Catawba County v. Environmental Protection Agency

United States Court of Appeals
FOR THE DISTRICT OF COLUMBIA CIRCUIT

Argued April 6, 2009 Decided July 7, 2009

No. 05-1064

CATAWBA COUNTY, NORTH CAROLINA, ET AL.,
PETITIONERS

v.

ENVIRONMENTAL PROTECTION AGENCY,
RESPONDENT

SIERRA CLUB,
INTERVENOR

Consolidated with 05-1065, 05-1067, 05-1068, 05-1069,
05-1071, 05-1072, 05-1073, 05-1075, 05-1076, 05-1077,
05-1078, 05-1184, 05-1190, 05-1196, 05-1200, 05-1202,
06-1049, 06-1052, 06-1083, 06-1088, 06-1102, 06-1172,
07-1412, 07-1417, 07-1418, 07-1428, 07-1465, 07-1467,
07-1530

On Petitions for Review of Final Actions
of the Environmental Protection Agency

Marc D. Machlin argued the cause and filed the briefs for
petitioner Oakland County, Michigan.
 2
Louis E. Tosi argued the cause for Industry Petitioners.
With him on the briefs were Paul E. Gutermann, Charles L.
Franklin, Douglas A. McWilliams, Allen A. Kacenjar,
Michael E. Born, Cheri Ann Budzynski, Gale Lea Rubrecht,
Kathy G. Beckett, David M. Flannery, and Edward L. Kropp.

Andrew M. Cuomo, Attorney General, Attorney General=s
Office of the State of New York, Michael J. Myers and Jacob
Hollinger, Assistant Attorney Generals, Mark Rudolph,
Senior Counsel, Office of Legal Services, West Virginia,
Steve Carter, Attorney General, Attorney General=s Office of
the State of Indiana, Steve D. Griffin and Valerie Tachtiris,
Deputy Attorneys General, and Thomas M. Fisher, Special
Counsel, were on the briefs of State Petitioners.

Karma Barsam Brown, Phillip L. Conner, Ronald E.
Cardwell, Ethan R. Ware, George William House, S. Kyle
Woosley, and Lewis S. Wiener were on the briefs for County
Petitioners.

Laurel A. Bedig and Monica Derbes Gibson, Attorneys,
and Jon M. Lipshultz, Senior Counsel, U.S. Department of
Justice, argued the cause for respondent. With them on the
brief were John C. Cruden, Deputy Assistant Attorney
General, Jessica O’Donnell, Attorney, and Geoffrey L.
Wilcox, Counsel, U.S. Environmental Protection Agency.
Kenneth C. Amaditz, Attorney, U.S. Department of Justice,
entered an appearance.

David S. Baron and Jennifer C. Chavez were on the brief
for intervenor. Howard I. Fox entered an appearance.

Before: TATEL, GARLAND, and GRIFFITH, Circuit Judges.

Opinion for the Court filed PER CURIAM.
 3
PER CURIAM: In these consolidated petitions, several
states, counties, and industrial entities challenge the
Environmental Protection Agency’s promulgation of area
designations for the annual national ambient air quality
standard applicable to fine particulate matter, a category of air
pollutants consisting of miniscule airborne particles known
to present adverse health risks. Insisting that EPA’s
methodology for designating areas as “nonattainment” for the
fine particulate matter standard violates section 107(d) of the
Clean Air Act, which governs such designations, and that this
methodology and the individual designations it produced are
otherwise arbitrary and capricious, petitioners ask us to vacate
the nonattainment designations and to send EPA back to the
drawing board. With one minor exception, we deny the
petitions for review. Faced with the complex task of
identifying those geographic areas that contribute to fine
particulate matter pollution, EPA both complied with the
statute and, for all but one of the 225 counties or partial
counties it designated as nonattainment, satisfied—indeed,
quite often surpassed—its basic obligation of reasoned
decisionmaking.

I.
Title I of the Clean Air Act charges EPA with
formulating national ambient air quality standards (NAAQS)
for air pollutants that may reasonably be anticipated to
endanger public health and welfare. 42 U.S.C. §§ 7408–09.
NAAQS set maximum ambient air concentrations for those
pollutants. Id. While each state has “primary responsibility
for assuring air quality” within its borders and, in particular,
for developing a state implementation plan (SIP) for
achieving and maintaining the NAAQS for each air pollutant,
42 U.S.C. § 7407(a), the Act triggers more or less stringent
requirements depending on the quality of an area’s ambient
air. Thus, before a state can design an appropriate SIP, it
 4
must know which areas within its boundaries comply with the
NAAQS and which do not.

This is where CAA section 107(d) comes in. It requires
EPA to designate areas as “attainment,” “nonattainment,” or
“unclassifiable” depending on their compliance with the
relevant NAAQS. “Attainment” areas are those that meet the
relevant standard; “nonattainment” areas are those that exceed
the standard or that “contribute[] to ambient air quality in a
nearby area” that exceeds the standard; “unclassifiable” areas
are those that permit no determination given existing data.
§ 7407(d)(1)(A)(i)–(iii). In nonattainment areas, the Act
requires stricter pollution controls. For instance, states must
implement controls that will achieve attainment “as
expeditiously as practicable” in nonattainment areas,
id. § 7502(a), (c)(1), whereas states need only implement
measures that will prevent “significant deterioration of air
quality” for attainment and unclassifiable areas, id. § 7471.

In addition to setting the criteria for attainment and
nonattainment, section 107(d)(1) prescribes the designation
process. Upon promulgation of new or revised NAAQS,
states must submit to EPA their own “initial designations” of
all areas within their borders. § 7407(d)(1)(A). EPA must
then promulgate the submitted designations or modify them as
it deems necessary. § 7407(d)(1)(B). Specifically, CAA
section 107(d)(1)(B)(ii) provides that:

In making the promulgations required . . . , the
Administrator may make such modifications
as the Administrator deems necessary to
the designations of the areas (or portions
thereof) submitted [by the states] under
subparagraph (A) (including to the boundaries
of such areas or portions thereof). Whenever
 5
the Administrator intends to make a
modification, the Administrator shall notify
the State and provide such State with
an opportunity to demonstrate why any
proposed modification is inappropriate. The
Administrator shall give such notification no
later than 120 days before the date the
Administrator promulgates the designation,
including any modification thereto.

§ 7407(d)(1)(B)(ii).

This case involves the NAAQS for fine particulate
matter. Known as PM2.5, fine particulate matter consists of
airborne particles that are 2.5 micrometers in diameter or
smaller—i.e., less than one-thirtieth the thickness of a human
hair. Air Quality Designations and Classifications for the
Fine Particles (PM2.5) National Ambient Air Quality
Standards (“PM2.5 Designations Rule”), 70 Fed. Reg. 944,
945 (Jan. 5, 2005) (codified at 40 C.F.R. pt. 81). A
“significant association” links elevated levels of PM2.5 with
adverse human health consequences such as premature death,
lung and cardiovascular disease, and asthma. Id. And
significantly for the primary issue before us—EPA’s method
for identifying the geographic origins of elevated ambient
PM2.5 concentrations—PM2.5 can travel hundreds or thousands
of miles.

In 1997, EPA abandoned its practice of regulating all
particulate matter, both coarse and fine, under a unified
standard. Instead it established specific PM2.5 NAAQS for the
first time. National Ambient Air Quality Standards for
Particulate Matter, 62 Fed. Reg. 38,652 (July 18, 1997).
EPA promulgated annual and 24-hour PM2.5 NAAQS, setting
the annual standard—the one at issue here—at 15 micrograms
 6
per cubic meter. Id. at 38,677; see also 40 C.F.R. § 50.7.
Although section 107(d) required EPA to promulgate area
designations for the new standard “as expeditiously as
possible,” § 7407(d)(1)(B)(i), litigation here and in the
Supreme Court waylaid the designation process until we
finally upheld the standard five years later. See Am. Trucking
Ass’ns v. EPA, 283 F.3d 355 (D.C. Cir. 2002), on remand
from Whitman v. Am. Trucking Ass’ns, 531 U.S. 457 (2001),
aff’g in part and rev’g in part Am. Trucking Ass’ns v. EPA,
175 F.3d 1027 (D.C. Cir. 1999). In the meantime, Congress
had passed legislation requiring the deployment of a
nationwide PM2.5 monitoring network and extending the
deadline for the designations until three years of monitoring
data had been collected. See Transportation Equity Act for
the 21st Century (“TEA-21”), Pub. L. No. 105-178, §§ 6101–
02, 112 Stat. 107, 463–65 (1998). Then, following our
decision upholding the PM2.5 NAAQS and EPA’s initiation of
the designation process, Congress amended the Clean Air Act
to add section 107(d)(6), which set a firm deadline for the
PM2.5 area designations. See Pub. L. No. 108-199, § 425(a),
118 Stat. 3, 417 (2004) (codified at § 7407(d)(6)). Thus,
amended section 107(d)(6)(A) now provides:

Notwithstanding any other provision of law,
not later than February 15, 2004, the Governor
of each State shall submit designations referred
in paragraph (1) for the July 1997 PM2.5
national ambient air quality standards for each
area within the State, based on air quality
monitoring data collected in accordance with
any applicable Federal reference methods for
the relevant area.

§ 7407(d)(6)(A). And section 107(d)(6)(B) provides:
 7
Notwithstanding any other provision of law,
not later than December 31, 2005, the
Administrator shall, consistent with paragraph
(1), promulgate the designations referred to in
subparagraph (A) for each area of each State
for the July 1997 PM2.5 national ambient air
quality standards.

§ 7407(d)(6)(B).

In April 2003, EPA issued a guidance document initiating
the PM2.5 designation process. Known as the Holmstead
Memo, the document explains the timeline for state
submissions and, more importantly, the criteria EPA would
employ in reviewing those submissions. Memorandum from
Jeffrey R. Holmstead, Assistant Administrator, to Regional
Administrators, Regions I–X (Apr. 1, 2003) (“Holmstead
Memo”). Noting “recent evidence that violations of the PM2.5
air quality standards generally include a significant urban-
scale contribution,” the Holmstead Memo explains EPA’s
“inten[t] to apply a presumption that the boundaries for urban
nonattainment areas should be based on Metropolitan Area
boundaries.” Id. at 2. In other words, the Memo announces
EPA’s view that if any area within a metropolitan area
exceeds the annual PM2.5 NAAQS, then all areas within the
metropolitan area presumptively “contribute” to that violation
within the meaning of section 107(d)(1)(A) and therefore
warrant “nonattainment” designations. The Holmstead Memo
further explains that EPA would define metropolitan
boundaries by reference to Office of Management and Budget
definitions of metropolitan statistical areas (MSAs) and
consolidated metropolitan statistical areas (CMSAs)
(interchangeably, “C/MSAs”). Id. Thus, under this approach,
a violation of the PM2.5 NAAQS in the District of Columbia,
for instance, would trigger the presumption that seventeen
 8
counties in northern Virginia and five counties in Maryland—
all inside the applicable MSA, though only some are
contiguous with the District—contribute to elevated PM2.5
levels in the city and warrant “nonattainment” designations.

