FOR PUBLICATION

UNITED STATES COURT OF APPEALS

FOR THE NINTH CIRCUIT

NATURAL RESOURCES DEFENSE No. 20-72794

COUNCIL,

Petitioner, EPA No.

EPA-HQ-OPP-

v. 2009-0308

U.S. ENVIRONMENTAL PROTECTION

AGENCY; MICHAEL S. REGAN, in his OPINION official capacity as Administrator of the United States Environmental Protection Agency, Respondents,

THE HARTZ MOUNTAIN

CORPORATION,

Respondent-Intervenor.

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

Argued and Submitted November 22, 2021 San Francisco, California

Filed April 20, 2022 Before: Mary H. Murguia, Chief Judge, and R. Guy Cole, Jr. * and Ronald M. Gould, Circuit Judges.

Opinion by Judge Gould

SUMMARY **

Pesticides

The panel vacated the U.S. Environmental Protection Agency’s (“EPA”) denial of Natural Resources Defense Council (“NRDC”)’s petition to cancel the registration of the pesticide tetrachlorvinphos (“TCVP”), and remanded for a revised EPA response within 120 days.

Under the Federal Insecticide, Fungicide, and Rodenticide Act, pesticides sold in the United States generally must be registered by the EPA. Private parties can petition the EPA to cancel the registration of a pesticide, and the EPA is required to resolve those petitions “within a reasonable time.” 5 U.S.C. § 555(b). The EPA last approved the use of TCVP pesticide in pet products in 2006. In 2009, NRDC petitioned the EPA to cancel the registration of TCVP for use in household pet products. After repeated delays, the EPA eventually denied NRDC’s petition.

* The Honorable R. Guy Cole, Jr., United States Circuit Judge for the U.S. Court of Appeals for the Sixth Circuit, sitting by designation. ** This summary constitutes no part of the opinion of the court. It has been prepared by court staff for the convenience of the reader.

NRDC V. USEPA 3

The panel held that the EPA’s denial of NRDC’s petition was not supported by substantial evidence. The EPA failed to provide a reasoned explanation for its denial of NRDC’s petition and made several arbitrary calculations. The EPA’s errors primarily impacted two calculations central to its denial of NRDC’s petition: (1) the amount of TCVP dust released by the pet collars, and (2) the assumption that pet owners will trim the collars by at least 20%. The EPA without any explanation rejected a central finding of a study – the Torison Study – that it repeatedly stated was a key to its determination. In its brief before this court, the EPA for the first time gave a justification for its decision to reject the Torison Study’s finding that 97.2% of the dust released from the collars comprises TCVP. The panel held that it could only uphold agency action based on the reasons the agency gave for its decision. The panel held further that the EPA’s assumption that only 14.6% of the dust released from the collars was TCVP – instead of the Torison Study’s measurement of 97.2% – was also troubling on the merits. The panel held that it would not defer to the EPA’s highly inaccurate calculation that pet owners will trim pet collars by 20% when fitting the collar onto a pet’s neck. The panel concluded that it was apparent that the EPA’s denial of NRDC’s petition was simply not supported by substantial evidence when considered on the record as a whole. Peter J. DeMarco (argued) and Aaron Colangelo, Natural Resources Defense Council, Washington, D.C.; Ian Fein, Natural Resources Defense Council, San Francisco, California; for Petitioner.

Gus Maxwell (argued), Attorney; Jean E. Williams, Acting Assistant Attorney General; Environment and Natural Resources Section, United States Department of Justice, Washington, D.C.; Benjamin Wakefield and Erin Koch, Attorneys, United States Environmental Protection Agency, Washington, D.C.; for Respondents.

Amanda Shafer Berman (argued), Kirsten L. Nathanson, and Michael Boucher, Crowell & Moring LLP, Washington, D.C., for Respondent-Intervenor.

OPINION

GOULD, Circuit Judge:

The Natural Resources Defense Council (“NRDC”) again asks us to intervene in its thirteen-year dispute with the Environmental Protection Agency (“EPA”) regarding whether EPA is fulfilling its statutory duty to properly assess the risks of the pesticide tetrachlorvinphos (“TCVP”). After we granted NRDC’s writ of mandamus in April 2020, which required EPA to issue a final response to NRDC within 90 days, EPA denied NRDC’s petition to cancel the registration of TCVP. NRDC now argues that EPA erred when it denied the petition because EPA failed to provide an adequate explanation for its decision and relied on several mistaken calculations. Because EPA’s denial of NRDC’s

NRDC V. USEPA 5

petition lacks substantial evidence, we vacate EPA’s denial and remand for a revised EPA response within 120 days.

