Natural Resources Defense Council v. Michael S. Regan, Administrator , U.s. Environmental Protection Agency

United States Court of Appeals
FOR THE DISTRICT OF COLUMBIA CIRCUIT
Argued January 27, 2023 Decided May 9, 2023
No. 20-1335
NATURAL R ESOURCES D EFENSE C OUNCIL,
P ETITIONER
v.
M ICHAEL S. R EGAN, ADMINISTRATOR , U.S. ENVIRONMENTAL
P ROTECTION A GENCY AND ENVIRONMENTAL P ROTECTION
AGENCY ,
R ESPONDENTS
AMERICAN W ATER W ORKS ASSOCIATION ,
INTERVENOR
On Petition for Review of a Final Action
of the Environmental Protection Agency
Sarah V. Fort argued the cause for petitioner. With her on
the briefs was Charles R. Corbett.
David J. Berger was on the brief for amici curiae
Academic Scientists in support of petitioner.

-- 1 of 31 --

2
Sarah A. Buckley, Senior Attorney, U.S. Department of
Justice, argued the cause for respondents. With her on the brief
was Todd Kim, Assistant Attorney General.
Corinne Snow argued the cause for respondent-intervenor
American Water Works Association. With her on the brief was
Ronald J. Tenpas. Jeremy C. Marwell entered an appearance.
Annie S. Amaral and Thomas C. Roberts were on the brief
for amici curiae American Chemistry Council and Western
Growers in support of respondents.
Before: P AN , Circuit Judge, and S ENTELLE and TATEL,
Senior Circuit Judges.
Opinion for the Court filed by Senior Circuit Judge
S ENTELLE.
Opinion concurring in the judgment filed by Circuit Judge
P AN .
S ENTELLE, Senior Circuit Judge: In 2011, the
Environmental Protection Agency (“EPA”) issued its “final
determination to regulate perchlorate in drinking water” under
the Safe Drinking Water Act. Drinking Water: Regulatory
Determination on Perchlorate, 76 Fed. Reg. 7,762, 7,762 (Feb.
11, 2011). That determination started a clock under the Safe
Drinking Water Act requiring EPA to propose regulations
within twenty-four months and promulgate regulations within
eighteen months of the proposal. See 42 U.S.C. § 300g-
1(b)(1)(E). But EPA never promulgated perchlorate
regulations. Instead, nine years later, the agency purported to
withdraw its regulatory determination. See Drinking Water:
Final Action on Perchlorate, 85 Fed. Reg. 43,990, 43,991 (July
21, 2020). Natural Resources Defense Council (“NRDC”)

-- 2 of 31 --

3
petitions for review of this action, arguing that EPA lacks the
authority to withdraw a regulatory determination under the Act
and that, even if EPA possesses such authority, it acted
arbitrarily and capriciously by doing so. EPA, joined by
Intervenor American Water Works Association, defends its
action. Because the Safe Drinking Water Act does not permit
EPA to withdraw a regulatory determination, we grant
NRDC’s petition, vacate EPA’s withdrawal of its regulatory
determination, and remand to the agency for further
proceedings.
I. Background
a. Statutory Framework
The Safe Drinking Water Act authorizes EPA to regulate
potentially harmful contaminants in the nation’s drinking
water. See 42 U.S.C. § 300g-1(b)(1)(A). First enacted in 1974,
the Act has since undergone several amendments. The 1986
amendments required EPA to select at least twenty-five new
contaminants for regulation every three years. Pub. L. No. 99-
339, § 101(b)(3)(C), (D), 100 Stat. 642, 644 (1986). Congress
apparently created this strict regulatory scheme, at least in part,
because it believed EPA had failed to regulate a sufficient
number of contaminants under the Act’s prior structure. See S.
Rep. No. 104-169, at 8, 12 (1995). When Congress amended
the Act in 1996 to create the present scheme, it replaced the
strict three-year regulatory requirement with a discretionary
scheme that allows EPA to determine when contaminants
warrant regulation. 42 U.S.C. § 300g-1(b)(1)(A); S. Rep. No.
104-169, at 12–13.
Under the current Act, every five years EPA must publish
a list of unregulated contaminants that may require future
regulation (the “Contaminant Candidate List”) and make a

-- 3 of 31 --

4
preliminary determination, subject to notice and comment, of
whether to regulate at least five listed contaminants. 42 U.S.C.
§ 300g-1(b)(1)(B)(i)(I), (ii)(I). After the comment period ends,
EPA must make its final regulatory determination. Id.
§ 300g-1(b)(1)(B)(ii)(I). The agency can only determine to
regulate a contaminant if it finds, based upon the “best
available public health information,” id.
§ 300g-1(b)(1)(B)(ii)(II), that:
(i) the contaminant may have an adverse effect on the
health of persons; (ii) the contaminant is known to
occur or there is a substantial likelihood that the
contaminant will occur in public water systems with a
frequency and at levels of public health concern; and
(iii) in the sole judgment of the Administrator,
regulation of such contaminant presents a meaningful
opportunity for health risk reduction for persons served
by public water systems,
id. § 300g-1(b)(1)(A)(i)–(iii).
The Act frontloads EPA’s discretion, allowing the agency
to create the list of contaminants that may require future
regulation, id. § 300g-1(b)(1)(B)(i)(I), select which of those
listed contaminants to consider for regulation, id.
§ 300g-1(b)(1)(B)(ii)(I), and determine whether the selected
contaminants meet the statutory criteria for regulating, id.
§ 300g-1(b)(1)(A)(iii). Once EPA makes its regulatory
determination, however, the Act balances that discretion with
a strict, mandatory scheme governing the regulatory process.
It instructs that, after determining the statutory criteria are met,
the EPA Administrator “shall, in accordance with the
procedures established by this subsection, publish a maximum
contaminant level goal and promulgate a national primary
drinking water regulation.” Id. § 300g-1(b)(1)(A) (emphasis