An attachment to the Holmstead Memo elaborates on
EPA’s basis for adopting the presumption, as well as the
circumstances that would warrant a departure. As to the
rationale for the presumption, the Holmstead Memo explains
that after “examin[ing] the geographic distribution of total
PM2.5 concentrations in and near many metropolitan areas,”
the agency had “found an association of higher PM2.5
concentrations with greater levels of urban activity” such as
“motor vehicle use and home heating[,] as well as industrial
activities.” Holmstead Memo, Attach. 2, Guidance on
Nonattainment Area Designations for PM2.5, at 4–5
(“Holmstead Memo Guidance”). Thus, “[the] presumption
reflects EPA’s view that, in the absence of evidence to the
contrary, violations of the PM2.5 NAAQS in urban areas may
be presumed attributable at least in part to contributions from
sources distributed throughout the Metropolitan Area.” Id. at
5. That said, the Holmstead Memo also recognizes that
appropriate boundaries of urban nonattainment areas may
well be smaller or larger than the applicable C/MSA. EPA
would therefore “consider requests for urban nonattainment
area definitions that deviate from OMB’s metropolitan area
definitions on a case-by-case basis.” Id. at 6. The Holmstead
Memo lists nine factors to guide that case-by-case analysis:
(1) emissions in the potentially contributing areas; (2) air
quality in those areas; (3) population density and degree of
urbanization in those areas; (4) traffic and commuting
patterns; (5) expected growth; (6) meteorology; (7) geography
and topography; (8) jurisdictional boundaries; and (9) level of
control of emissions sources. Id. at 7. The Memo encourages
states submitting designations that depart from the
 9
metropolitan presumption to justify such designations by
reference to all nine factors. Id.

With that, the designation process was underway. The
states submitted their initial designations; EPA responded
with its proposed modifications, as well as explanations of its
analysis under the nine-factor test; the states then submitted
any responses; and EPA promulgated the final area
designations in December 2004. See PM2.5 Designations
Rule, 70 Fed. Reg. at 946. In a technical support document
accompanying the Rule, EPA explained the basis for the
designations and the analytical tools that it had developed and
applied to assess the nine factors. See EPA, Office of Air
Quality Planning and Standards, Technical Support Document
for State and Tribal Air Quality Fine Particle (PM2.5)
Designations (2004) (“Technical Support Document”).
Recognizing that 2004 monitoring data would soon become
available, EPA also invited states to submit any new data that
might support an amended designation. PM2.5 Designations
Rule, 70 Fed. Reg. at 948. Based on the new data, EPA then
revised designations for eight areas from nonattainment to
attainment and four areas from unclassifiable to attainment.
See Air Quality Designations for the Fine Particles (PM2.5)
National Ambient Air Quality Standards—Supplemental
Amendments, 70 Fed. Reg. 19,844, 19,844 (Apr. 14, 2005)
(codified at 40 C.F.R. pt. 81).

In the end, EPA applied the C/MSA presumption so that
nonattainment boundaries were coextensive with metropolitan
boundaries (and unchanged by the nine-factor analysis) in
only seven of thirty-nine metropolitan areas. See EPA Br. 54
(citing relevant portions of the Technical Support Document).
In every other metropolitan area, applying the nine-factor
analysis and finding that the presumption misjudged the
nature of the PM2.5 problem, EPA designated the urban
 10
nonattainment area as either an area larger than the OMB-
defined metropolitan area (for instance, in Chattanooga,
Tennessee) or smaller (for instance, in Washington, DC).

Several states, local governments, and industrial entities
petitioned EPA for reconsideration of the designations; others
filed petitions for review directly in this court. We stayed
proceedings in this court while EPA considered the petitions
for reconsideration. Once EPA resolved the petitions for
reconsideration, we consolidated all petitions for review.

Petitioners’ challenges fall into four broad categories.
First, they raise procedural challenges to the Holmstead
Memo and to the Designations Rule, arguing that EPA
illegally bypassed notice and comment for each. Second, they
raise various objections to EPA’s statutory authority to adopt
and implement the C/MSA presumption and the nine-factor
test. Third, they argue that even if section 107(d) permits
EPA to adopt the C/MSA presumption and the nine-factor
test, EPA’s analysis nonetheless suffers from such serious
methodological deficiencies and inconsistencies as to render
the entire Designations Rule arbitrary and capricious. Finally,
as a last resort, petitioners request that we vacate certain area
designations that affect them, claiming that EPA at least acted
arbitrarily and capriciously in making these particular
nonattainment and unclassifiable designations. We review
petitioners’ challenges under section 307(d)(9) of the Clean
Air Act, which requires the court to set aside EPA’s final
actions when they are in excess of the agency’s statutory
authority or otherwise arbitrary and capricious. 42 U.S.C. §
7607(d)(9)(a).

II.
Before addressing petitioners’ arguments, however, we
think it helpful to review some technical background.
 11
Fine particulate matter includes both “primary” particles
(e.g., carbonaceous particles and so-called “crustal” particles
like dust) that pollution sources emit directly into the
atmosphere, as well as “secondary” particles (e.g., sulfate and
nitrate particles) that form in the atmosphere as a result of
chemical reactions between PM2.5 precursors that sources
emit. PM2.5 Designations Rule, 70 Fed. Reg. at 945. Power
plants, diesel and gasoline powered engines in mobile sources
like cars and trucks, and other industrial sources produce most
carbonaceous particles; agriculture, mining, and other
activities that cause soil or metals to be suspended in the
atmosphere account for the crustal component. See Technical
Support Document § 3.1; EPA, Office of Air Quality and
Planning Standards, The Particle Pollution Report 6 (2004),
available at http://www.epa.gov/air/airtrends/aqtrnd04/pm.
html. The chemical precursors to secondary PM2.5 include
sulfur dioxide (SO2), emitted in substantial part by power
plants; nitrogen oxides (NOx), emitted in substantial part by
mobile sources, power plants, and other industrial sources;
and ammonia, emitted from agricultural sources, mobile
sources, and power plants. See PM2.5 Designations Rule, 70
Fed. Reg. at 945; Technical Support Document § 3.1.
Atmospheric chemical reactions between these gases yield
secondary PM2.5 in the form of sulfate and nitrate particles.
PM2.5 Designations Rule, 70 Fed. Reg. at 945. The PM2.5
NAAQS set a 15µg/m3 annual limit for all fine particulate
matter without distinguishing among the various kinds (or
“species”) of PM2.5. Even so, “speciation data” that breaks
the total PM2.5 concentration into its constituent components
is quite useful for the area designation process. Because such
data reveals the kinds of particles (carbon, sulfate, nitrate,
crustal particles, etc.) that most account for an area’s PM2.5
problem, it suggests, by extrapolation, the kinds of sources
most responsible for the problem as well.
 12
While the Holmstead Memo announced EPA’s
methodology for determining whether an area with passable
PM2.5 concentrations nonetheless deserves a “nonattainment”
designation, EPA regulations set forth the technical
procedures for measuring ambient PM2.5 concentrations in the
first place. Exhaustive technical specifications regulate the
states’ operation of a network of air monitors that collect air
quality data for any given area. See 40 C.F.R. pt. 50, Apps.
L, N. These monitors measure ambient PM2.5 concentration,
what’s called the “design value,” on any given day. The
annual design value—which determines whether an area
complies with the PM2.5 standard—is then computed by
averaging every quarter’s worth of daily design value samples
(typically collected every third or sixth day), averaging those
quarterly numbers to obtain an annual average, and then
averaging three years of annual numbers to yield the final
annual design value. See 40 C.F.R. pt. 50, App. N. Samples
from these monitors can be further analyzed to yield the
speciation data described above.

The majority of the PM2.5 designations at issue here drew
on monitoring data collected from 2001 to 2003. Based on a
judgment that no petitioner challenges, EPA decided that “the
county boundary . . . [would] determin[e] the extent of the
area reflected by [a violating] PM2.5 monitor.” PM2.5
Designations Rule, 70 Fed. Reg. at 946; see also id. at 946–47
(“[I]f a PM2.5 monitor was violating the standard based on
the 2001–2003 data, at a minimum we designated the county
where that monitor is located as nonattainment. We made
exceptions . . . in a few very large western counties where a
significant geographic feature such as a mountain range
divided a county. . . .”). Thus, for instance, if a monitor in an
industrial area of downtown Detroit registered a violation, all
of Wayne County would be designated as nonattainment—
including even its more idyllic corners like the town of
 13
Grosse Pointe. Cf. id. at 980 (designating Wayne County
nonattainment). To be clear, this reflects quite a separate
judgment from that underlying the C/MSA presumption.
EPA’s selection of the county as the unit of analysis resolved
a problem inherent in the monitoring process, namely, that a
monitor only measures air quality in its immediate vicinity.
Because of this, EPA had to determine how much compass to
give any monitored measurement, which it did by choosing
the county as the unit of analysis. The C/MSA presumption,
by contrast, addresses a different problem, namely, how to
identify those areas that, although deemed to be meeting the
standard themselves, are contributing to nearby violations.

We’ve already described the nine factors that EPA
evaluated to determine whether to depart from the C/MSA
presumption, see supra at 8, and in the hundreds of pages that
comprise the Technical Support Document, EPA explained its
findings for each metropolitan area on each of the nine
factors. EPA assessed these factors with the help of several
analytical tools and models it had developed. See Technical
Support Document §§ 3.0–5.9 (explaining analytical tools).
We describe the most important ones here in general fashion,
reserving additional elaboration for those portions of the
opinion that require it.

To start, given the hundreds of miles that PM2.5 can travel
in the atmosphere, EPA thought it important to isolate the
portion of urban PM2.5 that originates from a metropolitan
area’s local sources as opposed to regional sources much
farther away. Thus, under its “urban excess” analysis, EPA
paired an urban monitoring site with an upwind rural
monitoring site—i.e., a rural site where prevailing winds
move in the direction of the metropolitan area—and
“subtract[ed] the rural concentration from the measured
urban concentration.” Id. § 3.1. True to name, this simple
 14
arithmetic allowed EPA to estimate the portion of urban PM2.5
levels that arises from urban activities by cancelling out the
“rural background” that would exist regardless of those
activities. Id. EPA calculated the urban excess for each
PM2.5 species and then summed those numbers to yield a total
urban excess measure. Id. § 3.2.

EPA next used the urban excess calculations to develop
“weighted emissions scores” for each county in a C/MSA. Id.
§ 4.0. The notion underlying these scores is intuitive: if the
urban excess numbers for the District of Columbia, for
instance, reveal that all local PM2.5 pollution is in the form of
carbon, it would make little sense to think that a C/MSA
county plays a significant part in the District’s monitored
violation if the county has zero carbon emissions. Thus, for
purposes of evaluating the first of the nine factors—the
“emissions in areas potentially included versus excluded,”
Holmstead Memo Guidance at 7—EPA determined that raw
emissions data is usually less suggestive of contribution than
data adjusted to account for the PM2.5 species that actually
comprise the urban excess.

Calculating weighted emissions scores required a number
of steps. See Technical Support Document §§ 4.1–4.3. First,
EPA determined the total metropolitan emissions of carbon,
SO2, NOx, and crustal particles by summing the counties’
individual emissions of each pollutant. For those pollutants,
EPA then divided each county’s emissions by total C/MSA
emissions, calculating each county’s percentage share of total
metropolitan emissions. For example, if County A emits 50
tons of carbon and total C/MSA carbon emissions equal 100
tons, then the ratio would be fifty percent. Next EPA
“weighted” these percentages by multiplying them by the
proportion of urban excess attributable to the relevant
pollutant. To continue with our hypothetical, then, if carbon
 15
accounted for forty percent of urban excess, EPA would
multiply fifty percent by four-tenths, and County A’s
weighted carbon score would be twenty. Having calculated a
county’s weighted scores for each PM2.5 species, EPA then
added these scores together to derive a county’s total
weighted emissions score. Importantly, because these scores
scale a county’s raw emissions based on attributes specific to
an individual C/MSA—i.e., the urban excess number and the
total level of metropolitan emissions—they only provide a
measure for comparing counties within the same C/MSA.