I. BACKGROUND

Under the Federal Insecticide, Fungicide, and Rodenticide Act, pesticides sold in the United States generally must be registered by EPA. See 7 U.S.C. § 136a(a). Private parties can petition EPA to cancel the registration of a pesticide. 40 C.F.R. § 154.10. EPA is required to resolve those petitions “within a reasonable time.” 5 U.S.C. § 555(b).

EPA last approved the use of TCVP pesticide in pet products in 2006. In 2009, NRDC petitioned EPA to cancel the registration of TCVP for use in household pet products. TCVP is a type of organophosphate pesticide. Organophosphates “were developed from nerve warfare agents used during World War II” and “pose recognized dangers to the neurodevelopment of children, causing reduced cognitive capacity, delays in motor development, and behavioral problems.” In re Nat. Res. Def. Council, Inc. (NRDC 2020), 956 F.3d 1134, 1136 (9th Cir. 2020). Of concern here, TCVP is used in pet collars to prevent fleas and ticks, and TCVP can be transferred to humans who come into contact with pets wearing these collars. TCVP pet collars are designed to release TCVP gradually onto pet fur in either liquid or dust form. NRDC is concerned that TCVP dust, which is more easily transferred to humans through contact with pets than TCVP liquid, poses a particular risk to young children who are more likely to put their hands in their mouths and ingest TCVP after petting an animal wearing a TCVP collar. Hartz Mountain Corporation (“Hartz”) manufactures pet collars containing TCVP, and EPA estimates that collars using alternative pesticides cost five to six dollars more a month. After filing its petition in 2009, NRDC waited five years for a response from EPA until it sought a writ of mandamus to compel EPA’s answer in 2014. In November 2014, EPA denied NRDC’s petition. NRDC then challenged the denial of its petition in this Court as unlawful. A few months later, EPA filed a motion for voluntary remand because it was completing a new risk assessment that could change its position regarding NRDC’s petition. EPA repeatedly represented that it “intend[ed] to issue a revised response to NRDC’s petition within 90 days after finalizing the [revised] risk assessment.” Id. at 1137. Over NRDC’s objections, we remanded without a deadline in June 2016. See Order, Nat. Res. Def. Council v. U.S. EPA, No. 15-70025, ECF No. 30 (June 9, 2016.

In December 2016, EPA issued a revised risk assessment, which found that TCVP exposure could lead to health risks for young children (the “2016 Risk Assessment”). The 2016 Risk Assessment “recognized that children could be exposed to TCVP through contact with pets using TCVP products and that such exposure posed considerable risks to their health.” NRDC 2020, 956 F.3d at 1137. The 2016 Risk Assessment found risks of concern for children “regardless of the ratio of liquid/dust assumed.” EPA represented that it would reply to NRDC’s petition within 90 days and issued a press release stating that it had “identified potential risks to people, including children, . . . which exceed the Agency’s level of concern.” Id. at 1142.

EPA, however, did not issue a response to NRDC’s petition within 90 days. Instead, two years after EPA’s self- imposed deadline, NRDC sought a second writ of mandamus from us to compel EPA to issue a final response to its petition filed a decade earlier. EPA defended its delay by stating that it needed additional data from Hartz, specifically

NRDC V. USEPA 7

the results of a torsion study, which EPA had originally requested in 2017. Five days after NRDC filed its second request for a writ of mandamus, EPA for the first time compelled Hartz to perform the torsion study (the “Torsion Study”), which was completed in August 2019.

In April 2020, we granted NRDC’s second petition for a writ of mandamus. We reprimanded EPA, stating

[T]he EPA’s years-long delay on this critical matter of public health has been nothing short of egregious. For more than a decade, the EPA has frustrated NRDC’s ability to seek judicial review by withholding final agency action, all the while endangering the wellbeing of millions of children and ignoring its core mission of protecting human health and the environment.

Id. at 1142–43 (internal citation omitted). We “order[ed] the EPA to issue a full and final response to” NRDC within 90 days. Id. at 1143. Exactly 90 days after our ruling, EPA again denied NRDC’s petition.

