented “the most
appropriate” basis for pretreatment standards because, in
combination with treatment occurring at the POTWs,
PSES-1-based limitations will “substantially reduce the
levels of all cokemaking pollutants of concern” and that
compliance would be “economically achievable.” Id.

(3) EPA increased the production-normalized flow rate
from the proposed 158 gallons per ton of production (“gpt”)
to 163 gpt. See Final Development Document at § 13.3.1;
Proposed Development Document at § 7.3. Because the
final effluent limitation was calculated as the product of the
concentration-based limits and the production-normalized
flow rate, this increase in the flow rate had the effect (all
else remaining the same) of increasing the amount of
pollutant allowed per ton of production.

(4) EPA elected to base its evaluation of PSES-1
treatment of naphthalene upon different data than was used
in the NOPR to evaluate PSES-1. As proposed, the
reduction of naphthalene by PSES-1 treatment was
evaluated based upon a single EPA-administered sampling
episode, identified as ESE03. See Proposed Development
Document at § 12.7.4. However, in the interim, EPA
learned that the facility from which the ESE03 data derived
had a history of CWA violations, and for this reason EPA
no longer considered this episode representative of PSES-1
naphthalene treatment.17 Instead, EPA evaluated PSES-1

processes; PSES-3, in contrast, was identical to BAT-1. See id.
17
See EPA Memorandum from Yu-Ting L. Guilaran, re.
Selection of BAT Facilities for Cokemaking Subcategory for Final
 13

naphthalene treatment based upon two EPA-administered
sampling episodes, ESE01 and ESE02, and one industry
self-monitoring episode, identified as ISM54.18 The
substitution of these datasets resulted in an increased
calculation of naphthalene reduction by well-operated
facilities using PSES-1 technology, which in turn resulted
in a more stringent requirement for naphthalene reduction
in pretreatment. See Final Development Document at §
14.10.

The Institute and certain affected companies petitioned the
court to vacate the Final Rule and then sought reconsideration
by EPA while its petition was stayed. The Institute argued
before EPA that the mandated level of benzo(a)pyrene reduction
could not be achieved by BAT-1, naphthalene does not “pass
through” well-operated POTWs, and the mandated level of
naphthalene reduction could not be achieved by PSES-1. In
support of the first ground, the Institute submitted new data from
two cokemaking facilities — United States Steel Company’s
Gary Works and Clairton plants — which the Institute
represented as demonstrating the required benzo(a)pyrene
reductions using BAT-1 or technologies that go “beyond BAT[-
1]” were not achievable. In support of the second and third
grounds, the Institute challenged the methodology of EPA’s
pass-through analysis (in particular, the mathematical interplay
between the “ten times minimum level” data-selection criteria
and the assignment of 10 µg/L values to non-detect datapoints
in the POTW Study) and submitted new data purporting to
demonstrate that naphthalene did not pass through well-operated

Rule, DCN IS10816 (April 29, 2002) (memorandum contained in
section 14.10 of the record); see also Final Rule, 67 Fed. Reg. at
64,234.
18
See Final Development Document at § 14.10.
 14

POTWs and that the limitations were not achievable by PSES-1
treatment technologies. EPA denied reconsideration,19 the
Institute filed a new petition for review, and the court
consolidated the Institute’s petitions.

II.

The Institute’s central contention is that the Final Rule is
not a logical outgrowth of the NOPR because EPA failed to
provide adequate notice of the derivation of the naphthalene
limitation from three episodes or datasets — ESE01, ESE02,
ISM54 — that were not identified in the NOPR as the proposed
basis for naphthalene PSES limits, and of the exclusion of the
one dataset — ESE03 — that was identified as the source of the
proposed naphthalene limit.20 This change resulted in an
increase in the effectiveness of PSES-1 as expressed in the LTA
for naphthalene, which in turn resulted in more stringent
naphthalene pretreatment limit. See Final Rule, 67 Fed. Reg. at
64,234, 64,263. For the same reasons, the Institute contends it
was denied adequate notice with regard to the naphthalene
variability factors. Because it was deprived a fair opportunity to
comment upon the exclusion of the ESE03 data and employment
of the alternative three episodes, the Institute maintains EPA has
failed to provide adequate notice and opportunity for comment
as required by 5 U.S.C. § 553(c).