Finally, EPA developed so-called “pollution roses”
that depict 2001–2003 monitoring and meteorological data
for each PM2.5 air monitor. See, e.g., Technical Support
Document 6-11 to -12. Each pollution rose consists of
concentric circles, with the circles’ center representing the
location of the air monitor. EPA then plotted dots around the
circles, with each dot representing one monitored reading, the
dot’s size representing the magnitude of the reading, the dot’s
spatial location representing the prevailing wind direction on
the day of the reading, and the dot’s distance from the center
of the circle representing the average wind speed on that day.

With this technical background in mind, we turn to
petitioners’ four primary arguments.

III.
Petitioners first lodge procedural challenges against
EPA’s promulgation of the final designations rule and the
Holmstead Memo. Petitioners argue that EPA violated the
Administrative Procedure Act, 5 U.S.C. § 553, by failing to
publish both the Rule and the Holmstead Memo for notice and
comment. They are in error as to both.
 16
A. Final Designations Rule
Designation of nonattainment areas is governed by
section 107(d) of the Clean Air Act. § 7407(d). The parties
disagree as to which subsection of section 107(d) provides
EPA’s authority to promulgate designations. Industry
petitioners argue that designations are promulgated under
section 107(d)(6), which states: “Notwithstanding any other
provision of law, not later than December 31, 2004, the
Administrator shall, consistent with paragraph (1), promulgate
the designations referred to in subparagraph (A) for each area
of each State for the July 1997 PM2.5 national ambient air
quality standards.” § 7407(d)(6)(B). By contrast, EPA argues
that the source of its authority is a provision of section
107(d)(1), entitled “Promulgation by EPA of designations,”
which states in relevant part:

Upon promulgation or revision of a national
ambient air quality standard, the Administrator
shall promulgate the designations of all areas
(or portions thereof) submitted under
subparagraph (A) as expeditiously as
practicable, but in no case later than 2 years
from the date of promulgation of the new or
revised national ambient air quality standard.

§ 7407(d)(1)(B)(i). The distinction between these provisions
is important because the statute exempts designations
under section 107(d)(1), among others, from the APA’s
section 553 notice-and-comment requirements; it does not,
however, exempt designations under section 107(d)(6). See
§ 7407(d)(2)(B) (“Promulgation or announcement of a
designation under paragraph (1), (4) or (5) shall not be subject
to the provisions of sections 553 through 557 of title 5 of the
United States Code (relating to notice and comment), except
 17
nothing herein shall be construed as precluding such public
notice and comment whenever possible.”).

EPA is entrusted with administering the Clean Air Act, of
which section 107 is a part, and thus we review the agency’s
construction of the statutory provisions under the familiar
two-step framework set out in Chevron U.S.A. Inc. v. Natural
Resources Defense Council, 467 U.S. 837 (1984). Under
Chevron step one, we ask “whether Congress has directly
spoken to the precise question at issue.” Id. at 842. If at that
point we determine that “the intent of Congress is clear, that is
the end of the matter; for the court, as well as the agency,
must give effect to the unambiguously expressed intent of
Congress.” Id. at 842–43. We proceed to Chevron’s second
step only “if the statute is silent or ambiguous with respect
to the specific issue.” Id. at 843. At the second step, we
determine “whether the agency’s answer is based on a
permissible construction of the statute.” Id.

Here we need go no further than the first step because
the intent of Congress is clear. Petitioners’ reliance
on section 107(d)(6) is misplaced. Subparagraph (B) of
section 107(d)(6) requires that, “not later than December 31,
2004, the Administrator shall . . . promulgate the designations
referred to in subparagraph (A).” § 7407(d)(6)(B).
Subparagraph (A), in turn, states: “Notwithstanding any other
provision of law, not later than February 15, 2004, the
Governor of each State shall submit designations referred to
in paragraph (1) [of § 7407(d)] for the July 1997 PM2.5
national ambient air quality standards . . . .” § 7407(d)(6)(A)
(emphasis added). Thus, contrary to petitioners’ claim,
section 107(d)(6)(B), when read in conjunction with section
107(d)(6)(A), shows that section 107(d)(6) does not itself
authorize the promulgation of designations. Rather, section
107(d)(6) merely governs the timing of PM2.5 designations,
 18
which are made under the authority contained in section
107(d)(1)—a provision that the statute expressly exempts
from notice-and-comment requirements. See § 7407(d)(2)(B).

B. Holmstead Memo
Petitioners’ argument that the Holmstead Memo had to
undergo notice and comment stems, in part, from their
erroneous belief that the final designations were subject to
notice and comment. Our determination above, that the
statute exempts the nonattainment designations from notice-
and-comment procedures, suggests that the Holmstead
Memo—which was simply the first step in the promulgation
of designations—is also exempt.

Petitioners’ further argument, that the Holmstead Memo
is a legislative rule that must undergo notice and comment
independent of the final rule to which it relates, is also
unavailing. For support, petitioners point to General Electric
Co. v. EPA, 290 F.3d 377 (D.C. Cir. 2002), in which we
explained that whether an agency action is the type of action
that must undergo notice and comment depends on “whether
the agency action binds private parties or the agency itself
with the ‘force of law,’” id. at 382—that is, whether “a
document expresses a change in substantive law or policy
(that is not an interpretation) which the agency intends to
make binding, or administers with binding effect,” id. at 382–
83 (quoting Robert A. Anthony, Interpretive Rules, Policy
Statements, Guidances, Manuals, and the Like—Should
Federal Agencies Use Them to Bind the Public?, 41 DUKE
L.J. 1311, 1355 (1992)). General Electric further explained
that “an agency pronouncement will be considered binding as
a practical matter if it either appears on its face to be
binding . . . or is applied by the agency in a way that indicates
it is binding.” Id. at 383 (internal citation omitted).
 19
In response, EPA argues that the Holmstead Memo is
merely a policy statement, not a legislative rule, because it
does not create or modify legally binding rights or
obligations. As EPA correctly notes, the APA expressly
exempts policy statements from notice-and-comment
requirements. See 5 U.S.C. § 553(b)(A) (specifying that,
except when required by statute, the section 553 requirements
for notice and comment do not apply “to interpretative rules,
general statements of policy, or rules of agency organization,
procedure, or practice”).

EPA has the better of this dispute. First, the Holmstead
Memo is not binding on its face. It specifies that it merely
“provides guidance to State and local air pollution control
agencies . . . on the process for designating areas for the
purpose of implementing the fine particle national ambient air
quality standards.” Holmstead Memo at 1. It then explicitly
states that it is “not binding” on the states or EPA and notes
that it provides only EPA’s “current views” on the
designation process, suggesting that those views are open to
revision. Id. at 2. Unlike the agency documents at issue in
General Electric and CropLife America v. EPA, 329 F.3d 876
(D.C. Cir. 2003), which petitioners also cite, the Holmstead
Memo does not impose binding duties on states or the agency.
It merely clarifies the states’ existing duties under the Clean
Air Act and explains the process EPA suggests for states to
follow in providing their initial designations. As we
explained above, see supra at 7–8, the Memo establishes a
rebuttable C/MSA presumption and outlines nine factors for
EPA to consider in its final designations, see Holmstead
Memo Guidance at 5–7.

Petitioners point to language in the Holmstead Memo that
they view as evidence of its binding character vis-à-vis the
states: “A demonstration supporting the designation of
 20
boundaries that are less than the full metropolitan area must
show both that violations are not occurring in the excluded
portions of the metropolitan area and that the excluded
portions are not source areas that contribute to the observed
violations.” Id. at 7. But this language does not create a new
burden on the states; it merely reiterates the statutory
requirements. See § 7407(d)(1)(A)(i) (requiring Governors to
designate as nonattainment “any area that does not meet (or
that contributes to ambient air quality in a nearby area that
does not meet)” the NAAQS).

Nor does the Holmstead Memo bind EPA. The Memo
announces the C/MSA presumption as a rebuttable
presumption, which preserves the agency’s discretion to
deviate from the boundaries of a C/MSA in the final
designations. See Panhandle Producers & Royalty Owners
Ass’n v. Econ. Regulatory Admin., 822 F.2d 1105, 1110 (D.C.
Cir. 1987) (“‘An agency pronouncement is not deemed a
binding regulation merely because it may have some
substantive impact, as long as it leave[s] the administrator free
to exercise his informed discretion.’ . . . Presumptions, so
long as rebuttable, leave such freedom.” (quoting Brock v.
Cathedral Bluffs Shale Oil Co., 796 F.2d 533, 537 (D.C. Cir.
1986) (internal quotation marks omitted))).

Further, EPA has not applied the Holmstead Memo in a
binding manner. Petitioners again cite General Electric for
the proposition that an agency document will be considered
binding if “the affected private parties are reasonably led to
believe that failure to conform will bring adverse
consequences.” Gen. Elec., 290 F.3d at 383 (quoting
Anthony, supra, at 1328). The Memo “encouraged” states to
address all nine factors EPA identified, but did not require
them to do so. Holmstead Memo Guidance at 7. Some states
did not address all or even any of the factors. See, e.g., Letter
 21
from Robert G. Burnley, Dep’t of Envtl. Quality,
Commonwealth of Virginia, to Donald S. Welsh, U.S. EPA
Region III (Feb. 13, 2004) (recommending that all of Virginia
be designated attainment without addressing any of the nine
factors); Letter from Stephanie R. Timmermeyer, West
Virginia Dep’t of Envtl. Prot., to Donald S. Welsh, U.S.
EPA Region III (Feb. 13, 2004) (recommending PM2.5
nonattainment areas to match the existing ozone
nonattainment areas without addressing any of EPA’s other
eight factors). EPA considered such submissions and did not
impose “adverse consequences,” notwithstanding the states’
failure to address the factors listed in the Holmstead Memo.
Compare State of West Virginia PM2.5 Designations—
Preliminary Recommendations, with PM2.5 Designations Rule,
70 Fed. Reg. at 1014–15 (showing that EPA designated as
attainment an area that West Virginia had proposed be
designated nonattainment, despite the fact that West Virginia
did not address eight of EPA’s nine factors).

In sum, we deny petitioners’ procedural claims because
EPA was not required to submit either the final designations
rule or the Holmstead Memo for notice and comment.

IV.
Petitioners next claim that EPA violated section 107(d) of
the Clean Air Act by applying the C/MSA presumption and
nine-factor test to identify areas that contribute to nearby
PM2.5 violations. As we explained above, we review EPA’s
interpretation of the Clean Air Act under Chevron, asking
whether Congress has “directly spoken to the precise
question at issue,” 467 U.S. at 842, and if so, whether
it has unambiguously foreclosed the agency’s statutory
interpretation, e.g., Sierra Club v. EPA, 536 F.3d 673, 677
(D.C. Cir. 2008). But if the statute is either silent or
 22
ambiguous on the specific question at issue, we defer to
EPA’s statutory interpretation so long as it is reasonable. Id.

