Opinion by Judge BETTY B. FLETCHER; Partial Concurrence and Partial Dissent by Judge SILER.
ORDER WITHDRAWING OPINION AND OPINION
ORDER
The opinion filed on November 15, 2007 and published at 508 F.3d 508 (9th Cir.2007), is hereby vacated and withdrawn. Respondents’ petition for rehearing with suggestion for rehearing en banc is denied *1178as moot. The opinion vacated and withdrawn is replaced by an opinion filed simultaneously with this order. We file a new opinion with the following changes:
1. At 508 F.3d at 514, delete <Therefore, we remand to NHTSA to promulgate new standards as expeditiously as possible and to prepare a full Environmental Impact Statements and replace with < Therefore, we remand to NHTSA to promulgate new standards as expeditiously as possible and to prepare either a revised Environmental Assessment or an Environmental Impact Statements
2. At 508 F.3d at 552, delete <3. NHTSA must prepare an Environmental Impact Statement > and replace with < <3. NHTSA must prepare either a revised Environmental Assessment or, as necessary, an Environmental Impact Statement >
3. At 508 F.3d at 553, after the citation «Blue Mountains Biodiversity Project, 161 F.3d at 1212 (quoting Save the Yaak Comm. v. Block, 840 F.2d 714, 717 (9th Cir.1988)); see also Nat’l Parks & Conservation Ass’n, 241 F.3d at 730,> insert the following sentences:
NHTSA’s EA is markedly deficient in its attempt to justify the refusal to prepare a complete EIS. As explained below, the agency’s FONSI is based primarily on its conclusory assertion — contradicted by evidence in the record — that the Final Rule will have no significant environmental impact because it authorizes CAFE standards that will result in a very small decrease in carbon dioxide emissions.
4. At 508 F.3d at 554, delete paragraph <We conclude that NHTSA’s FONSI is arbitrary and capricious and the agency must.... Nat’l Parks & Conservation Ass’n, 241 F.3d at 730.>
5. At 508 F.3d at 554, insert <Idaho Sporting Cong., 137 F.3d at 1149 (internal quotation marks omitted) > between < “human environmental factor,” > and particularly in light of the compelling scientific evidence >
6. At 508 F.3d at 558, delete the sentences < NHTSA has not provided a “statement of reasons why potential effects are insignificant,” much less a “convincing statement of reasons.” See Blue Mountains Biodiversity Project, 161 F.3d at 1211 (emphasis added) (internal quotation marks omitted). It asserts simply that the insignificance of the effects is “self-evident[ ].”> and replace with <Instead of providing the required “convincing statement of reasons,” Blue Mountains Biodiversity Project, 161 F.3d at 1211 (internal quotation marks omitted), NHTSA simply asserts that the insignificance of the effects is “self-evident[.]”>
7. At 508 F.3d at 558, delete the sentence <In order that the public and the agency be fully advised, we remand and order the agency to prepare a full EIS.> and insert the following paragraphs:
Finally, we must decide the appropriate remedy given NHTSA’s inadequate EA. We have previously recognized that preparation of an EIS is not mandated in all eases simply because an agency has prepared a deficient EA or otherwise failed to comply with NEPA. If, for example, an EA is so procedurally flawed that we cannot determine whether the proposed rule or project may have a significant effect, the court should remand for the preparation of a new EA. See Metcalf v. Daley, 214 F.3d 1135, 1146 (9th Cir.2000) (ordering remand for preparation of a new EA, where prior EA was prepared after the agency had already rendered a substantive decision on the permitting action). If an agency completely fails to prepare an EA before deciding that a proposed project or rule will have no significant environmental impact, remand for preparation of an EA is likewise the prop*1179er remedy. See Jones v. Gordon, 792 F.2d 821, 828-29 (9th Cir.1986) (remanding where agency failed to prepare any NEPA document before issuing permit). And where an agency determines that consideration of certain factors are legally irrelevant to the agency’s action, rendering it impossible for the reviewing court to determine the accuracy of the FONSI, we also remand for preparation of an EA on a complete record. San Luis Obispo Mothers for Peace v. Nuclear Regulatory Comm’n, 449 F.3d 1016, 1024, 1031, 1035 (9th Cir.2006) (ordering remand for reconsideration of EA where agency rejected consideration of terrorist acts as factor to be considered in its review of application for license to construct nuclear spent-fuel storage facility), cert. denied sub nom. Pac. Gas & Elec. Co. v. San Luis Obispo Mothers for Peace, — U.S. -, 127 S.Ct. 1124, 166 L.Ed.2d 891 (2007).
By contrast, if the court determines that the agency’s proffered reasons for its FONSI are arbitrary and capricious and the evidence in a complete administrative record demonstrates that the project or regulation may have a significant impact, then it is appropriate to remand with instructions to prepare an EIS. See, e.g., Nat’l Parks & Conservation Ass’n., 241 F.3d at 733-34; Idaho Sporting Cong., 137 F.3d at 1154.
The distinction — between cases where it is appropriate to order immediate preparation of an EIS and those where it is not — is implicit in this circuit’s NEPA jurisprudence, and has been explicitly recognized elsewhere. See O’Reilly v. U.S. Army Corps of Eng’rs, 477 F.3d 225, 238-39 (5th Cir.2007) (“It is also clear that a decision to forego preparation of an EIS may be unreasonable for at least two distinct reasons: (1) the evidence before the court demonstrates that, contrary to the FONSI, the project may have a significant impact on the human environment, or (2) the agency’s review was flawed in such a manner that it cannot yet be said whether the project may have a significant impact .... If the court finds that a project may have a significant impact, the court should order the agency to prepare an EIS. If the court finds, on the other hand, that the EA is inadequate in a manner that precludes making the determination whether the project may have a significant impact, the court should remand the case to the agency to correct the deficiencies in its analysis.”) (citations omitted). So, if there is uncertainty over whether the proposed project may have a significant impact, including uncertainty caused by an incomplete administrative record or an inadequate EA, the court should ordinarily remand for the agency to either prepare a revised EA or reconsider whether an EIS is required. Metcalf, 214 F.3d at 1146 (“On reflection, and in consideration of our limited role in this process, we have decided that it is appropriate only to require a new EA, but to require that it be done under circumstances that ensure an objective evaluation free of the previous taint.”); see also Hill v. Boy, 144 F.3d 1446, 1451 (11th Cir.1998) (remanding where agency’s failure to prepare an EIS was based on incorrect assumption); Nat’l Audubon Soc’y v. Hoffman, 132 F.3d 7, 18-19 (2d Cir.1997) (remanding due to incomplete administrative record).
Whether to require an EIS now is a very close question. Petitioners’ evidence demonstrates, overwhelmingly, the environmental significance of C02 emissions and the effect of those emissions on global warming. How NHTSA can, on remand, prepare an EA that takes proper account of this evidence and still conclude that the 2006 Final Rule has no significant environmental impact is questionable. See 40 C.F.R. § 1508.13 (FONSI is a document “presenting the reasons why an action ... will not have a significant effect on the *1180human environment and for which an [EIS] therefore will not be prepared” (emphasis added)). We nonetheless give the benefit of the doubt to NHTSA and decline to order the immediate preparation of an EIS for two reasons.
First, the EA’s primary deficiency lies with its conclusory assertion that a modest 0.2 percent decrease in carbon emissions renders the 2006 Final Rule environmentally insignificant. The EA provides no reasons or analysis in support of this conclusion, much less “convincing” reasons. Blue Mountains Biodiversity Project, 161 F.3d at 1212. Although Petitioners’ evidence is daunting we cannot, in the abstract, categorically decide that NHTSA’s reasons for this conclusion in a revised EA would be unconvincing.
Second, we have a significant reason to defer to the agency as to whether an EA will suffice or whether an EIS is necessary. During the pendency of this appeal, Congress enacted the Energy Independence and Security Act of 2007 (“EISA”). Pub.L. No. 110-140, 121 Stat. 1492. EISA requires NHTSA to increase fuel economy standards for passenger and non-passenger automobiles to reach a combined average of at least 35 mpg by model year 2020. See 49 U.S.C. § 32902(b)(2)(A). In connection with EISA, NHTSA has already begun preparation of a complete EIS to inform the “NEPA analysis relating to the CAFE standards for MY 2011-2015 automobiles,” including light trucks. Notice of Intent to Prepare an Environmental Impact Statement for New Corporate Average Fuel Economy Standards, 73 Fed.Reg. 16,615, 16,616 (Mar. 28, 2008) (to be codified at 49 C.F.R. pts. 531 & 533). With the exception of MY 2011, the EIS prepared as a result of the passage of EISA will not encompass the CAFE standards implemented by the 2006 Final Rule. However, information developed in preparation of the EISA-based EIS would undoubtedly inform NHTSA’s decision as to the necessity of an EIS in this case. See City of Los Angeles, 912 F.2d at 503 (Wald, C.J., dissenting) (stating that proper course was to remand without enjoining agency action because agency was “already conducting a [separate] programmatic EIS which may yet provide an adequate explanation for its finding of no significant impact”).
Taken together, these reasons lead us to conclude that the record is insufficiently complete for us to order the immediate preparation of an EIS. See O’Reilly, 477 F.3d at 240-41 (remanding without instructions to prepare EIS); Metcalf, 214 F.3d at 1146 (same); Hill, 144 F.3d at 1451 (same); Nat’l Audubon Soc’y, 132 F.3d at 18-19. We therefore remand to NHTSA to prepare a revised EA or, as necessary, a complete EIS.
8. At 508 F.3d at 558, delete <Thus, we remand to NHTSA for the preparation of a full EIS.> and replace with <Thus, we remand to NHTSA for the preparation of a revised EA or, as necessary, a full EIS.>
Our denial of the petition for rehearing with suggestion for rehearing en banc is made without prejudice to any party who may wish to file a petition for rehearing or petition for rehearing en banc with regard to the substituted opinion.
IT IS SO ORDERED.
OPINION
BETTY B. FLETCHER, Circuit Judge:Eleven states, the District of Columbia, the City of New York, and four public interest organizations petition for review of a rule issued by the National Highway Traffic Safety Administration (NHTSA) entitled “Average Fuel Economy Standards for Light Trucks, Model Years 20082011,” 71 Fed.Reg. 17,566 (Apr. 6, 2006) (“Final Rule”) (codified at 49 C.F.R. *1181pt. 533). Pursuant to the Energy Policy and Conservation Act of 1975 (EPCA), 49 U.S.C. §§ 32901-32919 (2007), the Final Rule sets corporate average fuel economy (CAFE) standards for light trucks, defined by NHTSA to include many Sport Utility Vehicles (SUVs), minivans, and pickup trucks, for Model Years (MYs) 2008-2011. For MYs 2008-2010, the Final Rule sets new CAFE standards using its traditional method, fleet-wide average (Unreformed CAFE). For MY 2011 and beyond, the Final Rule creates a new CAFE structure that sets varying fuel economy targets depending on vehicle size and requires manufacturers to meet different fuel economy levels depending on their vehicle fleet mix (Reformed CAFE).
Petitioners challenge the Final Rule under the EPCA and the National Environmental Policy Act of 1969 (NEPA), 42 U.S.C. §§ 4321-4347 (2007).1 First, they argue that the Final Rule is arbitrary, capricious, and contrary to the EPCA because (a) the agency’s cost-benefit analysis does not set the CAFE standard at the “maximum feasible” level and fails to give due consideration to the need of the nation to conserve energy; (b) its calculation of the costs and benefits of alternative fuel economy standards assigns zero value to the benefit of carbon dioxide (C02) emissions reduction; (c) its calculation of costs and benefits of alternative fuel economy standards fails to evaluate properly the benefit of vehicle weight reduction; (d) Reformed CAFE standards will depend on manufacturer fleet mix and not guarantee a minimum average fuel economy or “backstop”; (e) the transition period during which manufacturers may choose to comply with either Unreformed or Reformed CAFE is contrary to the “maximum feasible” requirement and unnecessary; (f) it perpetuates the “SUV loophole,” which allows SUVs, minivans, and pickup trucks to satisfy a lower fuel economy standard than cars; and (g) it excludes most vehicles rated between 8,500 and 10,000 pounds gross vehicle weight (comprised mostly of large pickup trucks) from any fuel economy regulation, even though these vehicles satisfy the statutory criteria for regulation.
Second, Petitioners argue that NHTSA’s Environmental Assessment is inadequate under NEPA because it fails to take a “hard look” at the greenhouse gas implications of its rulemaking and fails to analyze a reasonable range of alternatives or examine the rule’s cumulative impact. Petitioners also argue that NEPA requires NHTSA to prepare an Environmental Impact Statement.
NHTSA argues that the Final Rule is not arbitrary and capricious or contrary to the EPCA, the Environmental Assessment’s evaluation of the environmental consequences of its action is adequate, and an Environmental Impact Statement is not required.
We have jurisdiction under 49 U.S.C. § 32909(a) to review the Final Rule issued by NHTSA. We hold that the Final Rule is arbitrary and capricious, contrary to the EPCA in its failure to monetize the value of carbon emissions, failure to set a backstop, failure to close the SUV loophole, and failure to set fuel economy standards for all vehicles in the 8,500 to 10,000 gross *1182vehicle weight rating (“GVWR”) class. We also hold that the Environmental Assessment was inadequate and that Petitioners have raised a substantial question as to whether the Final Rule may have a significant impact on the environment. Therefore, we remand to NHTSA to promulgate new standards as expeditiously as possible and to prepare either a revised Environmental Assessment or an Environmental Impact Statement.
I. FACTUAL AND PROCEDURAL BACKGROUND
A. CAFE Regulation Under the Energy Policy and Conservation Act
In the aftermath of the energy crisis created by the 1973 Mideast oil embargo, Congress enacted the Energy Policy and Conservation Act of 1975, Pub.L. No. 94-163, 89 Stat. 871, 901-16. See H.R.Rep. No. 94-340 at 1-3 (1975), as reprinted in 1975 U.S.C.C.A.N. 1762, 1763-65. Congress observed that “[t]he fundamental reality is that this nation has entered a new era in which energy resources previously abundant, will remain in short supply, retarding our economic growth and necessitating an alteration in our life’s habits and expectations.” Id. at 1763. The goals of the EPCA are to “decrease dependence on foreign imports, enhance national security, achieve the efficient utilization of scarce resources, and guarantee the availability of domestic energy supplies at prices consumers can afford.” S.Rep. No. 94-516 (1975) (Conf. Rep.), as reprinted in 1975 U.S.C.C.A.N. 1956, 1957. These goals are more pressing today than they were thirty years ago: since 1975, American consumption of oil has risen from 16.3 million barrels per day to over 20 million barrels per day, and the percentage of U.S. oil that is imported has risen from 35.8 to 56 percent. NRDC Cmt. at ll;2 see also 71 Fed.Reg. at 17,644.
In furtherance of the goal of energy conservation, Title V of the EPCA establishes automobile fuel economy standards. An “average fuel economy standard” (often referred to as a CAFE standard) is “a performance standard specifying a minimum level of average fuel economy applicable to a manufacturer in a model year.” 49 U.S.C. § 32901(a)(6) (2007). Only “automobiles” are subject to fuel economy regulation, and passenger automobiles must meet a statutory standard of 27.5 mpg, 49 U.S.C. § 32902(b),3 whereas non-passenger automobiles must meet standards set by the Secretary of Transportation, id. § 32902(a). Congress directs the Secretary to set fuel economy standards at “the maximum feasible average fuel economy level that the Secretary decides the manufacturers can achieve in that model year.” Id. § 32902(a).4 Under this subsection, the Secretary is authorized to “prescribe separate standards for different classes of automobiles.” Id. Congress also provides that “[w]hen deciding maximum feasible average fuel economy under this *1183section, the Secretary of Transportation5 shall consider technological feasibility, economic practicability, the effect of other motor vehicle standards of the Government on fuel economy, and the need of the United States to conserve energy.” Id. § 32902(f).
Under the EPCA’s definitional scheme, vehicles not manufactured primarily for highway use and vehicles rated at 10,000 lbs. gross vehicle weight or more are excluded from fuel economy regulation altogether because they are not “automobiles.”6 An “automobile” is defined as:
a 4-wheeled vehicle that is propelled by fuel, or by alternative fuel, manufactured primarily for use on public streets, roads, and highways ..., and rated at—
(A) not more than 6,000 pounds gross vehicle weight;7 or
(B) more than 6,000, but less than 10,-000, pounds gross vehicle weight, if the Secretary decides by regulation that—
(i) an average fuel economy standard under this chapter for the vehicle is feasible; and
(ii) an average fuel economy standard under this chapter for the vehicle will result in significant energy conservation or the vehicle is substantially used for the same purposes as a vehicle rated at not more than 6,000 pounds gross vehicle weight.
49 U.S.C. § 32901(a)(3). Although NHTSA has the authority to regulate the fuel economy of vehicles up to 10,000 lbs. GVWR, see id. § 32901(a)(3)(B), the agency has excluded vehicles exceeding 8,500 lbs. (other than medium-duty passenger vehicles manufactured during MY 2011 or thereafter) from its definition of “automobile,” see 49 C.F.R. § 523.3(b).
The CAFE standards NHTSA sets for non-passenger automobiles or “light trucks,” as referred to by the agency in its regulations,8 are lower than the standards for passenger automobiles. Compare 49 C.F.R. § 533.5(a) (2007) with 49 C.F.R. § 531.5(a) (2007). A “passenger automobile” is defined as:
an automobile that the Secretary decides by regulation is manufactured primarily for transporting not more than 10 individuals, but does not include an automobile capable of off-highway operation that the Secretary decides by regulation—
(A) has a significant feature (except 4-wheel drive) designed for off-highway operation; and
(B) is a 4-wheel drive automobile or is rated at more than 6,000 pounds gross vehicle weight.
49 U.S.C. § 32901(a)(16).
The Final Rule sets CAFE standards for “light trucks,” defined by NHTSA to include many SUVs, vans, and pickup trucks, for MYs 2008-2011. See 71 Fed. Reg. at 17,568; 49 C.F.R. § 533.5(a), (g), (h). A “light truck” is:
an automobile other than a passenger automobile which is either designed for off-highway operation, as described in paragraph (b) of this section,9 or de*1184signed to perform at least one of the following functions: (1) Transport more than 10 persons; (2) Provide temporary living quarters; (3) Transport property on an open bed; (4) Provide greater cargo-carrying than passenger-carrying volume; or (5) Permit expanded use of the automobile for cargo-carrying purposes or other non-passenger-carrying purposes through [removable or folda-ble, stowable seats to create a flat floor].
49 C.F.R. § 523.5(a) (2007).
For MYs 1996 to 2004, Congress froze the light truck CAFE standard at 20.7 mpg. See 71 Fed.Reg. at 17,568. After the legislative restrictions were lifted, NHTSA set new light truck CAFE standards in April 2003: 21.0 mpg for MY 2005, 21.6 mpg for MY 2006, and 22.2 mpg for MY 2007. Light Truck Average Fuel Economy Standards Model Years 2005-2007, 68 Fed.Reg. 16,868, 16,871 (Apr. 7, 2003) (codified at 49 C.F.R. pt. 533).
