Midwest Ozone Group v. Environmental Protection Agency

United States Court of Appeals
FOR THE DISTRICT OF COLUMBIA CIRCUIT
Argued September 28, 2022 Decided March 3, 2023
No. 21-1146
M IDWEST OZONE GROUP ,
PETITIONER
v.
ENVIRONMENTAL PROTECTION AGENCY AND M ICHAEL S.
REGAN , ADMINISTRATOR , UNITED STATES ENVIRONMENTAL
PROTECTION AGENCY,
RESPONDENTS
APPALACHIAN M OUNTAIN CLUB, ET AL .,
I NTERVENORS
On Petition for Review of a Final Action
of the Environmental Protection Agency
David M. Flannery argued the cause for petitioner. With
him on the briefs were Kathy G. Beckett and Edward L. Kropp.
Chloe H. Kolman, Attorney, U.S. Department of Justice,
argued the cause for respondents. With her on the brief were
Todd Kim, Assistant Attorney General, and Daniel P.
Schramm, Attorney, U.S. Environmental Protection Agency.

-- 1 of 12 --

2
Sean M. Helle, Kathleen Riley, Ann Brewster Weeks,
Hayden Hashimoto, Zachary Fabish, and Graham McCahan
were on the brief for respondent-intervenors.
Letitia James, Attorney General, Office of the Attorney
General for the State of New York, Barbara D. Underwood,
Solicitor General, Steven C. Wu, Deputy Solicitor General,
Judith Vale, Assistant Deputy Solicitor General, Morgan A.
Costello and Claiborne E. Walthall, Assistant Attorneys
General of Counsel, Kathleen Jennings, Attorney General,
Office of the Attorney General for the State of Delaware,
Christian Douglas Wright, Director of Impact Litigation,
Valerie Satterfield Edge, Deputy Attorney General, Matthew J.
Platkin, Attorney General, Office of the Attorney General for
the State of New Jersey, Maura Healy, Attorney General,
Office of the Attorney General for the Commonwealth of
Massachusetts, David S. Frankel, Special Assistant Attorney
General, and Christopher G. King, Senior Counsel, New York
City Law Department, were on the brief for amici curiae in
support of respondents.
Before: WILKINS , RAO and CHILDS , Circuit Judges.
Opinion for the Court filed by Circuit Judge CHILDS .
CHILDS , Circuit Judge: Petitioner Midwest Ozone Group
(MOG), an association of companies, trade organizations, and
individual entities maintaining a collective interest in air
quality, petitions for review of the Environmental Protection
Agency’s (EPA) final action, 86 Fed. Reg. 23,054 (Apr. 30,
2021), entitled the Revised Cross-State Air Pollution Update
Rule (Revised Rule) for the 2008 Ozone National Ambient Air
Quality Standards (NAAQS), which EPA promulgated in
response to this Court’s remand in Wisconsin v. EPA, 938 F.3d
303 (D.C. Cir. 2019). In the Revised Rule, EPA addresses its

-- 2 of 12 --

3
failure to balance emissions obligations in accordance with
2008 ozone NAAQS and its prescribed date of attainment. Id.
at 315. In this appeal, MOG contends that the Revised Rule is
arbitrary and capricious, and that EPA failed to conduct a
legally and technically appropriate assessment as required by
the Good Neighbor Provision of the Clean Air Act (CAA). 42
U.S.C. § 7410(a)(2)(D)(i). We disagree. Instead, we hold that
the Revised Rule is an appropriate exercise of EPA’s statutory
authority under the “Good Neighbor Provision,” and deny the
petition on the merits.
I.
The CAA, codified at 42 U.S.C. §§ 7401–7671q,
authorizes EPA to adopt NAAQS to regulate air pollutants,
such as ozone.1 Id. § 7409(a), (b). Wind carries air pollution
from state to state, thereby disregarding state boundaries.
Upwind is the direction the wind is coming from and
downwind is the direction toward which the wind is blowing.
Emissions from upwind States can impact downwind states’
attainment of the NAAQS. To address this problem, the CAA
contains the Good Neighbor Provision which requires each
1 This Court is familiar with ozone’s status as a pollutant and
recognizes its harmful effects. See Clean Wis. v. EPA, 964 F.3d
1145, 1154 (D.C. Cir. 2020). The Court has also exhaustively
summarized the regulatory framework governing EPA’s
conduct in addition to providing the background for statutory
provisions and the agency proceedings relevant to this case.
See id. See also Sierra Club v. EPA, 21 F.4th 815 (D.C. Cir.
2021); Maryland v. EPA, 958 F.3d 1185 (D.C. Cir. 2020); New
York v. EPA, 781 F. App’x 4 (D.C. Cir. 2019); EME Homer
City Generation, LP v. EPA, 795 F.3d 118 (D.C. Cir. 2015).
We draw on those decisions and incorporate them herein by
reference.