Under the APA, “[n]otice requirements are designed (1) to

19
EPA Decision on Petition for Reconsideration or, in the
Alternative, Petition to Revise the Effluent Limitations Guidelines,
Pretreatment Standards, and New Source Performance Standards for
the Iron and Steel Manufacturing Point Source Categories
(“Reconsideration”) (April 29, 2005).
20
See Proposed Development Document at § 12.2.1.
 15

ensure that agency regulations are tested via exposure to diverse
public comment, (2) to ensure fairness to affected parties, and
(3) to give affected parties an opportunity to develop evidence
in the record to support their objections to the rule and thereby
enhance the quality of judicial review.” Int’l Union, United
Mine Workers of Am. v. Mine Safety & Health Admin., 407 F.3d
1250, 1259 (D.C. Cir. 2005) (citing Small Refiner Lead
Phase-Down Task Force v. EPA, 705 F.2d 506, 547 (D.C. Cir.
1983)). “An agency satisfies the notice requirement, and need
not conduct a further round of public comment, as long as its
final rule is a ‘logical outgrowth’ of the rule it originally
proposed.” Northeast Md. Waste Disposal Auth. v. EPA, 358
F.3d 936, 951-52 (D.C. Cir. 2004). If interested parties “should
have anticipated” that the change was possible, and thus
reasonably should have filed their comments on the subject
during the notice-and-comment period, then the rule is deemed
to constitute a logical outgrowth of the proposed rule. See City
of Waukesha v. EPA, 320 F.3d 228, 245 (D.C. Cir. 2003); see
also Envtl. Integrity Project v. EPA, 425 F.3d 992, 996 (D.C.
Cir. 2005); First Am. Discount Corp., 222 F.3d 1008, 1015
(D.C. Cir. 2000); Nat’l Mining Ass’n v. Mine Safety & Heath
Admin., 116 F.3d 520, 531 (D.C. Cir. 1997). A petitioner must
demonstrate that the agency’s violation of the APA’s notice and
comment procedures has resulted in “prejudice,” 5 U.S.C. §
706(2). See generally Chamber of Commerce of the United
States v. SEC, 443 F.3d 890, 904 (D.C. Cir. 2006).

The flaw in the Institute’s logical outgrowth objection is
that the right to comment under the APA upon a proposed rule
does not extend to a right to comment upon each application of
the agency’s announced rulemaking procedures, even if different
applications may have significant consequences for the
regulated industry, where the agency gave adequate notice of the
procedures it intended to use, the criteria by which it intended to
select data, and the range of alternative sources of data it was
 16

considering. Here, the NOPR gave notice of each, and the
Institute was afforded an opportunity to comment on each. A
comparison of the rulemaking record shows that EPA made no
relevant changes to the procedures it used to derive the effluent
limitations. See supra note 15. Although the preamble in the
NOPR did not state that EPA intended to reconsider the use of
data from episode ESE03, nor that the ESE03 site would be
reviewed for compliance with the general criteria of “good
operation,” EPA made clear in the Proposed Development
Document accompanying the NOPR that the model facilities
“had to demonstrate good operation of the treatment
component,” Proposed Development Document at § 12.1, and
thus the Institute had notice that data deriving from any plants
that failed to demonstrate “good operation” would not be
considered sources of information on the operation of any
proposed BAT. EPA stated unambiguously that it “might
reconsider the exclusion of the naphthalene sampling data from
sampling episode ESE01 and ESE02 and self-monitoring
episode ISM54.” Id. at 12.2.1. Between 2000 and 2002, EPA
explained, new information indicated that the ESE03 episode
had come from a facility that was emitting pollutant “in excess
of the permit,” see Final Rule, 67 Fed. Reg. at 64,234, and thus
that the episode did not meet its announced data criteria and
could not be accepted as coming from a BAT facility.

The First Circuit, in BASF Wyandotte Corp. v. Costle, 598
F.2d 637, 642-46 (1st Cir. 1979), was similarly confronted with
a CWA rulemaking in which new data submitted during the
comment period showed that treatment technologies for
pesticides were more effective than previously indicated. The
court aptly observed:

It is perfectly predictable that new data will come in
during the comment period, either submitted by the
public with comments or collected by the agency in a
 17

continuing effort to give the regulations a more
accurate foundation. . . . If data used and disclosed for
the interim regulations presented the issues for
comment, then there is no need to seek new comment
even though significant quantitative differences result.

Id. at 644-45. Although the resulting limit was more stringent
than had been announced, the First Circuit stated “[t]he agency
should be encouraged to use such information in its final
calculations without thereby risking the requirement of a new
comment period.” Id. Similarly, the court rejected the argument
that a new round of notice and comment was required under the
APA when the agency decided in the final rule to merge
subcategories of the pesticide industry after comments argued
that the subcategories in the proposed rule, based on the
different chemical structure of the pollutant, were irrational
because data showed there were no statistically relevant
differences between the treatment of the different chemicals. Id.
at 643. The First Circuit reasoned that although commenters
apparently hoped to persuade EPA to adopt more subcategories,
EPA’s decision to abandon the subcategories altogether was not
an unforeseen consequence of objections that undermined the
premise for using subcategories. Id. Given that the commenters
“had a fair opportunity to present their views on how the
industry ought to be subcategorized” and that the choice to
merge the subcategories was a foreseeable result of the EPA’s
solicitation of comment on rationality of the subcategory
scheme, there was no failure of notice or opportunity to
comment even though parties may not have been able to predict
how the EPA would choose to act upon their comments. See id.
at 644.