-- 4 of 31 --

5
added). The maximum contaminant level goal (“MCLG”) is
an unenforceable, aspirational level and is defined as “the level
at which no known or anticipated adverse effects on the health
of persons occur and which allows an adequate margin of
safety.” Id. § 300g-1(b)(4)(A). The national primary drinking
water regulations also normally include an enforceable
maximum contaminant level (“MCL”) that must be set “as
close to the maximum contaminant level goal as is feasible.”
Id. § 300g-1(b)(4)(B). In limited circumstances, EPA can issue
an alternative enforceable standard. See id. § 300g-1(b)(6)(A),
(7)(A). EPA must propose the MCLG and national primary
drinking water regulations within twenty-four months of
making its determination to regulate and must publish the
MCLG and promulgate the regulations within eighteen months
of the proposal, subject to a nine-month extension. Id.
§ 300g-1(b)(1)(E).
The statute also contains an “anti-backslide” provision that
ensures that, once issued, a regulation can only be revised in a
way that will “maintain . . . or provide for greater” health
protections. Id. § 300g-1(b)(9). In all decisions the agency
makes that are based on science, EPA is instructed to use “the
best available, peer-reviewed science.” Id.
§ 300g-1(b)(3)(A)(i).
b. Factual and Procedural Background
Perchlorate, the contaminant at issue in this case, is a
naturally occurring and manufactured chemical commonly
used in the aerospace and defense sectors. When ingested,
perchlorate can inhibit the thyroid’s ability to absorb iodide.
85 Fed. Reg. at 43,994. An iodide-deficient thyroid, in turn,
disrupts the production of thyroid hormones. Id. And
disruptions in thyroid hormone production can lead to adverse
neurodevelopmental outcomes in sensitive populations whose

-- 5 of 31 --

6
brains are still developing, including fetuses and children of
lactating women. Id.
Recognizing the potential health risks associated with
perchlorate, EPA added perchlorate to its Contaminant
Candidate List in 1998, categorizing it as a chemical “needing
additional health effects, treatment research, and occurrence
information.” Announcement of the Drinking Water
Contaminant Candidate List, 63 Fed. Reg. 10,274, 10,275,
10,282 (Mar. 2, 1998). The agency published its first
Unregulated Contaminant Monitoring Rule (“UCMR-1”) in
1999, requiring all large water systems and a sample of small
water systems to collect data on perchlorate contamination
between 2001 and 2005. 85 Fed. Reg. at 43,993.
In 2008, after applying a health reference level (the “level
of concern”) of 15 μg/L and evaluating the frequency of
perchlorate contamination at that level using the UCMR-1 data,
EPA issued a preliminary determination not to regulate
perchlorate after determining that regulation “would not
present a meaningful opportunity for health risk reduction” and
sought comment on that proposed action. Drinking Water:
Preliminary Regulatory Determination on Perchlorate, 73 Fed.
Reg. 60,262, 60,269, 60,280–81. (Oct. 10, 2008). EPA issued
a supplemental notice seeking comment on alternative health
reference levels the following year. See Drinking Water:
Perchlorate Supplemental Request for Comments, 74 Fed. Reg.
41,883, 41,889 (Aug. 19, 2009). Deviating from its
preliminary determination, the agency issued its “final
determination to regulate perchlorate” in 2011. 76 Fed. Reg.
at 7,762. That determination “initiate[d] the process to develop
a national primary drinking water regulation . . . for
perchlorate” and started the clock for EPA to propose the
MCLG and regulations within twenty-four months and to
promulgate the final MCLG and regulations within the

-- 6 of 31 --

7
following eighteen months. 76 Fed. Reg. at 7762–63; 42
U.S.C. § 300g-1(b)(1)(A), (E).
Consistent with its statutory obligation, see id.
§ 300g-1(e), EPA consulted with the Science Advisory Board
as it worked to develop perchlorate regulations, National
Primary Drinking Water Regulations: Perchlorate, 84 Fed.
Reg. 30,524, 30,527–28 (June 26, 2019). At the Board’s
urging, EPA revised the model it used to predict the effects of
perchlorate exposure, developing a “broader and more
comprehensive framework” that directly links iodide uptake
inhibition to changes in thyroid hormone levels, allowing the
agency to better analyze the neurodevelopmental effects
caused by perchlorate exposure. Id.
After EPA missed the statutory deadlines for proposing
and promulgating the MCLG and regulations, NRDC sued the
agency in 2016, seeking to compel the agency to regulate. 84
Fed. Reg. at 30,526. The parties entered a consent decree
requiring EPA to propose and promulgate the MCLG and final
regulations by 2020. NRDC v. EPA, No. 1:16-cv-01251-ER,
ECF Nos. 38, 60 (S.D.N.Y.); 84 Fed. Reg. at 30,526. In June
2019, EPA proposed setting the MCLG and MCL at 56 μg/L
or, in the alternative, at 18 μg/L or 90 μg/L. 84 Fed. Reg. at
30,525. Alternatively, the agency also considered withdrawing
its 2011 regulatory determination and not promulgating an
MCLG or national primary drinking water regulations. Id. The
agency sought comment on its proposal and the three
alternatives. Id.
In July 2020, after the comment period ended, EPA
announced it was withdrawing its determination to regulate and
issuing a “final determination not to regulate perchlorate.” 85
Fed. Reg. at 43,991. The agency explained that it had “re-
evaluated” whether perchlorate satisfied the statutory criteria

-- 7 of 31 --

8
for regulating using its updated model and concluded that
“perchlorate does not occur at a frequency and at levels of
public health concern” and that regulation “does not present a
meaningful opportunity for health risk reduction.” Id. at
43,998. In EPA’s view, because its “re-evaluat[ion]” showed
that the statutory criteria were not met, it lacked the authority
to regulate. Id.
II. Analysis
a. Statutory Authority
EPA makes two primary arguments in support of its
authority to withdraw a regulatory determination. First, it
argues that its withdrawal of the 2011 regulatory determination
was consistent with the statute. Its second argument is
premised on its “inherent authority” to change positions and
withdraw a determination to regulate, which it claims the Safe
Drinking Water Act does not abrogate. Resp. Br. at 22.
We start with EPA’s second argument, which rests on a
faulty premise. While we have often referred to agencies’
“inherent authority,” see, e.g., Ivy Sports Med., LLC v. Burwell,
767 F.3d 81, 86 (D.C. Cir. 2014), the term “inherent” is
misleading because “it is ‘axiomatic’ that ‘administrative
agencies may act only pursuant to authority delegated to them
by Congress,’” Clean Air Council v. Pruitt, 862 F.3d 1, 9 (D.C.
Cir. 2017) (quoting Verizon v. FCC, 740 F.3d 623, 632 (D.C.
Cir. 2014)). Thus, “the more accurate label” for the power EPA
describes “is ‘statutorily implicit.’” HTH Corp. v. NLRB, 823
F.3d 668, 679 (D.C. Cir. 2016) (quoting Ivy Sports Med., 767
F.3d at 93 (Pillard, J., dissenting)). And although the power to
decide is normally accompanied by the power to reconsider,
“Congress . . . undoubtedly can limit an agency’s discretion to
reverse itself.” New Jersey v. EPA, 517 F.3d 574, 582–83