EPA based its denial on a revised risk assessment (the “2020 Risk Assessment”) dated one day before it denied NRDC’s petition. The 2020 Risk Assessment concluded that all seven TCVP pet collars of Hartz then on the market were unacceptably dangerous to human health because they had a margin of exposure surpassing the threshold of unacceptable risk, known as the Level of Concern (“LOC”). In response, Hartz requested voluntary cancellation of one of its feline collars and mitigated the exposure of TCVP on its six remaining collars by amending their instructions, reducing the amount of TCVP in the collars, and implementing other design changes. With these changes, EPA determined that Hartz’s remaining and modified pet collars did not exceed the LOC and therefore denied NRDC’s petition. After Hartz’s modifications, TCVP comprised roughly 14.6% of the remaining collars’ weight.

EPA’s denial of NRDC’s petition relied heavily on two studies completed by Hartz. First, it relied on the Torsion Study, which in part determined that: (1) 0.38% of the pet collars’ total mass would be released as dust; (2) approximately 97.2% of the dust released from the collars consisted of TCVP; and (3) an average of 1.1% of the TCVP in the collars was released as dust before any twisting. Second, EPA relied on Hartz’s Normal Wear Study, which measured the amount of TCVP and non-pesticides released by the collars after they were worn by dogs for between one and twenty-one days. Researchers wiped the collars after each period of use, and then measured the amount of released TCVP dust. In the Normal Wear Study, 4% of the collars’ total observed weight loss was measured as TCVP after three days.

NRDC timely challenged EPA’s denial of its petition.

II. STANDARD OF REVIEW

We review the denial of NRDC’s petition for “substantial evidence when considered on the record as a whole.” 7 U.S.C. § 136n(b). Substantial evidence is “more than a mere scintilla but less than a preponderance; it is such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” Nat. Res. Def. Council v. U.S. EPA (NRDC 2013), 735 F.3d 873, 877 (9th Cir. 2013) (quoting Vasquez v. Astrue, 572 F.3d 586, 591 (9th Cir. 1989)).

NRDC V. USEPA 9

III. DISCUSSION

a. Legal Standard

“It is well-established that an agency’s action must be upheld, if at all, on the basis articulated by the agency itself.” Id. at 877 (quoting Motor Vehicle Mfrs. Ass’n of the U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 50 (1983)). Courts do not “accept appellate counsel’s post-hoc rationalizations for agency action.” Nat. Res. Def. Council v. U.S. EPA (NRDC 2017), 857 F.3d 1030, 1040 (9th Cir. 2017) (quoting Hernandez-Cruz v. Holder, 651 F.3d 1094, 1109 (9th Cir. 2011)). “If the agency did not meet its burden, we ‘should not attempt . . . to make up for such deficiencies’ and ‘may not supply a reasoned basis for the agency’s action that the agency itself has not given.’” Ctr. for Biological Diversity v. Haaland, 998 F.3d 1061, 1067 (9th Cir. 2021) (quoting State Farm, 463 U.S. at 43).

Administrative law does not require agency perfection. The agency’s decision, even if of “less than ideal clarity,” will be upheld “if the agency’s path may reasonably be discerned.” State Farm, 463 U.S. at 43. It is fundamental that the agency must have “considered the relevant factors and articulated a rational connection between the facts found and the choices made.” Ctr. for Biological Diversity, 998 F.3d at 1067 (quoting Alaska Oil & Gas Ass’n v. Pritzker, 840 F.3d 671, 675 (9th Cir. 2016)). “Unsubstantiated” or “bare assumptions” will not be credited. NRDC 2017, 857 F.3d at 1038, 1042; see also Nat’l Parks Conservation Ass’n v. U.S. EPA, 788 F.3d 1134, 1143 (9th Cir. 2015) (“[U]nexplained assertions . . . unsupported by any explained reasoning” were “arbitrary and capricious”); Greater Yellowstone Coal., Inc. v. Servheen, 665 F.3d 1015, 1020 (9th Cir. 2011) (requiring agency to “articulate a rational connection between the data in the record and its determination”). Conversely, a “mere[] . . . reasonable basis for disagreement” about the evidence will not disturb an agency’s otherwise reasonable findings. NRDC 2013, 735 F.3d at 880; see also ASARCO, Inc. v. Occupational Safety & Health Admin., 746 F.2d 483, 490 (9th Cir. 1984).