We likewise conclude that EPA was not required to provide
a new opportunity for comment upon its decision to exclude the
ESE03 data based upon new information obtained after the
 18

NOPR was issued and, in its place, to use the ESE01, ESE02, &
ISM54 data that it had previously excluded subject to
reconsideration. See Proposed Development Document at §
12.2.1. EPA’s decision to exclude the ESE03 data based upon
its announced “good operation” criteria and the new information
on the ESE03 plant’s non-compliance with its permits is similar
to EPA’s decision in BASF to merge the subcategories based
upon newly submitted data showing a lack of significant
differences in the effectiveness of treatment: each decision was
based upon newly submitted information that was analyzed
under previously announced criteria, and each had the effect of
significantly altering the scope of the database used to derive the
pollutant limitation. Similarly, EPA’s decision in the final rule
to use the three additional episodes expanded the scope of the
relevant database, but in light of EPA’s statement in the
Proposed Development Document that it might reconsider its
exclusion of these other datasets, this possibility was noticed and
made available for possible comment. That these three episodes
indicated that PSES-1 was far more effective than the ESE03
results indicated was a “perfectly predictable” result of the
notice and comment process, BASF, 598 F.2d at 644. None of
the core goals of the notice and comment procedures, see United
Mine Workers of Am., 407 F.3d at 1259, were compromised by
EPA’s decisions with regard to the naphthalene data. Although
the final limitation is more stringent than had been initially
proposed, the limitation is calculated according to the
announced procedure and thus is not “surprisingly distant” from
the limitation presaged in the NOPR, see id. at 1260.

To the extent the Institute contends that the EPA provided
inadequate notice that the production-normalized flow rate
would be derived from the “best performing mills” instead of the
proposed median flow rates of all relevant facilities, where EPA
indicated that the industry median would be “applicable and
achievable,” Proposed Development Document at § 7.3, its
 19

contention fares no better. Because the flow rate adopted was
higher than the proposed flow rate (thus leading to less stringent
limitations), the Institute seems to maintain that the flow rate
should have been higher still because EPA, in explaining that
the proposed flow rate was based upon the median flow rates of
all relevant facilities, committed itself irrevocably to that
method of deriving the flow rate rather than the method adopted
in the final rule that looked only to the best performing 30% of
the relevant facilities. See Final Development Document at §
13.2. EPA gave notice that it might consider “alternative [flow
rates]” for particular manufacturing operations, and the
Proposed Development Document put parties on notice that
EPA’s flow rate would be based upon “better performing mills,”
id., which put commentators on notice that EPA was considering
an alternative to the “industry median.” See Envtl. Integrity
Project, 425 F.3d at 996; United Mine Workers of Am., 407 F.3d
at 1260. Moreover, the Institute can show no prejudice as the
flow rate in the Final Rule is less stringent than the proposed
flow rate, and this difference (all else being equal) resulted in a
less stringent limitation across the board. See Chem. Mfrs. Ass’n
v. EPA, 870 F.2d 177, 202 (5th Cir. 1989).

The Institute’s objection that EPA gave inadequate notice
of the potential for increased compliance costs if the final
limitations were stricter than those proposed in the NOPR fails
to suggest what difference further comment on costs would have
made. The Institute does not object that EPA rejected the more
effective, but also more expensive, technologies originally
proposed (BAT-3 for direct discharge; PSES-3 for
pretreatment/indirect discharge). It presents no persuasive
argument that it could not have anticipated the magnitudes of
increased costs depending on the stringency of the limitations.
The cases cited by the Institute are inapposite. In Weyerhaeuser
Co., 590 F.2d at 1028-29, EPA conceded error in calculating
costs based on new evidence. The Institute does not claim EPA
 20

has made a miscalculation. Although Shell Oil Co. v. EPA, 950
F.2d 741, 749-50 (D.C. Cir. 1991), involved increased costs
where the scope of a regulation expanded without adequate
notice, the Final Rule here produced no comparable change as
no unexpected pollutants were subjected to effluent limitations.
Moreover, in protesting the “drastic[]” increase of the costs, the
Institute ignores that compliance costs would have been higher
had EPA adopted the more effective BAT-3 and PSES-3
technologies as the BAT and BAT-equivalents, that the Institute
was given a fair opportunity to comment on the potential costs
of these more expensive technological options, and that it was
given notice of the range of costs that it now represents as
surprising. Cf. BASF, 598 F.2d at 642-44; Chem. Mfrs. Ass’n,
870 F.2d at 207.

Finally, the Institute’s contention that it was deprived of its
comment rights under the CWA, 33 U.S.C. § 1251(e); see
Waterkeeper Alliance, Inc. v. EPA, 399 F.3d 486, 502 (2d Cir.
2005), is contradicted by the record evidence of the Institute’s
extensive participation in every stage of the rulemaking. See
supra note 11.