At the outset we observe that section 107(d) is replete
with the kinds of words that suggest a congressional intent to
leave unanswered questions to an agency’s discretion and
expertise, see Nat’l Cable & Telecomms. Ass’n v. Brand X
Internet Servs., 545 U.S. 967, 980 (2005) (statutory ambiguity
is delegation to the agency “to fill the statutory gap in
reasonable fashion”). Section 107(d) requires EPA to
designate an area as nonattainment if it “contributes to
ambient air quality in a nearby area” that exceeds the relevant
standard, § 7407(d)(1)(A)(i), yet the statute defines neither
“contributes” nor “nearby”—words that we have expressly
found ambiguous as used in other sections of the Act. See
Envtl. Def. Fund v. EPA, 82 F.3d 451, 459 (D.C. Cir. 1996)
(“contributes to” ambiguous in another section of the Clean
Air Act); Sierra Club v. EPA, 719 F.2d 436, 443–44 (D.C.
Cir. 1983) (same, as to “nearby”). It authorizes EPA
to revise state-submitted designations whenever it “deems”
such modifications “necessary,” yet it says nothing of
what precisely will render a modification “necessary.”
§ 7407(d)(1)(B)(ii). And section 107(d) requires states to
submit PM2.5 designations “based on air quality monitoring
data collected in accordance with any applicable Federal
reference methods,” § 7407(d)(6)(A), yet it fails to define
“based on” and “[t]here is no question that the phrase ‘based
on’ is ambiguous,” Sierra Club v. EPA, 356 F.3d 296, 305–06
(D.C. Cir. 2004); accord. Nuclear Energy Inst., Inc. v. EPA,
373 F.3d 1251, 1269 (D.C. Cir. 2004). “[A]mbiguities in
statutes within an agency’s jurisdiction to administer are
delegations of authority to the agency to fill the statutory gap
in reasonable fashion.” Brand X, 545 U.S. at 980. Because it
conveys no clear-cut approach for determining whether an
 23
area contributes to a nearby PM2.5 violation, section 107(d)’s
text is consistent with such a delegation.

To be sure, a statute may foreclose an agency’s preferred
interpretation despite such textual ambiguities if its structure,
legislative history, or purpose makes clear what its text leaves
opaque. Cf. Ariz. Pub. Serv. Co. v. EPA, 211 F.3d 1280,
1287 (D.C. Cir. 2000) (court must “exhaust[] traditional tools
of statutory construction” at Chevron step one).
Notwithstanding petitioners’ torrent of arguments to the
contrary, this is not such a case—indeed, it isn’t even close.

We start with the argument that petitioners judge to be
their best. See Oral Arg. at 1:16–3:00. Pointing to section
107(d)(4), petitioners insist that the statute’s express mandate
that EPA apply the C/MSA presumption in other contexts
conclusively proves that Congress intended to preclude its use
here. Enacted as part of the 1990 Amendments to the Clean
Air Act, section 107(d)(4) “revise[s] . . . by operation of law”
the boundaries of certain urban ozone or carbon monoxide
nonattainment areas “to include the entire metropolitan
statistical or consolidated metropolitan statistical area,” unless
EPA determined that some portions “do not contribute
significantly to the violation of the national ambient air
quality standard.” § 7407(d)(4)(A)(iv)–(v). In contrast,
section 107(d)(6) says nothing about the C/MSA presumption.
Instead it provides that the PM2.5 area designations must be
“based on air quality monitoring data” and promulgated in
accordance with section 107(d)(1)’s general provisions for
area designations. § 7407(d)(6)(A). Citing the familiar canon
of statutory interpretation that “[w]here Congress includes
particular language in one section of a statute but omits it
from another section of the same Act, it is generally presumed
that Congress acts intentionally and purposely in the disparate
inclusion or exclusion,” Russello v. United States, 464 U.S.
 24
16, 23 (1983) (internal quotation marks and brackets omitted),
petitioners insist that the statute’s exclusive inclusion of the
C/MSA presumption for the ozone and carbon monoxide
designations demonstrates Congress’ unambiguous intent to
preclude EPA from adopting the presumption here.

Although petitioners are correct that we construe statutes
to give meaning to the disparate inclusion of particular
language, that principle hardly compels the interpretation they
favor. When interpreting statutes that govern agency action,
we have consistently recognized that a congressional mandate
in one section and silence in another often “suggests not a
prohibition but simply a decision not to mandate any solution
in the second context, i.e., to leave the question to agency
discretion.” Cheney R. Co. v. ICC, 902 F.2d 66, 69 (D.C. Cir.
1990); see also Clinchfield Coal Co. v. Fed. Mine Safety &
Health Review Comm’n, 895 F.2d 773, 779 (D.C. Cir. 1990)
(“[W]here an agency is empowered to administer the statute,
Congress may have meant that in the second context the
choice should be up to the agency.”). Silence, in other words,
may signal permission rather than proscription. For that
reason, that Congress spoke in one place but remained silent
in another, as it did here, “rarely if ever” suffices for the
“direct answer” that Chevron step one requires. Cheney, 902
F.2d at 69 (internal quotation marks omitted); see also Am.
Forest & Paper Ass’n v. FERC, 550 F.3d 1179, 1181 (D.C.
Cir. 2008) (statute’s discrepant inclusion of the modifier
“competitive” to describe “markets” renders statutory
provision lacking the modifier ambiguous).

Undaunted, petitioners insist that the silence here is
unambiguously prohibitive. They point out that Congress not
only refused to treat PM2.5 like ozone and carbon monoxide
but chose an altogether different scheme in subsection
(d)(6)—one “based on air quality monitoring data,”
 25
§ 7407(d)(6)(A), rather than OMB-drawn metropolitan
boundaries and a multi-factor test. As indicated above,
however, Congress’s mere choice of different standards in
subsections (d)(4) and (d)(6) tells us nothing about whether
Congress wanted to mandate different approaches or to permit
them. Thus, for this argument to succeed, subsection (d)(6)
must itself preclude EPA from adopting a test for PM2.5 like
the one that Congress mandated for ozone and carbon
monoxide in subsection (d)(4). But nothing in subsection
(d)(6) even hints at such a prohibition. Subsection (d)(6)(A)
requires states to submit PM2.5 designations “based on air
quality monitoring data collected in accordance with any
applicable Federal Reference methods”; subsection (d)(6)(B)
in turn requires EPA to promulgate those designations in
accordance with subsection (d)(1)’s general provisions. For
the sake of argument, we shall assume that subsection (d)(6)’s
“based on” language unambiguously applies to EPA even
though it appears only in the particular provision governing
states. We shall also assume that the language substantively
constrains EPA’s discretion in determining nonattainment
boundaries. But even given these assumptions, binding
precedent, the statute’s purpose, and basic common sense
foreclose petitioners’ argument that section 107(d)(6) itself
precludes EPA from adopting the C/MSA presumption and
nine-factor test.

First, as noted above, we have repeatedly held that the
words “based on” are unquestionably ambiguous: they neither
compel the agency to rest its decisions “solely on” the
specified factor nor indicate the extent to which the agency
may rely on additional factors. Sierra Club, 356 F.3d at 305–
06; accord. Nuclear Energy Inst., 373 F.3d at 1269. Instead,
they simply constrain the agency from “abandon[ing]” or
“supplant[ing]” the specified factor altogether. Sierra Club,
356 F.3d at 306. We need ask only whether EPA’s PM2.5
 26
designations “can still reasonably be described as ‘based on’”
air quality monitoring data, id., and we have no doubt that the
contribution designations meet this modest standard. If
nothing else, the contribution designations are “based on” the
air quality monitoring data that demonstrate a nearby
violation of the NAAQS. Even were this insufficient, EPA’s
use of air quality monitoring data as part of its nine-factor
analysis—for instance, its use of urban excess and weighted
emissions scores based on speciation data—would surely
suffice.

Second, because subsection d(1) directs EPA to designate
some areas as nonattainment despite monitoring data that
provides no basis, on its own, to do so, see § 7407(d)(1)(A)(i),
we think it quite clear that the statute contemplates reliance on
factors other than monitoring data to determine contribution.
How could EPA possibly fulfill its statutory duty to
determine, for instance, whether emissions in Indiana
contribute to monitored violations in Chicago without
considering wind and emissions data from Indiana?
Obviously it couldn’t. That the statute fails to set forth the
additional criteria for EPA to consider in evaluating
contribution hardly forecloses EPA from developing such
criteria in order to accomplish Congress’s objectives. See
e.g., Entergy Corp. v. Riverkeeper, Inc., 129 S. Ct. 1498, 1508
(2009). Indeed, when a statute is “silent . . . with respect to
all potentially relevant factors, it is eminently reasonable to
conclude that [the] silence is meant to convey nothing more
than a refusal to tie the agency’s hands.” Id. (emphasis
added).

Still undaunted, petitioners advert to section 107(d)(3)—
which permits EPA to require a revision to an area
designation “on the basis of air quality data, planning and
control considerations, or any other air quality-related
 27
considerations the Administrator deems appropriate,”
§ 7407(d)(3)—as evidence that Congress “knew how to
authorize a flexible, ‘multi-factor’ approach to attainment
decisions when it so intended.” Counties’ Opening Br. 20.
But the Supreme Court rejected just this kind of argument in
Entergy Corp. v. Riverkeeper, Inc., which asked whether EPA
may engage in cost-benefit analysis under section 316(b) of
the Clean Water Act given that it says nothing of cost-benefit
analysis while other sections expressly authorize it. 129 S.
Ct. at 1508. Pointing out that section 316(b) “is silent not
only with respect to cost-benefit analysis but with respect to
all potentially relevant factors,” the Court rejected the claim
that the silence reflected a prohibition, for if it did, “then the
EPA could not consider any factors in implementing [section
316(b)]—an obvious logical impossibility.” Id. Here we face
just this kind of overwhelming statutory silence. Although
logic dictates that EPA must evaluate some factors in addition
to monitoring data to determine contribution, the statute says
nothing about which factors it should consider. We thus
have no difficulty rejecting the claim that the statute
unambiguously forecloses EPA from adopting the C/MSA
presumption or considering its nine factors in applying it.

The legislative history petitioners cite fails to rehabilitate
their claim that section 107(d)(6) unambiguously requires
EPA to base all PM2.5 designations—including nonattainment
designations for contribution—“on air quality monitoring
data” alone. Although petitioners are correct that in the
Transportation Equity Act for the 21st Century, Congress
provided for the deployment of a PM2.5 monitoring network to
produce adequate monitoring data for these designations, that
does nothing to dispel the ambiguity over what criteria EPA
should rely on to assess contribution. Indeed, the particular
provision of TEA-21 that petitioners emphasize actually
reinforces that ambiguity rather than resolves it. See TEA-21
 28
§ 6102(c)(1) (“Only data from the monitoring network . . .
shall be considered for such designations. Nothing in the
previous sentence shall be construed as affecting . . . the
Administrator’s authority to promulgate the designation of an
area as nonattainment, under section 107(d)(1) of the Clean
Air Act, based on its contribution to ambient air quality in a
nearby nonattainment area.”).