In response to a request from Congress, the National Academy of Sciences (NAS) published in 2002 a report entitled “Effectiveness and Impact of Corporate Average Fuel Economy (CAFE) Standards.”10 The NAS committee made several findings and recommendations. It found that from 1970 to 1982, CAFE standards helped contribute to a 50 percent increase in fuel economy for new light trucks. Id. at 14. In the subsequent decades, however, light trucks became more popular since domestic manufacturers faced less competition in the light truck category and could generate greater profits. Id. at 18-19. The “less stringent CAFE standards for trucks ... provide[d] incentives for manufacturers to invest in minivans and SUVs and to promote them to consumers in place of large ears and station wagons.” Id. at 18. When the CAFE regulations were originally promulgated in the 1970s, “light truck sales accounted for about 20 percent of the new vehicle market,” but now they account for about half. Id. at 88. This shift has had a “pronounced” effect on overall fuel economy. Id. at 19. As the market share of light trucks has increased, the overall average fuel economy of the new light duty vehicle fleet (light trucks and passenger automobiles) has declined “from a peak of 25.9 MPG in 1987 to 24.0 MPG in 2000.” Id. Vehicle miles traveled (VMT) by light trucks has also been growing more rapidly than passenger automobile travel. Id.
The NAS committee found that the CAFE program has increased fuel economy, but that certain aspects of the program “have not functioned as intended,” including “[t]he distinction between a car for personal use and a truck for work use/cargo transport,” which “has been stretched well beyond the original purpose.” Id. at 3. The committee also found that technologies exist to “significantly reduce fuel consumption,” for cars and light trucks and that raising CAFE standards would reduce fuel consumption. Id. at 3-4. Significantly, the committee found that of the many reasons for improving fuel economy, “[t]he most important ... is concern about the accumulation in the atmosphere of so-called greenhouse gases, principally carbon dioxide. Continued increases in carbon dioxide emissions are *1185likely to further global warming.” Id. at 2. In addition, the committee found “exter-nalities of about $0.30/gal of gasoline associated with the combined impacts of fuel consumption on greenhouse gas emissions and on world oil market conditions”11 that “are not necessarily taken into account when consumers purchase new vehicles.” Id. at 4.
B. National Environmental Policy Act
NEPA requires a federal agency “to the fullest extent possible,” to prepare “a detailed statement on ... the environmental impact” of “major Federal actions significantly affecting the quality of the human environment.” 42 U.S.C. § 4332(2)(C)(i) (2007); see also 40 C.F.R. § 1500.2 (2007). The purpose of NEPA is twofold: “ ‘ensure[ ] that the agency ... will have available, and will carefully consider, detailed information concerning significant environmental impacts[, and] guarantee[ ] that the relevant information will be made available to the larger [public] audience.’ ” Idaho Sporting Cong. v. Thomas, 137 F.3d 1146, 1149 (9th Cir.1998) (quoting Robertson v. Methow Valley Citizens Council, 490 U.S. 332, 349, 109 S.Ct. 1835, 104 L.Ed.2d 351 (1989)); see also 40 C.F.R. § 1500.1(b) (stating that environmental information must be provided “before decisions are made and before actions are taken.”). “NEPA expresses a Congressional determination that procrastination on environmental concerns is no longer acceptable.” Found, for N. Am. Wild Sheep v. U.S. Dep’t of Agric., 681 F.2d 1172, 1181 (9th Cir.1982). NEPA “is our basic national charter for protection of the environment.” 40 C.F.R. § 1500.1(a).
If there is a substantial question whether an action “may have a significant effect” on the environment, then the agency must prepare an Environmental Impact Statement (EIS). See, e.g., Blue Mountains Biodiversity Project v. Blackwood, 161 F.3d 1208, 1212 (9th Cir.1998) (internal quotation marks omitted). An EIS should contain a discussion of significant environmental impacts and alternatives to the proposed action. See 40 C.F.R. §§ 1502.1, 1502.14, 1508.7. As a preliminary step, an agency may prepare an Environmental Assessment (EA) in order to determine whether a proposed action may “significantly affect[ ]” the environment and thereby trigger the requirement to prepare an EIS. See 40 C.F.R. § 1508.9(a)(1) (2007). An EA is “a concise public document” that “[b]riefly provide[s] sufficient evidence and analysis for determining whether to prepare an environmental impact statement or a finding of no significant impact.”12 Id. An EA “[s]hall include brief discussions of the need for the proposal, of alternatives as required by sec. 102(2)(E), of the environmental impacts of the proposed action and alternatives, and a listing of agencies and persons consulted.” Id. § 1508.9(b).
Whether an action may “significantly affect” the environment requires consideration of “context” and “intensity.” Id. § 1508.27; see also Nat’l Parks & Conservation Ass’n v. Babbitt, 241 F.3d 722, 731 (9th Cir.2001). “Context ... delimits the scope of the agency’s action, including the interests affected.” Nat’l. Parks & Conservation Ass’n, 241 F.3d at 731. Intensity refers to the “severity of impact,” which includes both beneficial and adverse impacts, “[t]he degree to which the proposed action affects public health or safety,” “[t]he degree to which the effects on the quality of the human environment are *1186likely to be highly controversial,” “[t]he degree to which the possible effects on the human environment are highly uncertain or involve unique or unknown risks,” and “[w]hether the action is related to other actions with individually insignificant but cumulatively significant impacts.” 40 C.F.R. § 1508.27(b)(2), (4), (5), (7).
C. NHTSA’s Proposed Rulemaking and Draft Environmental Assessment
On December 29, 2003, NHTSA published an advance notice of proposed rulemak-ing (ANPRM) that solicited comments on several proposed regulatory changes intended to increase fuel economy, including a proposal to modernize the light truek/ear distinction and a proposal to increase the GVWR limit on vehicles subject to CAFE standards. 68 Fed.Reg. 74,908 (Dec. 29, 2003). NHTSA acknowledged that its regulations define passenger and non-passenger vehicles “by the type of use to which they were generally put in the mid-1970s,” id. at 74,909, and that “[t]he markets for, and designs of, cars and light trucks have changed substantially,” with “some light trucks ... used primarily to transport passengers,” id. at 74,913. NHTSA noted that in its original NPRM promulgated in December 1976, it concluded that “Congress intended that passenger automobiles be defined as those used primarily for the transport of individuals and that all other vehicles would fall within the category of non-passenger automobiles.” Id. at 74,926. NHTSA did not present any specific proposals for reforming the CAFE program, but it presented two options for including vehicles under 10,000 lbs. GVWR in the program; (1) regulating medium-duty passenger vehicles (MDPVs),13 which are vehicles between 8,500 and 10,000 lbs. GVWR designed primarily for the transportation of persons, and (2) regulating all vehicles between 8,500 and 10,000 lbs. GVWR. Id. at 74,930.
On August 30, 2005, NHTSA issued proposed CAFE standards for light trucks MYs 2008-2011 of 22.5 mpg for MY 2008, 23.1 mpg for MY 2009, and 23.5 mpg for MY 2010.14 70 Fed.Reg. 51,414, 51,424 (Aug. 30, 2005). NHTSA determined that these were the “maximum feasible” standards using a marginal cost-benefit analysis. See id. For MY 2011 and beyond, NHTSA proposed to adopt a “Reformed CAFE” system, which would set different CAFE standards for vehicles based on size, measured by the vehicle’s footprint (the product of multiplying wheelbase by track width). Id. at 51,414, 51,429-41. NHTSA proposed six footprint categories (a step function), id. at 51,430, and it pro*1187posed a transition period (MY 2008-2010) to Reformed CAFE, during which manufacturers could choose to comply with either Reformed or Unreformed CAFE. NHTSA also proposed not to change the criteria by which vehicles are classified as passenger automobiles or light trucks, id. at 51,422, and it proposed to regulate only MDPVs within the 8,500 to 10,000 lb. vehicle class as light trucks, id. at 51,455-56.
NHTSA issued a Draft Environmental Assessment in August 2005. The Draft EA integrated much of the text from the Final EA that accompanied NHTSA’s light truck rulemaking for MYs 2005-2007 released in April 2003. See Draft Environmental Assessment, NHTSA Proposed Corporate Average Fuel Economy (CAFE) Standards 9 (Aug. 2005) (Draft EA). The Draft EA analyzed three alternatives to the proposed rule. Alternative A (“No Action”) would extend the MY 2007 standard of 22.2 mpg through MY 2011. Alternative B would be Unreformed CAFE in MY 2008-2010 and Reformed CAFE in MY 2011. Alternative C would be Reformed CAFE set at equalized cost with Unreformed CAFE in MY 2008-2010 and Reformed CAFE in MY 2011. Id.
The Draft EA noted that “C02 ... has started to be viewed as an issue of concern for its global climate change potential.” Id. at 18. With regard to biological resources, the Draft EA stated, “emissions of criteria pollutants and greenhouse gases could result in ozone layer depletion and promote climate change that could affect species and ecosystems.” Id. at 19. The projected lifetime fuel savings for MY 2008-2011 light trucks under Alternatives B and C would “rang[e] from 1.3% to 1.7% of their fuel compared to the baseline, corresponding to 4.7-6.0 billion gallons.” Id. at 25. The estimated lifetime emissions of C02 ranged from 1,341.4 million metric tons (mmt) under baseline to 1,306.4 and 1,304.0 mmt under Alternatives B and C, respectively. Id. at 29. The Draft EA concluded that the proposed standards would “result in reduced emissions of C02, the predominant greenhouse gas emitted by motor vehicles,” “reductions in contamination of water resources,” and “minor reductions in impacts to biological resources.” Id. at 30-31. In addition, “the cumulative effects estimated to result from both the 2005-2007 and 2008-2011 light truck rulemakings over the lifetimes of the vehicles they would affect are projected to be very small.” Id. at 34.
NHTSA received over 45,000 comments on the NPRM and Draft EA from states, consumer and environmental organizations, automobile manufacturers and associations, members of Congress, and private individuals. See 71 Fed.Reg. at 17,577. Manufacturers argued that reliance on a cost-benefit analysis might not “adequately account for the capabilities of the industry.” Id. They also generally opposed subjecting vehicles greater than 8,500 lbs. GVWR to CAFE regulation, arguing that those vehicles are used in a different manner than lighter vehicles and that their regulation would not result in significant fuel savings. Id. at 17,577-78. The states and environmental and consumer organizations generally argued that:
• The need of the nation to conserve energy and national security require more stringent standards, and such standards are feasible and practicable. E.g., NRDC Cmt. at 4-5; Environmental Defense Cmt. at 1 — 7;15 Public Citizen Cmt. at 1-2.16 For example, the Alliance to Save Energy — American Council for an Energy-Efficient Economy (ACEEE) argued that “[t]he *118810 billion gallons of fuel that NHTSA claims will be saved through the new standards over the three-decade life of model year 2008-2011 vehicles amount to less than one month’s supply of gasoline for U.S. light-duty vehicles. These savings are also insufficient to offset the expected growth in gasoline usage for even the four-year period covered by the rule.” ACEEE Cmt. at l.17
• NHTSA’s use of marginal cost-benefit analysis unlawfully overemphasizes cost at the expense of technological feasability and energy conservation and is not “technology-forcing,” as EPCA intended. E.g., NRDC Cmt. at 14-16; Environmental Defense Cmt. at 4-5; Public Citizen Cmt. at 1-2.
• Even if NHTSA’s cost-benefit analysis is permissible, the “maximum feasible” standard cannot be determined properly without taking environmental impacts into account, and the failure to monetize certain benefits such as greenhouse gas (GHG) emissions underestimates benefits of stricter standards. E.g., CBD Cmt. at 1-4;18 NRDC Cmt. at 8 (suggesting specific figures and sources for the value per ton of C02 emissions avoided, from $8/ton to $26.50/ton); Environmental Defense Cmt. at 5-6; Environmental Defense Cmt. Re: Carbon Costs at 1-3 (citing new studies from the United Kingdom that value carbon at $96-174/ton carbon).19
• Reformed CAFE “rewards fuel economy laggards while penalizing industry leaders,” Sierra Club Cmt. at 4,20 and, like Unreformed CAFE, promotes the manufacture of larger, less fuel-efficient vehicles. E.g., App. G to NRDC Cmt. at 3-4;21 States Cmt. at 5;22 Environmental Defense Cmt. at 12-13.23
*1189• NHTSA’s analysis of the adverse safety effects of vehicle weight reduction is flawed and confounds size and weight. E.g., Sierra Club Cmt. at 8-10; App. C to Environmental Defense Cmt. at 1-4; Public Citizen Cmt. at 12-19; App. B to Public Citizen Cmt. at 1-16.
• Since the Reformed CAFE standard for a particular manufacturer depends on its fleet mix, NHTSA should include a “backstop” or guarantee that average fuel economy levels will not fall below the “maximum feasible” level. E.g., NRDC Cmt. at 24-25; ACEEE Cmt. at 5; see also App. E to NRDC Cmt. at 6-12 (analysis of gaming scenarios and upsizing trends); Environmental Defense Cmt. at 13-14 (same); App. G. to Environmental Defense Cmt. at 1-14 (same).
• The transition period (MY 2008-2010) to Reformed CAFE is unnecessary and undercuts fuel economy savings. E.g., NRDC Cmt. at 27-28; ACEEE Cmt. at 2; Environmental Defense Cmt. at 8-9; UCS Cmt. at 9.
• All Class 2b trucks between 8,500-10,000 lbs. GVWR should be regulated because fuel economy standards for them are feasible, would result in significant energy conservation, and they are used for substantially the same purposes as vehicles 6,000 lbs. or less. Environmental Defense Cmt. at 9-11; App. F to Environmental Defense Cmt. at 1-2; and Polk Study.24
• Higher fuel economy standards would help domestic automakers remain competitive in the long term and protect U.S. jobs. App. D to NRDC Cmt. at 22.25 The California Energy Resources Conservation and Development Commission commented that “[upgrading CAFE requirements could enhance jobs in the United States, especially in the automobile manufacturing sector.... Increasing light truck CAFE to 26.9 mpg in 2010 and 31 mpg in 2015 (with corresponding changes for passenger cars) would increase net jobs up to 346,000.” California Energy Commission Cmt. at 9-10.26
• NHTSA’s draft EA is inadequate and fails to consider the proposed rule’s impact on climate change. States Cmt. at 1-11; CBD Cmt. at 5-12.
See also 71 Fed.Reg. at 17,578-79 (summarizing comments).
Commenters also submitted to NHTSA numerous scientific reports and studies regarding the relationship between climate change and greenhouse gas emissions and the expected impacts on the environment.27 *1190Emissions from light trucks make up about eight percent of annual U.S. greenhouse gas emissions. Final EA at 22 (citing EPA, EPA-430-R-05-003, Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990-2004 (Draft 2006)). The transportation sectors account for about 31 percent of human-generated C02 emissions in the U.S. economy. NAS Report at 14. “Overall, U.S. light-duty vehicles [passenger cars and light trucks] produce about 5 percent of the entire world’s greenhouse gases.” Id. at 20. The NAS committee concluded, “Since the United States produces about 25 percent of the world’s greenhouse gases, fuel economy improvements could have a significant impact on the rate of C02 accumulation in the atmosphere.” Id. at 14.
The Intergovernmental Panel on Climate Change (IPCC)’s “Third Assessment Report,” published in 2001, presented the consensus view of hundreds of scientists on key issues relating to climate change. The IPCC concluded that “C02 concentrations increasing over [the] 21st century[are] virtually certain to be mainly due to fossil-fuel emissions,” and that “[stabilization of atmospheric C02 concentrations at 450, 650, or 1,000 ppm would require global anthropogenic C02 emissions to drop below year 1990 levels, within a few decades, about a century, or about 2 centuries, respectively, and continue to decrease steadily thereafter to a small fraction of current emissions.”28 Id. The average earth surface temperature has increased by about 0.6 degree Celsius since the late 19th century, see Technical Summary of IPCC Working Group I Report at 26; snow and ice cover have decreased about 10 percent since the late 1960s, id. at 30; and global average sea level has risen between 10 to 20 cm during the 20th century, id. at 31. The IPCC also developed a range of emissions scenarios as its basis for predicting the environmental effect of increased emissions. Id. at 62-63.29
More recent evidence shows that there have already been severe impacts in the Arctic due to warming, including sea ice decline. See Attachments J, L, & Q to CBD Cmt. Global warming has already affected plants, animals, and ecosystems around the world. See, e.g., Attachment M to CBD Cmt. at 15-16. Some scientists predict that “on the basis of mid-range climate-warming scenarios for 2050, that 15-37% of species in our sample of regions and taxa will be ‘committed to extinction.’ ” *1191Attachment 0 to CBD Cmt (Thomas, Extinction Risk from, Climate Change, 427 Nature at 145). In addition, there will be serious consequences for human health, including the spread of infectious and respiratory diseases, if worldwide emissions continue on current trajectories. See, e.g., Attachment F to CBD Cmt. at 32-64. Sea level rise and increased ocean temperatures are also associated with increasing weather variability and heightened intensity of storms such as hurricanes. Id. at 21-24. Past projections have underestimated sea level rise. See id. at 20. Several studies also show that climate change may be non-linear, meaning that there are positive feedback mechanisms that may push global warming past a dangerous threshold (the “tipping point”). Id. at 26-27; see also Technical Summary of IPCC Working Group I Report at 46-53; Attachment F to CBD Cmt. at 26-27; IPCC Report at 14-16; States Cmt. 9.
D. The Final Rule: CAFE Standards for Light Trucks MYs 2008-2011
NHTSA issued the Final Rule on April 6, 2006. 71 Fed.Reg. at 17,566. NHTSA set the CAFE standards for MY 2008-2010 (Unreformed CAFE) at the same levels as proposed in the NPRM.30 Unreformed CAFE sets a fleet-wide average fuel economy standard “with particular regard to the ‘least capable manufacturer with a significant share of the market.’ ” 71 Fed.Reg. at 17,580. NHTSA has reformed the structure of the CAFE program for light trucks, effective MY 2011 (Reformed CAFE). Under Reformed CAFE, fuel economy standards are based on a truck’s footprint, with larger footprint trucks subject to a lower standard and smaller footprint trucks subject to higher standards.31 71 Fed.Reg. at 17,566. Instead of six footprint categories (a step function) as proposed in the NPRM, Reformed CAFE would be based on a continuous function, meaning a separate fuel economy target for each vehicle of a different footprint. See id. at 17,595-96. “A particular manufacturer’s compliance obligation for a model year will be calculated as the harmonic average of the fuel economy targets for the manufacturer’s vehicles, weighted by the distribution of manufacturer’s production volumes among the footprint increments.” Id. at 17,566. A manufacturer’s CAFE compliance obligation will vary with its fleet mix. A manufacturer that produces more large footprint light trucks will have a lower required CAFE standard than one that produces more small footprint light trucks.32
During MYs 2008-2010, manufacturers may choose to comply with Unreformed CAFE or Reformed CAFE. See id. at 17,593-94.