-- 3 of 12 --

4
upwind state to prevent its air pollutant emissions from
contributing significantly to nonattainment in any other
downwind state. See 42 U.S.C. § 7410(a)(2)(D)(i).
In Wisconsin v. EPA, we held that EPA, in implementing
the predecessor of the Revised Rule, the Cross-State Air
Pollution Rule Update for the 2008 Ozone NAAQS (CSAPR
Update), 81 Fed. Reg. 74,504 (Oct. 26, 2016), acted unlawfully
and violated its statutory authority under the Good Neighbor
Provision. We remanded the CSAPR Update because it
improperly allowed upwind states to continue polluting beyond
statutory deadlines which were still applicable to downwind
states. Wisconsin, 938 F.3d at 309, 336.
EPA devised the Revised Rule using the four-step method
for evaluating Good Neighbor Provision obligations. See
Maryland v. EPA, 958 F.3d 1185, 1188 (D.C. Cir. 2020).
At the first step, EPA “performed air quality modeling
coupled with ambient measurements in an interpolation
technique to project ozone concentrations at air quality
monitoring sites in 2021.” 86 Fed. Reg. at 23,057. Linear
interpolation is a mathematical method of using the equation of
a line to find a new data point, based on an existing set of data
points. EPA observed that “in this case the known data are the
2016 measured-based and 2023 modeling-based ozone
concentrations.” Id. at 23,058. EPA acknowledged evaluating
“2021 projected ozone concentrations at individual monitoring
sites[, referred to as nonattainment and/or maintenance
receptors,] and consider[ing] current ozone monitoring data at
these sites to identify receptors that [we]re anticipated to have
problems attaining or maintaining the 2008 ozone NAAQS.”
Id.
At step two, EPA “used an air quality modeling-based

-- 4 of 12 --

5
technique to quantify the contributions in 2021 from upwind
states to ozone concentrations at individual monitoring sites.”
Id. Once the contributions were quantified, EPA “then
evaluated these contributions relative to a screening threshold
of 1 percent of the NAAQS (i.e., 0.75 [parts per billion]) for
those monitoring sites identified as nonattainment and/or
maintenance receptors in step [one].” Id. “States with
contributions that equal[ed] or exceed[ed] 1 [%] of the NAAQS
were identified as warranting further analysis for significant
contribution to nonattainment or interference with
maintenance.” Id. “States with contributions below 1 [%] of
the NAAQS were considered to not significantly contribute to
nonattainment or interfere with maintenance of the NAAQS in
downwind states.” Id. As a result of its air quality and
contribution analysis for the analytic year 2021, EPA
concluded that Illinois, Indiana, Kentucky, Louisiana,
Maryland, Michigan, New Jersey, New York, Ohio,
Pennsylvania, Virginia, and West Virginia had ozone
contributions that equaled or surpassed the 2008 NAAQS
thereby warranting further analysis for significant contribution
to nonattainment or interference with maintenance. Id. For the
nine remaining states of Alabama, Arkansas, Iowa, Kansas,
Mississippi, Missouri, Oklahoma, Texas, and Wisconsin, EPA
found that they were not linked to 2021 downwind air quality
problems. Id. at 23,057.
At step three, EPA applied a multifactor test which
evaluated “cost, available emission reductions, and downwind
air quality impacts to determine the amount of linked upwind
states’ emissions that ‘significantly’ contribute to downwind
nonattainment or maintenance receptors.” Id. at 23,058. EPA
applied the multifactor test to both electricity generating units
and non-electricity generating source categories and “assessed
potential emission reductions in all years for which there [wa]s
a potential remaining interstate ozone transport problem (i.e.,