III.

Turning to the Institute’s contentions that EPA’s
promulgation of the Final Rule was arbitrary and capricious, we
begin by observing the limited nature of the court’s review. An
agency action is “arbitrary and capricious,” 5 U.S.C. §
706(2)(A), if it “has relied on factors which Congress has not
intended it to consider, entirely failed to consider an important
aspect of the problem, offered an explanation for its decision
that runs counter to the evidence before the agency, or is so
implausible that it could not be ascribed to a difference in view
or the product of agency expertise.” Motor Vehicle Mfrs. Ass’n
of the U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29,
 21

43 (1983). The court owes particular deference to EPA when its
rulemakings rest upon matters of scientific and statistical
judgment within the agency’s sphere of special competence and
statutory jurisdiction. See, e.g., West Virginia v. EPA, 362 F.3d
861, 871 (D.C. Cir. 2004); Small Refiner Lead Phase-Down
Task Force v. EPA, 705 F.2d 506, 535 (D.C. Cir. 1983);
Kennecott v. EPA, 780 F.2d 445, 447-48 (4th Cir. 1985).

The Institute makes two objections to EPA’s conclusion
that naphthalene “passes though” well operated POTWs.21 First,
it maintains that EPA inappropriately relied upon the POTW
Study, which is over twenty years old and characterized by the
Institute as “outdated,” in coming to the conclusion that POTWs
can remove up to 95% of the naphthalene from treated
wastewater. However, EPA explained that because the
technology has not changed since the study was completed, it
remains the best source of POTW data.22 At best, the Institute
claims that later POTW data — especially with respect to

21
The Institute challenges to EPA’s decision to regulate
naphthalene on the ground that naphthalene does not satisfy EPA’s
criteria as a pollutant of concern. Even if this contention is not waived
because raised only in the reply brief, see, e.g., Charter Oil Co. v. Am.
Employers’ Ins. Co., 69 F.3d 1160, 1170-71 (D.C. Cir. 1995), there is
no indication that this challenge was raised before EPA, and so it is
waived, see Amgen, Inc. v. Smith, 357 F.3d 103, 117 (D.C. Cir. 2004);
Nat’l Wildlife, 286 F.3d at 562. In any event, the NOPR generally
identified as pollutants of concern “numerous . . . volatile organic
compounds and polynuclear aromatic compounds” that were released
into byproduct wastewater. Naphthalene is a volatile organic and an
aromatic polynuclear compound.
22
The Institute’s claim that EPA ignored later POTW data is
incorrect. See Proposed Development Document at § 11.9.1; Final
Development Document at § 12.2.2; see also EPA Comment Response
Document at 13-25.
 22

POTW treatment of high-naphthalene concentration influent —
more accurately represents the effectiveness of typical POTW
treatment of naphthalene in cokemaking wastewater. But EPA’s
position that the CWA requires it to consider POTW treatment
capabilities on a nationwide basis, see EPA Comment Response
Document at 13-25, is a permissible interpretation of Section
307 of the CWA, 33 U.S.C. § 1317(b)(1), which requires EPA
to promulgate national standards for pretreatment. See Chevron
USA, Inc. v. Nat. Res. Def. Council, 467 U.S. 837, 844-45
(1984). Section 307 refers to POTWs in the most general sense
with no qualifications that would suggest it would be
unreasonable for EPA to understand Congress to intend a
uniform standard based on the average, not exceptionally
effective POTW.

Second, the Institute challenges EPA’s pass-through
analysis. The Institute’s claim that there is no record evidence
that naphthalene actually passes through POTWs into navigable
waters is incorrect. Although naphthalene, as a highly volatile
compound, often escapes from wastewater directly into the air
and the level of naphthalene after POTW treatment is at a non-
detectable level, the record indicates naphthalene is sometimes
present even after POTW treatment. Cf. NOPR, 65 Fed. Reg. at
82,012-13; Proposed Development Document at § 11.9.1.
However, the Institute also challenges EPA’s assignment of a
substitution value of 10 µg/L to non-detect results when that
assumption, by EPA’s own acknowledgment, could have
resulted in an underestimate of the naphthalene removal
efficiency of POTWs relative to PSES-1, and thus might have
increased the likelihood that EPA would find that naphthalene
“passes-through” POTWs. EPA acknowledged the possibility
of bias as a result of the assignment of the 10 µg/L level to a
non-detect result. However, EPA also recognized that the use
of the “ten times minimum level” criteria may have biased the
calculations the other way. See supra note 10. Given the
 23

limitations of record data on the effectiveness of PSES-1 and
POTWs at removing naphthalene, EPA was confronted with a
situation in which it was not possible, given the current state of
technology, to establish conclusively the relative effectiveness
of POTW and PSES-1 treatment. Nonetheless, the Institute
points to nothing that would indicate that EPA’s decision to
adopt the ten-times minimum level criteria and to assign the 10
µg/L value to nondetect results, and to base its pass-though
analysis upon calculations affected by these choices, was an
unreasonable way to implement the statutory command to
regulate pollutants that pass-through POTWs. See 33 U.S.C. §
1317(b); Chem. Mfrs., 870 F.2d at 244; see also Reynolds
Metals Co. v. EPA, 760 F.2d 549, 553, 567 (4th Cir. 1985).
Such decisions involve expert statistical and scientific
judgments to which this court properly defers. See Penn. Dept.
of Envt’l Prot. v. EPA, 429 F.3d 1125, 1128-29 (D.C. Cir. 2005);
Int’l Fabricare Inst. v. EPA, 972 F.2d 384, 389 (D.C. Cir.
1992).