-- 8 of 31 --

9
(D.C. Cir. 2008); see Ivy Sports Med., 767 F.3d at 86 (“[A]ny
inherent reconsideration authority does not apply in cases
where Congress has spoken.”).
EPA, then, has no inherent authority. It has only the
authority given it by the Safe Drinking Water Act. The
question, then, is whether that authority includes the authority
to withdraw a regulatory determination.
To answer that question, we look to the statutory text. The
Act instructs that “[t]he [EPA] Administrator shall, in
accordance with the procedures established by this subsection,
publish a maximum contaminant level goal and promulgate a
national primary drinking water regulation for a contaminant
. . . if the Administrator determines that” the statutory criteria
have been met. 42 U.S.C. § 300g-1(b)(1)(A) (emphasis
added). Elsewhere in the statute, Congress repeated the
directive three additional times that EPA “shall” regulate. “For
each contaminant that the Administrator determines to regulate
under subparagraph (B), the Administrator shall publish
maximum contaminant level goals and promulgate, by rule,
national primary drinking water regulations . . . .” Id.
§ 300g-1(b)(1)(E) (emphasis added). Within 24 months of
making the determination to regulate, “[t]he Administrator
shall propose the maximum contaminant level goal and
national primary drinking water regulation.” Id. (emphasis
added). And within 18 months of proposing the MCLG and
regulations, “[t]he Administrator shall publish a maximum
contaminant level goal and promulgate a national primary
drinking water regulation.” Id. (emphasis added).
It is well established that “[t]he word ‘shall’ generally
indicates a command that admits of no discretion on the part of
the person instructed to carry out the directive.” Assoc. of
Civilian Technicians, Mont. Air Chapter No. 29 v. FLRA, 22

-- 9 of 31 --

10
F.3d 1150, 1153 (D.C. Cir. 1994). That is even more true
where, as here, the statute explicitly grants the agency
significant discretion at the outset but later instructs the agency
that it “shall” act. Cf. Anglers Conservation Network v.
Pritzker, 809 F.3d 664, 671 (D.C. Cir. 2016) (“[W]hen a
statutory provision uses both ‘shall’ and ‘may,’ it is a fair
inference that the writers intended the ordinary distinction.”).
In 2011, EPA determined that perchlorate satisfied the
statutory criteria for regulating. 76 Fed. Reg. at 7,763. Under
the statute, then, EPA has one authorized course of action: it
“shall” propose and promulgate the MCLG and regulations,
and it “shall” do so by the statutory deadlines. See 42 U.S.C.
§ 300g-1(b)(1)(A), (E). EPA recognized as much when it
issued its 2011 “final regulatory determination.” See 76 Fed.
Reg. at 7,763 (“Once EPA makes a determination to regulate a
contaminant in drinking water, [the Act] requires that EPA
issue a proposed [regulation] within 24 months and a final
[regulation] within 18 months of proposal.” (emphasis added)).
Intervenors likewise admit the statute imposes “a duty to issue
regulations where the Administrator has determined to
regulate.” Intervenor’s Br. at 35. To read into the statute
another course of action—one that allows EPA to withdraw its
regulatory determination entirely and decide that it “shall not”
regulate—would be to contravene the statute’s clear language
and structure and “nullif[y] textually applicable provisions
meant to limit [EPA’s] discretion.” New Jersey v. EPA, 517
F.3d 574, 583 (D.C. Cir. 2008) (quoting Whitman v. Am.
Trucking Ass’ns, 531 U.S. 457, 485 (2001)). Because EPA
lacked the authority to withdraw its regulatory determination
in the first instance, its argument that its authority to do so is
not time-limited by the statutory deadlines for proposing and
promulgating regulations is of no import.
EPA urges the Court to overlook the statute’s clear
language, first by arguing that its 2011 determination to

-- 10 of 31 --

11
regulate was only a preliminary, interlocutory step in the
regulatory process that did not bind the agency to issue future
regulations. This argument contradicts the plain language of
the statute and is easily disposed of. The Safe Drinking Water
Act does anticipate that the agency will make a preliminary
regulatory determination. See 42 U.S.C.
§ 300g-1(b)(1)(B)(ii)(I) (“[T]he Administrator shall, after
notice of the preliminary determination and opportunity for
public comment, . . . make determinations of whether or not to
regulate such contaminants.”). But the preliminary
determination precedes the notice and comment period. Once
that period ends, the agency makes its regulatory
determination, and that determination is final. See id.
§ 300g-1(b)(1)(E). EPA issued its preliminary determination
in 2008 and then, after the period for public comments, issued
its final determination in 2011. Compare 73 Fed. Reg. at
60,263 (presenting “EPA’s preliminary regulatory
determination on perchlorate”), with 76 Fed. Reg. at 7,762
(presenting “EPA’s final determination to regulate
perchlorate”).
EPA cites several other provisions of the statute that it
argues implicitly give the agency the authority to withdraw a
regulatory determination. None of these provisions, however,
negate the statute’s clear directive that the agency “shall”
propose and promulgate regulations after making a regulatory
determination. The first provision EPA cites requires the
agency to consider the “best available public health
information” when evaluating whether the statutory criteria for
regulating are met, id. § 300g-1(b)(1)(B)(ii)(II), while a similar
provision requires it to apply “the best available, peer-reviewed
science” to all decisions it makes that are based on science, id.
§ 300g-1(b)(3)(A)(i). EPA argues, and Intervenor agrees, the
statute required the agency to use its new model, which it
considers to be the “best available public health information,”

-- 11 of 31 --

12
to re-evaluate whether the statutory criteria for regulating were
satisfied. Because it did so and concluded that two of the
criteria were no longer met—specifically, that “perchlorate
does not occur at a frequency and at levels of public health
concern, and that regulation of perchlorate does not present a
meaningful opportunity for health risk reduction”—EPA and
Intervenor argue the agency lacked the authority to regulate
and had to withdraw its regulatory determination. 85 Fed. Reg.
at 43,998.
This argument rests on a faulty dichotomy: that EPA must
either disregard the statutory requirements and issue the
regulations, or that it can adhere to the statutory requirements
but is then required to withdraw its regulatory determination.
Instead, the statute compels a third option. EPA must apply the
“best available public health information” to determine
whether the statutory criteria for regulating are satisfied. Id.
§ 300g-1(b)(1)(B)(ii)(II). Once that determination is made, it
is final. EPA’s obligation then is to consider and apply the
“best available, peer-reviewed science,” including any new
developments, to set the substance of the regulations—not to
reevaluate whether to regulate. See id. § 300g-1(b)(3)(A)(i).
EPA claims that this reading of the statute will
“hamstr[i]ng” its decision-making and result in outdated,
scientifically unsupported regulations. 85 Fed. Reg. at 43,992.
But this takes an “all-or-nothing” approach and ignores the
statutory requirements. If the science changes after the agency
makes its determination to regulate but before it issues the
regulations, EPA can—and must—account for those changes
when setting the appropriate regulatory level.
Undeterred, EPA argues that several other provisions
militate against our understanding that “shall” means “shall.”
One such provision, known as the “anti-backslide” provision,