-- 5 of 12 --

6
through 2025), in order to ensure a full remedy in accordance
with the Wisconsin decision.” Id.
Finally, at step four of the four-step framework, EPA
specified enforceable measures in Federal Implementation
Plans (FIP) for Illinois, Indiana, Kentucky, Louisiana,
Maryland, Michigan, New Jersey, New York, Ohio,
Pennsylvania, Virginia, and West Virginia to accomplish
required emission reductions in these states. Id. at 23,059.
EPA proposed the Revised Rule on October 30, 2020. 85
Fed. Reg. 68,964. EPA published the Revised Rule on April
30, 2021, with an effective date of June 29, 2021. 86 Fed. Reg.
23,054.
II.
MOG challenges EPA on three of the four steps of the
Good Neighbor Provision evaluation method. MOG asserts
that “EPA deviated from its past practice of performing state-
of-the-science photochemical air quality modeling2 for the
analytical year of 2021 . . . in favor of using a linear
interpolation technique to predict air quality concentrations at
monitors in 2021,” at the first step of the four-step framework.
Pet’r’s Br. 7. MOG asserts that “EPA’s linear interpolation
methodology resulted in a significantly higher estimate of 2021
ozone design values than was appropriate,” id. at 25, and “was
executed even though the Courts have gone to great lengths to
uphold EPA non-linear modeling in connection with prior
2 “Photochemical modeling is the central element of the air
quality modeling process and is used to simulate and predict
pollutant concentrations.” Tex. Comm’n on Env’tal Quality,
https://www.tceq.texas.gov/airquality/airmod/overview/am_p
m.html (last visited Oct. 14, 2022).

-- 6 of 12 --

7
Good Neighbor Provision rules.” Id. at 18–19 (citing, e.g.,
Wisconsin, 938 F.3d at 310–11). MOG labels EPA’s action “a
mathematical and analytical shortcut” that should not have
been used “to determine mandatory state obligations.” Id. at
10, 11. As a result, EPA’s actions are “arbitrary and
capricious” because “‘the assumptions and the methodology
used’ [we]re inconsistent with prior modeling upheld by this
Court.” Id. at 11. MOG argues that EPA should have used
photochemical modeling to assess the analytic year of 2021,
but instead chose to use “modeling [that] did not include legal
emission reduction requirements in effect for downwind
sources and failed to consider the impact of exceptional events
on the impacted monitors.” Id. at 12.
As additional criticism of EPA’s approach, MOG cites to
New Jersey v. Wheeler, 475 F. Supp. 3d 308 (S.D.N.Y. 2020).
There, the court ordered EPA, in the context of FIPs for upwind
states Illinois, Indiana, Michigan, Ohio, Pennsylvania,
Virginia, and West Virginia, “to promulgate a complete-
remedy rulemaking addressing . . . EPA’s outstanding statutory
obligations by March 15, 2021.” Id. at 313. MOG asserts that
to meet the Wheeler court’s deadline, EPA used existing
modeling data rather than conduct new modeling, shortened
notice and/or comment periods, refused to extend said periods,
and would not allow a redefinition of nonattainment and
maintenance receptors.
MOG further argues that EPA’s adoption of the Revised
Rule is arbitrary because (1) eleven of the twelve states
identified were considered significant pollution contributors
based on flawed data, (2) EPA’s modeling failed to consider
official regulatory programs and/or other emission reduction
requirements applicable to sources in downwind states that
could contribute to improving ambient air quality, and (3) EPA
failed to account for the impact of exceptional events such as