Courts give special deference to agency interpretations of scientific issues. NRDC 2017, 857 F.3d at 1036 (internal citation omitted) (“When, as in this case, the agency is making predictions, within its area of special expertise, at the frontiers of science . . . a reviewing court must generally be at its most deferential.”); United States v. Alpine Land & Reservoir Co., 887 F.2d 207, 213 (9th Cir. 1989). Even so, we do not credit “arbitrary and highly inaccurate calculations.” Trs. of Cal. State Univ. v. Riley, 74 F.3d 960, 967 (9th Cir. 1996), as amended (Feb. 20, 1996).

b. Substantial evidence does not support EPA’s denial of NRDC’s Petition

For the following reasons, we conclude that EPA’s denial of NRDC’s petition is not supported by substantial evidence. EPA failed to provide a reasoned explanation for its denial of NRDC’s petition and made several arbitrary calculations. EPA’s errors primarily impact two calculations central to its denial of NRDC’s petition: (1) the amount of TCVP dust released by the collars, and (2) the assumption that pet owners will trim the collars by at least 20%. Further, we cannot consider EPA’s post-hoc rationalizations, which themselves suffer from numerous flaws.

i. The amount of TCVP dust released by the collars

When calculating the amount of TCVP dust released by Hartz’s pet collars, EPA relied on only some of the Torsion

NRDC V. USEPA 11

Study’s findings. In particular, EPA accepted the Torsion Study’s finding that 0.38% of the collars’ total weight was released as dust (and the rest released as liquid). However, EPA did not use one of the Torsion Study’s other main findings: that about 97.2% of the dust released from the collars consisted of TCVP. Instead, EPA assumed that the dust lost by the collars contained only 14.6% TCVP—an amount equivalent to the percent of TCVP in the collars as a whole. This second assumption, which directly contradicted the results of the Torsion Study and was unexplained, had the effect of significantly lowering EPA’s estimate of how much TCVP dust was released by the collars.

EPA’s selective use of the Torsion Study is problematic for several reasons. Before its denial of NRDC’s petition— including in its briefing to us—EPA repeatedly emphasized the importance of the Torsion Study, variously referring to it as the “best means,” the “optimal method,” and “most promising solution” for determining the physical form of the TCVP released from the collars. 1 In assessing the study, EPA further noted that “the purpose of the [Torsion Study] was to measure the amount of solids (dust/powder) released from a flea and tick collar and the active ingredient content

1 See EPA’s Opp. to NRDC Pet. for a Writ of Mandamus, NRDC v. U.S. EPA, No. 19-71324, ECF No. 13 (Sept. 9, 2019), 15 (“EPA has concluded that the best means of determining the form of TCVP released from the collars is through a mechanical torsion study where the collar is twisted and stretched repeatedly to exaggerate the movement of the collar against the animal.”), 16 (“[T]he Agency continues to believe that a mechanical torsion study is needed.”), 20 (“[T]he Agency has determined that the optimal method for reducing the uncertainty relating to the physical form of TCVP is to require the registrant to conduct a composition study in the form of a mechanical torsion test.”), Reaves Decl. ¶ 23 (“EPA has determined that the most promising solution for identifying the physical form of TCVP released from each pet collar is to require the registrant to conduct . . . a mechanical torsion study.”). of the released solids when the collar is exposed to mechanical torsion and stress.” However, in sharp contrast to its earlier statements, in its denial of NRDC’s petition and in the 2020 Risk Assessment, EPA provides no explanation for why it rejected the Torsion Study’s calculation of TCVP dust released from the collars. This is despite its earlier statements that the Torsion Study was the best method for measuring the form of TCVP released by the collars. Without any explanation, EPA’s rejection of a central finding of a study it repeatedly stated was key to its determination is troubling. See NRDC 2013, 735 F.3d at 877.

In its brief, EPA for the first time gives a justification for its decision to reject the Torsion Study’s finding that 97.2% of the dust released from the collars comprises TCVP. EPA now claims that it rejected this measurement because the Torsion Study used “exaggerated twisting conditions . . . [which] likely overrepresented the amount of [TCVP] in dust extruded from the collars in normal use on pets.” Even assuming EPA’s current rationale withstands scrutiny, it is simply not contained in its denial of NRDC’s petition or in its 2020 Risk Assessment. The law is clear that we can only uphold agency action based on the reasons the agency gave for its decision. See SEC v. Chenery Corp., 332 U.S. 194, 196 (1947) (“[A] reviewing court, in dealing with a determination or judgment which an administrative agency alone is authorized to make, must judge the propriety of such action solely by the grounds invoked by the agency.”); Fed. Power Comm’n v. Texaco Inc., 417 U.S. 380, 397 (1974).