-- 12 of 31 --

13
requires EPA to review regulations at least every six years and
only permits revisions to those regulations that will “maintain,
or provide for greater, [health] protection.” Id. § 300g-1(b)(9).
In EPA’s view, because this provision applies only to existing
regulations, the agency is free to withdraw a regulatory
determination at any time before it promulgates final
regulations. EPA also argues that another provision that
renders a determination not to regulate subject to judicial
review, see id. § 300g-1(b)(1)(B)(ii)(IV), implicitly means that
a determination to regulate is not subject to judicial review and
can be withdrawn. Both arguments ignore the rigidity of the
statute, which creates only two possible courses of action after
the agency considers a contaminant for regulation: EPA can
either determine not to regulate, or it can determine to regulate
and then promulgate the regulations. If EPA takes the first
approach, there are no regulations to which the anti-backslide
provision can apply, and its determination is subject to judicial
review. If EPA takes the second approach, it must promulgate
regulations, to which the anti-backslide provision applies, and
which are themselves subject to judicial review. EPA’s attempt
to create a third option, one in which there are no regulations
to which the anti-backslide provision applies but whereby the
agency still evades judicial review, is inconsistent with the
statutory scheme.
Having run out of provisions in the statute that it views as
favorable, EPA flips the script and cites to the absence of a
provision governing how the agency must withdraw a
regulatory determination, arguing that because the agency did
not “contravene any express statutory command” or “avoid any
otherwise applicable statutory process” for withdrawing a
determination to regulate, its withdrawal was permissible.
Resp. Br. at 26–27 (internal citation omitted). But this just
repackages the already rejected argument that the agency
possesses an “inherent” authority to change its mind. Congress

-- 13 of 31 --

14
did not create a process for EPA to withdraw a regulatory
determination because it seemingly did not want EPA to have
the power to do so. “Regardless of how serious the [purported]
problem an administrative agency seeks to address, . . . it may
not exercise its authority ‘in a manner that is inconsistent with
the administrative structure that Congress enacted into law.’”
FDA v. Brown & Williamson Tobacco Corp., 529 U.S. 120,
125 (2000) (quoting ETSI Pipeline Project v. Missouri, 484
U.S. 495, 517 (1988)). Further, EPA’s withdrawal of its
regulatory determination did contravene an express statutory
command: the command that the agency “shall” regulate.
EPA eventually abandons the statute altogether, turning
instead to the Act’s legislative history. In its view, allowing
the agency to withdraw a regulatory determination would be
consistent with Congress’s intent in passing the 1996
amendments, which it argues was to grant the agency
additional discretion to decide when regulation is warranted
and to eliminate wasteful spending on regulations without
significant health benefits. EPA runs into two problems with
this argument. First, and most fundamentally, EPA’s
interpretation of what Congress intended in the statute cannot
overcome the statute’s directive that the agency “shall”
regulate. See Azar v. Allina Health Servs., 139 S. Ct. 1804,
1815 (2019) (“[M]urky legislative history . . . can’t overcome
a statute’s clear text and structure.”). And second, the history
EPA cites is not inconsistent with our interpretation. The 1996
amendments undoubtedly gave the agency more discretion in
determining which contaminants to consider for regulation and
whether to regulate them in the first instance. But they also
balanced that discretion with a mandatory scheme requiring
EPA to regulate after it determines to do so.
Because the Safe Drinking Water Act requires that the
agency “shall” regulate after making a regulatory

-- 14 of 31 --

15
determination, EPA lacks authority to withdraw that
determination and decide that it “shall not” regulate.
b. Arbitrary and Capricious Challenges
NRDC also challenges EPA’s decision-making process
that led to the withdrawal of its regulatory determination,
arguing that EPA acted arbitrarily and capriciously by setting
“levels of public health concern” that fail to protect against
adverse health effects and by relying on selectively updated
data to assess the frequency of perchlorate contamination.
Because we conclude that EPA lacked statutory authority to
withdraw its regulatory determination, we do not reach these
issues. See New York Stock Exch. LLC v. SEC, 962 F.3d 541,
559 (D.C. Cir. 2020). Having been apprised of NRDC’s
contentions and our concurring colleague’s views, EPA is, of
course, free to take those into consideration when it develops
perchlorate regulations.
III. Conclusion
We grant NRDC’s petition, vacate EPA’s withdrawal of
its determination to regulate, and remand to the agency for
further proceedings.

-- 15 of 31 --

P AN , Circuit Judge, concurring in the judgment: In 2011,
the EPA determined that perchlorate met the statutory factors
for regulation under the Safe Drinking Water Act. For
unexplained reasons, the agency missed the 24-month deadline
to propose the required regulation and the 42-month deadline
to promulgate it. Eight years passed. When the EPA finally
commenced the notice-and-comment process to limit
perchlorate in drinking water, the agency proposed establishing
a Maximum Contaminant Level Goal (“MCLG”) associated
with a two-point drop in the average IQ of the most sensitive
population. The agency also sought comments about
alternative MCLGs associated with one- or three-point drops
in the benchmark IQ. Moreover, the EPA requested comments
about whether perchlorate should be regulated at all, “in light
of new considerations . . . including information on lower
levels of occurrence of perchlorate than the EPA had
previously believed to exist and new analysis of the
concentration [of perchlorate] that represents a level of health
concern.” National Primary Drinking Water Regulations:
Perchlorate, 84 Fed. Reg. 30,524, 30,525 (June 26, 2019).
After the notice-and-comment period expired, the EPA decided
that the statutory factors for regulation were not met, and
therefore “withdr[e]w” its 2011 determination to regulate
perchlorate, based on an updated understanding of the “best
available public health information.” Drinking Water: Final
Action on Perchlorate, 85 Fed. Reg. 43,990, 43,992 (July 21,
2020). The agency’s withdrawal of its regulatory
determination relied on a MCLG associated with a one-point
drop in the average IQ of the most sensitive population, and on
a partial update of the data that the agency used to measure the
prevalence of perchlorate in the nation’s drinking water.
In my view, under the circumstances presented, the EPA
had authority to withdraw its initial regulatory determination
based on changes in the best available, peer-reviewed science.
But the agency’s ultimate decision not to regulate perchlorate
was arbitrary, capricious, and not in accordance with law,