-- 7 of 12 --

8
wildfires on the ozone design values of the air quality monitors.
Finally, MOG contends that at step three of the four-step
framework, EPA arbitrarily “determined control requirements
for the units subject to th[e] Rule” when the Court in Wisconsin
v. EPA did not require EPA to perform this task and did so
using data from “states not affected by the Rule,” which
“resulted in EPA assessing units that exhibit different
characteristics . . . .” Id. at 47–48, 54.
In response to MOG’s arguments, EPA admits that it
adjusted its traditional step one methodology to finish the
Revised Rule before the July 20, 2021 serious attainment date
for downwind states, as required by the Court in Wisconsin v.
EPA. EPA contends that it used linear interpolation
methodology “to determine how much of the ozone
improvement between the 2016 base year and the 2023
projected year could be expected to occur by 2021,” but the
2021 air quality values were derived from a full set of air
quality modeling emission inventories for 2023. Resp.’s Br. 8–
9 (citing 86 Fed. Reg. at 23,078–80). Moreover, EPA contends
that it conducted additional testing and those outcomes showed
that MOG’s preferred approach would not have led to a
different regulatory result. In this regard, EPA asserts that
despite its revised methodology, MOG has not demonstrated
that its preferred photochemical air quality modeling
methodology would have changed which states were affected
by the Revised Rule.
III.
This Court has jurisdiction to review EPA’s Revised Rule
pursuant to 42 U.S.C. § 7607(b)(1). Because “we apply the
same standard of review under the [CAA] as we do under the
Administrative Procedure Act,” Allied Loc. & Reg’l Mfrs.
Caucus v. EPA, 215 F.3d 61, 68 (D.C. Cir. 2000), this Court

-- 8 of 12 --

9
will uphold EPA’s action unless it is “arbitrary, capricious, an
abuse of discretion, or otherwise not in accordance with law . .
. .” Id. § 7607(d)(9)(A). Our review is narrow; if an action is
not contrary to law, “agency action simply [must] be
‘reasonable and reasonably explained.’” Cmtys. for a Better
Env’t v. EPA, 748 F.3d 333, 335 (D.C. Cir. 2014) (citation
omitted). Generally, a reviewing court “must affirm . . . EPA’s
rules if the agency has considered the relevant factors and
articulated a ‘rational connection between the facts found and
the choice made.’” Allied Loc. & Reg’l Mfrs. Caucus v. EPA,
215 F.3d 61, 68 (D.C. Cir. 2000) (quoting Motor Vehicle Mfrs.
Ass’n v. State Farm Mut. Auto. Ins., 463 U.S. 29, 43 (1983)).
Under this standard, “[a]gency determinations based upon
highly complex and technical matters are ‘entitled to great
deference,’” Appalachian Power Co. v. EPA, 249 F.3d 1032,
1051–52 (D.C. Cir. 2001) (citation omitted), because “many
agency actions having the force of law require expertise the
courts lack and involve policy choices more appropriately
overseen by a politically accountable branch of the
government.” Edwards, Harry T., Post Publication Update for
Federal Standards of Review, 119 (2022); see also Huls Am.
Inc. v. Browner, 83 F.3d 445, 452 (D.C. Cir. 1996) (“[W]e will
give an extreme degree of deference to the agency when ‘it is
evaluating scientific data within its technical expertise.’”)
(citation omitted); Maryland, 958 F.3d at 1196 (“[A reviewing
court] must give an extreme degree of deference to . . . EPA’s
evaluation of scientific data within its technical expertise,
especially where . . . EPA’s administration of the complicated
provisions of the [CAA is under review.]”) (citations and
quotation marks omitted).
Statistical analysis has been described as “perhaps the
prime example of an area of technical wilderness into which
judicial expeditions are best limited to ascertaining the lay of