NHTSA used the manufacturers’ preexisting product plans as the baseline for its analyses of technical and economic feasibility under both Unreformed and Reformed CAFE. Id. at 17,579. NHTSA made adjustments to the product plans by applying additional technologies in a “cost-minimizing fashion,”33 id. at 17,582, and stopping at the point where marginal costs equaled marginal benefits, id. at 17,597. NHTSA *1192considered the cost of new technologies and the benefits of fuel savings over the lifetime of the vehicle as the costs and benefits of higher fuel economy standards. Id. at 17,585-87, 17,622-23. NHTSA monetized some externalities such as emission of criteria pollutants during gasoline refining and distribution and crash and noise costs associated with driving. See Final Regulatory Impact Analysis, Corporate Average Fuel Economy and CAFE Reform for MY 2008-2011 Light Trucks at VIII-60, VIII-74-80 (March 2006) (FRIA). However, NHTSA did not monetize the benefit of reducing carbon dioxide emissions, which it recognized was the “the main greenhouse gas emitted as a result of refining, distribution, and use of transportation fuels.” FRIA at VIII-61 to 62.34 NHTSA acknowledged the estimates suggested in the scientific literature, see 71 Fed.Reg. at 17,638; FRIA at VIII-63, but concluded:
[T]he value of reducing emissions of C02 and other greenhouse gases [is] too uncertain to support their explicit valuation and inclusion among the savings in environmental externalities from reducing gasoline production and use. There is extremely wide variation in published estimates of damage costs from greenhouse gas emissions, costs for controlling or avoiding their emissions, and costs of sequestering emissions that do occur, the three major sources for developing estimates of economic benefits from reducing emissions of greenhouse gases.
71 Fed.Reg. at 17,638; see also FRIA at VIII-64 to 65.
In its cost-benefit analysis, NHTSA also excluded weight reduction for vehicles between 4,000 and 5,000 lbs. curb weight as a potential measure that manufacturers could use to increase fuel economy. 71 Fed.Reg. at 17,627. NHTSA accepted the possibility of weight reduction for vehicles over 5,000 lbs. curb weight as a cost-effective technology35 that would not reduce overall safety. Id. NHTSA relied on a study by Dr. Charles Kahane36 for this 5,000 lb. figure:
[T]he net safety effect of removing 100 pounds from a light truck is zero for light trucks with a curb weight greater than 3,900 lbs. However, given the significant statistical uncertainty around that figure, we assumed a confidence bound of approximately 1,000 lbs. and used 5,000 lbs. as the threshold for considering weight reduction.
Id. (footnotes omitted). By “net safety effect,” NHTSA means that 3,900 lbs. is the breakeven point: “the point where the total effect of reducing all vehicles heavier than the breakeven weight by an equal amount is zero.” Id. at 17,628. In the FRIA, NHTSA explained that it chose the approximately 1,000 lb. confidence bound *1193based on additional empirical work found in Kahane’s study:
Kahane estimated a crossover weight37 of 5,085 lbs. if manufacturers changed both weight and footprint, and the interval estimated ranged from 4,224 lbs. to 6,121 lbs[.], i.e., an interyal +/-1000 lbs[.] around the point estimate. Although the crossover weight differs from the point of zero net impact, they would both tend to have similar sampling errors. We applied this interval to the 3,900 lbs. point of zero net impact (which is based on the assumption that footprint is held constant); therefore, the agency felt it would be prudent to limit weight reductions to those vehicles above 5,000 lbs. curb weight.
FRIA at V-15 (internal citation omitted).
NHTSA rejected the idea of a “backstop” under Reformed CAFE. 71 Fed.Reg. at 17,592; id. at 17,617. NHTSA stated that a backstop, or a required fuel economy level applicable to a manufacturer if its required level under Reformed CAFE fell below a certain minimum, “would essentially be the same as an Unreformed CAFE standard.” Id. at 17,592. NHTSA argued that “EPCA permits the agency to consider consumer demand and the resulting market shifts in setting fuel economy standards,” id. at 17,593, and that a backstop “would essentially limit the ability of manufacturers to respond to market shifts arising from changes in consumer demand. If consumer demand shifted towards larger vehicles, a manufacturer potentially could be faced with a situation in which it must choose between limiting its production of the demanded vehicles, and failing to comply with the CAFE light truck standard.” Id.
Finally, NHTSA declined to change the regulatory definition of cars and light trucks to close the SUV loophole and refused to regulate vehicles between 8,500 and 10,000 lbs. GWVR, other than MDPVs. See id. at 17,574.
II. STANDARD OF REVIEW
The Administrative Procedure Act (APA), 5 U.S.C. §§ 701-706 (2007), provides that agency action must be set aside by the reviewing court if it is “ ‘arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.’ ” Competitive Enter. Inst. v. NHTSA (CEI III), 45 F.3d 481, 484 (D.C.Cir.1995) (quoting 5 U.S.C. § 706(2)(A)) (applying the APA to review a rulemaking under the EPCA). The scope of review is narrow, but “the agency must examine the relevant data and articulate a satisfactory explanation for its action including a ‘rational connection between the facts found and the choice made.’ ” Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983) (citation omitted). An agency rule would normally be arbitrary and 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 before the agency, or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise.
Id. The reviewing court “ ‘may not supply a reasoned basis for the agency’s action that the agency itself has not given.’ ” Id. (quoting SEC v. Chenery Corp., 332 U.S. 194, 196, 67 S.Ct. 1575, 91 L.Ed. 1995 (1947)).
*1194If Congress has spoken directly to the “precise question at issue,” then we must give effect to Congress’s “unambiguously expressed intent.” Chevron U.S.A., Inc. v. NRDC, 467 U.S. 837, 842-43, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). However, “if the statute is silent or ambiguous with respect to the specific issue, the question for the court is whether the agency’s answer is based on a permissible construction of the statute.” Id. at 843, 104 S.Ct. 2778. We “must reject administrative constructions which are contrary to clear congressional intent.” Id. at 843 n. 9, 104 S.Ct. 2778.
NHTSA’s compliance with NEPA is reviewed under an arbitrary and capricious standard pursuant to the APA. See, e.g., Nat’l Parks & Conservation Ass’n, 241 F.3d at 730. With respect to NEPA documents, the agency must take a “hard look” at the impacts of its action by providing “ ‘a reasonably thorough discussion of the significant aspects of the probable environmental consequences.’ ” Thomas, 137 F.3d at 1149 (quoting Or. Nat. Res. Council v. Lowe, 109 F.3d 521, 526 (9th Cir.1997)). We must determine whether the EA “ ‘foster[s] both informed decision-making and informed public participation.’ ” Native Ecosystems Council v. U.S. Forest Serv., 418 F.3d 953, 960 (9th Cir.2005) (quoting California v. Block, 690 F.2d 753, 761 (9th Cir.1982)).
III. DISCUSSION
A. Energy Policy and Conservation Act Issues
1. NHTSA’s use of marginal cost-benefit analysis to determine “maximum feasible average fuel economy level”
With respect to non-passenger automobiles (i.e., light trucks), the fuel economy standard “shall be the maximum feasible average fuel economy level that the Secretary decides the manufacturers can achieve in that model year.” 49 U.S.C. § 32902(a). “Maximum feasible” is not defined in the EPCA. However, the EPCA provides that “[w]hen deciding maximum feasible average fuel economy under this section, the Secretary of Transportation shall consider technological feasibility, economic practicability, the effect of other motor vehicle standards of the Government on fuel economy, and the need of the United States to conserve energy.” Id. § 32902(f).
Petitioners argue that the meaning of “maximum feasible” is plain, and that NHTSA’s decision to maximize economic benefits is contrary to the plain language of the EPCA because “feasible” means “ ‘capable of being done,’ ” not economically optimal. But even if “feasible” means ‘“capable of being done,’” technological feasibility, economic practicability, the effect of other motor vehicle standards, and the need of the nation to conserve energy must be considered in determining the “maximum feasible” standard. American Textile Manufacturers Institute v. Donovan does not support Petitioners’ interpretation of “feasible.” 452 U.S. 490, 101 S.Ct. 2478, 69 L.Ed.2d 185 (1981). In that case, no other language in the statute modified the phrase at issue: “to the extent feasible.” Id., 452 U.S. at 508-11, 101 S.Ct. 2478. Here, “maximum feasible” standards are to be determined in light of technological feasibility, economic practicability, the effect of other motor vehicle standards, and the need of the nation to conserve energy.38
*1195The EPCA clearly requires the agency to consider these four factors, but it gives NHTSA discretion to decide how to balance the statutory factors — as long as NHTSA’s balancing does not undermine the fundamental purpose of the EPCA: energy conservation. In Center for Auto Safety v. NHTSA, the D.C. Circuit considered whether NHTSA gave “impermissible weight to shifts in consumer demand” in setting the MY 1985 and 1986 standards for light trucks. 793 F.2d 1322, 1338 (D.C.Cir.1986). Petitioners in that case challenged NHTSA’s rule that revised the standards downward. Id. at 1323-24. The court held that since Congress had not directly spoken to the issue of consumer demand, the court must determine whether the agency’s interpretation represented a “ ‘reasonable accommodation of conflicting policies that were committed to the agency’s care by the statute.’ ” Id. at 1338 (quoting Chevron, 467 U.S. at 845, 104 S.Ct. 2778). The court reasoned that:
Congress intended energy conservation to be a long term effort that would continue through temporary improvements in energy availability. Thus, it would clearly be impermissible for NHTSA to rely on consumer demand to such an extent that it ignored the overarching goal of fuel conservation. At the other extreme, a standard with harsh economic consequences for the auto industry also would represent an unreasonable balancing of EPCA’s policies.
Id. at 1340 (footnote omitted). The court concluded that NHTSA’s consideration of consumer demand was permissible because Congress did not speak to the precise issue, and “it specifically delegated the process of setting light truck fuel economy standards with broad guidelines concerning the factors that the agency must consider. NHTSA has remained within the reasonable range permitted by those factors.” Id. at 1341; see also Pub. Citizen v. NHTSA, 848, F.2d 256, 265 (D.C.Cir.1988) (R. Ginsburg, J.).
In Public Citizen, the petitioners challenged the NHTSA’s lowering of the fuel economy standard for passenger cars for MY 1986. 848 F.2d at 259. They argued that NHTSA’s determination that the statutory 27.5 mpg standard was not economically practicable improperly elevated consumer demand and market forces, subordinated the statute’s technology-forcing design, and ignored the need of the nation to conserve energy. Id. at 264. The court held that NHTSA’s “consideration of the likelihood of economic hardship within its assessment of ‘economic praeticability[]’ must be accorded due weight.” Id. at 264-65. Based on economic analyses supplied by other governmental agencies, “NHTSA concluded that the industry-wide economic effects of the higher CAFE standard would be severe,” id. at 265, “including sales losses well into the hundreds of thousands, and job losses well into the tens of thousands,” id. at 264; see also 49 Fed.Reg. 41,250, 41,252 (Oct. 22, 1984).
Petitioners cite Public Citizen for the proposition that consideration of “economic practicability” allows lowering fuel economy standards only if a higher standard would cause substantial economic hardship to a manufacturer with a substantial share of the market. But that is not precisely what Public Citizen held. Rather, that court concluded that given the extensive *1196evidence in the record showing that severe economic hardship would result from a higher standard, NHTSA’s decision to lower the standard under those circumstances was not devoid of rational support. Pub. Citizen, 848 F.2d at 265.
The Public Citizen court held that NHTSA’s balancing of the statutory factors in 49 U.S.C. § 32902(f) was reasonable given that the possible energy savings from the higher standard did not outweigh the severe economic costs, id. at 265, since “the maximum potential increase in annual fuel consumption attributable to th[e] rule would amount to less than 0.1 percent of current consumption,” id. at 268. The court observed, “NHTSA found the maximum yearly impact of the lower (26.0 mpg) standard on U.S. gasoline consumption to be 210 million gallons, 0.3% of annual U.S. gasoline consumption and 0.09% of annual U.S. petroleum consumption. That savings, NHTSA stated, was not commensurate with ‘potential sales losses to the industry in the hundreds of thousands, "job losses in the tens of thousands, or the unreasonable restriction of consumer choices.’ ” Pub. Citizen, 848 F.2d at 260-61 (citation omitted). In sum, Congress did not “offer[ ] a more precise balancing formula for the agency to apply to the four ... factors,” and “[i]n the absence of a sharper congressional delineation,” the court could not conclude that, under the circumstances presented there, NHTSA’s decision was not a reasonable accommodation of conflicting statutory policies. Id. at 265.
In this rulemaking, NHTSA does not set forth its interpretation of the four factors in 49 U.S.C. § 32902(f). It simply states that in determining the “maximum feasible” fuel economy level, NHTSA “assesses what is technologically feasible for manufacturers to achieve without leading to adverse economic consequences, such as a significant loss of jobs or the unreasonable elimination of consumer choice.” 70 Fed. Reg. at 51,425; 71 Fed.Reg. at 17,585 (citing Pub. Citizen, 848 F.2d at 264). NHTSA “balance[s]” the four factors in § 32902(f), “along with other factors such as safety,” in determining the CAFE standards. 71 Fed.Reg. at 17,588, 17,655. In earlier rulemakings, NHTSA interpreted “technological feasibility” to mean “whether particular methods of improving fuel economy will be available for commercial application in the model year for which a standard is being established,” “economic practicability” to mean “whether the implementation of projected fuel economy improvements is within the economic capability of the industry,” “effect of other Federal motor vehicle standards on fuel economy” to mean “an analysis of the unavoidable adverse effects on fuel economy of compliance with emission, safety, noise, or damageability standards,” and “the need of the Nation to conserve energy” to mean “the consumer cost, national balance of payments, environmental, and foreign policy implications39 of our need for large quantities of petroleum, especially imported petroleum.” 42 Fed.Reg. 63,184, 63,188 (Dec. 15, 1977) (emphasis added); see also Ctr. for Auto Safety, 793 F.2d at 1325 n. 12.
NHTSA “reeognize[s] that [it] in the past has expressed its belief that the statutory consideration of economic practicability differs from, but does not preclude consideration of, cost/ benefit analysis.” 70 Fed.Reg. at 51,435. In its final rule establishing passenger automobile CAFE standards for MYs 1981-1984, NHTSA stated, “not equating cost-benefit considerations with economic practicability is consistent *1197with the goal of achieving maximum feasible fuel economy by allowing economically and technologically possible standards which will improve fuel economy but which an analysis, subject to many practical limitations, might indicate are not cost-beneficial.” See 42 Fed.Reg. 33,534, 33,536 (1977). The agency further opined, “A cost-benefit analysis would be useful in considering [economic practicability], but sole reliance on such an analysis would be contrary to the mandate of the Act.”40 Id. at 33,537. In this rulemaking, however, NHTSA states that “the cosVbenefit anal-yses conducted today ... are substantially more robust than those conducted in decades past and provide a more substantial basis for consideration of economic practicability.” 70 Fed.Reg. at 51,435.
We agree with NHTSA that “EPCA neither requires nor prohibits the setting of standards at the level at which net benefits are maximized.” Id. at 51,435. The statute is silent on the precise question of whether a marginal cost-benefit analysis may be used. See Chevron, 467 U.S. at 843, 104 S.Ct. 2778. Public Citizen and Center for Auto Safety persuade us that NHTSA has discretion to balance the oft-conflicting factors in 49 U.S.C. § 32902(f) when determining “maximum feasible” CAFE standards under 49 U.S.C. § 32902(a).
To be clear, we reject only Petitioners’ contention that EPCA prohibits NHTSA’s use of marginal cost-benefit analysis to set CAFE standards. Whatever method it uses, NHTSA cannot set fuel economy standards that are contrary to Congress’s purpose in enacting the EPCA — energy conservation. We must still review whether NHTSA’s balancing of the statutory factors is arbitrary and capricious. Additionally, the persuasiveness of the analysis in Public Citizen and Center for Auto Safety is limited by the fact that they were decided two decades ago, when scientific knowledge of climate change and its causes were not as advanced as they are today.41 The need of the nation to con*1198serve energy is even more pressing today than it was at the time of EPCA’s enactment. See, e.g., NRDC Cmt. at 4, 11 (“When fuel economy legislation was first enacted, America consumed 16.3 million barrels of oil per day and 35.8 percent of U.S. oil came from imports. In the nearly 30 years since then, oil consumption has risen to over 20 million barrels per day and 56 percent of U.S. oil is imported. If fuel economy standards are not strengthened, these trends are only expected to get worse, with transportation oil use driving 80 percent of U.S. oil demand growth through 2025 and imports rising to 68 percent of U.S. oil demand. The light duty vehicle fleet currently consumes 8.3 million barrels per day, and in the absence of stronger standards, that is projected to grow to 12.45 million barrels by 2025.”); NAS Report at 13-14, 20. What was a reasonable balancing of competing statutory priorities twenty years ago may.not be a reasonable balancing of those priorities today.42
2. Failure to monetize benefits of greenhouse gas emissions reduction
Even if NHTSA may use a cost-benefit analysis to determine the “maximum feasible” fuel economy standard, it cannot put a thumb on the scale by undervaluing the benefits and overvaluing the costs of more stringent standards. NHTSA fails to include in its analysis the benefit of carbon emissions reduction in either quantitative or qualitative form. It did, however, include an analysis of the employment and sales impacts of more stringent standards on manufacturers. See 71 Fed.Reg. at 17,590-91.
To determine the “maximum feasible” CAFE standards, NHTSA began with the fuel economy baselines for each of the seven largest manufacturers — that is, “the fuel economy levels that manufacturers were planning to achieve in those years.” Id. at 17,581. NHTSA then “add[ed] fuel saving technologies to each manufacturer’s fleet until the incremental cost of improving its fuel economy further just equal[ed] the incremental value of fuel savings and other benefits from doing so.” Id. at 17,-596. The standard is further adjusted “until industry-wide net benefits are maximized. Maximization occurs when the incremental change in industry-wide compliance costs from adjusting it further would be exactly offset by the resulting incremental change in benefits.” Id. NHTSA claims that this “cost-benefit analysis carefully considers and weighs all of the benefits of improved fuel savings,” and that “there is no compelling evidence that the *1199unmonetized benefits would alter our assessment of the level of the standard for MY2011.” Mat 17,592.