-- 16 of 31 --

2
because it relied on a MCLG that did not meet the statutory
standard, as well as on a biased dataset that was selectively
updated. I would vacate and remand on those alternative
grounds, and therefore respectfully concur in the judgment.
I. Background
In 2011, the EPA published a final determination that
perchlorate met the requirements for regulation under the Safe
Drinking Water Act. See Drinking Water: Regulatory
Determination on Perchlorate, 76 Fed. Reg. 7,762 (Feb. 11,
2011). Specifically, it found that perchlorate (1) “may have an
adverse effect on the health of persons”; (2) “is known to occur
or there is a substantial likelihood that [perchlorate] will occur
in public water systems with a frequency and at levels of public
health concern”; and (3) “in the sole judgment of the
Administrator[ of the EPA], regulation of [perchlorate]
presents a meaningful opportunity for health risk reduction for
persons served by public water systems.” Id. at 7,762–63; see
also 42 U.S.C. § 300g-1(b)(1)(A)(i)–(iii). The statute required
the agency to propose a MCLG and an accompanying
regulation that would limit perchlorate in drinking water within
24 months after making that regulatory determination — i.e.,
by February 11, 2013. See 42 U.S.C. § 300g-1(b)(E).
But the EPA missed that deadline. In February of 2016,
the Natural Resources Defense Council (“NRDC”) brought suit
to compel the agency to issue the tardy perchlorate regulation.
See Compl., NRDC v. EPA, No. 2:16-cv-1251 (S.D.N.Y. Feb.
19, 2016), ECF No. 7. The parties entered a consent decree
that required the EPA to propose the MCLG and accompanying
regulation by May 28, 2019, and to finalize the MCLG and
accompanying regulation by June 19, 2020. See Consent
Decree and Extensions, NRDC v. EPA, No. 2:16-cv-1251
(S.D.N.Y.), ECF No. 38, 57, 59, 60.

-- 17 of 31 --

3
Pursuant to the consent decree, the EPA issued its proposal
for limiting the amount of perchlorate in drinking water in
2019. See 84 Fed. Reg. at 30,565. The agency proposed setting
both the MCLG and Maximum Contaminant Level (“MCL”)
at 56 micrograms per liter (μg/L), a level associated with “a 2
percent decrease in IQ” in the most sensitive population. Id. at
30,540. The proposal also requested public comments on three
alternatives: (1) setting the MCLG and MCL at 18 μg/L, a
level associated with a one-point drop in IQ; (2) setting the
MCLG and MCL at 90 μg/L, a level associated with a three-
point drop in IQ; or (3) withdrawing the agency’s 2011
determination to regulate perchlorate altogether, “based on new
information that indicates that perchlorate does not occur in
public water systems with a frequency and at levels of public
health concern and there may not be a meaningful opportunity
for health risk reduction.” Id. at 30,524, 30,541.
In 2020, the EPA withdrew the 2011 determination to
regulate perchlorate. The agency based its decision on a new
understanding of the scientific evidence regarding
perchlorate’s prevalence in drinking water and its effects on
human health. The agency “recognize[d] that the [Safe
Drinking Water Act] does not include a provision explicitly
authorizing withdrawal of a regulatory determination,” but
concluded that “such authority is inherent in the authority to
issue a regulatory determination . . . particularly given the
requirement that such determination be based on the ‘best
available public health information.’” 85 Fed. Reg. at 43,992.
The EPA explained that “new data and analysis developed by
the Agency as part of the 2019 proposal demonstrate that the
occurrence and health effects information used as the basis for
the 2011 determination no longer constitute ‘best available
information,’ are no longer accurate, and no longer support the
Agency’s prioritization of perchlorate for regulation.” Id.

-- 18 of 31 --

4
According to the EPA, the best available scientific
evidence had changed in two ways. First, the EPA refined its
understanding of the concentrations at which perchlorate
causes health problems. In 2011, the agency identified “levels
of public health concern” that “range[d] from 1 μg/L to 47
μg/L.” See 76 Fed. Reg. at 7,764. But the publication of
various studies in the following years shed further light on the
relationship between exposure to perchlorate and iodide
deficiencies in pregnant women that cause “a variety of adverse
neurodevelopmental outcomes” in their fetuses, including
decreases in IQ. See 84 Fed. Reg. at 30,531; 85 Fed. Reg. at
44,000 (identifying “the fetus of the iodide deficient pregnant
mother” as the most sensitive population). Using a new study
that the agency identified as “the most rigorous analysis
available in the literature to date,” see 84 Fed. Reg. at 30,534,
the EPA developed a new model that caused it to reconsider the
levels at which perchlorate is detrimental to health. Based on
that model, the agency increased the relevant “levels of public
health concern” from between 1 and 47 μg/L to between 18 and
90 μg/L. See 85 Fed. Reg. at 43,992.
Second, the EPA revised its data showing the prevalence
of perchlorate in the nation’s water supply. The agency based
its 2011 determination to regulate in part on the UCMR-1
study, a nationwide survey of perchlorate occurrence in
drinking water conducted between 2001 and 2005. See 76 Fed.
Reg. at 7,764–65. After the 2001–2005 period of data
collection for the UMCR-1 study, California and
Massachusetts enacted enforceable state-level perchlorate
drinking-water standards. See 85 Fed. Reg. at 43,995. The
EPA used compliance-monitoring information from those
states to update some of the UMCR-1 data points. Id. The
updates for California, in particular, had the potential to
disproportionately affect the national picture of perchlorate
occurrence: “In the original UCMR 1 dataset . . . 320 of 540