The Institute also challenges the rationality of the effluent
limits in the Final Rule. These challenges fail.

Benzo(a)pyrene. The Institute’s challenge that the final
limit is arbitrary and capricious because EPA failed to identify
any BAT-1 facility that was capable of meeting this limit is
based solely on Tanners’ Council of America, Inc. v. Train, 540
F.2d 1188, 1192-93 (4th Cir. 1976), which is distinguishable in
one key respect: unlike the instant rulemaking, in Tanners’, EPA
did not develop the limitations based upon data from the
regulated industry, but instead relied upon data from another
industry without an adequate explanation of how the regulated
industry would be able to adopt the treatment technology
identified as the relevant model technology. Id. at 1192. Here,
EPA not only developed the limitation based upon data from
existing BAT-1 facilities treating benzo(a)pyrene in cokemaking
 24

wastewater, but EPA took steps to ensure that its concentration-
based limitations did not result in a limit that overstated the
capabilities of the BAT-1 technology. For example, after
deriving the maximum daily concentration limit for
benzo(a)pyrene, EPA cross-checked the limitation against the
raw data in order to verify that the statistical assumptions
provided a reasonable “fit” with the data. See Final
Development Document at § 14.8.4. EPA thus took reasonable
steps to ensure that the limit represents BAT-1 treatment
capabilities and its choices are not arbitrary or capricious. See
Kennecott, 780 F.2d at 450-51; Reynolds Metals Co. v. EPA,
760 F.2d 549, 560 (4th Cir. 1985); cf. Chem. Mfns. Ass’n v.
EPA, 28 F.3d 1259, 1265-66 (D.C. Cir. 1994).

To the extent the Institute suggests EPA must be able to
point to data from a specific facility showing no instances where
the benzo(a)pyrene limit was ever breached, it cites no authority
for this proposition. Given that the purpose of the CWA is “to
achieve the complete elimination of all discharges of pollutants
into the nation’s waters,” Am. Frozen Food Inst. v. Train, 539
F.2d 107, 124 (D.C. Cir. 1976), EPA may surely make statistical
choices about how to treat data based on the capabilities of the
model technologies with this goal in mind, cf. Natural Res.
Defense Council, Inc. v. EPA, 859 F.2d 156, 201-202 (D.C. Cir.
1988); Kennecott, 780 F.2d at 450-51; Reynolds Metals, 760
F.2d at 558-59. Although EPA’s decision to develop the daily
and monthly concentration limits by employing the “ten times
minimum level” data-selection criteria, and then to aggregate
data from multiple facilities as either a single mean or median
figure, see supra note 15, could have produced a limit that no
BAT-1 facility actually achieved, the choice of effluent data-
editing criteria and statistical methodology falls within EPA’s
special area of expertise. EPA’s determinations do not appear
arbitrary or capricious in light of the technological limitations on
measuring very low concentration levels and EPA’s reasonable
 25

explanations of its procedures as a means to capture more
accurately the capability of BAT-1 in the general run of
applications to the cokemaking industry. Cf. Kennecott, 780
F.2d at 450-51; Reynolds Metals, 760 F.2d at 558-59.23

Naphthalene. For the reasons previously discussed
regarding benzo(a)pyrene, the Institute’s contention that EPA
failed to identify any PSES-1 facility that was capable of
meeting the indirect discharge limitation fails. The notion that
EPA was capricious by departing from its position that, in
general, it would not regulate volatile pollutants misreads the
NOPR, which did not state that EPA will not regulate volatile
pollutants, such as naphthalene, see 65 Fed. Reg. at 82,010, but
rather generally observed that certain pollutants need not be
monitored because many pollutants may be surrogates for
others. In any event, the NOPR identified naphthalene as a
pollutant in cokemaking wastewater, 65 Fed. Reg at 81,975, that
might be subject to the effluent limitations.