Under this methodology, the values that NHTSA assigns to benefits are critical. Yet, NHTSA assigned no value to the most significant benefit of more stringent CAFE standards: reduction in carbon emissions. Petitioners strongly urged NHTSA to include this value in its analysis, and they cited peer-reviewed scientific literature in support. NRDC cited figures for the benefit of carbon emissions reduction ranging from $8 to $26.50 per ton C02, based on values assigned by the California Public Utilities Commission, the Idaho Power Company, and the European Union (EU) carbon trading program.43 NRDC Cmt. at 8. NRDC also cited a study published by the National Commission on Energy Policy, which “found that measures mitigating climate change emissions have estimated benefits of $3-19 per ton of carbon dioxide equivalent. The Commission recommends a price of $7 per ton beginning in 2010 and then rising 5 percent each year.” Id. at 23 (footnote omitted). Environmental Defense and the Union of Concerned Scientists recommended a minimum value of $50 per ton carbon (or $13.60 per ton C02), which reflects a mean marginal damage cost developed in 28 peer-reviewed studies.44 Environmental Defense Cmt. at 6, A-4;45 UCS Cmt. at 16. Valuing carbon emissions at $50 per ton carbon translates into approximately $0.15 per gallon of gasoline saved. UCS Cmt. at 16. The NAS committee, on which NHTSA relies for other aspects of its analysis, also valued the benefit of carbon emissions reduction at $50 per ton carbon. NAS Report at 85.
NHTSA acknowledged that “[c]onserv-ing energy, especially reducing the nation’s dependence on petroleum, benefits the U.S. in several ways. [It] has benefits for *1200economic growth and the environment, as well as other benefits, such as reducing pollution and improving security of energy supply.” 71 Fed.Reg. at 17,644. NHTSA also acknowledged the comments it received that recommended values for the benefit of carbon emissions reduction; however, the agency refused to place a value on this benefit. See id. at 17,638.46 NHTSA stated:
The agency continues to view the value of reducing emissions of C02 and other greenhouse gases as too uncertain to support their explicit valuation and inclusion among the savings in environmental externalities from reducing gasoline production and use. There is extremely wide variation in published estimates of damage costs from greenhouse gas emissions, costs for controlling or avoiding their emissions, and costs of sequestering emissions that do occur, the three major sources for developing estimates of economic benefits from reducing emissions of greenhouse gases. Moreover, ... commenters did not reliably demonstrate that the un-monetized benefits, which include C02, and costs, taken together, would alter the agency’s assessment of the level of the standard for MY 2011. Thus, the agency determined the stringency of that standard on the basis of monetized net benefits.
Id.; see also FRIA, at VIII-64 to 65.47
NHTSA’s reasoning is arbitrary and capricious for several reasons. First, while the record shows that there is a range of values, the value of carbon emissions reduction is certainly not zero. NHTSA conceded as much during oral argument when, in response to questioning, counsel for NHTSA admitted that the range of values begins at $3 per ton carbon. NHTSA insisted at argument that it placed no value on carbon emissions reduction rather than zero value. We fail to see the difference. The value of carbon emissions reduction is nowhere accounted for in the agency’s analysis, whether quantitatively or qualitatively. This position also contradicts NHTSA’s own explanation in the Final Rule that “the agency determined the stringency of [the MY 2011] standard on the basis of monetized net benefits.” 71 Fed.Reg. at 17,638 (emphasis added).48 By presenting a scientifically-supported range of values that does not begin at zero, Petitioners have shown that it is possible to monetize the benefit of carbon emissions reduction.
Second, NHTSA gave no reasons why it believed the range of values presented to it was “extremely wide”; in fact, several *1201commenters and the NAS committee recommended the same value: $50 per ton carbon. The NAS committee selected the value of $50 per ton carbon although it acknowledged the wide range of values in the literature and the potential controversy in selecting a particular value. NAS Report at 85. NHTSA argues that the problem was not simply “the ultimate value to be assigned, but the wide variation in published estimates of the three major underlying costs of carbon dioxide emissions — the cost of damages caused by such emissions, the costs of avoiding or controlling such emissions, and the costs of sequestering resulting emissions.” NHTSA Br. at 49. But NHTSA fails to explain why those three “underlying costs” are relevant to the question of how carbon emissions should be valued. We are convinced by Petitioners’ response:
To monetize the benefits of reducing C02 emissions from automobiles, NHTSA did not need to calculate the “costs of sequestering emissions.” Carbon capture and sequestration, though a feasible means of reducing emissions from large stationary sources such as coal-fired power plants, was not within the range of actions at issue in this automobile fuel economy rulemaking. Nor were “costs for controlling or avoiding [C02] emissions” a genuine methodological barrier here: NHTSA already performed an elaborate analysis of the costs of mandating increases in fuel economy. For purposes of this rule-making, that was the relevant category of control costs.
EPCA Reply Br. at 10-11.49 In sum, there is no evidence to support NHTSA’s conclusion that the appropriate course was not to monetize or quantify the value of carbon emissions reduction at all.
Citizens for Clean Air v. EPA 959 F.2d 839 (9th Cir.1992), which NHTSA cites to support its contention that agencies may decline to adopt a “particular monetary value” when the “costs and benefits are too uncertain,” NHTSA Br. at 48, is inappo-site. In Citizens for Clean Air, petitioners filed for administrative review of a state agency’s grant of a permit for construction of a solid waste incinerator. Citizens for Clean Air, 959 F.2d at 841. EPA denied the petitions, and this court held that the decision of the EPA not to consider recycling as a possible “best available control technology” under the Clean Air Act was not arbitrary or capricious. Id. at 841-42. The EPA noted in its proposed rule that “it was ‘unable to reliably quantify the emission reductions attributable to materials separation when a[ ] [waste incinerator] is equipped with highly efficient at-the-stack air pollution control devices.’ ” Id. at 844 (citation omitted). Petitioners submitted “no hard evidence” that recycling would reduce air pollution when the waste incinerators are already equipped with “state-of-the-art pollution control equipment” (e.g., scrubbers). Id. at 847-48. In addition, the Clean Air Act required “that the proposed technology [i.e., recycling] be the best available control technology, and in the absence of anything specific or quantifiable in support ... we conclude that EPA’s decision not to consider recycling in permitting Spokane’s incinerator was not arbitrary or capricious.” Id. at 848. The petitioners in Citizens for Clean Air had to satisfy such a high statutory threshold (“best available control technology”), and they could not satisfy that threshold without hard evidence. By con*1202trast, Petitioners here provided substantial evidence of the value of carbon emissions reduction, and they do not have to satisfy a high statutory threshold.
Third, NHTSA’s reasoning is arbitrary and capricious because it has monetized other uncertain benefits, such as the reduction of criteria pollutants, crash, noise, and congestion costs, see FRIA at VIII-73 to 80, and “the value of increased energy security,” 71 Fed.Reg. at 17,592. Dr. Michael Wang of the Center for Transportation Research at Argonne National Laboratory stated in his peer review of the CAFE compliance and effect model50 used by NHTSA in its rulemaking that the wide range of dollar values per ton of C02 “is not a good reason that CO[2] dollar values are not included .... The' same can be said [of] dollar values for criteria pollutants. Yet, monetary values for criteria pollutant emissions are included in the model.” Wang Cmt. at 6.51
Fourth, NHTSA’s conclusion that com-menters did not “reliably demonstrate” that monetizing the value of carbon reduction would have affected the stringency of the CAFE standard “ ‘runs counter to the evidence’ ” before it. NRDC v. U.S. Forest Serv., 421 F.3d 797, 806 (9th Cir.2005) (citation omitted). The Union of Concerned Scientists concluded that “including [a $50/tC value] in the determination of cost-efficient fuel economy could increase the 2011 targets by an average of 0.4-1.1 mpg.” UCS Cmt. at 16. Given that the CAFE standards set by NHTSA increase only 1.5 mpg from MY 2008 to 2011,52 an additional 0.4 to 1.1 mpg increase by MY 2011 is significant. In addition, Environmental Defense “calculate^] the benefits of the cumulative reductions at $50/tC and 3% discount rate at $19.7 billion by 2020 and $28.4 billion by 2030 (current dollars).” Environmental Defense Cmt. at 6.
We agree with Petitioners that the values they suggest, 10-22 cents per gallon of gasoline in NHTSA’s estimation, would not be a small benefit. Under NHTSA’s own calculation that Reformed CAFE will save 2.8 billion gallons of gasoline for MY 2011 light trucks, see 71 Fed.Reg. at 17,619, 10-22 cents a gallon of carbon benefits “would yield hundreds of millions of dollars in benefits even after discounting — benefits that by themselves would be substantial in relation to the net benefits that NHTSA calculated for the rule.” EPCA Reply Br. at 12 (citing 71 Fed.Reg. at 17,623 (showing net benefits of $461 million for MY 2011 under Reformed CAFE)). NHTSA simply did not “ ‘examine the relevant data and articulate a satisfactory explanation for its action including a rational connection between the facts found and the choice made.’ ” Motor Vehicle Mfrs. Ass’n, 463 U.S. at 43, 103 S.Ct. 2856 (quoting Burlington Truck Lines v. United States, 371 U.S. 156, 168, 83 S.Ct. 239, 9 L.Ed.2d 207 (1962)).
Finally, there is no merit to NHTSA’s unfounded assertion that if it had accounted for the benefit of carbon emissions reduction, it would have had to account for the adverse safety effects of downweight-ing, and the two would have balanced out, resulting in no change to the final CAFE standards. No evidence supports this assertion. The assertion is also based on the controversial assumption that higher fuel *1203economy standards for light trucks causes adverse safety effects from downweight-ing.
Thus, NHTSA’s decision not to monetize the benefit of carbon emissions reduction was arbitrary and capricious, and we remand to NHTSA for it to include a monetized value for this benefit in its analysis of the proper CAFE standards.
3. Vehicle safety analysis
Petitioners argue that NHTSA acted arbitrarily and capriciously when it omitted weight reduction for vehicles between 4,000 and 5,000 lbs. curb weight53 as a cost-effective measure manufacturers could use to increase fuel economy. The NAS committee found that weight reduction for vehicles greater than 4,000 lbs. curb weight would result in a safety benefit: “total fatalities in a hypothetical fleet of relatively modern passenger vehicles would be reduced by about 0.26 percent if all pickups and SUVs weighing more than 4,000 lbs. were replaced with pickups and SUVs weighing 3,500 to 4,000 lbs.”54 NAS Report at 72. The NAS committee also noted that such weight reduction would improve safety in part because it improves “crash compatibility,” or the variance in vehicle weights between large and small vehicles. Id. at 5.55
The study by Dr. Charles Kahane, cited by the NAS committee and NHTSA, did not find a safety benefit resulting from reducing the weight of light trucks between 4,000 to 5,000 lbs., but it found that “[ojverall, light trucks weighing 3,870 pounds or more are involved in fatal crashes that result in a total of 14,705 fatalities per year. A 100-pound reduction would not significantly change those fatalities.” Kahane Study at 161.56
*1204In the Final Rule, NHTSA applied a “confidence bound” of approximately 1,000 lbs. to the 3,900 figure, “given the significant statistical uncertainty around that figure,” and “used 5,000 lbs. as the threshold for considering weight reduction.” 71 Fed.Reg. at 17,627. The source of the 1,000 lbs. confidence bound is “additional empirical work found in Kahane’s study” dealing with the crossover weight. FRIA, at V-15. NHTSA recognized that the crossover weight and the point of zero net impact (3,900 lbs.) differ, but it concluded that “they would both tend to have similar sampling errors.” Id.
Petitioners argue that NHTSA’s selection of a 1,000 lbs. confidence bound is arbitrary and capricious because Kahane’s 4,000 lbs. figure already accounts for uncertainty, and the confidence bound is taken from a different part of Kahane’s study. They also argue that NHTSA arbitrarily excluded from its uncertainty analysis “the benefits of reducing crash incompatibility resulting from weight reductions in the 3,900-5,000 pound class, and whether these might offset any statistically uncertain safety costs.” EPCA Reply Br. at 15.
Petitioners’ arguments have some merit, and we believe this is a close question. However, we conclude that NHTSA’s explanation for using the confidence interval for the crossover weight — that both the crossover weight and the point of zero net impact would have similar sampling errors — does not “rel[y] on factors which Congress has not intended it to consider, entirely fail[] to consider an important aspect of the problem, ... [or] run[ ] counter to the evidence before the agency, [and is not] so implausible that it could not be ascribed to a difference in view or the product of agency expertise.” Motor Vehicle Mfrs. Ass’n, 463 U.S. at 43, 103 S.Ct. 2856. Petitioners have not sufficiently established the deficiencies in NHTSA’s reasoning.
4. Backstop for Reformed CAFE
Under Reformed CAFE, a manufacturer’s required CAFE level would depend on its fleet mix. See 71 Fed.Reg. at 17,587-88, 17,595-96.57 Reformed CAFE (setting individual fuel economy targets for vehicles of every footprint size) plus a backstop (overall fleet-wide average) would prevent manufacturers from upsizing their vehicles or producing too many large footprint vehicles, if the backstop were set high enough. Under Unreformed CAFE, manufacturers had to meet only a fleet-wide average, which means that they could increase the number of small vehicles (with higher fuel economy) they produced in order to balance out the larger vehicles (with lower fuel economy) and achieve the required CAFE standard. NHTSA argues that Reformed CAFE will alleviate the problem of downweighting because there will no longer be a large gap between the CAFE targets for passenger cars and light trucks. See 71 Fed.Reg. at 17,620 (“Reformed CAFE, as adopted today, links the level of the average fuel economy targets to the size of footprint so that there is an incentive to reduce weight only to the extent one can do so while also preserving size.”).
Petitioners generally agree that Reformed CAFE, with its progressive fuel economy targets based on vehicle foot*1205print, is an improvement over Unreformed CAFE. However, they argue that Reformed CAFE must include a “backstop” so that the “minimum level of average fuel economy applicable to a manufacturer in a model year” would not be determined solely by the manufacturer’s fleet mix. See 49 U.S.C. § 32901(a)(6). They argue that the statutory language — “maximum feasible average fuel economy level,” id. § 32902(a) (emphasis added), “minimum level of average fuel economy applicable to a manufacturer in a model year,” id. § 32902(a)(6) (emphasis added) — and the statutory structure contemplate a fixed minimum CAFE standard for light trucks.
NHTSA argues that a backstop would unduly limit consumer choice and perpetuate the problems with Unreformed CAFE. See 71 Fed.Reg. at 17,592-93. It argues that the statutory requirement that there be a “minimum” level of average fuel economy applicable to a manufacturer does not necessarily mean a fixed minimum and is consistent with a minimum standard applicable to a manufacturer based on that manufacturer’s fleet mix.
Neither the EPCA’s language nor structure explicitly requires NHTSA to adopt a backstop.58 The issue is whether it was arbitrary or capricious in not adopting a backstop. Under Reformed CAFE, manufacturers would still be required to meet a minimum average fuel economy level— there would simply be no corporate minimum average fuel economy level. That is, each vehicle of a particular footprint would be required to meet a minimum average fuel economy level, but there would be no fleet-wide minimum. See 71 Fed.Reg. at 17,608-09. The corporate or fleet-wide minimum would depend entirely on the number of vehicles of each footprint that the manufacturer decided to produce. See id.
Although Congress has not directly spoken on this issue, it has directed the agency to set the average fuel economy level for light trucks at the “maximum feasible” level, 49 U.S.C. § 32902(a), considering technological feasibility, economic practicability, the need of the nation to conserve energy, and the effect of other motor vehicle standards of the government, id. § 32902(f). NHTSA did not consider these factors in deciding whether to adopt a backstop. See 71 Fed.Reg. at 17,592-93. Instead, the agency explained:
The intent of the CAFE program is not to preclude future mix shifts and design changes in response to consumer demand. A backstop would likely have this influence.... Such a system would be in opposition to congressional intent to establish a regulatory system that does not unduly limit consumer choice.
71 Fed.Reg. at 17,617. NHTSA may consider consumer demand, but “it would clearly be impermissible for NHTSA to rely on consumer demand to such an extent that it ignored the overarching goal of fuel conservation.” Ctr. for Auto Safety, 793 F.2d at 1340. We believe that NHTSA has committed this error here. Although EPCA is not intended to “unduly limit[] consumer choice,” H.R.Rep. No. 94-340, at 87 (1975), as reprinted in 1975 U.S.C.C.A.N. 1762, 1849, energy conservation is the fundamental purpose of the statute and an explicit statutory factor that NHTSA “shall” consider, see 49 U.S.C. *1206§ 32902(f). “An agency may not ignore factors Congress explicitly required be taken into account.” Earth Island Inst. v. Hogarth, 494 F.3d 757, 765 (9th Cir.2007) (citation omitted). NHTSA did not adequately consider the “need of the nation to conserve energy,” as it was required to do under 49 U.S.C. § 32902(f), and it has not argued that a backstop would be technologically infeasible or economically impracticable. See 71 Fed.Reg. at 17,592-93, 17,-617-18.
We do agree with NHTSA that the continuous function of Reformed CAFE will likely reduce the incentive to upsize because there will no longer be only two categories of CAFE standards (light trucks and passenger cars). See 71 Fed. Reg. at 17,621. But, Petitioners raise well-founded concerns (given the historical trend) that a floating fleet-mix-based standard would continue to permit upsizing— which is not just a function of consumer demand, but also a function of manufacturers’ own design and marketing decisions. We remand to NHTSA for it to reconsider under the proper standard whether to adopt a backstop based on the factors in 49 U.S.C. § 32902(f).
5. Transition period
The Final Rule permits manufacturers to choose to comply with Unreformed CAFE or Reformed CAFE in MYs 2008-2010. 71 Fed.Reg. at 17,593-95, 17,-639. Beginning in MY 2011, manufacturers must comply with Reformed CAFE. Petitioners argue that the transition period offered by NHTSA to manufacturers is impermissible under EPCA since NHTSA must prescribe a single maximum feasible average fuel economy level. They argue that NHTSA has “[i]n effect ... granted an automatic exemption to manufacturers unwilling to meet the Reformed CAFE standards between 2008 and 2010,” and that this is impermissible because the statute grants exemptions only on an individual basis and if the manufacturer meets certain statutory criteria. EPCA Br. at 43-44 (citing 49 U.S.C. § 32902(d)).
We conclude that the transition period is not prohibited by EPCA, and NHTSA’s decision to allow it is not arbitrary or capricious. First, NHTSA has not granted an “exemption” to manufacturers during the transition period, because manufacturers are still required to comply with the CAFE program — they just have a choice as to which standard with which to comply.
Second, and more importantly, NHTSA has provided a reasoned explanation for the transition period: it “will minimize the potential for unintended compliance burdens that may be experienced by a manufacturer as the result of shifting to a new regulatory structure,” it is “critical given that this is the first comprehensive reform of the light truck CAFE program since its inception,” and “the structure of the Reformed CAFE might require some manufacturers to revise their compliance strategies,” since it “minimizes the ability of manufacturers to offset the low fuel economy performance of larger vehicles by increasing the production of smaller vehicles with higher fuel economies. Manufacturers that relied on such a compliance strategy in the past might need to revise their product plans in order to comply with the Reformed CAFE standard.” 71 Fed.Reg. at 17,593. NHTSA also noted that “Manufacturers develop product plans for their fleets at least 5 years in advance, plans which incorporate consideration of CAFE compliance.” Id. at 17,594.
6. Changing the definition of passenger and non-passenger automobiles in order to close the SUV loophole
Petitioners challenge NHTSA’s decision not to reform the SUV loophole. They argue that this decision is arbitrary *1207and capricious because it runs counter to the evidence showing that the majority of SUV's, minivans, and pickup trucks function solely or primarily as passenger vehicles, and because NHTSA has not provided a reasoned explanation for why the transition to Reformed CAFE could not be accomplished at the same time as a revision in the definitions.