-- 19 of 31 --

5
samples in which perchlorate was detected were in California.”
EPA Br. 52; see also Env’t Prot. Agency, EPA 816-R-19-003,
Perchlorate Occurrence and Monitoring Report App. D at D-
3 (May 2019). Based on the updated data, the EPA concluded
that perchlorate did not occur in public water systems at the
requisite levels of public health concern to justify regulation.
See 85 Fed. Reg. at 43,992.
II. Authority to Withdraw a Regulatory
Determination
Under the Safe Drinking Water Act, the EPA’s issuance of
a regulatory determination triggers a duty to propose and
promulgate an appropriate regulation. As my colleagues in the
majority note, the statute imposes that duty by repeatedly using
the word “shall.” See Maj. Op. 9–10. Specifically, the EPA
Administrator (1) “shall, in accordance with the procedures
established by this subsection, publish a [MCLG] and
promulgate a[n accompanying] regulation for a contaminant . .
. if the Administrator determines that” the statutory factors are
met, 42 U.S.C. § 300g-1(b)(1)(A) (emphasis added); (2) “shall
publish [MCLGs] and promulgate, by rule,” accompanying
regulations “[f]or each contaminant that the Administrator
determines to regulate,” id. § 300g-1(b)(1)(E) (emphasis
added); (3) “shall propose” the MCLG and accompanying
regulation within 24 months of the determination to regulate,
id. (emphasis added); and (4) “shall publish” the MCLG and
accompanying regulation within 18 months of the proposal
(with a possible nine-month extension), id. (emphasis added).
Based on this language, the majority concludes that when
the EPA initially determined to regulate perchlorate in 2011,
the issuance of regulations became mandatory — full stop. See
Maj. Op. 9–10. But in my view, the question here is not
whether the existence of a regulatory determination gives rise

-- 20 of 31 --

6
to a duty to actually regulate. It undoubtedly does. Instead, the
salient question is whether the agency may withdraw its
determination to regulate based on changed circumstances,
thereby vitiating the agency’s obligation to proceed with
regulation. It is the regulatory determination that kicks off all
the statutory timelines and imposes on the agency a firm
obligation to regulate; if that determination is withdrawn, those
attendant requirements are no longer operative.
The withdrawal of a determination to regulate in this
context appears to be unprecedented. This may be explained
by the 24-month statutory deadline to propose a regulation: A
significant change in the underlying science is unlikely to occur
in that relatively short timeframe. But here, eight years passed
after the EPA issued its initial determination. At that point,
after undergoing a notice-and-comment procedure, the agency
made a new determination based on updated information. The
issue before us is not whether the EPA should have violated the
Safe Drinking Water Act’s statutory deadline — it should not
have. But, now that it has, we must consider whether the EPA
has the power to withdraw a regulatory determination when
changed circumstances justify such a withdrawal. In my view,
the agency surely has that authority.
Contrary to my colleagues’ view, nothing in the Safe
Drinking Water Act forbids the EPA from withdrawing a
determination to regulate. The statute is silent on that issue.
But reading such a prohibition into the Safe Drinking Water
Act would force the EPA to violate another statutory provision.
Specifically, the statute obligates the EPA to use “the best
available, peer-reviewed science and supporting studies
conducted in accordance with sound and objective scientific
practices” in every action under the Safe Drinking Water Act
that “is based on science.” 42 U.S.C. § 300g-1(b)(3)(A)(i). If
new information comes to light before the EPA proposes a

-- 21 of 31 --

7
MCLG and accompanying regulation, and the “best available,
peer-reviewed science” makes clear that the initial regulatory
determination is no longer supported by the evidence, then
proceeding to regulate despite that new evidence would violate
this provision. We obviously should not adopt an interpretation
of the statute that discounts or ignores the EPA’s duty to rely
on the best available science. See Loughrin v. United States,
573 U.S. 351, 358 (2014) (explaining the “‘cardinal principle’
of interpretation that courts ‘must give effect, if possible, to
every clause and word of a statute’” (quoting Williams v.
Taylor, 529 U.S. 362, 404 (2000))); Del. Dep’t of Nat. Res. &
Env’t Control v. EPA, 895 F.3d 90, 99 (D.C. Cir. 2018) (“[W]e
strive to construe statutes ‘so that effect is given to all its
provisions, so that no part will be inoperative or superfluous,
void or insignificant.’” (quoting Corley v. United States, 556
U.S. 303, 314 (2009))).
The majority asserts that the agency can reconcile these
statutory mandates by “apply[ing] the ‘best available, peer-
reviewed science,’ including any new developments, to set the
substance of the regulations — not to reevaluate whether to
regulate.” Maj. Op. 12; see also NRDC Br. 34–35 (making a
similar argument). But that fails to account for circumstances
where, as here, the agency concludes that the best available,
peer-reviewed science does not support regulating a
contaminant at all. Under the majority’s approach, the agency
is forced to regulate anyway. The regulation of a contaminant
entails setting a MCLG and a MCL, which in turn triggers
potentially costly testing requirements. See 42 U.S.C.
§ 300g-3(c)(1)(A)(i) (requiring public water systems to notify
customers of “any failure” to “comply with an applicable
maximum contaminant level”); 40 C.F.R. § 141.23 (requiring
water systems to “conduct monitoring to determine compliance
with . . . maximum contaminant levels”). The majority’s
interpretation gives the agency no choice but to impose a

-- 22 of 31 --

8
pointless burden on water systems to test for a substance that
the agency does not even think should be regulated. Congress
could not have intended “such an illogical result.” Tri-State
Hosp. Supply Corp. v. United States, 341 F.3d 571, 577 (D.C.
Cir. 2003); see also Mova Pharm. Corp. v. Shalala, 140 F.3d
1060, 1068 (D.C. Cir. 1998) (highlighting “the long-standing
rule that a statute should not be construed to produce an absurd
result,” i.e., a “result [that] is contrary to common sense” and
“inconsistent with the clear intentions of the statute’s
drafters”).
For its part, NRDC contends that the “best available, peer-
reviewed science” provision requires the EPA to consider only
the best evidence that was available “at the time” of the original
determination, i.e., in 2011. See NRDC Br. 35–36. Citing our
decision in Chlorine Chemistry Council v. EPA, NRDC relies
on our statement “that the action [should] be taken on the basis
of the best available evidence at the time of the rulemaking.”
206 F.3d 1286, 1291 (D.C. Cir. 2000) (emphasis in original).
But Chlorine Chemistry Council involved very different facts.
In that case, the EPA refused to establish a MCLG supported
by the best available evidence “because of the possibility of
contradiction in the future by evidence unavailable at the time
of the action.” Id. at 1290–91. In this case, the EPA is not
arguing that the currently available evidence might be
contradicted in the future, but that the currently available
evidence does contradict the agency’s past understanding of the
science. Moreover, although the EPA made a regulatory
determination in 2011, the “time of the rulemaking” in this case
was in 2019, when the EPA sought comments about its
proposed MCLG and accompanying regulation. See 5 U.S.C.
§ 551(5) (“‘[R]ule making’ means agency process for
formulating, amending, or repealing a rule[.]”). If the best
available evidence at that later time revealed that the statutory
prerequisites for regulation were not met, then the agency’s