Further casting doubt on its position, EPA’s brief repeatedly argues that the Normal Wear Study provided more reliable information about the amount of TCVP dust that would be shed by the collars, citing this study

NRDC V. USEPA 13

approximately 20 times. Again, though, EPA’s reliance on the Normal Wear Study is new to its briefing: EPA only cited the Normal Wear Study once (as part of a string cite in a footnote with no explanation) in its petition denial and 2020 Risk Assessment. EPA’s newly-discovered explanations appear to be “post-hoc rationalizations” that cannot support a finding of substantial evidence. See NRDC 2017, 857 F.3d at 1040 (quoting Hernandez-Cruz, 651 F.3d at 1109).

EPA’s assumption that only 14.6% of the dust released from the collars is TCVP—instead of the Torsion Study’s measurement of 97.2%—is also troubling on the merits for several reasons. First, EPA ignored obvious evidence suggesting its assumption is an underestimate. The TCVP in the collar is specifically designed to be released; other parts of the collar, like the plastic buckle, are unlikely to release dust. So, it seems reasonable to expect that TCVP will make up a disproportionate amount of the dust released by collars. Further, the Normal Wear Study which EPA now praises, showed that after only three weeks of use the percent of TCVP remaining in the collars had decreased by nearly half, while the total weight of the collars had declined by much less than half. This again strongly supports the common-sense conclusion that the collars would likely release a disproportionate amount of TCVP.

Second, EPA does not explain why, if it believed that the Torsion Study exaggerated the amount of TCVP released by the collars, it still relied on the Torsion Study’s other finding regarding the total amount of dust released from the collars. If we were to accept EPA’s reasoning, both measurements should be suspect: they should be equally affected by the Torsion Study’s supposedly exaggerated twisting.

Third, in its brief EPA seeks to buttress its assumption that only 14.6% of the dust released from collars is TCVP by pointing to the Normal Wear Study’s finding that only 4% of the collars’ total observed weight loss was measured as TCVP after three days. According to EPA, this result “validated” its assumption that the Torsion Study exaggerated the amount of TCVP in the dust. Once again, EPA’s rationale is not contained in its petition denial and is at best an improper post-hoc rationalization. See NRDC 2017, 857 F.3d at 1040. Further, EPA’s reasoning suffers from the obvious flaw that the Normal Wear Study only measured the composition of fifteen percent of the total observed weight loss; thus, EPA assumes that the remaining uncollected 85% of the observed weight loss would not change its analysis.

Another display of irrationality shown by EPA’s use of the Torsion Study is its assumption that only 0.38% of the TCVP in the collars will be released as dust. However, the Torsion Study undermines this assumption because it found that an average of 1.1% of the TCVP on the collars—roughly three times as high as EPA’s estimate—was released as dust before any twisting occurred.

ii. Assumption that owners will trim pet collars by 20%

EPA also assumed that pet owners will always trim the collars by at least 20% when fitting the collar onto a pet’s neck. This assumption reduces the expected amount of TCVP to which individuals would be exposed: all else equal, less collar means less TCVP. 2 EPA’s assumption that

2 In its brief, EPA argues that because longer collars are more likely to be used on larger pets, longer collars disperse collars’ TCVP across a greater surface area of fur and actually reduce TCVP exposure in humans. However, even if larger pets disperse collars’ TCVP across a

NRDC V. USEPA 15

owners will trim 20% of collars is drawn from a Hartz laboratory study of a different type of collar which does not contain TCVP. In this study (the “Efficacy Study”), researchers trimmed 20–43% of collars when fitting them to dogs’ necks to test the collars’ efficacy at protecting pets from ticks, fleas, and mosquitos. EPA assumed that this study was applicable to consumers using Hartz’s pet collars. And, “to provide a conservative assumption of how much collar might be removed during use,” used the lowest figure that the researchers trimmed from collars in the Efficacy Study (i.e. 20%).