In sum, we conclude that neither section 107(d)’s
requirement that PM2.5 designations be “based on air quality
monitoring data” nor its mandate that EPA apply the C/MSA
presumption and a multi-factor test for pollutants other than
PM2.5 unambiguously reveals Congress’s intent to prevent
EPA from using the presumption and the nine-factor test to
determine contribution here. Our rejection of petitioners’
purportedly strongest argument is a fair harbinger of the fate
of their remaining statutory complaints.

To begin with, petitioners waived two of their statutory
arguments by failing to raise them in their opening briefs. See
New York v. EPA, 413 F.3d 3, 20 (D.C. Cir. 2005) (petitioners
waive those arguments that they fail to raise in their opening
briefs). Specifically, not until their reply briefs did they
present a statutory challenge to EPA’s interpretation that a
“nearby” area under section 107(d)(1)(A) may include non-
contiguous areas, or to its conclusion that an area’s future
reductions in emissions qualifies as a relevant factor for
assessing contribution. This leaves just two general
challenges: petitioners’ claim that the C/MSA presumption
and the nine-factor test run afoul of the statutory term
“contribute”; and their claim that the presumption
impermissibly encroaches on states’ statutory prerogative to
have first-say on area designations within their borders.
 29
As to the textual claim, petitioners insist that the verb
“contribute” necessarily connotes a significant causal
relationship, meaning that EPA may not designate a county as
contributing to nonattainment if “corrective measures in [the
county] will do nothing to address the problem or help
achieve compliance in the nonattainment area.” Counties’
Opening Br. 25. We reject both the major and the minor
premise. Although petitioners cite one dictionary that
supports the claim that the adverb “significantly” is implicit in
the verb “contribute,” EPA cites other dictionaries that define
“contribute” without reference to any threshold level of
significance. This alone suggests an ambiguity that fatally
undermines petitioners’ Chevron step one argument. See
Cellular Telecomms. & Internet Ass’n v. FCC, 330 F.3d 502,
509 (D.C. Cir. 2003) (“[D]ueling over dictionary definitions
is pointless, for it fails to produce any plain meaning of the
disputed word.”). But even were we to think that “contribute”
unambiguously means “significantly contribute,” we still
disagree that “significantly contribute” unambiguously means
“strictly cause.” Cf. Michigan v. EPA, 213 F.3d 663, 667–68
(D.C. Cir. 2000) (“significant” is ambiguous). Given that the
statute uses the word “contribute” and that a contribution may
simply exacerbate a problem rather than cause it, we see no
reason why the statute precludes EPA from determining that a
county’s addition of PM2.5 into the atmosphere is significant
even though a nearby county’s nonattainment problem would
still persist in its absence. In fact, a contrary interpretation of
“contribute” would effectively preclude a nonattainment
designation for any attaining county when the cause of the
violation is metropolitan-wide. We may not interpret
“contribute” in a way that does such violence to section
107(d)’s very purpose.

We also reject petitioners’ argument that EPA violated
the statute by failing to articulate a quantified amount of
 30
contribution that would trigger a nonattainment designation.
Petitioners apparently prefer a bright-line, “objective” test of
contribution, see, e.g., Counties’ Opening Br. 28, but it is the
statute, not petitioners’ preferences, that constrains EPA’s
discretion. And nothing in the statute compels EPA to
quantify a uniform amount of contribution below which
counties will automatically escape nonattainment designations
or to quantify similar thresholds for the nine factors EPA
evaluated in making those determinations. Section 107(d) is
ambiguous as to how EPA should measure contribution and
what degree of contribution is sufficient to deem an area
nonattainment, as even petitioners seem to concede, see
Counties’ Reply Br. 13 (“EPA was supposed to be defining
and deciding ‘contributes.’”). Thus, reasonably exercising the
discretion that Congress delegated to it, EPA interpreted
“contribute” to mean “sufficiently contribute,” and then
applied the C/MSA presumption and nine-factor test precisely
to identify those areas that meet that definition. Petitioners
offer no plausible reason to think that the statute forecloses
this approach.

Nor do we agree with petitioners that EPA’s failure to
quantify its analysis somehow rendered its interpretation of
“contribute” arbitrary and capricious and therefore
unreasonable under Chevron step two. Cf. Northpoint Tech.,
Ltd. v. FCC, 412 F.3d 145, 151 (D.C. Cir. 2005) (statutory
interpretation that is arbitrary and capricious is unreasonable
under Chevron step two). An agency is free to adopt a
totality-of-the-circumstances test to implement a statute that
confers broad discretionary authority, even if that test lacks a
definite “threshold” or “clear line of demarcation to define an
open-ended term.” PDK Labs., Inc. v. DEA, 438 F.3d 1184,
1195 (D.C. Cir. 2006) (internal quotation marks omitted). To
be reasonable, such an “all-things-considered standard” must
simply define and explain the criteria the agency is applying,
 31
id. at 1194, something the Holmstead Memo and certainly the
Technical Support Document did in spades. Of course, EPA
may have applied its nine-factor test inconsistently, resulting
in similar counties being treated dissimilarly—a question we
address in Part V. EPA may also have applied it so
erroneously in a particular case that it could not have
reasonably concluded that a county was contributing to
nearby violations—an issue we consider in Part VI. But
EPA’s use of a flexible multi-factor analysis is not in and of
itself unreasonable just because it lacks quantitative standards.
See id. at 1194–95.

We are equally unimpressed by petitioners’ last general
argument: that the C/MSA presumption unlawfully “deprived
states of the deference to which their designations were
entitled” under section 107(d). States’ Opening Br. 1. To the
extent petitioners are claiming that EPA owes the states a
measure of procedural deference under section 107(d),
we agree that EPA must wait its turn before it makes any
individual county designations. Indeed, in contrast to its
many ambiguities, section 107(d)(1) clearly provides
that states submit their “initial designations” first,
§ 7407(d)(1)(A), and only then does EPA promulgate or
modify the designations as it “deems necessary,”
§ 7407(d)(1)(B)(ii). Not only is that precisely what happened
here, but nothing in section 107(d)(1) prevents EPA from
developing general principles to govern its exercise of
discretion when the time comes, or from announcing those
general principles before the states submit their initial
designations. To the extent petitioners think that EPA owes
the states a measure of substantive deference under section
107(d)(1)—a claim that seems implicit in their objection that
the C/MSA presumption somehow alters states’ “burden” in
the designation process, see States’ Opening Br. 24—we
disagree. Though EPA may, of course, go along with states’
 32
initial designations, it has no obligation to give any quantum
of deference to a designation that it “deems necessary” to
change. See, e.g., Pa. Dep’t of Envtl. Prot. v. EPA, 429 F.3d
1125, 1129 (D.C. Cir. 2005) (recognizing that section 107(d)
gives “deference” to states’ initial designations provided EPA
deems no modification necessary). In short, EPA had
authority to apply the C/MSA presumption at the time it
rejected the states’ submissions. We see no reason why
section 107(d)(1) would force EPA to bite its tongue until
then.

Having rejected petitioners’ arguments that section
107(d) unambiguously precludes EPA from adopting the
C/MSA presumption and nine-factor test, we can easily
conclude that EPA reasonably interpreted the statute as
permitting it to do so. Even if we read section 107(d)
favorably to petitioners, it requires only that EPA designate,
based on air quality monitoring data, nonattainment areas that
either violate or contribute to violations of the PM2.5 NAAQS.
Acting on evidence that urban PM2.5 violations usually stem
from metropolitan-wide activities, EPA adopted a
presumption that designates all metropolitan areas as
nonattainment when at least one metropolitan area registers a
PM2.5 violation, as well as a specifically defined multi-factor
analysis to assess when that presumption fails to reflect the
realities of a given metropolitan area. Petitioners give us
every reason to think they would prefer another system of
analysis—specifically, one that would allow them to escape
certain nonattainment designations—but they give us no
reason to think the system EPA selected is unreasonable.

V.
Petitioners next argue that the Designations Rule is
arbitrary and capricious because it is riddled with
methodological flaws and inconsistencies. They challenge the
 33
Rule in four regards: EPA’s designation of noncontiguous
townships as nonattainment, the agency’s refusal to account
for some potential emissions reductions, the so-called carbon
error in some weighted emissions scores, and the way EPA
applied the nine-factor test.

“[W]e apply the same standard of review under the Clean
Air Act as we do under the Administrative Procedure Act
(APA), 5 U.S.C. § 706(2)(A),” Allied Local & Reg’l Mfrs.
Caucus v. EPA, 215 F.3d 61, 68 (D.C. Cir. 2000), and will set
aside the Designations Rule only if it is “arbitrary, capricious,
an abuse of discretion, or otherwise not in accordance with
law,” § 7607(d)(9)(A). We must affirm the Rule if the record
shows EPA considered all relevant factors and articulated a
“rational connection between the facts found and the choice
made.” Burlington Truck Lines v. United States, 371 U.S.
156, 168 (1962). Of particular note in this challenge, we give
an “extreme degree of deference to [EPA] when it is
evaluating scientific data within its technical expertise,” City
of Waukesha v. EPA, 320 F.3d 228, 247 (D.C. Cir. 2003)
(internal quotation marks omitted). Such deference is
especially appropriate in our review of EPA’s administration
of the complicated provisions of the Clean Air Act. See Nat’l
Ass’n of Clean Air Agencies v. EPA, 489 F.3d 1221, 1229
(D.C. Cir. 2007).

In basing its designation decisions on a rigorous analysis
of each county’s particular attributes, EPA satisfied the
requirements of reasoned decisionmaking. Given our highly
deferential standard of review, these four challenges provide
no basis to question EPA’s general analysis or to upset the
entire Designations Rule.
 34
A. Designation of Noncontiguous Townships
As we have described above, EPA’s designation analysis
starts with two presumptions. See supra at 12–13. First, “if a
PM2.5 monitor [is] violating” the NAAQS, “at a minimum
[EPA will] designate the entire county where that monitor is
located as nonattainment” because the “county boundary” is
“the basic jurisdictional boundary for determining the extent
of the area reflected by the PM2.5 monitor.” PM2.5
Designations Rule, 70 Fed. Reg. at 946–47. Second, all
counties within the C/MSA contribute to that violation. See
id. at 947; Holmstead Memo Guidance at 4. When these
presumptions operate in tandem, a violating monitor within a
C/MSA will result in a single contiguous block of
nonattainment counties that includes both the county with the
violating monitor and the other C/MSA counties that are
deemed to contribute to that violation. Likewise, when EPA
includes out-of-C/MSA counties in a nonattainment block,
those counties typically adjoin violating counties. But EPA
also invites states to recommend smaller PM2.5 designation
areas on a “case-by-case basis” if they “provide an adequate
justification demonstrating that a smaller area would include
the full area that is violating the standards and all nearby
source areas that contribute to the violation.” Holmstead
Memo Guidance at 6. Some states took up EPA’s offer and
asked that a township rather than a county be designated as
nonattainment when a power plant in the township was the
overwhelming source of the county’s contribution to a nearby
PM2.5 violation. See PM2.5 Designations Rule, 70 Fed. Reg. at
947; Technical Support Document § 6.5.4.4 (Discussion).

To borrow petitioners’ imagery, this approach created an
island of nonattainment—the township—surrounded by a sea
of attainment throughout the remainder of the county.
Industry petitioners argue that creating such islands of
nonattainment arbitrarily deviates from the presumption that
 35
designations should be made at the county level. Because
emissions from a power plant in a township do not skip over
all other parts of the surrounding county and come to rest in a
nearby area with a violating monitor, petitioners argue that
violating areas designated nonattainment should be
contiguous to those areas contributing to the violation.