-- 23 of 31 --

9
only way forward was to withdraw its earlier decision to the
contrary. Otherwise, the EPA would be simultaneously
forbidden yet compelled to rely on “the best available, peer-
reviewed science.” 42 U.S.C. § 300g-1(b)(3)(A)(i).
Notably, Congress clearly knew how to limit the agency’s
ability to change its mind and chose to do so only later in the
Safe Drinking Water Act’s regulatory process. Specifically,
the statute’s anti-backsliding provision applies after
promulgation of the MCLG and accompanying regulation. See
id. § 300g-1(b)(9). It mandates that “[a]ny revision of a
national primary drinking water regulation . . . shall maintain,
or provide for greater, protection of the health of persons.” Id.
(emphasis added). That provision is inoperative here because
the EPA made (and then withdrew) only a determination to
regulate. Nevertheless, Congress’s enactment of a specific
limitation on the EPA’s ability to revisit and alter drinking-
water regulations under the Safe Drinking Water Act highlights
the absence of any explicit limitation on changing regulatory
determinations in this context. The statutory text thus strongly
suggests that there is no implicit constraint on the agency’s
ability to reconsider a regulatory determination. See Jama v.
Immigr. & Customs Enf’t, 543 U.S. 335, 341 (2005) (“We do
not lightly assume that Congress has omitted from its text
requirements that it nonetheless intends to apply, and our
reluctance is even greater when Congress has shown elsewhere
in the same statute that it knows how to make such a
requirement manifest.”).
My conclusion that the EPA may withdraw a regulatory
determination is consistent with the ordinary rule that agencies
may “use the same procedures when they amend or repeal a
rule as they used to issue the rule in the first instance.” Perez
v. Mortg. Bankers Ass’n, 575 U.S. 92, 101 (2015) (citing FCC
v. Fox Television Stations, Inc., 556 U.S. 502, 515 (2009));

-- 24 of 31 --

10
Hickman & Pierce, Administrative Law Treatise § 4.5.1 (6th
ed. 2019) (“In the ordinary course, legislative rules must be
promulgated using notice and comment procedures and can
only be modified or replaced using notice and comment
procedures.”). Indeed, it is a core principle of administrative
law “that an agency must be given ample latitude to ‘adapt their
rules and policies to the demands of changing circumstances.’”
See Motor Vehicles Ass’n of U.S., Inc. v. State Farm Mut. Auto.
Ins. Co., 463 U.S. 29, 42 (1983) (citation omitted)). If
Congress had intended to depart from these well-established
principles in the present context, it would have spoken to that
issue directly. Cf. Ivy Sports Med., LLC v. Burwell, 767 F.3d
81, 86 (D.C. Cir. 2014). As it did not, the EPA changed its
mind in the way that agencies routinely do: It made its 2011
determination to regulate after notice and comment; then, eight
years later, it went through another round of notice and
comment before deciding to change course.1
Finally, it bears mention that if, after “employing
traditional tools of statutory construction,” the Safe Drinking
Water Act remained ambiguous about whether the EPA can
withdraw a regulatory determination, the agency’s
interpretation ordinarily would be entitled to deference. See
1 This case does not fall within the exception that applies when
Congress, by providing an alternative statutory mechanism to correct
mistakes, restricts the means through which an agency can change
course. See, e.g., Ivy Sports, 767 F.3d at 86 (holding that agency
lacked authority to reconsider prior decision where Congress
“creat[ed] . . . a specific statutory mechanism to correct alleged . . .
errors”); New Jersey v. EPA, 517 F.3d 574, 582–83 (D.C. Cir. 2008)
(similar); Am. Methyl Corp. v. EPA, 749 F.2d 826, 835 (D.C. Cir.
1984) (similar). To the contrary, denying the EPA the ability to
withdraw its regulatory determination under the present
circumstances leaves the agency with no mechanism at all to alter
what it later concluded was an incorrect decision.

-- 25 of 31 --

11
Chevron, U.S.A., Inc. v. NRDC, 467 U.S. 837, 843 n.9 (1984).
But because the EPA did not cite Chevron in its brief and
avoided relying on it at oral argument, see Oral Arg. Tr. 27:15–
18, I decline to consider the applicability of Chevron here. See
HollyFrontier Cheyenne Refin., LLC v. Renewable Fuels
Ass’n, 141 S. Ct. 2172, 2180 (2021) (“[T]he government does
not [invoke Chevron.] . . . We therefore decline to consider
whether any deference might be due its regulation.”).
In short, under the circumstances presented, it was
permissible for the EPA to reconsider and withdraw a
determination to regulate a contaminant under the Safe
Drinking Water Act. The agency had not yet proposed and
promulgated a final regulation when it made a new finding that
the best available, peer-reviewed science no longer supported
its prior regulatory determination. In my view, the EPA may
appropriately reverse a decision to regulate based on a change
in scientific evidence, after engaging in notice-and-comment
procedures.
III. Withdrawal of the 2011 Determination to
Regulate Perchlorate
Although the EPA was empowered to reconsider its initial
regulatory determination based on changes in the best
available, peer-reviewed science, the agency’s ultimate
decision not to regulate perchlorate in drinking water was
arbitrary, capricious, and otherwise not in accordance with law.
See 5 U.S.C. § 706(2)(A); see also Fox Television Stations, 556
U.S. at 514–16; Nat’l Ass’n of Home Builders v. EPA, 682 F.3d
1032, 1037–38 (D.C. Cir. 2012). An agency action is arbitrary
or capricious “if the agency has relied on factors which
Congress has not intended it to consider, entirely failed to
consider an important aspect of the problem, offered an
explanation for its decision that runs counter to the evidence

-- 26 of 31 --

12
before the agency, or is so implausible that it could not be
ascribed to a difference in view or the product of agency
expertise.” State Farm, 463 U.S. at 43. An agency action that
“violates [a statute] is ‘not in accordance with law’ within the
meaning of 5 U.S.C. § 706(2)(A).” Chrysler Corp. v. Brown,
441 U.S. 281, 318 (1979); see also Ethyl Corp. v. EPA, 306
F.3d 1144, 1150 (D.C. Cir. 2002) (holding that agency action
that violated authorizing statute was “not in accordance with
law”).
A. Proposed MCLGs
The Safe Drinking Water Act requires the EPA to “set [the
MCLG] at the level at which no known or anticipated adverse
effects on the health of persons occur and which allows an
adequate margin of safety.” 42 U.S.C. § 300g-1(b)(4)(A)
(emphasis added). But the EPA sought comments on MCLGs
that permitted levels of perchlorate associated with one-, two-,
and three-point decreases in the average IQ of the most
sensitive population; and then used those MCLGs as the “levels
of public health concern” by which it evaluated the need to
limit perchlorate in drinking water. See 85 Fed. Reg. at 43,995
(“[T]he EPA used these potential MCLGs as the levels of
public health concern in assessing the frequency of occurrence
of perchlorate in this regulatory determination.”).2 The EPA
2 The EPA characterizes its proposed MCLGs of 18 μg/L,
56 μg/L, or 90 μg/L as the levels of perchlorate that “avoid” IQ
decreases of one, two, or three points, respectively. See 85 Fed. Reg.
at 43,994, 43,995, 43,999; EPA Br. 2, 11, 16, 33, 34, 39, 45. That
characterization is at best confusing and at worst misleading. The
proposed MCLGs are the levels of perchlorate associated with
decreases in IQ of one, two, or three points — not the levels at which
those cognitive harms do not occur. See 84 Fed. Reg. at 30,536
(explaining that the agency based its calculations on the daily dose
of perchlorate “associated with a 1, 2, or 3 point decrease from the