The EPCA defines “passenger automobile” as “an automobile that the Secretary decides by regulation is manufactured primarily for transporting not more than 10 individuals,” excluding “an automobile capable of off-highway operation that the Secretary decides ... has a significant feature except 4-wheel drive designed for off-highway operation” and is 4-wheel drive or more than 6,000 lbs. GVWR. 49 U.S.C. § 32901(a)(16). “Non-passenger automobiles” are thus defined by exclusion. NHTSA defines an automobile other than a passenger automobile as a “light truck,” a term not used in the statute. 49 C.F.R. § 528.5 (2007). Under 49 U.S.C. § 32901(a)(16), the Secretary has discretion to decide what constitutes a “passenger automobile” within the confines of the listed criteria.
NHTSA initially sought input on ways to revise the regulatory distinction because the passenger automobile/light truck distinction had become obsolete: “The application of the regulation to the current vehicle fleet (designed with the regulatory distinctions in mind) less clearly differentiates between passenger cars and light trucks than it did in the 1970s.” 68 Fed. Reg. at 74,927 (ANPRM). However, in the NPRM, NHTSA decided not to:
chang[e] those classification regulations at this time in part because [NHTSA] believe[s] an orderly transition to Reformed CAFE could not be accomplished if [NHTSA] simultaneously change[s] which vehicles are included in the light truck program and because, as applied in MY 2011, Reformed CAFE is likely to reduce the incentive to produce vehicles classified as light trucks instead of as passenger cars.
70 Fed.Reg. at 51,422. Ultimately, NHTSA did not change the light truck definition other than by expanding the flat floor provision to include vehicles with folding seats, if the vehicles have at least three rows of designated seating. See 49 C.F.R. § 523.5(a)(5); 71 Fed.Reg. at 17,-650-52.
We conclude that NHTSA’s decision not to otherwise revise the passenger automobile/light truck definitions is arbitrary and capricious. First, NHTSA has not provided a reasoned explanation of why an orderly transition to Reformed CAFE could not be accomplished at the same time that the passenger automobile/light truck definitions are revised.
Second, NHTSA asserts that it reasonably decided to look to the purpose for which a vehicle is manufactured instead of consumers’ use of a vehicle because it is a more objective way of differentiating between passenger and non-passenger automobiles. But this overlooks the fact that many light trucks today are manufactured primarily for transporting passengers, as NHTSA itself has acknowledged: “Many vehicles produced today, while smaller than many other passenger cars, qualify as light trucks because they have been designed so that their seats can be easily removed and their cargo carrying capacity significantly enhanced.” 68 Fed.Reg. at 74,927 (emphasis added); see also 71 Fed. Reg. at 17,621 n. 102 (“NAS Report ... noted that [the passenger automobile/light truck fuel economy] gap created an incentive to design vehicles as light trucks instead of cars.”). Today’s design differences, which capitalize on the lower light truck CAFE standard, are the very reason that NHTSA sought input on ways to revise the regulatory distinction “in light of *1208the current and emerging motor vehicle fleet.” See 68 Fed.Reg. at 74,927.
In addition, NHTSA’s new focus on the purpose for which automobiles are manufactured conflicts with its earlier assertion that “Congress intended that passenger automobiles be defined as those used primarily for the transport of individuals.” Id. at 74,926 (emphasis added); see also id. at 74,913 (“The market suggests that while some light trucks may be used primarily to transport passengers, their ‘peak use or value’ capability (towing boats, hauling heavy loads, etc.) may be a critical factor in the purchase decision.” (emphasis added)).
Third, NHTSA’s decision runs counter to the evidence showing that SUVs, vans, and pickup trucks are manufactured primarily for the purpose of transporting passengers and are generally not used for off-highway operation. The NAS committee found that:
The less stringent CAFE standards for trucks did provide incentives for manufacturers to invest in minivans and SUVs and to promote them to consumers in place of large cars and station wagons. ... By shifting their product development and investment fpcus to trucks, they created more desirable trucks with more carlike features: quiet, luxurious interiors with leather upholstery, top-of-the-line audio systems, extra rows of seats, and extra doors.
NAS Report at 18 (emphasis added); see also id. at 23 (noting the exploding demand for light trucks such as minivans and “four-door SUVs and pickup trucks with passenger-friendly features such as extra rows of seats”). Consumers use light trucks primarily for passenger-carrying purposes in large part because that is precisely the purpose for which manufacturers have manufactured and marketed them. See, e.g., App. A to Public Citizen Cmt. (Kathleen Kerwin, “You Call This the Family Car? Pickups with Roomy Cabs Become a Status Accessory,” Business Week, Apr. 26, 1999.). A pickup truck usage study conducted by R.L. Polk & Co.59 showed that 73% of light pickup users use their trucks to carry passengers on a daily or weekly basis, 68% use them for personal trips on a daily or weekly basis, 58% use them for commuting on a daily or weekly basis, 59% never use them for towing, and 69% never use them for driving off-road. Polk Study at 11. Seventy-three percent of medium pickup users use them for carrying passengers on a daily or weekly basis, 65% use them for commuting on a daily or weekly basis (61% daily), and 64% never use them for driving off-road. Polk Study at 12. Even among heavy pickup users, 76% use them for carrying passengers on a daily or weekly basis, and 52% never use them for driving off-road. Polk Study at 13.60 The NAS Committee further found:
When CAFE regulations were originally formulated, different standards were set for passenger vehicles and for work/cargo vehicles .... because [work/cargo vehicles] needed extra power, different gearing, and less aerodynamic body configurations to carry out their utilitarian, load-carrying functions. ... [But this] working definition distinction between a car for personal use and a truck for work use/cargo transport] has broken down, initially with minivans, and more recently with sport utility vehicles and other “cross-over” vehicles that may be *1209designed for peak use but which are actually used almost exclusively for personal transport.... The car/truck distinction has been stretched well beyond its original purpose.
NAS Report at 88 (internal quotation marks and citation omitted). One of the changes the NAS committee recommended to alleviate this problem was to “tighten” the definition of a light truck, a step the EPA has already taken for emissions standards purposes. Id. We agree with Petitioners that NHTSA’s decision not to do the same was arbitrary and capricious, especially in light of EPCA’s overarching goal of energy conservation. Thus, we remand to NHTSA to revise its regulatory definitions of passenger automobile and light truck or provide a valid reason for not doing so.
7. Exclusion of 8,500-10,000 lb. pickup trucks from CAFE regulation
Petitioners argue that NHTSA’s decision not to regulate the fuel economy of vehicles between 8,500 and 10,000 lbs. GVWR (generally referred to as Class 2b trucks),61 other than MDPVs, is arbitrary and capricious because fuel economy standards for these vehicles are feasible and will result in significant energy conservation. See 49 U.S.C. § 32901(a)(3). We agree.
All 4-wheeled, fuel and alternative fuel-propelled vehicles manufactured for use on roads and highways that are 6,000 lbs. gross vehicle weight or less are automobiles. See 49 U.S.C. § 32901(a)(3)(A). Vehicles more than 6,000 but less than 10,000 lbs. gross vehicle weight are “automobile[s]” for the purpose of fuel economy regulation “if the Secretary decides by regulation that — (i) an average fuel economy standard under this chapter for the vehicle is feasible; and (ii) an average fuel economy standard ... for the vehicle will result in significant energy conservation or the vehicle is substantially used for the same purposes as a vehicle rated at not more than 6,000 pounds gross vehicle weight.” Id. § 32901(a)(3)(B). Since 1978, NHTSA has defined vehicles 8,500 lbs. GVWR or less as automobiles. See 49 C.F.R. § 523.3(b)(2)(iii).
The ANPRM presented two options under which the fuel economy of vehicles with a GWVR of up to 10,000 lbs. could be regulated. See 68 Fed.Reg. at 74,930. One option was to include MDPVs, vehicles with a GVWR of greater than 8,500 but less than 10,000 lbs. that are designed primarily for the transportation of persons. NHTSA explained that “[t]his definition would essentially make SUVs between 8,500 and 10,000 lbs. GVWR subject to CAFE, while continuing to exclude most medium and heavy-duty pickups and most medium and heavy-duty cargo vans that are primarily used for agricultural and commercial purposes.” Id. Another option was to make all vehicles between 8,500 and 10,000 lbs. GVWR subject to CAFE standards. Id. NHTSA invited comments on these and any other proposals to regulate vehicles between 8,500 and 10,000 lbs. GVWR. Id. The NPRM discussed the inclusion of MDPVs, but it did not address the proposal to regulate all vehicles between 8,500 and 10,000 lbs. GVWR. See 70 Fed.Reg. at 51,455-56.
*1210The Final Rule incorporates MDPVs into the definition of “automobile” such that these vehicles would be regulated as light trucks beginning in MY 2011. 49 C.F.R. § 523.3(b)(3); 71 Fed.Reg. at 17,-648. NHTSA declined to regulate other vehicles between 8,500 and 10,000 lbs. because unlike MDPVs, they:
are not subject to EPA testing that provides the data necessary to determine compliance with the CAFE program. Inclusion of the heavier-rated-non-MDPVs would increase the test burden for manufacturers. These vehicles would be subject to a whole new testing regime. Moreover, because these vehicles are not subject to comparable testing requirements, there is not sufficient data to estimate a fuel economy baseline. Without a reliable baseline, the agency is unable to determine fuel economy targets that would result in required fuel economy levels that are economically practicable and technologically feasible.
71 Fed.Reg. at 17,650.
We conclude that this is not a reasoned explanation for excluding Class 2b trucks from CAFE regulation. First, Petitioners presented compelling evidence that setting fuel economy standards for Class 2b trucks is feasible. For example, a Department of Energy research planning study included estimates:
based on detailed simulation modeling of both the city and highway driving cycles working from a baseline Class 2b truck; baseline estimates were 21.1 mpg city, 15.4 mpg highway, and 13.6 mpg combined .... [The study] identified technology options capable of yielding substantial improvements, including 50% higher fuel economy with technologies available over a 7-year horizon and, with use of hybrid, engines in diesel versions of the vehicles, a near doubling of the fuel economy of a baseline gasoline Class 2b pickup would be a feasible “stretch goal.”
App. F to Environmental Defense Cmt. at 1. An Argonne National Laboratory study identified numerous technological improvements that could be applied to Class 2b trucks, including “aerodynamic improvements,” “lower tire rolling resistance,” “improved transmissions,” “turbocharging for diesel engines,” “other engine refinements,” “integrated starter-generator,” and “hybrid-electric powertrains.” Id. Another study, published by the American Council for an Energy-Efficient Economy, found that the potential fuel economy improvement was “37% ... over a baseline full-size pickup.” Id. at 2. Overall, while Class 2b trucks are “designed with heavier frames and higher capacities, and therefore larger powertrains and other components, [they] (primarily heavy-duty pickups) entail substantially the same engineering as vehicles under 8500 lbs GVWR, and in many cases share components.” Id. at 1; see also UCS Cmt. at 34.
Second, Petitioners presented substantial evidence that setting CAFE standards for Class 2b trucks would result in significant energy conservation and that these vehicles are substantially used for the same purposes as a vehicle 6,000 lbs. GVWR or less.62 Class 2b trucks constitute the majority of vehicles in the 8,500 to 10,000 lb. GVWR class. See ACEEE Cmt. to ANPRM at 10 (“[P]ickups constitute about 85% of vehicles in the 8,500-10,000 lb. weight range.”).63 One of the Petitioners argued:
*1211EPA estimates sales of class 2b trucks at 931,000 per year. Given higher per vehicle oil consumption, we estimate that 2b trucks consume 13% of overall demand from trucks under 10,000 pounds GVWR.
If class 2b trucks were to improve their fuel economy by 4% per year ... over MY 2008-2011, it would save 47,000 barrels of gasoline and diesel fuel per day by 2020 and reduce GHG emissions by a cumulative 16 mmtC over that time frame. Even at comparable fuel economy improvements to those NHTSA proposes for the largest class 2a trucks regulated under the Reformed system— roughly 2% per year, the country would save 24,000 barrels of gasoline and diesel fuel per day by 2020. These amounts are significant: A 24,000 barrel per day (bpd) saving would be equivalent to $700 million of annual savings at a relatively modest shadow price of $1.90/gallon. This far exceeds the $100 million threshold for a “significant energy action” under [Executive Order] 13211.
Environmental Defense Cmt. at 10; see also UCS Cmt. at 34 (estimating that “[i]f these vehicles had been held to the same fuel economy standard as other light-duty trucks, the total fuel consumption by trucks under 10,000 lbs GVWR would be approximately 890 million gallons less in 2005, for a savings of nearly 60,000 barrels of oil per day. This corresponds to about 18% of Class 2b fuel use.”).
The evidence also shows that Class 2b trucks are “substantially used for the same purposes” as vehicles not more than 6,000 lbs. GVWR. The Polk Study showed that 76% of heavy pickup truck owners use them for carrying passengers on a daily or weekly basis, 57% use them for personal trips on a daily or weekly basis, 49% use them for commuting on a daily basis, and 52% never use them for driving off-road. Polk Study at 13; see also Environmental Defense Cmt. at 10 (citing a 2002 vehicle inventory and use survey conducted by the U.S. Census Bureau for the argument that “[i]n the intervening decades [since NHTSA revised its definition of ‘automobile’ to include Class 2a trucks but exclude Class 2b trucks], trucks of all sizes have increasingly shifted from commercial uses to personal uses.”).64
NHTSA did not address any of this evidence in the Final Rule, and it does not argue that setting CAFE standards for Class 2b trucks would not be feasible,65 that it would not result in significant energy conservation, or that Class 2b trucks are not substantially used for the same purposes as smaller trucks. See 71 Fed. Reg. at 17,649-50; NHTSA Br. at 77-81. Instead, NHTSA makes the bold assertion that “the agency is not obliged to justify exclusion of such very large vehicles, as the statute has already excluded them, *1212subject to NHTSA’s discretionary decision to include such vehicles.” NHTSA Br. at 79 (emphasis in original). This is clearly wrong. The statute gives NHTSA some discretion in deciding whether the 49 U.S.C. § 32901(a)(3) factors are met for Class 2b trucks, but if these factors are satisfied, then they are “automobiles” for which NHTSA must set fuel economy standards.66
NHTSA asserts that without EPA testing data, it cannot set CAFE standards for Class 2b trucks. 71 Fed.Reg. at 17,650. But EPA already subjects most Class 2b trucks to the city and highway fuel economy tests (i.e., city and highway chassis dynamometer testing) that NHTSA asserts are needed for it to determine CAFE standards. See 40 C.F.R. § 86.101 (2007) (applying test procedures to gasoline-fueled trucks above 8,500 lbs. GVWR (“Otto-cycle complete heavy-duty vehicles”)). NHTSA does not dispute this fact.
Moreover, NHTSA has given no explanation of why it would be infeasible to set standards for Class 2b trucks without EPA’s tests. Instead, NHTSA’s position is merely that imposing one set of tests “minimize[s]” the “test burden to manufacturers.” See 71 Fed.Reg. at 17,649. This concern has no relevance to any of the statutory factors under 49 U.S.C. § 32901(a)(3)(B).
Finally, NHTSA’s reasoning is arbitrary because it decided that it is feasible to set CAFE standards for MDPVs even though they are not currently subject to EPA testing. See id. (“MDPVs are not currently required to undergo chassis dynamome-ter testing”). EPA will begin phasing in city chassis dynamometer testing for MDPVs in MY 2008, but MDPVs are exempted from highway chassis dynamome-ter testing. Id. Yet, having “determined that this additional testing will not be burdensome for the manufacturers,” NHTSA required highway tests as a result of including MDPVs in its “automobile” definition. Id.
In sum, NHTSA’s decision not to set average fuel economy standards for all vehicles between 8,500 and 10,000 lbs. GVWR is arbitrary and capricious. That Class 2b trucks have never been regulated by NHTSA is not a reason for not regulating them now. We remand to NHTSA to revisit this issue and promulgate average fuel economy standards for these vehicles, or to provide a validly reasoned basis for continuing to exclude them from the regulation.
B. National Environmental Policy Act
1. The EPCA does not limit NHTSA’s NEPA obligations
NHTSA argues both that it has broad discretion to balance the factors of 49 U.S.C. § 32902(f) in setting fuel economy standards and that the EPCA constrains it from considering more stringent alternatives in the EA. NHTSA can’t have it both ways. Its hands are not tied, as demonstrated by its discretionary, substantive decisions to, among other things, value the benefit of carbon emissions reduction at zero, 71 Fed.Reg. at 17,638, peg its Unreformed CAFE standard to the least capable manufacturer with a substantial share of the market, id. at 17,568, *1213apply technologies only until marginal cost equals marginal benefit,67 id. at 17,589, 17,597, reject weight reduction as a cost-effective technology for vehicles between 4,000 and 5,000 lbs. curb weight, id. at 17,627, and not adopt a backstop, id. at 17,593.
NHTSA relies on Department of Transportation v. Public Citizen, 541 U.S. 752, 124 S.Ct. 2204, 159 L.Ed.2d 60 (2004), for its contention that it did not have to consider the effect of its rule on climate change. Public Citizen is inapposite.
In Public Citizen, the Supreme Court held that the Federal Motor Carrier Safety Administration (FMCSA) did not need to consider the environmental effects of cross-border operations of motor carriers in its EA, since it had no ability to prevent those operations. 541 U.S. at 770, 124 S.Ct. 2204. The Court reasoned, “where an agency has no ability to prevent a certain effect due to its limited statutory authority over the relevant actions, the agency cannot be considered a legally relevant ‘cause’ of the effect.” Id. The “critical feature” of the case was the fact that “FMCSA has no ability to countermand the President’s lifting of the moratorium or otherwise categorically to exclude Mexican motor carriers from operating within the United States.” Id. at 766, 124 S.Ct. 2204. “FMCSA has only limited discretion regarding motor vehicle carrier registration: It must grant registration to all domestic or foreign motor carriers that are ‘willing and able to comply with’ the applicable safety, fitness, and financial-responsibility requirements. FMCSA has no statutory authority to impose or enforce emissions controls or to establish environmental requirements unrelated to motor earner safety.” Id. at 758-59, 124 S.Ct. 2204 (emphasis added) (citation omitted). Moreover, neither of the purposes of NEPA — informing the public and ensuring agency consideration of the environmental impacts of its actions — would be served. Knowledge of the environmental impacts would not affect FMCSA decisionmaking since FMCSA “simply lacks the power to act on whatever information might be contained in the EIS.” Id. at 768, 124 S.Ct. 2204.