We find no fault in EPA’s conclusion that it would be
unreasonable to designate as nonattainment those areas that
do not, in fact, contribute to violations. EPA designated as
nonattainment a township with an emitting source—rather
than the entire county—because “it would be inappropriate to
include other portions of a county, merely because those
portions lay between the large stationary source and the
remainder of the designated nonattainment area.” PM2.5
Designations Rule, 70 Fed. Reg. at 947. Although contiguity
of nonattainment areas may follow from the operation of the
Rule’s twin presumptions, the designations process does not
mandate that result when the scientific data show that a source
contributing to a violation is not in an area contiguous to the
county with the violating monitor. According to the
Technical Support Document, EPA designated townships
rather than entire counties as nonattainment only when a
power plant in the township was responsible for the vast
majority of the county’s PM2.5 emissions. For example, the
Conesville Plant in Franklin Township emitted 99% of the
SO2, 90% of the NOx, 78% of the carbon, and 87% of the
crustal emissions for all of Coshocton County. See Technical
Support Document § 6.5.4.4 (Factor 1). By designating only
Franklin Township as nonattainment, EPA accounted for the
lion’s share of Coshocton County’s emissions that “are
contributing to the violation in the Columbus Metropolitan
Area,” id. (Discussion). EPA presented comparable statistics
for the townships designated nonattainment in Adams and
Gallia Counties. Id. § 6.5.4.6 (Factor 1) (providing a similar
 36
rationale for designating Monroe and Sprigg Townships in
Adams County and Cheshire Township in Gallia County).
EPA reasonably concluded there was no need to designate as
nonattainment any other part of these counties.

Industry petitioners also argue that EPA failed to use
monitoring or modeling data in designating townships as
nonattainment. But petitioners fail to identify any township
designations lacking such support. And although we have no
obligation to comb through the voluminous record in this case
to determine the merits of an argument for which petitioners
offer no support, see FED. R. APP. P. 28(a)(9)(A) (requiring
briefs to contain “citations to the authorities and parts of the
record on which the appellant relies”), we note at least one
instance in which we easily found that the record contradicts
their assertion. EPA used air quality modeling to determine
that the power plant in Franklin Township was responsible for
the vast majority of Coshocton County’s contribution to PM2.5
violations in the Columbus Metropolitan Area. See
Memorandum from Brian Timin & Richard Damberg, EPA
Office of Air Quality Planning & Standards, Air Quality
Modeling To Assess Power Plant Impacts 2 (Jan. 20, 2006);
Technical Support Document § 6.5.4.4 (Discussion, Factor 1,
Factor 2); see also PM2.5 Designations Rule, 70 Fed. Reg. at
947 (explaining that EPA uses speciation data from monitors
to determine which sources contribute to violations).

Finally, industry petitioners contend that because EPA
designated only contiguous areas as nonattainment for
excessive levels of ozone, it acted arbitrarily in not doing so
for PM2.5. But ozone was the subject of a different
designations process and a separate rulemaking, and nothing
compels EPA to use the same approach for PM2.5. Petitioners
emphasize that PM2.5 and ozone are both pollutants that can
travel long distances. That similarity alone, however, is not
 37
enough to force EPA to treat the two pollutants as if they pose
the same threat to public health and welfare. When EPA set
forth the criteria it would use during the PM2.5 designations
process, it explained that “unlike ozone,” PM2.5 “can arise on
a very localized basis. For example, violations can be caused
by the emissions from a single major source or set of
sources.” Holmstead Memo Guidance at 6. Given this
critical difference, EPA was well within its discretion to
consider state recommendations for smaller PM2.5
designations on a “case-by-case basis,” id.

B. Future Reductions in Emissions
When it is evident that federally enforceable pollution
controls will yield significant near-term reductions in
emissions, EPA accounts for those forecasted reductions in
estimating an area’s emissions levels for the purpose of
evaluating contribution. See Letter from Stephen L. Johnson,
Adm’r, EPA, to David M. Flannery, Counsel For Midwest
Ozone Group et al., Attach. at 13 (Aug. 16, 2007) (“Johnson
Attach.”). Industry petitioners argue that EPA overestimated
emissions levels and thus made mistaken designations by
failing to account for future reductions from two federal
programs: the Clean Air Interstate Rule (CAIR) and the NOx
State Implementation Plan (also known as the “NOx SIP
Call”). See generally North Carolina v. EPA, 531 F.3d 896,
902–03 (D.C. Cir.) (describing programs), modified, 550 F.3d
1176 (D.C. Cir. 2008). We find no error in EPA’s refusal to
consider estimates of lower emissions levels that might result
from these two programs.

EPA promulgated CAIR to reduce SO2 and NOx
emissions from upwind sources in 28 states and the District of
Columbia that contribute to nonattainment levels for ozone
and PM2.5 in downwind states. Id. at 903. When EPA made
its PM2.5 designations, it reasonably decided emissions
 38
reductions from CAIR were, at the time, uncertain and should
not be a factor in estimating emissions levels for the
designations process. See Johnson Attach. at 13 (explaining
that EPA would include projected emissions reductions only
for federally enforceable agreements that were “in place by
the time that EPA was required to promulgate the
designations in December 2004”). As discussed above,
designations for C/MSAs, counties, and townships turn in part
on contributions from identifiable sources. When EPA made
its designation decisions in December 2004, there was no
assurance that a state’s compliance with CAIR, which did not
become effective until March 2005, would result in reduced
PM2.5 emissions for specific sources. Indeed, CAIR did not
require states to submit to EPA their recommendations as to
which power plants would reduce SO2 and NOx emissions and
how they would do so until September 2006—nearly two
years after the agency had designated areas under the PM2.5
NAAQS. See Rule To Reduce Interstate Transport of Fine
Particulate Matter and Ozone (Clean Air Interstate Rule), 70
Fed. Reg. 25,162, 25,162 (May 12, 2005); Rule To Reduce
Interstate Transport of Fine Particulate Matter and Ozone
(Clean Air Interstate Rule): Reconsideration, 71 Fed. Reg.
25,304, 25,305 (Apr. 28, 2006) (“Each State covered by
CAIR may independently determine which emission sources
to control, and which control measures to adopt.”). Even
EPA’s provisional compliance regime—designed to start
reductions in emissions levels before states were required to
file their plans—did not take effect until more than a year
after the agency had completed designating areas. See
Rulemaking on Section 126 Petition from North Carolina To
Reduce Interstate Transport of Fine Particulate Matter and
Ozone, 71 Fed. Reg. 25,328, 25,328 (Apr. 28, 2006); see also
North Carolina, 531 F.3d at 903 (describing program). It was
therefore reasonable for EPA to disregard as too speculative
any claimed emissions reductions that might come from
 39
CAIR when it promulgated the PM2.5 designations in
December 2004. See Johnson Attach. at 8 n.11.

It was likewise reasonable for EPA to disregard
forecasted reductions from the NOx SIP Call. Created to
reduce ozone pollution, this program requires 22 upwind
states and the District of Columbia to decrease NOx
emissions, a significant precursor to PM2.5, but it has nothing
to do with reducing SO2, another significant PM2.5 precursor.
See Finding of Significant Contribution and Rulemaking for
Certain States in the Ozone Transport Assessment Group
Region for Purposes of Reducing Regional Transport of
Ozone, 63 Fed. Reg. 57,356, 57,356 (Oct. 27, 1998). The
power plants in question are often among the largest sources
of SO2 in their respective areas. See Johnson Attach. at 8.
EPA concluded that reducing NOx but not SO2 was a step in
the right direction for reducing PM2.5, but it fell short of what
EPA required to consider forecasted reductions in its
designations. See id. at 14. Lacking evidence of significant
near-term reductions in SO2, EPA reasonably decided not to
alter nonattainment designations based only on forecasted
reductions in NOx. See, e.g., Technical Support Document
§ 6.4.4.1 (Factor 9) (showing EPA designated as attainment
Stokes County, North Carolina, because a power plant there
installed controls certain to reduce significantly both NOx and
SO2).

C. Carbon Error
In making nonattainment designations, EPA relied upon a
mistaken estimate of carbon emissions by power plants that
burn bituminous coal. Industry petitioners seize on this error
as evidence that the Designations Rule is arbitrary and
capricious. We disagree. EPA used the best available
information, and the mistaken estimate of carbon had no
effect on the reasonableness of the challenged designations.
 40
Among the analytical tools EPA uses to make
designation determinations is a county’s weighted emissions
score, which allows the agency to compare SO2, NOx, carbon,
and crustal emissions across counties within a C/MSA. See
PM2.5 Designations Rule, 70 Fed. Reg. at 947. This score is
not based on measurements of actual emissions by a particular
source. Instead, EPA uses emissions estimates from the
National Emissions Inventory (NEI) to calculate total PM2.5
emissions, as well as SO2, NOx, carbon, and crustal emissions
for each county. See What Is the National Emissions
Inventory (NEI)?, http://www.epa.gov/ttn/chief/net/neiwhatis.
html (last visited June 11, 2009). The NEI is a database
assembled by EPA’s Emission Inventory and Analysis Group
that houses estimates of the kinds and amounts of substances
emitted by particular sources, including point sources like
power plants and mobile sources like automobiles. Based on
these estimates, EPA creates “speciation profiles,” which
describe the chemicals that make up the emissions associated
with a particular type of source. See Speciation: Emissions
Modeling Clearinghouse, http://www.epa.gov/ttn/chief/emch/
speciation/ (last visited June 11, 2009).

The speciation profile EPA used for large electric
generating units (EGUs) estimated that carbon makes up 21%
of their PM2.5 emissions. As it turns out, that estimate was
wrong for plants that burn bituminous coal. Carbon accounts
for only 2.9% of their PM2.5 emissions. See J.A. at 2848–49.
EPA updated the speciation profile for these plants in its 2006
revision of the PM2.5 NAAQS, but retained the old profile for
plants that burn primarily lignite coal.

Industry petitioners argue that because the 2004
designations were based on a flawed EGU profile that vastly
overestimated carbon emissions, EPA’s nonattainment
designations for counties with large electric power plants that
 41
burn bituminous coal—particularly the 23 identified in the
petition for reconsideration—are likewise flawed. EPA
responds that its ultimate designations did not turn on any one
estimate of a single chemical component of PM2.5. Rather,
the agency relied on numerous data points, including
emissions levels and county rankings of weighted emissions
scores within a C/MSA, that were largely unaffected by the
lower carbon estimate. According to EPA, changes in the
speciation profile for plants that burn bituminous coal did not
substantially lower total PM2.5 emissions estimates. Instead,
the proportions of pollutants emitted by these sources
changed. Specifically, the estimate of crustal particles, also a
precursor to PM2.5, increased as the carbon estimate
decreased. See Johnson Attach. at 4.