Also without merit is the Institute’s contention that EPA
acted capriciously in promulgating the naphthalene limitation
because it had previously justified its decision to regulate
naphthalene as an indicator of the effectiveness of biological
treatment, which was a component of the co-proposed PSES-3

23
The Institute has waived its challenge to EPA’s decision to
adopt the naphthalene variability factors as the benzo(a)pyrene
variability factors because, although the decision was set forth in the
Proposed Development Document at § 12.7.5 (Table 5), the Institute
did not claim during the administrative proceedings that
benzo(a)pyrene has a different variability than naphthalene, or that the
chemical structure of the two pollutants would lead one to expect any
difference in their variability. See Nat’l Wildlife, 286 F.3d at 562;
Appalachian Power Co. v. EPA, 251 F.3d 1026, 1036 (D.C. Cir.
2001).
 26

model. According to the Institute, because EPA decided in the
Final Rule to model indirect discharge limits upon PSES-1
technology, which does not employ a biological treatment
component, its justification for the regulation of naphthalene
was obviated. By this argument, the Institute would suggest that
by originally justifying regulation of naphthalene as an indicator
of the effectiveness of biological treatment, EPA had forever
committed itself to this justification as the sole legally sufficient
basis for its action. This is incorrect. The NOPR gave notice
naphthalene was potentially subject to PSES limitations. The
record evidence supports EPA’s conclusion that the employment
of the PSES-1 treatment technology could significantly reduce
naphthalene discharge into POTWs. Hence, EPA did not act
capriciously by promulgating its naphthalene limitations even
though it had originally justified its choice to regulate
naphthalene on other grounds.24

Cyanide. The Institute repeats its incapability challenge by
arguing that the model BAT-1 facilities used by EPA were not
able to achieve the cyanide reduction required by the Final Rule.
The Institute relies on Tanners’, 540 F.2d at 1193, and points to
data from the model facility from which the two relevant
episodes (ESE01 & ISM50) were drawn — United States Steel
Company’s Clairton Works plant — showing the facility was

24
The Institute claimed that nothing in the record indicated
that PSES-1 technology was capable of reducing naphthalene, and
thus EPA would be capricious in promulgating a rule requiring
naphthalene reductions modeled upon a BAT that had no effect upon
naphthalene. However, as clarified during oral argument and in
supplemental filings in response to the court’s order, the record shows
EPA had data that unambiguously indicated that PSES-1 reduced
naphthalene. For example, a 1992 study in the rulemaking record, but
not initially in the record filed in this court, indicates that ammonia
stripping was extremely effective in reducing naphthalene in
cokemaking wastewater.
 27

not able “consistently” to avoid exceeding the maximum
cyanide limit. Many of these objections fail because they are
predicated either upon data that was never submitted to EPA or
that does not accord with the data that before EPA; with regard
to the data that remains, EPA reasonably explained that certain
anomalous information need not be the basis for the final
cyanide limitation.

It is well established that “the focal point for judicial
review” under the 5 U.S.C. § 706(2)(A) “should be the
administrative record already in existence, not some new record
made initially in the reviewing court.” Camp v. Pitts, 411 U.S.
138, 142 (1973). After excluding from consideration data not
before EPA, only one Clairton self-monitoring episode (ISM50)
is at issue. EPA points out that although the ISM50 episode
indicates that 7 out of 265 days recorded indicate that the
Clairton facility exceeded the daily cyanide concentration limit,
four of the data-points coincide with the failure of the plant’s
gas handling and chemical recovery system. The remaining
three samples represent only 1.15% of the relevant dataset.
Given that the maximum daily limit was intended by the EPA to
reflect the 99th percentile distributional basis of the actual
results of a well-operated plant employing BAT-1, see Final
Development Document at § 14.6.2, EPA reasonably concluded
that these three ISM50 data-points cannot be taken to indicate
that Clairton is not capable of achieving the cyanide limitations
at all times. Similarly, according to EPA, because the maximum
monthly average limitation was intended to represent the 95th
percentile distribution, and the three instances in which the
monthly average fall outside of this distribution, the monthly
average data fails to indicate that Clairton is incapable of
achieving the cyanide limitations.

The court will not second-guess EPA’s expertise with
regard to what the maximum effluent limits represent. See Nat’l
 28

Wildlife, 286 F.3d at 571-73. As EPA explains in the Final
Development Document, the daily and monthly average effluent
limitations are not promulgated with the expectation that a plant
will operate with an eye toward barely achieving the limitations.
Final Development Document at § 14.6.2. Should a plant do so,
it could be expected to exceed these limits frequently because of
the foreseeable variation in treatment effectiveness. Rather, the
effluent limitations are promulgated with the expectation that
plants will be operated with an eye towards achieving the
equivalent of the LTA for the BAT-1 model technology. Id.
However, even operated with the goal of achieving the BAT-1
LTA, a plant’s actual results will vary. EPA’s maximum daily
limitations are designed to be forgiving enough to cover the
operations of a well-operated model facility 99% of the time,
while its maximum monthly average limitations are designed to
be forgiving enough to accommodate the operations of a well-
operated model facility 95% of the time. See id. EPA’s choice
of percentile distribution represented by its maximum effluent
limitation under the CWA represents an expert policy judgment
that is not arbitrary or capricious. See Nat’l Wildlife, 286 F.3d
at 573; Kennecott, 780 F.2d at 449.25