EPA’s assumption that pet owners will remove 20% of collars reversed its earlier assumption: when preparing its 2016 Risk Assessment, EPA noted that because it could not determine the amount of collars owners would trim, it assumed that owners would not trim any of the collar. EPA’s assumption that owners would trim 20% of the collars also conflicts with its 2012 Standard Operating Procedures for Residential Pesticide Exposure Assessment, which states that “[b]ecause the trimmed length and corresponding active ingredient loss cannot be determined, the maximum application rate of the [pet] collar as labeled should be assumed for assessment of post-application risk.” Far from providing a sound justification for departing from its previous assumption and standard operating procedure, though, EPA’s denial of NRDC’s petition blandly stated that “[a]ccounting for the percentage of the pet collar removed is believed to better represent typical usage of the product as it is fit to the treated animal.” EPA provides no explanation for its reliance on the Efficacy Study, which did not use TCVP pet collars and did not mimic consumer behavior.

greater surface area of fur, an untrimmed collar would disperse more TCVP on a pet than a collar trimmed by 20% on that same pet. Nor did EPA explain how the results of the Efficacy Study, which only examined medium-sized dogs, could inform how pet owners are likely to use TCVP pet collars for dogs of all sizes and cats. EPA’s reliance on this assumption is further undermined by Hartz’s own representations. Hartz advertises that at least one of its collars fits dogs with necks that measure up to 26 inches, but the longest collar it sells is only 27 inches. In other words, it would be impossible for the owner of a dog with a 26-inch neck to remove 20% of the collar. We do not defer to such a “highly inaccurate” calculation. See Riley, 74 F.3d at 967.

iii. Lack of substantial evidence

EPA’s denial of NRDC’s petition suffers from numerous flaws. EPA does not explain its selective use of the Torsion Study, instead relying now on post-hoc rationalizations which we do not consider and which suffer from obvious flaws. See Ctr. for Biological Diversity, 998 F.3d at 1068 (“[A]n agency must provide its ‘reasoned explanation’ in a form that can adequately be examined on judicial review, not simply present arguments in its briefing how the decision might have been reached.”). EPA’s selective use of the Torsion Study reveals further inconsistences and weaknesses in its decision to deny NRDC’s petition. Further, and very importantly, many justifications now urged by EPA are not given as a part of its basis for its regulatory decision. We cannot consider contentions raised in EPA’s briefs that were not given previously as the basis for EPA’s decision. Finally, EPA’s decision to abandon its own guidance regarding how much pet owners will remove from a collar, without a discernable rationale, and instead rely on an assumption from a different study which is clearly inapplicable to at least some of Hartz’s pet collars, is not a viable basis for EPA’s decision. See Riley, 74 F.3d at 967.

NRDC V. USEPA 17

In combination, it is apparent to us that EPA’s denial of NRDC’s petition is simply not supported by “substantial evidence when considered on the record as a whole.” See 7 U.S.C. § 136n(b). 3, 4

IV. CONCLUSION

At times, NRDC’s efforts to receive a reasoned response from EPA have seemed Sisyphean as the agency consistently delayed its decision. After NRDC had doggedly pursued this matter for more than a dozen years, when EPA finally did reach a decision under pressure of a mandamus order of our court, NRDC was justified in expecting a rational, supported, and reasoned response from EPA. EPA, though, did not provide a well-reasoned or reasonable decision. Instead, its stated reasons were cursory and often at odds with EPA’s own prior assumptions and statements. Then, in response to this lawsuit by NRDC, EPA has provided many arguments in its briefing that were never given in the record as the basis for the administrative decision. Because EPA’s denial of NRDC’s petition is not supported by substantial evidence, we VACATE EPA’s denial of NRDC’s petition

3 Because we grant NRDC’s requested relief without considering the Declaration of Miriam Rotkin-Ellman (which was attached to NRDC’s Opening Brief), we do not consider whether this declaration was properly before us. 4 Hartz and EPA have demonstrated compelling reasons to maintain under seal only the Efficacy Study and the Petting Study; conversely, they have not met their burden to overcome the presumption in favor of open access to court records for the Torsion Study and the Normal Wear Study. See Ctr. for Auto Safety v. Chrysler Grp., LLC, 809 F.3d 1092, 1096 (9th Cir. 2016). We order the unsealing of the Torsion Study and the Normal Wear Study. and REMAND to EPA to issue a revised response to NRDC’s petition within 120 days.

VACATED AND REMANDED.