We hold that EPA was not obligated to upend the
designations process when it discovered a mistake in its
speciation profile for certain power plants. EPA used the best
information available in making its designations, and that is
all our precedent requires. In American Iron & Steel Institute
v. EPA, 115 F.3d 979, 1006 (D.C. Cir. 1991), the petitioners
argued that EPA’s estimate for the mercury concentration
permitted in the Great Lakes was flawed because the agency
used inaccurate data that had since been corrected. Relying
on ICC v. Jersey City, 322 U.S. 503 (1944), and Vermont
Yankee Nuclear Power Corp v. Natural Resources Defense
Council, 435 U.S. 519 (1978), we held that EPA did not act
arbitrarily by using the older data in its calculation. “The
agency was not obliged to stop the entire process because a
new piece of evidence emerged. If this were true then the
administrative process could never be completed. An agency
does, however, have an obligation to deal with newly
acquired evidence in some reasonable fashion.” Am. Iron,
115 F.3d at 1007.
 42
Here, EPA dealt with newly acquired evidence in a
reasonable fashion by explaining why it would not have
changed the challenged designations. EPA is correct that
revisions to the speciation profile for plants that burn
bituminous coal altered only one component of the weighted
emissions score, which itself is only one of numerous
analytical tools used to assess the first of nine factors EPA
considers in determining contribution to PM2.5 violations.
Speciation profiles for power plants are by no means the
exclusive or even the primary basis for EPA’s designations,
but merely make up one part of a much larger and multi-
factored decisionmaking process. Johnson Attach. at 6–7.

Petitioners nonetheless contend that the carbon error is
particularly important because carbon typically makes up a
large portion of the urban excess and therefore weighs heavily
in these counties’ weighted emissions scores, which
themselves weigh heavily in the contribution analysis because
of the presence of large power plants. That may be so, but
EPA granted a March 2006 request to recalculate the
weighted emissions scores using the revised estimates and
concluded

that even if [the agency] were to reconsider the
designations, the area by area evaluation of
counties with emissions scores or activities
contributing to violations of the NAAQS
would not result in a different outcome. Of
the counties [petitioners] identified in
[their] petition, EPA sees no change in the
rank or magnitude of sources relative to
other counties in the areas that would negate
the appropriateness of inclusion of the
counties within their respective designated
nonattainment areas.
 43

Letter from Stephen L. Johnson, Adm’r, EPA, to David M.
Flannery, Counsel For Midwest Ozone Group et al. 2–3 (Aug.
16, 2007); see also J.A. at 2881–83 (showing initial and
revised scores for the 23 counties identified by petitioners);
Johnson Attach. at 10–12 (explaining why particular
designations would not have changed with new data). Even
with a change in the estimated proportion of carbon emitted
by plants that burn bituminous coal, the overall level of
pollutants emitted by those EGUs generally stayed the same,
as did county rankings of weighted emissions scores. For
example, EPA explained that using the new speciation data
for a power plant in Jefferson County, Indiana, would not
have changed its nonattainment designation despite industry
petitioners’ claim that it is among the most problematic of
EPA’s determinations. See J.A. at 2826 (Midwest Ozone
Petition for Reconsideration). Although the weighted
emissions score for Jefferson County would have been lower,
it still would have been higher than surrounding counties’
scores because of the significant levels of SO2 and NOx that
the county’s power plant continued to emit. Johnson Attach.
at 12. EPA dealt with the so-called carbon error in a
reasonable fashion.

D. Application of the Nine-Factor Test
As explained above, EPA uses nine factors—including
things like air quality, population density, and traffic
patterns—to determine the boundaries of areas contributing to
nearby PM2.5 violations. See PM2.5 Designations Rule, 70
Fed. Reg. at 947; see also supra at 8. State petitioners argue
generally that EPA arbitrarily applied its nine-factor test by
treating similarly situated counties differently without
adequately explaining the allegedly divergent outcomes. In
each of their challenges, petitioners seize upon discrete data
points and ignore the very nature of the nine-factor test, which
 44
is designed to analyze a wide variety of data on a “case-by-
case basis,” Holmstead Memo Guidance at 6. It is EPA’s
holistic assessment of numerous factors that drives the
process—no single factor determines a particular designation.
And although petitioners seek to paint a picture of system-
wide inconsistencies, their challenges really amount to an
attack on EPA’s designations of a small group of New York
counties. Although we address such individual challenges in
Part VI, we conclude here that with respect to the system as a
whole, EPA consistently applied its nine-factor test and
adequately explained its decisions based on record evidence.

1. First Factor:
EPA’s Characterizations of County Emissions
The first of the nine factors EPA uses to designate areas
calls for the agency to consider how emissions levels
contribute to nearby PM2.5 violations. In describing these
levels, EPA characterizes a county’s emissions as low, high,
significant, insignificant, and so forth. State petitioners argue
that EPA characterized county emissions inconsistently,
providing further evidence that the designations were
arbitrary. For example, petitioners claim it is manifestly
arbitrary to designate as attainment counties with emissions
levels EPA characterized as “low” based on weighted
emissions scores of 9.4 (Sevier County, Tennessee) and 6.3
(Jasper County, Georgia), while designating as nonattainment
counties with lower scores of 4.5 (Orange County, New
York), 3.7 (Westchester County, New York), and 1.9
(Rockland County, New York). States’ Opening Br. 34. As
explained above, a weighted emissions score reflects only a
county’s share of a C/MSA’s total emissions. Weighted
emissions scores cannot be used in any meaningful way to
compare emissions levels between counties in different
C/MSAs. Because cumulative emissions scores for all
 45
counties within a C/MSA must total 100, areas with few
counties (like the Athens, Georgia, MSA) will invariably have
relatively larger county-level scores than areas with numerous
counties (like the NY-NJ-CT-PA C/MSA), even though
emissions levels in the smaller C/MSA may be lower.

Likewise, petitioners argue that two instances in which
EPA revised a county designation from nonattainment to
attainment show that its characterization of emissions data
and subsequent designations were arbitrary. EPA originally
described the emissions levels in Woodford County,
Kentucky, as “significant” and designated it as nonattainment.
Technical Support Document § 6.4.3.3 (Factor 1). But EPA
later concluded—without any change in emissions levels—
that the county “has relatively low emissions,” id.
(Justifications for Changes to EPA Recommendations
Contained in the June 24, 2004, Letters to States), and revised
its designation to attainment. These changes, petitioners
argue, reflect the flawed manner in which EPA applied the
first factor. But EPA adequately explained the change. After
the initial designation, Kentucky submitted evidence that
PM2.5 violations in nearby Fayette County were due to local
sources, not emissions from Woodford County as originally
thought. Id. Petitioners reply that EPA used the new data to
explain the change in Woodford County’s designation, not the
change in EPA’s assessment of emissions levels. To the
extent that is even true, though, we can reasonably discern
EPA’s path. Given that the weighted emissions score is a
rough estimate of a county’s relative (and relevant) emissions
in the first place, EPA simply interpreted the numeric score
differently when the new data suggested its facial significance
was inaccurate.

In the same vein, petitioners contend EPA acted
arbitrarily in revising its designation for Jasper County,
 46
Georgia. Finding “significant emissions” that potentially
contribute to PM2.5 violations in other parts of the region,
EPA originally designated the county as nonattainment. See
id. § 6.4.2.2 (Factor 1). Data later submitted by Georgia,
however, showed that emissions from a source in Jasper
County were actually insignificant, prompting EPA to
redesignate the county as attainment. See id. § 6.4.2.1
(Justification for Changes to EPA Recommendations
Contained in the June 24, 2004, Letters to States); Letter from
Ron Methier, Chief, Air Prot. Branch, Ga. Dep’t of Natural
Res., to Beverly Bannister, Dir., Air, Pesticides & Toxics
Mgmt. Div., EPA Region 4, at 2 (Nov. 1, 2004). Far from
being arbitrary, these revised designations demonstrate the
reasonableness of EPA’s case-by-case approach to applying
the first factor.

2. First Factor:
EPA Region 1’s Use of a Bright-line Test
As explained in greater detail in Part VI, EPA Region 1
used a bright-line test to determine which counties within a
C/MSA would be designated as attainment. The test worked
like this: EPA Region 1 ranked each of its counties from
highest to lowest according to their weighted emissions
scores. Starting from the top of the list, EPA Region 1 added
each county’s score and stopped when the sum hit 80%.
Counties above the 80% cut-off point were presumed to be
nonattainment; those below were designated as attainment,
provided they did not have a violating monitor and were not
among those recommended for nonattainment status by a
state. See Technical Support Document § 6.1.1 (Factor 1).
State petitioners argue that application of this 80% test led to
inconsistent area designations, and they are right in one
instance. Rockland County, New York, which is not in EPA
Region 1 and was designated as nonattainment, would have
been designated as attainment under the 80% test. Such an
 47
inconsistency is evidence of an arbitrary designation, but as
Part VI explains, EPA’s mistaken designation of Rockland
County is an aberration, and petitioners fail to identify any
other designation that might have changed had EPA applied
the 80% test elsewhere. Nothing about the way EPA Region
1 applied the first factor reveals a fundamental problem with
EPA’s evaluation of emissions in potentially contributing
areas.

3. Second Factor: Use of Design Values
Under the second of the nine factors, which looks to air
quality, EPA uses monitoring data to compute a “design
value” to describe the concentration of ambient PM2.5 in a
county. See supra at 12. The agency then compares the
design value to the annual NAAQS to help determine whether
the county is violating PM2.5 standards or contributing to
violations nearby. See id.; EPA Br. 14. State petitioners
argue that EPA arbitrarily designated some counties as
nonattainment despite their relatively low design values and
other counties as attainment despite their higher design
values. For example, EPA designated Hardin County,
Kentucky, as attainment even though its design value of 14.1
was higher than the 12.5 design value for Westchester
County, New York, which was designated as nonattainment.
See Technical Support Document § 6.4.3.2 (Factor 2); id.
§ 6.2.2 (Factor 2).

As EPA explained, however, design values alone do not
determine designations based on contribution. Indeed, they
are merely one component, albeit an important one, of a
complex process that ultimately yields designations.
Petitioners’ argument ignores the multiple factors EPA uses in
making case-by-case assessments of counties’ contributions to
nearby violations. See Holmstead Memo Guidance at 4.
Some areas, like Hardin County, may have relatively high
 48
design values but still fall short of violating the annual PM2.5
standard of 15 micrograms per cubic meter. And many
such counties are not contributing areas because they
have low rankings for other factors like population,
traffic, and emissions levels. See, e.g., Technical Support
Document § 6.4.3.2 (explaining Hardin County’s attainment
designation). The three areas state petitioners use as
examples of inconsistent treatment—Nassau, Suffolk, and
Westchester Counties in New York—may have had lower
design values when compared to other attainment areas, but
each ranked high for emissions levels, population, and
number of commuters, all of which support the determination
that they contributed to nearby violations. See id. § 6.2.2. In
short, EPA had ample evidence upon which it based its
designation of these counties as nonattainment, despite their
relatively low design values.