Here, in contrast, NHTSA clearly has statutory authority to impose or enforce fuel economy standards, 49 U.S.C. § 32902(a), (c), and it could have, in exercising its discretion, set higher standards if an EIS contained evidence that so warranted. See also Sierra Club v. Mainella, 459 F.Supp.2d 76, 105 (D.D.C.2006) (“The holding in Public Citizen extends only to those situations where an agency has ‘no ability’ because of lack of ‘statutory authority’ to address the impact. [National Park Service], in contrast, is only constrained by its own regulation from considering impacts on the Preserve from adjacent surface activities.” (emphasis in original)). Although NEPA does not demand substantive environmental outcomes, Public Citizen, 541 U.S. at 756, 124 S.Ct. 2204, NHTSA possesses the power to act on whatever information might be contained in an EIS. This court has recognized that “NEPA’s legislative history reflects Congress’s concern that agencies might attempt to avoid any compliance with NEPA by narrowly construing other statutory directives to create a conflict with NEPA. Section 102(2) of NEPA therefore requires government agencies to comply ‘to the fullest extent possible.’ ” Forelaws on Board v. Johnson, 743 F.2d 677, 683 (9th Cir.1985); id. at 685 (holding that the Bonneville Power Administration *1214interpreted its statutory authority too narrowly and “[g]iven BPA’s statutory duty both to conserve energy use and to preserve fish and wildlife, and the multitude of alternative proposals suggested by government agencies and citizen groups, the failure to prepare an EIS demonstrating that the agency has considered all significant alternatives violates both NEPA and the APA.” (citations omitted)).
Moreover, the CAFE standard will affect the level of the nation’s greenhouse gas emissions and impact global warming. See Ctr. for Auto Safety, 793 F.2d at 1334-35 (“If setting a higher standard cannot result in vehicles with increased fuel efficiency, then the entire regulatory scheme is pointless.”); NAS Report at 3 (“The CAFE program has clearly contributed to increased fuel economy of the nation’s light-duty vehicle fleet during the past 22 years.”). NHTSA does not dispute that light trucks account for a significant percentage of the U.S. transportation sector, that the U.S. transportation sector accounts for about six percent of the world’s greenhouse gases, and that “fuel economy improvements could have a significant impact on the rate of C02 accumulation in the atmosphere,” which would affect climate change. NAS Report at 14; cf. Massachusetts, 127 S.Ct. at 1457 (“EPA does not dispute the existence of a causal connection between man-made greenhouse gas emissions and global warming. At a minimum, therefore, EPA’s refusal to regulate such emissions ‘contributes’ to Massachusetts’ injuries.”).68
In sum, the EPCA does not limit NHTSA’s duty under NEPA to assess the environmental impacts, including the impact on climate change, of its rule. EPCA’s goal of energy conservation and NEPA’s goals of “helping] public officials make decisions that are based on understanding of environmental consequences, and take actions that protect, restore, and enhance the environment,” 40 C.F.R. § 1500.1(c), and “insur[ing] that environmental information is available to public officials and citizens before decisions are made and before actions are taken,” id. § 1500.1(b), are complementary. NEPA prohibits uninformed agency action. Robertson v. Methow Valley Citizens Council, 490 U.S. 332, 351, 109 S.Ct. 1835, 104 L.Ed.2d 351 (1989). “The procedures included in § 102 of NEPA are not ends in *1215themselves. They are intended to be ‘action forcing.’ The unequivocal intent of NEPA is to require agencies to consider and give effect to the environmental goals set forth in the Act, not just to file detailed impact studies which will fill governmental archives.” Envtl. Def. Fund, Inc. v. Corps of Eng’rs of the U.S. Army, 470 F.2d 289, 298 (8th Cir.1972) (citation omitted).
2. Sufficiency of the Environmental Assessment
We examine the EA with two purposes in mind: to determine whether it has adequately considered and elaborated the possible consequences of the proposed agency action when concluding that it will have no significant impact on the environment, and whether its determination that no EIS is required is a reasonable conclusion.
Even though an EA need not “conform to all the requirements of an EIS,” it must be “sufficient to establish the reasonableness of th[e] decision” not to prepare an EIS. Found. for N. Am. Wild Sheep, 681 F.2d at 1178 n. 29 (1982);69 see also 40 C.F.R. § 1508.9(a)(1). An EA “[s]hall include brief discussions of the need for the proposal ... [and] the environmental impacts of the proposed action and alternatives.” 40 C.F.R. § 1508.9(b). An EA “must in some circumstances include an analysis of the cumulative impacts of a project.... An EA may be deficient if it fails to include a cumulative impact analysis .... ” Native Ecosystems Council v. Dombeck, 304 F.3d 886, 895 (9th Cir.2002); see also Klamath-Siskiyou Wildlands Ctr. v. Bureau of Land Mgmt., 387 F.3d 989, 993-94 (9th Cir.2004); Kern v. U.S. Bureau of Land Mgmt., 284 F.3d 1062, 1076-78 (9th Cir.2002).
a. Cumulative impacts of greenhouse gas emissions on climate change and the environment
A cumulative impact is defined as “the impact on the environment which results from the incremental impact of the action when added to other past, present, and reasonably foreseeable future actions regardless of what agency ... or person undertakes such other actions. Cumulative impacts can result from individually minor but collectively significant actions taking place over a period of time.” 40 C.F.R. § 1508.7. In Klamath-Siskiyou Wildlands Center, this court held that:
Cumulative impacts of multiple projects can be significant in different ways. The most obvious way is that the greater total magnitude of the environmental effects ... may demonstrate by itself that the environmental impact will be significant. Sometimes the total impact from a set of actions may be greater than the sum of the parts.
387 F.3d at 994.
The EA catalogues the total tonnage of C02 emissions for light trucks for MYs 2005-2011. Final EA at 35-39. Table 4-5 of the Final EA lists the amount of fuel *1216consumption and emissions of criteria pollutants and C02 emissions. Id. at 35. For example, it shows that under Unreformed CAFE, the lifetime C02 emissions for light trucks MY 2005-2011 would be 4,979 million metric tons (mmt). Id. Under Reformed CAFE, including MDPYs in MY 2011, C02 emissions would be 4,966 million metric tons. Id. NHTSA estimated that:
together with the previous action raising MY 2005-07 light truck CAFE standards, the various alternatives for the current action will reduce lifetime carbon dioxide (C02) emissions from MY 2005-11 light trucks by 122 to 196 million metric tons, or by 2.4 to 3.8 percent from their level if neither action had been taken. ... MY 2008-11 light truck CAFE standards are projected to result in cumulative reductions from the previous and current actions ranging from 0.2 to 0.3 percent of U.S. greenhouse gas emissions over the lifetimes of MY 2005-11 light trucks.
Id. at 36-37 (emphasis added).
We conclude that the EA’s cumulative impacts analysis is inadequate. While the EA quantifies the expected amount of C02 emitted from light trucks MYs 2005-2011, it does not evaluate the “incremental impact” that these emissions will have on climate change or on the environment more generally in light of other past, present, and reasonably foreseeable actions such as other light truck and passenger automobile CAFE standards.70 The EA does not discuss the actual environmental effects resulting from those emissions or place those emissions in context of other CAFE rulemakings. This is a similar deficiency as that found in the Bureau of Land Management’s EA in Klamathr-Siskiyou Wildlands Center, where this court held that the BLM’s cumulative impacts analysis was inadequate because “[a] calculation of the total number of acres to be harvested in the watershed is a necessary component of a cumulative effects analysis, but it is not a sufficient description of the actual environmental effects that can be expected from logging those acres” and “stating the total miles of roads to be constructed is similar to merely stating the sum of the acres to be harvested — it is not a description of the actual environmental effects.” 387 F.3d at 995.
NHTSA does not dispute that the CAFE standard will have an effect on global warming due to an increase in greenhouse gas emissions. The new rule will not actually result in a decrease in carbon emissions, but potentially only a decrease in the rate of growth of carbon emissions. NHTSA concedes that “the new CAFE standards will not entirely offset the projected effect of increases in the number of light trucks.” NHTSA Br. at 92. However, NHTSA contends that Congress is “the cause of that shortfall,” not the agency, since it “is Congress’s decision in EPCA to require that CAFE standards be technologically feasible and economically practicable.” Id. NHTSA concludes from this that it has no obligation to assess the cumulative impact of its rule on climate change.
This argument is without merit for the reasons already discussed. NHTSA has the power to change the CAFE standards based on information contained in an EIS. We agree with Petitioners that “[b]y allowing particular fuel economy levels, which *1217NHTSA argues translate directly into particular tailpipe emissions, NHTSA’s regulations are the proximate cause of those emissions just as EPA Clean Air Act rules permitting particular smokestack emissions are the proximate cause of those air pollutants and are unquestionably subject to NEPA’s cumulative impacts requirements.” NEPA Gray Br. at 22; cf. Massachusetts, 127 S.Ct. at 1455-58. Thus, the fact that “climate change is largely a global phenomenon that includes actions that are outside of [the agency’s] control ... does not release the agency from the duty of assessing the effects of its actions on global warming within the context of other actions that also afreet global warming.” States’ Gray Br. at 15 (emphasis added). The cumulative impacts regulation specifically provides that the agency must assess the “impact of the action when added to other past, present, and reasonably foreseeable future actions regardless of what agency (Federal or non-Federal) or person undertakes such other actions.” 40 C.F.R. § 1508.7; see also Res. Ltd., Inc. v. Robertson, 35 F.3d 1300, 1306 (9th Cir.1994) (“The Forest Service says that cumulative impacts from non-Federal actions need not be analyzed because the Federal government cannot control them. That interpretation is inconsistent with 40 C.F.R. § 1508.7, which specifically requires such analysis.”).
The impact of greenhouse gas emissions on climate change is precisely the kind of cumulative impacts analysis that NEPA requires agencies to conduct. Any given rule setting a CAFE standard might have an “individually minor” effect on the environment, but these rules are “collectively significant actions taking place over a period of time.” 40 C.F.R. § 1508.7; see also Native Ecosystems Council, 304 F.3d at 897 (holding that the Forest Service’s road density standard amendments must be subject to cumulative impacts analysis because otherwise, “the Forest Service will be free to amend road density standards throughout the forest piecemeal, without ever having to evaluate the amendments’ cumulative environmental impacts.”); City of Los Angeles v. NHTSA 912 F.2d 478, 501 (D.C.Cir.1990) (Wald, C.J., dissenting) (“[W]e cannot afford to ignore even modest contributions to global warming. If global warming is the result of the cumulative contributions of myriad sources, any one modest in itself, is there not a danger of losing the forest by closing our eyes to the felling of the individual trees?”), overruled on other grounds by Fla. Audubon Soc. v. Bentsen, 94 F.3d 658 (D.C.Cir.1996). Thus, NHTSA must provide the necessary contextual information about the cumulative and incremental environmental impacts of the Final Rule in light of other CAFE rulemakings and other past, present, and reasonably foreseeable future actions, regardless of what agency or person undertakes such other actions.
b. Reasonable alternatives
NHTSA must “Vigorously explore and objectively evaluate all reasonable alternatives.” 40 C.F.R. § 1502.14(a). The alternatives section is the “heart” of an EIS. Id. § 1502.14; Ilio'laokalani Coalition v. Rumsfeld, 464 F.3d 1083, 1095 (9th Cir.2006); NRDC v. U.S. Forest Serv., 421 F.3d 797, 813 (9th Cir.2005). Although “an agency’s obligation to consider alternatives under an EA is a lesser one than under an EIS,” Native Ecosystems Council v. U.S. Forest Serv., 428 F.3d 1233, 1246, 1245 (9th Cir.2005), “NEPA requires that alternatives ... be given full and meaningful consideration,” whether the agency prepares an EA or an EIS, id. at 1245 (alteration in original; internal quotation marks omitted). The agency must “provide sufficient evidence and analysis for determining whether to prepare an environmental impact statement or a find*1218ing of no significant impact.” 40 C.F.R. § 1508.9.
In the EA, NHTSA considered a very narrow range of alternatives. All the alternatives evaluated were derived from NHTSA’s cost-benefit analysis. 71 Fed. Reg. at 17,672 n. 265 (“The evaluated alternatives represent standards set under the traditional Unreformed CAFE process and under the marginal cost-benefit analysis previously described.”). The alternatives presented in the EA are as follows:
• Alternative A (“Baseline”): MY 2007 standard of 22.2 mpg for MYs 2008-2011.
• Alternative B: Unreformed CAFE for MYs 2008-201071 and Reformed CAFE for MY 2011 with fuel economy targets set using continuous function. MPDVs included in MY 2011 only.
• Alternative C: Reformed CAFE for MYs 2008-2011 with fuel economy targets set using continuous function. Impacts are shown including MDPVs for MY 2011 only and excluding MDPVs altogether.
• Alternative D: Reformed CAFE as described in NPRM, with fuel economy targets set using step function (six footprint categories). Entirely excludes MDPVs.
• Alternative E: Reformed CAFE described in NPRM, with revised fuel economy targets for each of its six footprint categories. Impacts are shown including MDPVs for MY 2011 only and excluding MDPVs altogether.
Final EA at vii, 8-15. The alternative NHTSA ultimately chose is a mix between Alternatives B and C: NHTSA adopted Reformed CAFE beginning in MY 2011 and Unreformed CAFE for MYs 2008-2010, but it allowed a transition period to Reformed CAFE in MY 2011, so manufacturers may choose to continue to follow Unreformed CAFE in MYs 2008-2010. The adopted alternative includes MDPVs for MY 2011 only.
These alternatives are hardly different from the option that NHTSA ultimately adopted. The EA states that “Alternatives C and E are projected to result in the largest reductions in fuel consumption, energy use, and environmental effects among these alternatives,” id. at vii, but Alternative C is close to what NHTSA adopted (the only difference being no transition period), and Alternative E barely results in less CO 2 emissions than Alternative C. See id. at x (change from baseline for Alternative C and Alternative E with MDPVs in MY 2011 are the same (-73 mmt), and change in baseline without MDPVs is -70 mmt for Alternative C and -71 mmt for Alternative E). The entire range of alternatives considered in the EA ranged from “22.2 to 22.7 mpg for MY 2008, 22.2 to 23.3 mpg for MY 2009, and 22.2 to 23.6 mpg for MY 2010.” Id. at 16. The estimated lifetime fuel and energy use by MY 2008-2011 light trucks under the alternatives ranged from a 1.8 to 2.6 percent decrease from “baseline,” id. at 27-28, and the estimated lifetime emissions of C02 ranged from 2,767 to 2,840 mmt, id. at 29, which is extremely small compared to the overall volume of emissions.72
NHTSA acknowledged that “the range of impacts from the considered alternatives is very narrow and minimal.” 71 Fed.Reg. at 17,672. However, the agency justified its choice of range and refusal to consider other alternatives on the ground *1219that “standards more stringent than those represented by the alternatives would not satisfy the statutory requirement to establish standards ... that are both technologically feasible and economically practicable.... NEPA’s requirements must be applied in light of the constraints placed on the agency by EPCA.” Final EA at viii. Once again, NHTSA falls back on its contention that it had no discretion to consider setting higher CAFE standards. As before, we conclude that this argument is flawed.
NHTSA also erroneously contends that Petitioners have not identified any specific alternative the agency should have considered. To the contrary, Environmental Defense submitted a detailed appendix to its comment titled, “Revised Benefit-Cost Analysis for Calculating Optimal CAFE Targets.” See Environmental Defense Cmt.App. A.73 In this document, Environmental Defense performed a marginal cost-benefit analysis, using a variety of different assumptions and inputs. Table A-l set forth 28 different possible CAFE standards for MY 2011 (including NHTSA’s figure). On the basis of its calculations, it recommended a final rule that would increase CAFE standards at a rate of 4% per year and achieve a standard of 26 mpg by MY 2011. Environmental Defense Cmt., Letter from Kevin Mills, Director, Clean Car Program, Environmental Defense, to Jacqueline Glassman, Acting Administrator, NHTSA (Nov. 22, 2005).
We also disagree with NHTSA that Petitioners’ suggested alternatives would not be reasonably related to the project’s purpose. The purpose of the Final Rule is to set CAFE standards for light trucks for MYs 2008-2011. 71 Fed.Reg. at 17,566. NHTSA itself describes the scope of the EA as “analyz[ing] the environmental impacts associated with various alternatives to the existing CAFE program.” Final EA at 4. Since EPCA’s overarching goal is energy conservation, consideration of more stringent fuel economy standards that would conserve more energy is clearly reasonably related to the purpose of the CAFE standards. Energy conservation and environmental protection are not coextensive, but they often overlap. The Supreme Court has recently recognized as much. See Massachusetts, 127 S.Ct. at 1462 (“[T]hat DOT sets mileage standards in no way licenses EPA to shirk its environmental responsibilities. EPA has been charged with protecting the public's ‘health’ and ‘welfare,’ 42 U.S.C. § 7521(a)(1), a statutory obligation wholly independent of DOT’S mandate to promote energy efficiency. The two obligations may overlap, but there is no reason to think the two agencies cannot both administer their obligations and yet avoid inconsistency.” (citation omitted)).
3. NHTSA must prepare either a revised Environmental Assessment or, as necessary, an Environmental Impact Statement
An agency must prepare an EIS “if ‘substantial questions are raised as to whether a project ... may cause significant degradation of some human environmental factor.’ ” Idaho Sporting Cong. v. Thomas, 137 F.3d 1146, 1149 (9th Cir.1998). Petitioners “need not show that significant effects will in fact occur," but only that there are “substantial questions whether a project may have a significant effect.” Id. at 1150 (internal quotation marks omitted); see also Blue Mountains Biodiversity Project v. Blackwood, 161 F.3d 1208, 1212 (9th Cir.1998); Nat’l *1220Parks & Conservation Ass’n v. Babbitt, 241 F.3d 722, 730 (9th Cir.2001). “If an agency decides not to prepare an EIS, it must supply a ‘convincing statement of reasons’ to explain why a project’s impacts are insignificant. ‘The statement of reasons is crucial to determining whether the agency took a “hard look” at the potential environmental impact of a project.’ ” Blue Mountains Biodiversity Project, 161 F.3d at 1212 (quoting Save the Yaak Comm. v. Block, 840 F.2d 714, 717 (9th Cir.1988)); see also Nat’l Parks & Conservation Ass’n, 241 F.3d at 730. NHTSA’s EA is markedly deficient in its attempt to justify the refusal to prepare a complete EIS. As explained below, the agency’s FONSI is based primarily on its conclusory assertion — contradicted by evidence in the record — that the Final Rule will have no significant environmental impact because it authorizes CAFE standards that will result in a very small decrease in carbon dioxide emissions.
“Whether there may be a significant effect on the environment requires consideration of two broad factors: ‘context and intensity.’” Nat’l Parks & Conservation Ass’n, 241 F.3d at 731 (quoting 40 C.F.R. § 1508.27). A number of factors should be considered in evaluating intensity, including, “[t]he degree to which the proposed action affects public health or safety,” “[t]he degree to which the effects on the quality of the human environment are likely to be highly controversial,” “[t]he degree to which the possible effects on the human environment are highly uncertain or involve unique or unknown risks,” “[t]he degree to which the action may establish a precedent for future actions with significant effects or represents a decision in principle about a future consideration,” “[wjhether the action is related to other actions with individually insignificant but cumulatively significant impacts,” and “[tjhe degree to which the action may adversely affect an endangered or threatened species or its habitat.” 40 C.F.R. § 1508.27(b)(2), (4), (5), (6), (7), (9). An action may be “significant” if one of these factors is met. Ocean Advocates v. U.S. Army Corps of Eng’rs, 361 F.3d 1108, 1125 (9th Cir.2004); see also Nat’l Parks & Conservation Ass’n, 241 F.3d at 731 (either degree of uncertainty or controversy “may be sufficient to require preparation of an EIS in appropriate circumstances.”).