-- 27 of 31 --

13
found that “even at the most stringent regulatory level
considered in the 2019 proposal” — i.e., the level associated
with a one-point drop in IQ — “perchlorate does not occur in
public water systems ‘with a frequency and at levels of public
health concern.’” Id. at 43,992.
The EPA’s proposed MCLGs plainly violated the statutory
mandate to reflect “the level at which no known or anticipated
adverse effects on the health of persons occur.” 42 U.S.C.
§ 300g-1(b)(4)(A). A decrease in average IQ of even one point
is undoubtedly an “adverse effect[] on the health of persons.”
Id. Rather than debate that self-evident conclusion, the EPA
chooses the path of obfuscation, essentially arguing that in this
“complicated technical area,” the court must defer to the
agency’s chosen approach to regulation. EPA Br. 34, 37.
Specifically, the EPA says that we should defer to its reliance
on the agency’s “‘Benchmark Dose Guidance,’ which
supported using a 1% effect [on IQ] as a starting point.” Id. at
34. That guidance focuses not on “what individual levels [of a
contaminant] can be considered adverse” but instead on “the
level of change in the endpoint [here, IQ] at which the effect is
considered to become biologically significant (as determined
by expert judgment or relevant guidance documents).” See
Env’t Prot. Agency, Benchmark Dose Technical Guidance x
(June 2012). In other words, the agency relied on its own
judgment about whether an adverse health effect is
standardized mean IQ”). At oral argument, the EPA contended that
setting MCLGs at these levels would “avoid” the relevant IQ
decreases because regulating at those levels of perchlorate would
prevent water from containing those levels of perchlorate. See Oral
Arg. Tr. 37:24–38:14, 39:7–9. But the truth appears to be the
opposite: Setting a maximum of 18 μg/L would not avoid water
having that much perchlorate, it would permit that level of
perchlorate — a level that, according to the EPA’s own data, is
associated with a one-point decline in IQ.

-- 28 of 31 --

14
“biologically significant” instead of adhering to the statutory
standard, which requires setting the MCLG “at the level at
which [there are] no known or anticipated adverse effects on
the health of persons.” 42 U.S.C. § 300g-1(b)(4)(A). Notably,
the EPA used the Benchmark Dose Guidance to replace the
“NOAEL/LOAEL [no or low observed adverse effect level]
approach,” which had been “used for many years” and which
mirrors what the statute requires. See Benchmark Dose
Technical Guidance viii. One need not be a scientist to
understand that by rejecting the “no observed adverse effect
level” approach, the EPA eschewed what the Safe Drinking
Water Act demands. Compare id., with 42 U.S.C.
§ 300g-1(b)(4)(A).
Beneath the technical jargon and puffery about agency
expertise, the EPA is not really arguing that it complied with
the statute. Instead, the agency appears to contend that the
statute’s requirements are not the best way to go about making
policy in this area, and that its own judgment should control.
Of course, that position finds no support in the law. Congress
directed the EPA to set MCLGs “at the level at which no known
or anticipated adverse effects on the health of persons occur.”
42 U.S.C. § 300g-1(b)(4)(A). The agency did not act in
accordance with that law when it used a MCLG associated with
a one-point drop in IQ — which plainly is an “adverse effect[]
on the health of persons,” id. — as the basis for withdrawing
its determination to regulate perchlorate.
B. UCMR-1 Data
As previously noted, the EPA relied on data from the
UCMR-1 — a nationwide sampling of public water systems for
perchlorate — when it made its determination to regulate in
2011. See 76 Fed. Reg. at 7,764. The UCMR-1 survey
detected “perchlorate at levels greater than or equal to 4 μg/L .

-- 29 of 31 --

15
. . [in] approximately 1.9 percent of the” samples collected
during the 2001–2005 study period. Id. After the UCMR-1
study, however, Massachusetts and California passed
enforceable state-level standards on perchlorate in drinking
water. As a result, statewide monitoring data of perchlorate
have become available in those states. See 85 Fed. Reg. at
43,995.
When the EPA reconsidered its determination to regulate
in 2019, it updated the UCMR-1 data to reflect some, but not
all, of the newly available information from Massachusetts and
California. Instead of replacing all the data points from
Massachusetts and California in the UCMR-1 dataset, the
agency updated only those samples where perchlorate was
detected during the 2001–2005 data collection. See id. In other
words, it revised any samples that were positive for perchlorate
in the 2001–2005 data, but left untouched those samples that
were negative for perchlorate. Thus, as NRDC argues, the
“EPA set up a one-way ratchet: [C]ontaminated water could
become clean, but clean water could not become
contaminated.” NRDC Br. 62. If the purpose of adjusting the
UCMR-1 data was to obtain the most current systematic picture
of perchlorate occurrence, there was no apparent reason (and
the agency proffers none) to selectively update the data from
Massachusetts and California instead of using all the new data
from those states. The EPA’s evident failure “to consider [that]
important aspect of the problem” was arbitrary and capricious.
State Farm, 463 U.S. at 43.
IV. Conclusion
For the foregoing reasons, I would hold that the EPA has
authority to withdraw a determination to regulate a
contaminant under the Safe Drinking Water Act, prior to the
promulgation of a MCLG and accompanying regulation, when

-- 30 of 31 --

16
the best available science supports the agency’s conclusion that
the required factors for regulation are no longer met. But in my
view, the EPA’s 2020 decision not to regulate perchlorate was
arbitrary, capricious, and contrary to law because it was based
on a MCLG that did not comply with a statutory directive, and
relied on selectively updated data concerning the prevalence of
perchlorate in drinking water. On those grounds, I would
vacate the agency’s withdrawal of its 2011 regulatory
determination and remand for further proceedings. I therefore
concur in the judgment.

-- 31 of 31 --