-- 9 of 12 --

10
the land.” Appalachian Power Co. v. EPA, 135 F.3d 791, 802
(D.C. Cir. 1998). “Although computer models are ‘a useful and
often essential tool for performing the Herculean labors
Congress imposed on EPA in the [CAA],’ their scientific
nature does not easily lend itself to judicial review.” Id.
(internal citation omitted). Thus, “[a reviewing court] do[es]
not look at the decision as would a scientist, but only to ensure
that EPA adheres to certain minimal standards of rationality.”
Cmtys. for a Better Env’t, 748 F.3d at 336 (citation and
quotation marks omitted). The reviewing court also “will not
take it upon [itself], as nonstatisticians, to perform [its] own
statistical analysis—a job more properly left to the agency to
which it was delegated.” Appalachian Power Co., 135 F.3d at
802. “[I]t is only when the model bears no rational relationship
to the characteristics of the data to which it is applied that [the
reviewing court] will hold that the use of the model was
arbitrary and capricious.” Id. (citations omitted).
IV.
We have considered MOG’s arguments as to the
arbitrariness and capriciousness of the Revised Rule and
observe that the Court has never required EPA to use a
particular modeling method to generate its data or adhere to
past practice, but rather that EPA “consider[s] all of the
relevant factors, and demonstrate[s] a reasonable connection
between the facts on the record and its decision.” Id. (quoting
Ethyl Corp. v. EPA, 51 F.3d 1053, 1064 (D.C. Cir. 1995)).
Thus, when an agency has not otherwise acted contrary to law,
we will conclude that its choice of model is arbitrary and
capricious if “the model is so oversimplified that the agency’s
conclusions from it are unreasonable.” Appalachian Power,
249 F.3d at 1052 (quoting Small Refiner Lead Phase–Down
Task Force v. EPA, 705 F.2d 506, 535 (D.C. Cir. 1983)).

-- 10 of 12 --

11
Based on the record before us, EPA appears to have chosen
analytical techniques rationally connected to the Revised Rule
and appropriately explained its use of the linear interpolation
and subsequent methods for establishing the Revised Rule. In
addition, EPA’s methodology did also incorporate
photochemical modeling, MOG’s preferred technique, as the
“foundation for its projections” and “merely layered an
additional mathematical function, linear interpolation” over the
original projected data to generate 2021 ozone concentrations.
Resp.’s Br. at 19. EPA then performed further data analysis by
checking its 2021 interpolated projection against both a
sensitivity analysis3 and engineering analytics approach.4
These tools produced consistent results and MOG has not
proven that different states would have been regulated
differently under any other method, including a purely
photochemical modeling approach.
Against the backdrop of MOG’s complaints and our
directive in Wisconsin, EPA also was cognizant of the CAA’s
statutory directive that emissions reductions should be done “as
expeditiously as practicable.” 42 U.S.C. § 7511(a)(1). We
therefore conclude that EPA reasonably believed it should
address upwind states’ significant contributions before the next
downwind attainment deadline, which was the serious
attainment deadline of July 20, 2021. See, e.g., 86 Fed. Reg. at
23,072. Given the limited amount of time EPA had to complete
the rulemaking for the Revised Rule, we discern that EPA
3 Using the North American Emissions Modeling Platform,
EPA sensitivity analysis projected 2021 emissions numbers
based on a comprehensive assessment of emissions expected.
See, e.g., 86 Fed. Reg. at 23,075.
4 This analytical approach estimated 2021 power plant
emissions based on historical emissions and known fleet
changes.

-- 11 of 12 --

12
reasonably chose to use existing air quality modeling and
contribution information to derive an appropriately reliable
projection of air quality conditions and contributions in 2021.
In reaching this determination, the Court does not disregard
MOG’s technical data presentation depicting higher ozone
NOx emissions resulting from use of the linear interpolation
methodology, as opposed to photochemical modeling.
However, in the context of the deferential standard afforded
EPA, MOG has not established that EPA’s linear interpolation
method is oversimplified or that the agency has produced
unreasonable results. See id. at 23,080–81. See also
Appalachian Power Co., 135 F.3d at 802 (“[S]o long as EPA
‘acted within its delegated statutory authority, . . . we will not
interfere with its conclusion.” (quoting Ethyl Corp., 51 F.3d at
1064)).
V.
For the reasons stated above, MOG fails to demonstrate
that EPA’s promulgation of the Revised Rule was arbitrary,
capricious, or promulgated in violation of its statutory authority
under the Good Neighbor Provision. Accordingly, we deny
MOG’s petition.
So ordered.

-- 12 of 12 --