4. Sixth and Seventh Factors:
Distance Between a Power Plant and a Violating Monitor
EPA takes account of the distance between a power plant
and a violating monitor to help determine whether an area’s
meteorological features (the sixth factor) and its particular
geography and topography (the seventh factor) will increase
the likelihood that emissions from the plant will contribute to
a violation. State petitioners argue that EPA used this
distance inconsistently. For example, the agency designated
Clearfield County, Pennsylvania, as attainment even though it
has a power plant 60 miles from a violating monitor. See id.
§ 6.3.4.2 (Discussion). By contrast, EPA designated Orange
County, New York, as nonattainment even though its power
plant is 50 miles from the closest violating monitor. See id.
§ 6.2.2. But EPA never placed the type of weight on the
distance factor that petitioners’ argument assumes. Although
EPA considered the distances, it weighed other factors as
well. For example, Orange County has very high emissions
 49
levels, and meteorological data show winds blowing from the
county toward violating monitors elsewhere. See id.
Clearfield County, on the other hand, has mountainous terrain
and other geographic features limiting emissions travel.
See id. § 6.3.4.2 (Discussion). Petitioners claim that the
attainment designation for Jasper County, Georgia,
demonstrates inconsistency because it has a power plant 45
miles from the nearest violating monitor. But the plant’s
emissions are relatively insignificant, and other factors—such
as low population and few commuters—support the county’s
designation. See id. § 6.4.2.1 (Justification for Changes to
EPA Recommendations Contained in the June 24, 2004,
Letters to States); Letter from Ron Methier to Beverly
Bannister at 2. Once again, seizing upon a single factor
misapprehends the purpose of the nine-factor test.

5. Multiple Factors
Finally, petitioners compare weighted emissions scores,
design values, population density, number of commuters, and
population growth for four attainment counties (Lee County,
Alabama; Russell County, Alabama; Sevier County,
Tennessee; and Fulton County, Ohio) with the same data for
one nonattainment county—New York’s Orange County—to
illustrate that EPA applied the nine factors inconsistently,
rendering the designation process unpredictable and arbitrary.
See Technical Support Document §§ 6.4.2.5, 6.4.6.2, 6.5.4.9.
Although this argument acknowledges, where the others do
not, that EPA considers how the various factors might work
together, it fails for two reasons. First, as we have already
explained, see supra at 15, comparisons of weighted
emissions scores between counties in different C/MSAs are
meaningless. And second, even though Orange County
ranked relatively low on some factors, which might suggest it
was a good candidate for an attainment designation, it also
has emissions levels of PM2.5, SO2, and NOx that far exceed
 50
the levels of the four attainment counties petitioners point to.
See Technical Support Document § 6.2.2. Orange County’s
designation, which is adequately justified, is yet another
illustration of the case-by-case analysis that the Designations
Rule calls for and the nine-factor test achieves.

VI.
Petitioners finally argue that even if the PM2.5
nonattainment designations are reasonable as a general matter,
certain individual county designations are independently
arbitrary and capricious. With but one exception, we reject
these challenges as well.

A. New York County Designations
New York challenges the nonattainment designations of
five counties surrounding New York City: Suffolk,
Westchester, Nassau, Orange, and Rockland—collectively the
“outer counties.” We find New York’s claims without merit,
except for its challenge to the designation of Rockland
County, which we remand to EPA for additional explanation.

New York makes several broad challenges to the
designations of the outer counties. Most generally, the state
argues that EPA’s designations lack a rational basis and that
EPA failed to respond to the comments New York submitted.
To justify the nonattainment designations for Westchester,
Suffolk, and Nassau Counties, EPA cites their high emissions,
population, traffic, and commuting patterns. For Orange
County, EPA relies on high emissions, as indicated by a
weighted emissions score ranking it fifth in the CMSA. EPA
also relies on meteorological data to show that wind blows
from each of the outer counties toward the violating monitors
some percentage of the time. Technical Support Document
at 6-24 to -36. EPA considered and responded in some
 51
detail to New York’s comments challenging the designations.
Responses to Comments EPA Region 2 at 2-9 to -14. We
find that EPA’s basis for designating Suffolk, Westchester,
Nassau, and Orange Counties can reasonably be discerned
from the record. See Motor Vehicles Mfrs. Ass’n v. State
Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983) (“We
will . . . ‘uphold a decision of less than ideal clarity if the
agency’s path may reasonably be discerned.’” (quoting
Bowman Transp., Inc. v. Arkansas-Best Freight Sys., Inc., 419
U.S. 281, 286 (1974))).

New York next contends that EPA did not correctly take
into account commuting data for the outer counties and points
to its own data showing that only 13% of outer county
commuters drive to New York City. States’ Opening Br. 39.
EPA responds that the relevant factor on which it relied was
not the percentage of commuters but rather the raw number of
commuters. New York correctly highlights the inconsistency
in EPA’s treatment of Rockland County’s commuter data,
which is discussed below, but as to the other counties, New
York’s arguments are not persuasive. The Technical Support
Document shows that Westchester and Nassau Counties each
have over 100,000 commuters to New York and Bronx
Counties. Technical Support Document at 6-29. Suffolk
County has fewer commuters to New York and Bronx
Counties (roughly 44,000), but it has higher Vehicle Miles
Traveled (VMT) than either Westchester or Nassau Counties,
likely due to its location on eastern Long Island. Id. New
York is correct that Orange County’s commuter numbers and
VMT are far lower than any of the other three counties, id.,
but EPA based Orange County’s nonattainment designation
on its high emissions, not its commuter numbers, id. at 6-24.
EPA’s reliance on commuting data is supported by the record,
which shows significant numbers of commuters and VMT for
Westchester, Nassau, and Suffolk Counties.
 52
New York also maintains that EPA impermissibly
changed its theory of outer county contribution from
contribution to a violating monitor in New Haven,
Connecticut, to contribution to violating monitors in
Manhattan and the Bronx. EPA’s initial rationale for
designation in the Technical Support Document reads:
“Nassau County ranks high for emissions, population, traffic,
and commuting patterns. In addition, an analysis of pollution
roses and back trajectories to New Haven, CT showed a
contribution from Nassau County.” Id. Based on comments
submitted by New York and Connecticut, EPA agreed that the
violating New Haven monitor was “not representative of
community exposure” and thus should not be the basis of
designations. See id. at 6-35. EPA did not, however, change
the outer counties’ designations on this ground.

New York’s claim that EPA’s continued nonattainment
designations were arbitrary fails because EPA’s rationale for
the designations can be discerned from the Technical Support
Document, which lists numerous reasons for designation other
than contribution to the New Haven monitor. It is a
reasonable reading of the Technical Support Document to
attribute the list of factors, such as emissions and commuters,
as referring to contribution to the violating monitors in
Manhattan and the Bronx, while the initial rationale for
contribution to New Haven was back trajectories and
pollution roses. More important, New York is protesting the
iterative process of revision that the CAA itself mandates:
EPA revised its position in response to New York’s
comments. New York’s underlying complaint is that the
iterations should have continued, perhaps ad infinitum. But
such a process is inconsistent with the CAA: Congress
imposed deadlines on EPA and thus clearly envisioned an end
to the designation process.
 53
Although we reject New York’s challenge to the
designations of Westchester, Suffolk, Nassau, and Orange
Counties, the state’s challenge to Rockland County’s
nonattainment designation has merit. EPA appears to have
acted inconsistently in designating Rockland County as a
nonattainment area.

First, New York persuasively shows that Rockland’s
treatment was inconsistent with the treatment of other
counties in the same CMSA that fall within a different EPA
region. New York notes that EPA Region 1, which
encompasses the CMSA’s New England counties, employed
what New York terms an “80% test” in deciding which
counties would be designated nonattainment. The Region 1
materials do not reference such a test, but New York is correct
in its description of the process EPA Region 1 apparently
employed. Region 1 “dropped” from nonattainment
consideration Litchfield and Middlesex Counties in
Connecticut and Hampden and Berkshire Counties in
Massachusetts because: “(1) none of these counties contain
violating PM2.5 monitors, (2) none were recommended for
nonattainment designation by the state, and (3) all have
emissions scores ≤2.5.” Id. at 6-6. Region 1 arrived at the 2.5
cutoff by ranking all of the CMSA counties from highest to
lowest weighted emissions score, summing the weighted
emissions scores from top to bottom, and drawing a line after
the county at which the cumulative emissions score equaled
80%—thus, the “80% test.” Any counties below the 80% line
were dropped from further consideration if they did not
contain a violating monitor and were not recommended by the
state for a nonattainment designation. See id. at 6-4 to -6.
Although this may be a reasonable approach in the abstract,
New York’s complaint is that, if this process had
been applied in Region 2, which includes the New York
counties, Rockland County would have been dropped from
 54
consideration instead of designated nonattainment. EPA
responds that the “80% test” was no test at all. But the
agency’s characterization is unavailing because no matter
how Region 1’s process is characterized, the fact remains that
Rockland County would have been designated attainment if it
had been in Region 1, but was designated nonattainment by
EPA Region 2. Such inconsistent treatment is the hallmark of
arbitrary agency action.

Second, EPA’s rationale for designating Rockland
County changed between the initial designation and the final
designation, with no apparent change in data. The only
rationale EPA cited in its initial designation of Rockland was
that Rockland “is contiguous to . . . Orange and Westchester
Counties,” both of which EPA designated as nonattainment.
Id. at 6-24. In the initial designations, EPA characterized
Rockland County’s commuter numbers as “low,” id. at 6-31;
when it revised its designations, EPA characterized Rockland
County’s commuter numbers as “significant,” though there
was no intervening change in data, id. at 6-35.

Third, and relatedly, New York argues that EPA treated
Rockland County differently than Dutchess County in New
York and Ocean County in New Jersey, both of which were
designated attainment. New York notes that both counties
have similar or worse values than Rockland on most or all of
the factors EPA assessed. EPA’s response, laid out in its brief
to this court, is that of the factors New York cites, only
commuting was significant for Rockland, and that Rockland’s
numbers of commuters to violating counties are three times
the same statistic for Dutchess and Ocean Counties. EPA also
notes that Rockland has large power plants, while Ocean
County does not. EPA Br. 154.
 55
EPA’s attempt to distinguish Rockland County does not
withstand close inspection. First, EPA cannot rely on
Rockland’s power plants to distinguish Rockland from Ocean
County because power plants are solely a proxy for emissions,
and Rockland has lower emissions than either Ocean or
Dutchess County. Second, EPA is correct that, while
Rockland has three to four times the number of commuters to
violating counties as the other two counties, EPA initially
characterized Rockland’s commuter numbers as “low.” The
agency’s later rhetorical revision of its characterization to
“significant” is not justified by any change in the underlying
data, which renders suspect EPA’s reliance on commuters as
the sole basis for distinguishing Rockland from the other two
counties.

In sum, Rockland County’s nonattainment designation is
troubling because of the apparent inconsistency in EPA’s
approach to designations in different EPA regions, EPA’s
varying characterizations of Rockland’s statistics, and EPA’s
treatment of Rockland as compared to Dutchess and Ocean
Counties. In light of the agency’s scientific expertise and the
complexity of the designation process, we remand to give
EPA another opportunity to provide a coherent explanation
for its designation. See, e.g., North Carolina v. EPA, 550
F.3d 1176 (D.C. Cir. 2008).

B. Other Individual County Challenges
We have considered the other individual county
challenges lodged by Petitioners and conclude that none of
them has merit. Our standard of review is deferential. EPA
“must examine the relevant data and articulate a satisfactory
explanation for its action including a ‘rational connection
between the facts found and the choice made.’” State Farm,
463 U.S. at 43 (quoting Burlington Truck Lines, 371 U.S. at
 56
168). The record before us shows that EPA considered
numerous relevant factors for each challenged county, and the
evidence supports the nonattainment designations EPA
promulgated. In the accompanying judgment, we deny the
petitions for review of the remaining individual county
designations.

VII.
Having considered petitioners’ remaining arguments and
finding them without merit, we deny the petitions for review
in all respects save one: the designation of Rockland County
is remanded to EPA.

So ordered.