25
The Institute also maintains that, when calculated according
to the weight rather than the concentration of the pollutant discharged,
there are actually nine instances where Clairton exceeded the cyanide
limitations in the ISM50 dataset; see Grey Appendix D.1 & D.2]; such
a difference is of little moment. The Institute appears implicitly to
assume that at the time the daily samples were taken the Clairton
facility was operating at a particular flow rate. The data does not
appear to suggest anything of the sort, and the calculation of the
“nine” instances in which Clairton may have exceeded the daily mass
limitations is founded upon the application of flow rates that are,
presumably, either averages for Clairton or else averages for similar
facilities. In either event, the Institute fails to explain sufficiently and
substantiate its objections on this score.
 29

Ammonia-N. The Institute also contends that the ammonia-
N PSES limitations are not achievable by PSES-1 technology.
Citing both Tanners’, 540 F.2d at 1193, which is inapposite, and
Chemical Manufacturers, 28 F.3d at 1265-66, the Institute
contends that EPA promulgated the ammonia-N pretreatment
limit on the basis of an industry data, ISM54, that demonstrated
the ammonia-N limit being breached multiple times. In
Chemical Manufacturers, the court reviewed an EPA model for
determining an exposure level above which a hazardous air
pollutant may be said to pose a high risk. Id. at 1264.
Observing that “the normal criterion by which to evaluate a
model is not the accuracy of the assumptions from which it
proceeds but the utility of the results it produces,” id. at 1265,
the court noted that EPA “need not justify the model on an ad
hoc basis for every chemical to which the model is applied, even
when faced with data indicating that it is not a perfect fit,” id.
The court will “reverse the agency’s application of the generic
air dispersion model as arbitrary and capricious if there is
simply no rational relationship between the model and the
known behavior of the hazardous air pollutant to which it is
applied.” Id. (citing State Farm, 463 U.S. at 43). In that case,
the court found no rational relationship between the data and the
air dispersion model when EPA was faced with “specific
scientific evidence” demonstrating that the model fundamentally
failed to represent the physical properties of the chemical at
issue, and where EPA sought to explain away any discrepancies
with unsubstantiated speculation and conclusory assertions of
fact.

Chemical Manufacturers is apposite because the cumulative
effect of the data editing criteria, the assignment of a 10 µg/l
level to nondetect results, and the various statistical choices
made in the Final Rule may be treated as analogous to the
generic air dispersion model examined in that case; in both
instances, there is a risk that the conceptual tools used to predict
 30

the behavior of a range of physical processes might, in a
particular instance, fail so significantly that the conceptual tools
must be revised in the face of reality. Here, EPA recognized as
much by specifically incorporating a cross-check of the
concentration limits against the data used to derive those limits,
see Final Development Document at § 14.8.4. As the court had
observed, a “perfect fit” is not required, see Chemical
Manufacturers, 28 F.3d at 1265, and deference to EPA’s
judgment is appropriate where minor differences between the
record data and the agency’s representation of that data may be
explained as matters of judgment as to the statistical relevancy
of apparently anomalous information, see Nat’l Wildlife, 286
F.3d at 573; Kennecott, 780 F.2d at 450-51; Reynolds Metals,
760 F.2d at 559.

The Institute does not question whether EPA’s analysis of
the ISM54 ammonia-N data bears a rational relationship to the
data before the agency. Again, the Institute relies solely upon
the argument that — calculated according to weight of
ammonia-N discharged — the ISM54 data indicate three
instances where samples exceeded the daily maximum and six
instances in which the maximum monthly average limitation
would have been breached. As explained supra note 24, there
is no reason the court would take on faith the Institute’s
calculation of the weight of ammonia-N discharged based upon
data in the record indicating only the concentration of ammonia-
N in cokemaking effluent, where such a calculation rests upon
unexplained and unsubstantiated assumptions about flow rates.
Unlike the situation in Chemical Manufacturers, 28 F.3d at
1266, it is not the agency but the Institute that relies upon
unsubstantiated speculation in order to make the record data fit
its perception of reality. EPA indicated as part of the
rulemaking record that the concentration of ammonia-N never
exceeded the maximum daily limit in the relevant ISM54 data,
and the monthly limit was exceeded only once out of the twelve
 31

monthly averages represented in the dataset. The single monthly
average above the maximum limitations represents 8.33% of the
monthly average data. Thus, the limit might be said to represent
the 92d percentile distributional basis of the ISM54 data.
Although the maximum monthly limit does not quite represent
the 95th percentile distributional that was EPA’s goal in
establishing this limit, see Final Development Document at §
14.6.2, the maximum average monthly limitation “fits” the
ISM54 data, see Chemical Manufacturers, 28 F.3d at 1264; the
small deviation between the distribution of the data and EPA’s
distributional goal is insufficient to cause the court to second-
guess EPA’s expert judgment as to the relevancy of a single
potentially anomalous instance.