NHTSA’s finding of no significant impact (FONSI) stated that the agency determined that its Final Rule “will not have a significant effect on the human environment. This finding of no significant impact is based on the attached Final Environmental Assessment (EA).... ” Finding of No Significant Environmental Impact for Model Year 2008-2011 Light Truck Fuel Economy Standards, NHTSA Docket No. 2006-24309-3 (Mar. 28, 2006). In the Final EA, NHTSA explained that compared to the “baseline” alternative of extending the MY 2007 light truck CAFE standard through MYs 2008-2011, its evaluated alternatives would have a minor beneficial impact on various environmental resources. Final EA at 26-33, 39-42. NHTSA concluded that “the final rule would produce, compared to U.S. emissions of C02, a small decrease in emissions of C02, the primary component of greenhouse gas emissions, under the selected alternative. Accordingly, the agency determined that the action we are adopting today will not have a significant impact on the environment.” 71 Fed.Reg. at 17,673 (citing Final EA at 32).
Petitioners argue that the evidence raises a substantial question as to whether the Final Rule may have a significant impact on the environment and that NHTSA failed to provide a convincing statement of reasons for why a small decrease (rather than a larger decrease) in the growth of CO 2 emissions would not have a signifi*1221cant impact on the environment. Petitioners note that NHTSA has never evaluated the impacts of carbon emissions from light trucks or other vehicles, much less the effect of any reduction or increase in those emissions on climate change. Petitioners presented evidence that continued increase in greenhouse gas emissions may change the climate in a sudden and non-linear way. Without some analysis, it would be “impossible for NHTSA to know ... whether a change in GHG emissions of 0.2% or 1% or 5% or 10% ... will be a significant step toward averting the ‘tipping point’ ” and irreversible adverse climate change. States’ Gray Br. at 6.
NHTSA argues that its “conclusion that a 0.2 percent decrease in carbon dioxide emissions will not have a significant impact upon the environment is self-evidently reasonable and consistent” with City of Los Angeles v. NHTSA, 912 F.2d 478 (D.C.Cir.1990), and Public Citizen v. NHTSA, 848 F.2d 256 (D.C.Cir.1988). NHTSA Br. at 111. NHTSA also argues that the impact of the rule on global warming is too speculative to warrant NEPA analysis.
Petitioners have raised a “substantial question” as to whether the CAFE standards for light trucks MYs 2008-2011 “may cause significant degradation of some human environmental factor,” Idaho Sporting Cong., 137 F.3d at 1149 (internal quotation marks omitted) particularly in light of the compelling scientific evidence concerning “positive feedback mechanisms” in the atmosphere.74 Among the evidence Petitioners presented to the agency was the following:
• The IPCC Third Assessment Report, which discusses the history of anthropogenic interference with the climate system, the projected increase in climate variability and extreme weather events, and the projected effects on various ecological systems. See IPCC Third Assessment Report at 2-33. The IPCC found:
Changes in climate could increase the risk of abrupt and non-linear changes in many ecosystems, which would affect their function, biodiversity, and productivity. The greater the magnitude and rate of the change, the greater the risk of adverse impacts. For example:
Changes in disturbance regimes and shifts in the location of suitable climatically defined habitats may lead to abrupt breakdown of terrestrial and marine ecosystems with significant changes in composition and function and increased risk of extinctions.
Sustained increases in water temperatures of as little as 1°C, alone or in combination with any of several stresses ... can lead to corals ejecting their algae (coral bleaching) and the eventual death of some corals.... Inertia is a widespread inherent characteristic of the interacting climate, ecological, and socioeconomic systems. Thus some impacts of anthropogenic climate change may be slow to become apparent, and some could be irreversible if climate change is not limited in both rate and magnitude before associated thresholds, whose positions may be poorly known, are crossed. IPCC Third Assessment Report at 16 (emphasis added); see also id. at 15 (Table SPM-2 shows “[ejxamples of climate variability and extreme climate events and examples of their impacts.”).
• The IPCC Working Group I Technical Summary provided: “The possibility *1222for rapid and irreversible changes in the climate system exists, but there is a large degree of uncertainty about the mechanisms involved and hence also about the likelihood or time-scales of such transitions. The climate system involves many processes and feed-backs that interact in complex nonlinear ways. This interaction can give rise to thresholds in the climate system that can be crossed if the system is 'perturbed sufficiently.” Technical Summary of IPCC Working Group I Report at 53 (emphasis added); see also id. at 46-53 (discussion of positive feedback mechanism).
• “The American Meterological Society, the American Geophysical Union, and the American Association for Advancement of Science, among many, many other scientific organizations have all concluded that the evidence of human induced warming is compelling.... In an April 2004 article, leading NASA and Department of Energy scientists stated that emissions of carbon dioxide and other heat-trapping gases have warmed the oceans and led to an energy imbalance that is causing and will continue to cause, significant warming, increasing the urgency of reducing C02 emissions.” States’ Cmt. at 9 (citing essay that reviewed 928 peer-reviewed scientific papers).
• The Climate Change Futures Report published by the Center for Health and the Global Environment at Harvard Medical School, which analyzed in detail climate change scenarios that “will affect the health of humans as well as the ecosystems and species on which we depend.” Climate Change Futures Report at 5; see generally id. at 32-90 (case studies involving infectious and respiratory diseases, extreme weather events, and natural and managed systems).
Finally, Petitioners have satisfied several of the “intensity” factors listed in 40 C.F.R. § 1508.27(b) for determining “significant effect.” For example, the Final Rule clearly may have an “individually insignificant but cumulatively significant” impact with respect to global warming. Evidence that Petitioners submitted in the record also shows that global warming will have an effect on public health and safety. Climate Change Futures Report at 6-90. Petitioners do not claim (nor do they have to show) that NHTSA’s Final Rule would be the sole cause of global warming, and that is NHTSA’s only response on this point.
Petitioners have also satisfied the “controversy” factor. See 40 C.F.R. § 1508.27(b)(4); see Blue Mountains Biodiversity Project, 161 F.3d at 1212 (“ ‘controversial’ is ‘a substantial dispute [about] the size, nature, or effect of the major Federal action rather than the existence of opposition to a use.’” (alteration in original)). NHTSA received over 45,000 individual submissions on its proposal. See 71 Fed.Reg. at 17,577; see also Nat’l Parks & Conservation Ass’n, 241 F.3d at 736 (four-hundred and fifty comments, 85% of which opposed the agency’s preferred alternative was “more than sufficient to meet the ‘outpouring of public protest’ discussed in [pri- or case law]. More important, to the extent the comments urged that the EA’s analysis was incomplete, ... they cast substantial doubt on the adequacy of the Parks Service’s methodology and data.” (footnote and citation omitted)). We reject NHTSA’s argument that “petitioners’ controversy does not concern the ‘size, nature, or effect’ of the new CAFE standards, but rather the desire of some commenters for different regulations that they have not described in any detail.” NHTSA Br. at 117. The entire dispute between Petitioners and NHTSA centers on the stringency *1223of the MY 2008-2011 light truck CAFE standards — their “size” or “effect.”75
In light of the evidence in the record, it is hardly “self-evident” that a 0.2 percent decrease in carbon emissions (as opposed to a greater decrease) is not significant. NHTSA’s conclusion that a small reduction (0.2% compared to baseline) in the growth of carbon emissions would not have a significant impact on the environment was unaccompanied by any analysis or supporting data, either in the Final Rule or the EA. See, e.g., 71 Fed.Reg. at 17,673; Final EA at 32.
Nowhere does the EA provide a “statement of reasons” for a finding of no significant impact, much less a “convincing statement of reasons.” For example, the EA discusses the amount of C02 emissions expected from the Rule, but does not discuss the potential impact of such emissions on climate change. In the “Affected Environment” section of the EA, NHTSA states that “[i]nereasing concentrations of greenhouse gases are likely to accelerate the rate of climate change.” Final EA at 22. The agency notes that “[t]he transportation sector is a significant source of greenhouse gas (GHG) emissions, accounting for approximately 28 percent of all greenhouse gas emissions in the United States.” Id. From this, NHTSA jumps to the conclusion that “[c]oupled with the effects resulting from the 2003 light truck rule, the effects resulting from the agency’s current action are expected to lessen the GHG impacts discussed above.” Id.
Table 3-2 of the EA, which shows the potential health effects of criteria air pollutants, is similarly devoid of meaningful analysis or a statement of reasons why the effects would be insignificant. The potential health effect for C02 is described: “Increase in greenhouse gases can lead to climate change. Hot temperatures can lead to cardiovascular problems, heat exhaustion, and some respiratory problems. There may be an increased risk of infectious diseases due to increased temperatures. Heat can also increase the concentration of ground-level ozone.” Id. at 20.
Nor is there any analysis or statement of reasons in the section of the EA that discusses environmental impacts. The EA states that reduction in fuel production and consumption would reduce “contamination of water resources,” acid rain, risk of oil spills and contamination, and “lead to minor reductions in impacts to biological resources ... including] habitat encroachment and destruction, air and water pollution, greenhouse gases, and oil contamination from petroleum refining and distribution.” Id. at 32-33; see also id. at 39 (Table 4-7 compares the impacts under the baseline CAFE standard of 22.2 mpg and the analyzed alternatives. It cursorily summarizes the impacts as “slower rate of growth in fuel consumption for light trucks,” “reduction of GHG emissions,” and “minor benefit [to water and biological resources] from reductions in energy consumption, GHG emissions and extremely small changes in criteria pollutant emissions.”); id. at 32-33 (citing no supporting data for its conclusions regarding impacts on water and biological resources).
NHTSA’s EA “shunted aside[significant questions] with merely conclusory statements,” failed to “directly address[ ]” “substantial questions,” and most importantly, “provide[d] no foundation” for the important inference NHTSA draws between a decrease in the rate of carbon emissions growth and its finding of no significant impact. Found, for N. Am. Wild Sheep, 681 F.2d at 1179. NHTSA makes “vague *1224and conclusory statements” unaccompanied by “supporting data,” and the EA “do[es] not constitute a ‘hard look’ at the environmental consequences of the action as required by NEPA.” Great Basin Mine Watch v. Hankins, 456 F.3d 955, 973 (9th Cir.2006). Thus, the FONSI is arbitrary and capricious. See Klamath-Siskiyou Wildlands Center, 387 F.3d at 994 (“[T]he problem with the entire table is that it does not provide any objective quantification of the impacts. Instead, the reader is informed only that a particular environmental factor will be ‘unchanged,’ ‘improved,’ or ‘degraded’ and whether that change will be ‘minor’ or ‘major.’ The reader is not told what data the conclusion was based on, or why objective data cannot be provided.”).
The only reason NHTSA provided for why the environmental impact of the Final Rule would be insignificant is that it results in a decreased rate of growth of GHG emissions compared to the light truck CAFE standard for MY 2007. But simply because the Final Rule may be an improvement over the MY 2007 CAFE standard does not necessarily mean that it will not have a “significant effect” on the environment.
NHTSA has not explained why its rule will not have a significant effect.76
In light of the emergent consensus on global warming, Chief Judge Wald’s reasoning in her dissent in City of Los Ange-les is not only prescient but persuasive:
While NHTSA did the calculations necessary to determine how much extra carbon dioxide would be emitted, it failed completely to discuss in any detail the global warming phenomenon itself, or to explain the benchmark for its determination of insignificance in relation to that environmental danger. Had the emissions been slightly over one percent, would that have been significant? Without some articulated criteria for significance in terms of contribution to global warming that is grounded in the *1225record and available scientific evidence, NHTSA’s bald conclusion that the mere magnitude of the percentage increase is enough to alleviate its burden of conducting a more thorough investigation cannot carry the day.
912 F.2d at 500.
Petitioners have raised a substantial question of whether the Final Rule may significantly affect the environment. NHTSA acknowledges that carbon emissions contribute to global warming, and it does not dispute the scientific evidence that Petitioners presented concerning the significant effect of incremental increases in greenhouses gases. Instead of providing the required “convincing statement of reasons,” Blue Mountains Biodiversity Project, 161 F.3d at 1211 (internal quotation marks omitted), NHTSA simply asserts that the insignificance of the effects is “self-evident[.]”
Finally, we must decide the appropriate remedy given NHTSA’s inadequate EA. We have previously recognized that preparation of an EIS is not mandated in all cases simply because an agency has prepared a deficient EA or otherwise failed to comply with NEPA. If, for example, an EA is so procedurally flawed that we cannot determine whether the proposed rule or project may have a significant effect, the court should remand for the preparation of a new EA. See Metcalf v. Daley, 214 F.3d 1135, 1146 (9th Cir.2000) (ordering remand for preparation of a new EA, where prior EA was prepared after the agency had already rendered a substantive decision on the permitting action). If an agency completely fails to prepare an EA before deciding that a proposed project or rule will have no significant environmental impact, remand for preparation of an EA is likewise the proper remedy. See Jones v. Gordon, 792 F.2d 821, 828-29 (9th Cir.1986) (remanding where agency failed to prepare any NEPA document before issuing permit). And where an agency determines that consideration of certain factors are legally irrelevant to the agency’s action, rendering it impossible for the reviewing court to determine the accuracy of the FONSI, we also remand for preparation of an EA on a complete record. San Luis Obispo Mothers for Peace v. Nuclear Regulatory Comm’n, 449 F.3d 1016, 1024, 1031, 1035 (9th Cir.2006) (ordering remand for reconsideration of EA where agency rejected consideration of terrorist acts as factor to be considered in its review of application for license to construct nuclear spent-fuel storage facility), cert, denied sub nom. Pac. Gas & Elec. Co. v. San Luis Obispo Mothers for Peace, — U.S.-, 127 S.Ct. 1124, 166 L.Ed.2d 891 (2007).
By contrast, if the court determines that the agency’s proffered reasons for its FONSI are arbitrary and capricious and the evidence in a complete administrative record demonstrates that the project or regulation may have a significant impact, then it is appropriate to remand with instructions to prepare an EIS. See, e.g., Nat’l Parks & Conservation Ass’n., 241 F.3d at 733-34; Idaho Sporting Cong., 137 F.3d at 1154.
The distinction — between cases where it is appropriate to order immediate preparation of an EIS and those where it is not — is implicit in this circuit’s NEPA jurisprudence, and has been explicitly recognized elsewhere. See O’Reilly v. U.S. Army Corps of Eng’rs, 477 F.3d 225, 238-39 (5th Cir.2007) (“It is also clear that a decision to forego preparation of an EIS may be unreasonable for at least two distinct reasons: (1) the evidence before the court demonstrates that, contrary to the FONSI, the project may have a significant impact on the human environment, or (2) the agency’s review was flawed in such a manner that it cannot yet be said whether the project may have a significant im*1226pact.... If the court finds that a project may have a significant impact, the court should order the agency to prepare an EIS. If the court finds, on the other hand, that the EA is inadequate in a manner that precludes making the determination whether the project may have a significant impact, the court should remand the case to the agency to correct the deficiencies in its analysis”) (citations omitted). So, if there is uncertainty over whether the proposed project may have a significant impact, including uncertainty caused by an incomplete administrative record or an inadequate EA, the court should ordinarily remand for the agency to either prepare a revised EA or reconsider whether an EIS is required. Metcalf, 214 F.3d at 1146 (“On reflection, and in consideration of our limited role in this process, we have decided that it is appropriate only to require a new EA, but to require that it be done under circumstances that ensure an objective evaluation free of the previous taint.”); see also Hill v. Boy, 144 F.3d 1446, 1451 (11th Cir.1998) (remanding where agency’s failure to prepare an EIS was based on incorrect assumption); Nat’l Audubon Soc’y v. Hoffman, 132 F.3d 7, 18-19 (2d Cir.1997) (remanding due to incomplete administrative record).
Whether to require an EIS now is a very close question. Petitioners’ evidence demonstrates, overwhelmingly, the environmental significance of C02 emissions and the effect of those emissions on global warming. How NHTSA can, on remand, prepare an EA that takes proper account of this evidence and still conclude that the 2006 Final Rule has no significant environmental impact is questionable. See 40 C.F.R. § 1508.13 (FONSI is a document “presenting the reasons why an action ... will not have a significant effect on the human environment and for which an [EIS] therefore will not be prepared” (emphasis added)). We nonetheless give the benefit of the doubt to NHTSA and decline to order the immediate preparation of an EIS for two reasons.
First, the EA’s primary deficiency lies with its conclusory assertion that a modest 0.2 percent decrease in carbon emissions renders the 2006 Final Rule environmentally insignificant. The EA provides no reasons or analysis in support of this conclusion, much less “convincing” reasons. Blue Mountains Biodiversity Project, 161 F.3d at 1212. Although Petitioners’ evidence is daunting we cannot, in the abstract, categorically decide that NHTSA’s reasons for this conclusion in a revised EA would be unconvincing.
Second, we have a significant reason to defer to the agency as to whether an EA will suffice or whether an EIS is necessary. During the pendency of this appeal, Congress enacted the Energy Independence and Security Act of 2007 (“EISA”). Pub.L. No. 110-140, 121 Stat. 1492. EISA requires NHTSA to increase fuel economy standards for passenger and non-passenger automobiles to reach a combined average of at least 35 mpg by model year 2020. See 49 U.S.C. § 32902(b)(2)(A). In connection with EISA, NHTSA has already begun preparation of a complete EIS to inform the “NEPA analysis relating to the CAFE standards for MY 2011-2015 automobiles,” including light trucks. Notice of Intent to Prepare an Environmental Impact Statement for New Corporate Average Fuel Economy Standards, 73 Fed.Reg. 16,615, 16,616 (Mar. 28, 2008) (to be codified at 49 C.F.R. pts. 531 & 533). With the exception of MY 2011, the EIS prepared as a result of the passage of EISA will not encompass the CAFE standards implemented by the 2006 Final Rule. However, information developed in preparation of the EISA-based EIS would undoubtedly inform NHTSA’s decision as to the necessity of an EIS in this case. See City of Los Angeles, 912 F.2d at 503 (Wald, C.J., *1227dissenting) (stating that proper course was to remand without enjoining agency action because agency was “already conducting a [separate] programmatic EIS which may yet provide an adequate explanation for its finding of no significant impact”).
Taken together, these reasons lead us to conclude that the record is insufficiently complete for us to order the immediate preparation of an EIS. See O’Reilly, 477 F.8d at 240-41 (remanding without instructions to prepare EIS); Metcalf, 214 F.3d at 1146 (same); Hill, 144 F.3d at 1451 (same); Natl Audubon Soc’y, 132 F.3d at 18-19. We therefore remand to NHTSA to prepare a revised EA or, as necessary, a complete EIS.