IV.

Finally, the Institute contends that, in light of the additional
information it submitted after the rulemaking record closed
relevant to the benzo(a)pyrene and naphthalene limitations, EPA
erred in denying reconsideration. Given EPA’s decision on
reconsideration to address the Institute’s claims on the merits,
we review the Institute’s contentions pursuant to 5 U.S.C. §
706(2)(A); see Prof’l Pilots Fed’n v. FAA, 118 F.3d 758, 763
(D.C. Cir. 1997), and find them lacking.

Benzo(a)pyrene. The Institute challenged the
benzo(a)pyrene limitation on the basis of new data from United
States Steel Company’s Gary Works, a facility employing
technologies very similar to BAT-1, and Clairton plants. EPA
pointed out that Gary Works does not use BAT-1 technology
and therefore the data from this facility is irrelevant to
limitations in the Final Rule based on BAT-1 technology. EPA
described differences between Gary Works’ system and BAT-1
systems, such as Gary Works’ biological treatment, which
combines into one “unit” or piece of equipment the biological
 32

treatment and clarification stages, where in BAT-1 these stages
are accomplished in separate units. Nonetheless, the Institute
maintains that because both BAT-1 and Gary Works use an
“activated sludge treatment” followed by aerobic nitrification
(no matter how these steps are technically accomplished) the
Gary Works data is decisive evidence that BAT-1 cannot
achieve the benzo(a)pyrene limitations. This challenge fails in
light of EPA’s plausible distinction between the BAT-1 and
Gary Works systems.

As to the few times when Clairton exceeded the daily and
monthly concentration-based limits in two three-month periods
in 1999 and 2001, EPA explained that the high 1999 numbers
resulted from unusually high rainfall in Pittsburgh and surmised
that similar causes lay at the root of the high summer 2001
numbers. By contrast, the self-monitoring data from the
following three years (fall 2001 to fall 2004) established that
Clairton consistently met the limitations. With regard to other
instances where the Clairton facility occasionally exceeded the
benzo(a)pyrene limits, EPA reasonably maintains that these
instances would fall into the 99th percentile distribution and do
not of themselves upset the limit established. Cf. Nat’l Wildlife,
286 F.3d at 572.

The Institute also contends that the EPA arbitrarily and
capriciously refused to reconsider its benzo(a)pyrene limitation
in light of the newly-submitted self-monitoring data from Gary
Works and Clairton that, when combined with existing record
data, would have allowed EPA to calculate a benzo(a)pyrene
variability factor directly and avoid borrowing the naphthalene
variability factor for the benzo(a)pyrene analysis. Assuming the
Institute’s additional data would have allowed a direct
calculation of the benzo(a)pyrene variability factor, the Institute
could show “prejudicial error,” 5 U.S.C. § 706(2); see Chamber
of Commerce, 443 F.3d at 904, only if it could plausibly show
 33

that a directly calculated benzo(a)pyrene variability factor would
be higher than the naphthalene variability factor; only a higher
variability factor would result in a less stringent benzo(a)pyrene
limitation. The Institute makes no such showing and EPA
concluded that the extra-record evidence indicated that
benzo(a)pyrene’s variability might be lower.

Naphthalene. The Institute’s renewed challenge to EPA’s
pass-through analysis fails because it is based on data from a
facility that does not employ the PSES-1 technology, and thus
can be reasonably excluded from consideration by EPA. While
the Institute maintains that EPA’s pass-through analysis was
illegitimate because there was no statistically significant
difference between the removal of naphthalene in well-operated
POTWs and naphthalene removal by PSES-1 facilities, given
that high naphthalene levels were treated by the PSES-1
facilities and relatively low naphthalene levels were treated by
POTWs, EPA points to various errors in the Institute’s
arguments and, regardless, this dispute involves EPA’s expert
judgments about statistical relevance to which the court properly
defers. Similarly, whether the median value of a set of averages
may be treated as representative of the set is a choice that falls
squarely within EPA’s area of expertise. The Institute’s focus
on EPA’s decision not to regulate phenol, which also passes into
POTWs is, at best, weakly relevant where other cost factors and
technological capabilities would weigh in EPA’s evaluation.
EPA has taken the reasonable position that, unlike naphthalene,
phenols are easily biodegraded and hence are more readily
treated by POTWs (which use biological treatment processes).26

26
See EPA Comment Response Document at 13-3; Effluent
Limitations Guidelines, Pretreatment Standards, and New Source
Performance Standards for the Iron and Steel Manufacturing Point
Source Category, Extension of Comment Period, 66 Fed. Reg. 10,253,
10,257 (Feb. 21 2001).
 34

See Chem. Mfrs., 870 F.2d at 247.

Accordingly, we deny the petitions for review.