IV. CONCLUSION
NHTSA’s failure to monetize the value of carbon emissions in its determination of the MY 2008-2011 light truck CAFE standards, failure to set a backstop, failure to revise the passenger automobile/light truck classifications, and failure to set fuel economy standards for all vehicles in the 8,500 to 10,000 lb. GVWR class, was arbitrary and capricious and contrary to the EPCA. We therefore remand to NHTSA to promulgate new standards consistent with this opinion as expeditiously as possible and for the earliest model year practicable.
We also hold that the EA was inadequate and Petitioners have raised a substantial question as to whether the Final Action may have a significant impact on the environment. Thus, we remand to NHTSA for the preparation of a revised EA or, as necessary, a full EIS.
REVERSED AND REMANDED.
. Petitioners also argued in their opening briefs that the EPCA does not preempt California's Clean Air Act motor vehicle greenhouse gas emissions standards. They raised this argument in response to NHTSA’s assertion in the preamble of the Final Rule that the EPCA preempts state laws and regulations regarding fuel economy standards. See 71 Fed.Reg. at 17,654-70. We do not address this issue since the parties agreed in their response briefs and at oral argument that the preemption discussion in the preamble of the Final Rule is not final agency action and thus not currently reviewable.
. Natural Resources Defense Council-Comments, NHTSA Docket No. 2005-22223-1705 (Nov. 23, 2005).
. The Secretary of Transportation may prescribe regulations amending the standard for passenger automobiles "to a level that the Secretary decides is the maximum feasible average fuel economy level for that model year.” 49 U.S.C. § 32902(c) (2007).
. "Non-passenger automobiles. At least 18 months before the beginning of each model year, the Secretary of Transportation shall prescribe by regulation average fuel economy standards for automobiles (except passenger automobiles) manufactured by a manufacturer in that model year. Each standard shall be the maximum feasible average fuel economy level that the Secretary decides the manufacturers can achieve in that model year. The Secretary may prescribe separate standards for different classes of automobiles.” Id. § 32902(a).
. The Secretary of Transportation delegated authority to promulgate average fuel economy regulation to NHTSA. See 49 C.F.R. § 1.50(f) (2007).
. For example, the Hummer HI is more than 10,000 lbs. GVWR and thus not subject to CAFE regulation. UCS Cmt. at 33 n. 14 (Union of Concerned Scientists-Comments, NHTSA Docket No. 2005-22223-1978 (Nov. 25, 2005)).
. A "GVWR'' is "the value specified by the manufacturer as the loaded weight of a single vehicle.” 49 C.F.R. § 523.2 (2007).
. See, e.g., 49 C.F.R. § 523.5.
. 49 C.F.R. § 523.5(b) provides:
An automobile capable of off-highway operation is an automobile—
(l)(i) That has 4-wheel drive; or
(ii) Is rated at more than 6,000 pounds gross vehicle weight; and
*1184(2) That has at least four of the following characteristics [affecting off-road capability relating to approach angle, breakover angle, departure angle, running clearance, and front and rear axle clearances],
. U.S. DOT/NHTSA-Report, Effectiveness and Impact of Corporate Average Fuel Economy (CAFE) Standards, NHTSA Docket No. 2005-22223-14 (Aug. 31, 2005) (Committee on the Effectiveness and Impact of Corporate Average Fuel Economy (CAFE) Standards, National Research Council (2002)) (hereinafter “NAS Report”).
. The committee identified the environmental cost of carbon emissions as $50/tonne carbon (tC), or $0.12 of this $0.30/gal figure. NAS Report at 85.
. A "finding of no significant impact" is known as a "FONSI.”
. NHTSA proposed adopting the Environmental Protection Agency's (EPA) definition of MDPVs (used in setting emissions standards):
Medium-duty passenger vehicle (MDPV) means any heavy-duty vehicle (as defined in this subpart) with a gross vehicle weight rating (GVWR) of less than 10,000 pounds that is designed primarily for the transportation of persons. The MDPV definition does not include any vehicle which:
(1) Is an "incomplete truck" as defined in this subpart; or
(2) Has a seating capacity of more than 12 persons; or
(3) Is designed for more than 9 persons in seating rearward of the driver’s seat; or (4)Is equipped with an open cargo area (for example, a pickup truck box or bed) of 72.0 inches in interior length or more. A covered box not readily accessible from the passenger compartment will be considered an open cargo area for purposes of this definition.
40 C.F.R. § 86.1803-01. EPA defines "heavy-duty vehicle” as "any motor vehicle rated at more than 8,500 pounds GVWR or that has a vehicle curb weight of more than 6,000 pounds or that has a basic vehicle frontal area in excess of 45 square feet.” Id.
. NHTSA requires manufacturers to meet these average fuel economy standards on a fleet-wide basis.
. Environmental Defense-Comments, NHTSA Docket No. 2005-22223-1805 (Nov. 22, 2005).
. Public Citizen-Comments, NHTSA Docket No. 2005-22223-2188 (Dec. 13, 2005).
. American Council for an Energy-Efficient Economy-Comments, NHTSA Docket No. 2005-22223-1711 (Nov. 23, 2005).
. Center for Biological Diversity-Comments and Attachments A through E, NHTSA Docket No. 2005-22223-1638 (Nov. 22, 2005). Among other things, the Center for Biological Diversity argued, “An estimate of the true costs of the carbon emissions is one of the most important inputs into the NHTSA’s algorithm for determining the maximum feasible average fuel economy level. Estimates of the monetary benefits ... [are] readily available. ... Excluding a monetization of the greenhouse gas emissions from the NHTSA's light truck fuel economy rulemaking on the basis that the future costs of global warming are uncertain is arbitrary and capricious.... The NHTSA cannot dismiss these costs as • ‘uncertain’ while simultaneously relying upon the uncertain projections of claimed economic hardship recited by the automobile industry for an estimate of the cost of increasing fuel economy.” Id. at 3-4.
. Environmental Defense-Comments, NHTSA Docket No. 2005-22223-2249 (Mar. 13, 2006); Environmental Defense-Report-The Social Costs of Carbon Review: Methodological Approaches for Using SCC Estimates in Policy Assessment, NHTSA Docket No. 2005-22223-2251 (Mar. 13, 2006) (Paul Watkiss, et al., The Social Cost of Carbon (SCC) Review — Methodological Approaches for Using SCC Estimates in Policy Assessment, Final Report (Nov.2005)).
. Sierra Club, U.S. Public Interest Research Group, and National Environmental Trust-Comments, NHTSA Docket No. 2005-22223-1636 (Nov. 22, 2005).
. Natural Resources Defense Council-Appendices E-I, NHTSA Docket No. 2005-22223-1710 (Nov. 23, 2005).
. California Department of Justice-Comments, NHTSA Docket No. 2005-22223-1637 (Nov. 22, 2005) (Comments of the Attorneys General of the States of California, Massachusetts, New York, Connecticut, New Jersey, Maine, Oregon, Vermont and the Corporation Counsel for the City of New York).
. Environmental Defense noted: "From 1988 to 2003, the average wheelbase of the combined car and truck fleet grew at an aver*1189age rate of 0.3% per year. Continuing this trend would result in a 0.14 mpg drop for the MY2011 light truck fuel economy requirement ..., and if the trend continued through MY2030, a lost oil savings of 227,000 barrels per day year — or 30% of the oil savings that would be achieved from the proposed MY08-11 standards in that year.” Environmental Defense Cmt. at 13.
. Environmental Defense-Attachment 5-Pickup Truck Usage Study, NHTSA Docket No. 2005-22223-1703 (Nov. 23, 2005) (R.L. Polk & Co., Pickup Truck Usage Study).
. Natural Resources Defense Council-Appendix D, NHTSA Docket No. 2005-22223-1709 (Nov. 22, 2005).
. Joseph F. Desmond-Comments, NHTSA Docket No. 2005-22223-1557 (Nov. 21, 2005).
. See generally Attachment A to CBD Cmt. (Global Warming and Its Impacts); Attachment B to CBD Cmt. (Albritton, D.L., et al., Technical Summary, Climate Change 2001: The Scientific Basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental panel on Climate Change (IPCC) (2001)); Attachment F to CBD Cmt. (Epstein, P.R. and E. Mills (eds.), Climate Change Futures: Health, Ecological and Economic Dimensions (2005)); Attachment G to CBD Cmt. (Intergovernmental Panel on Climate Change (IPCC), Climate Change 2001: Synthesis Report (Summary for Policy*1190makers) (2001)); Attachment L to CBD Cmt. (Overpeck, J.T., et al., "Arctic System on Trajectory to New, Seasonally Ice-free State,” EOS (2005)); Attachment M to CBD Cmt. (Parmesan, C. and H. Galbraith, Pew Center on Global Climate Change, Observed Impacts of Global Climate Change in the U.S. (Sept. 2004)); Attachment O to CBD Cmt. (Thomas, C.D., et al., "Extinction Risk from Climate Change,” 427 Nature 145 (Jan. 8, 2004)); Attachment P to CBD Cmt. (World Health Organization, The World Health Report 2002 (2002)); Attachment Q to CBD Cmt. (Arctic Climate Impact Assessment, Impacts of a Warming Arctic: Highlights (2004)).
. "The atmospheric concentration of C02 has increased from 280 ppm in 1750 to 367 ppm in 1999.... Today’s C02 concentration has not been exceeded during the past 420,-000 years and likely not during the past 20 million years. The rate of increase over the past century is unprecedented, at least during the past 20,000 years.” Technical Summary of IPCC Working Group I Report at 39.
. The draft of the IPCC Fourth Assessment Report, "Climate Change 2007,” was published recently (containing reports of the Working Groups and Technical Summaries) (available at http://www.ipcc.ch). "The summary ... said that efforts to rein in the billions of tons of annual releases of carbon dioxide and other heat-trapping gases would have to begin soon to limit risks of large changes in the climate and their impact on humans and nature.” Andrew C. Revkin, Climate Panel Sees Need for New Steps on Emissions, N.Y. Times, Apr. 27, 2007, at A20.
. MY 2008: 22.5 mpg; MY 2009: 23.1 mpg; MY 2010: 23.5 mpg.
. The NPRM proposed a step function for Reformed CAFE, with six different footprint categories. The Final Rule establishes target fuel economy levels for each value of vehicle footprint, referred to as a “continuous function.” 71 Fed.Reg. at 17,587. A continuous function reduces the incentive to enlarge the footprints of light trucks in order to shift them into a higher bracket with a lower fuel economy standard. See id. at 17,609.
. See 71 Fed.Reg. at 17,608-09 (description of Reformed CAFE formula).
. This means adding technologies "in order of lower to higher costs.” 71 Fed.Reg. at 17,582; see also FRIA at VI-13.
. NHTSA recognized that "[c]arbon dioxide emissions account for more than 97% of total greenhouse gas emissions from the refining and use of transportation fuels.” FRIA at VIII-62 n. 83.
. Petitioners presented a study "based on real-world examples suggesting] that the cost per pound of weight reduced through the use of high strength steel and advanced engineering techniques has been as low as, or lower than, 31 cents per pound reduced. This means that for a large light truck with curb weight of 4000 pounds, the cost of material substitution per percentage fuel economy improvement would be roughly $20, ranking it among the most cost-effective fuel economy improvement options available.” Environmental Defense Cmt. at 4.
.U.S. DOT/NHTSA-Report: Vehicle Weight, Fatality Risk and Crash Compatibility of Model Year 1991-99 Passenger Cars and Light Trucks, NHTSA Docket No. 2005-22223-17 (Aug. 31, 2005) (Kahane, C.J., DOT HS 809 662 (Oct.2003)) (henceforth "Kahane Study”).
. “Crossover weight” is "the weight at which a reduction in weight would produce a zero effect on safety. Each and every light truck weighing more than the crossover weight would experience a net benefit from reduced weight. All those below the crossover weight would experience a net loss in safety.” FRIA at V-15 n. 41.
. Petitioners also cite an earlier NHTSA rulemaking, for light trucks MYs 1992-1994, to support their interpretation of "feasible.” In that rulemaking, the agency stated that it "has in the past-interpreted ‘feasible’ to refer to whether something is capable of being done.” 55 Fed.Reg. 3608, 3616 (Feb. 2, *11951990). But NHTSA further explained, “a standard set at the maximum feasible average fuel economy level must: (1) Be capable of being done and (2) be at the highest level that is capable of being done, taking account of what manufacturers are able to do in light of technological feasibility, economic practicability, how other Federal motor vehicle standards affect average fuel economy, and the need of the nation to conserve energy.” Id. (emphasis added).
. See, e.g., App. A to NRDC Cmt. at 4-12 (Natural Resources Defense Council-Appendix A, NHTSA Docket No. 2005-22223-1706 (Nov. 23, 2005) (issue paper examining how oil dependence affects the American economy and national security)).
. One of the Petitioners noted that "[w]hile previous standards have utilized cost-benefit analysis as part of the regulatory impact analysis after the standard was set, the proposed reforms put the cost-benefit analysis front and center.” App. G to NRDC Cmt. at 3.
. See Massachusetts v. EPA, 549 U.S. 497, 127 S.Ct. 1438, 1447-49, 1455, 167 L.Ed.2d 248 (2007) (describing how "the scientific understanding of climate change [has] progressed” since the 1970s and discussing the evidence showing that "[t]he harms associated with climate change are serious and well recognized.”); see generally Attachment A to CBD Cmt. (Global Warming and Its Impacts); Attachment B to CBD Cmt. (Albritton, D.L., et al., Technical Summary, Climate Change 2001: The Scientific Basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental panel on Climate Change (IPCC) (2001)); Attachment F to CBD Cmt. (Epstein, P.R. and E. Mills (eds.), Climate Change Futures: Health, Ecological and Economic Dimensions (2005)); Attachment G to CBD Cmt. (Intergovernmental Panel on Climate Change (IPCC), Climate Change 2001; Synthesis Report (Summary for Policymakers) (2001)); Attachment J to CBD Cmt. (National Snow and Ice Data Center, Sea Ice Decline Intensifies (2005)); Attachment L to CBD Cmt. (Overpeck, J.T., et al., "Arctic System on Trajectory to New, Seasonally Ice-free State,” EOS (2005)); Attachment M to CBD Cmt. (Parmesan, C. and H. Galbraith, Pew Center on Global Climate Change, Observed Impacts of Global Climate Change in the U.S. (Sept.2004)); Attachment O to CBD Cmt. (Thomas, C.D., et al., “Extinction Risk from Climate Change,” 427 Nature 145 (Jan. 8, 2004)); Attachment P to CBD Cmt. (World Health Organization, The World Health Report 2002 (2002)); Attachment Q to CBD Cmt. (Arctic Climate Impact Assessment, Impacts of a Warming Arctic: Highlights (2004)) Apps. A-B to NRDC Cmt.; Scoping Uncertainty in the Social Cost of Carbon, Final Project Report, NHTSA Docket No. 2005-22223-2250 (Mar. 13, 2006) (Thomas E. Downing, et al., Social Cost of Carbon: A Closer Look at Uncertainty (July 2005)); Environmental Defense-Report-The Social Cost *1198of Carbon Review: Methodological Approaches for Using SCC Estimates in Policy Assessment; (Paul Watkiss, et al.. The Social Cost of Carbon (SCC) Review — Methodological Approaches for Using SCC Estimates in Policy Assessment, Final Report (Nov.2005)).
. Public Citizen is also factually distinguishable. The Public Citizen court based its conclusion that NHTSA's balancing was reasonable on the evidence in the record showing that (a) severe economic consequences would result from a higher standard and (b) the potential fuel savings from a higher standard would be minor in comparison. Pub. Citizen, 848 F.2d at 265. Neither of those things are true here. First, NHTSA has provided no evidence that the auto industry would suffer severe economic consequences as a result of higher CAFE standards for light trucks MYs 2008-2011. Second, Petitioners calculated that "standards increasing steadily to an equivalent level of 26 mpg in 2011 would save 940,000 barrels per day of oil by 2020 and achieve a cumulative reduction of 304 million metric carbon equivalent tons (mmtC) by that date, more than double the amounts offered by NHTSA’s current proposal." Environmental Defense Cmt. at 2 (emphasis added); see also App. A to NRDC Cmt. at 4 (arguing that the U.S. uses over 20 million barrels of oil per day and that a higher CAFE standard that would save 940,000 barrels per day would amount to 4.7% of U.S. consumption per *1199day). This is far higher than the 0.09% figure in Public Citizen.
. The EU has established carbon emission limits for industrial emitters, thereby creating a market price for carbon emission allowances. NRDC Cmt. at 8; see, e.g., http:// www.pointcarbon.com.
. Environmental Defense also submitted an additional comment letter after the comment period closed noting a recent study from the Social Cost of Carbon project sponsored by the U.K. Department for Environment, Food and Rural Affairs. Environmental Defense Cmt. Re: Carbon Costs at 1-3. The study found that " '[a] lower benchmark of 35 £/tC [about $60 per ton of carbon] is reasonable for a global decision context committed to reducing the threat of dangerous climate change and includes a modest level of aversion to extreme risks, relatively low discount rates and equity weighting.’ ” Id. at 1-2; see Social Cost of Carbon Review. The report concluded, " ‘we believe that a value of £55/tC in 2000 [about $95/tC], but rising more sharply than the current guidance (i.e. at a higher rate than the current £l/tC per year), would seem to capture the evidence using a pragmatic approach.’ ” Id. at 2. Environmental Defense concluded:
These results support monetizing the carbon benefits of the light truck fuel economy rule using values in the range of $96 to $174 per ton of carbon [$26 to $47 per ton of C02] (at current exchange rates).... These values translate into shadow values of 30 to 54 cents per gallon.... Calculating the benefits of these savings at the new values consistent with the SCC study recommendations would yield present value benefits of $54 billion by 2020 ... and $82 billion by 2030 ... calculated using a 3% discount rate. On an annual basis, benefits would grow from $1.1 billion in 2011 to $5.9 billion in 2030.[¶] These benefits are substantial in relation to the costs estimated by NHTSA for its proposal. Yet [they] were entirely omitted from NHTSA's calculations ....
Id. at 3.
.Citing R.S.J. Tol, The Marginal Damage Costs of Carbon Dioxide Emissions: An Estimate of the Uncertainties, 33 Energy Pol’y 2064, 2074 (2005).
. NHTSA erroneously states that Environmental Defense expressed its recommended value as $50 per ton C02, rather than $50 per ton carbon. See 71 Fed.Reg. at 17,638. Fifty dollars per ton carbon is equivalent to $13.60 per ton C02, which is within the range that NRDC suggested. See id.
. In the Final Rule, NHTSA did not exclude the value of carbon reduction from its analysis on the ground that it now asserts on appeal: "EPCA does not compel NHTSA to set CAFE levels with reference to carbon dioxide emissions specifically, or environmental effects generally.” NHTSA Br. at 47. We "may only sustain an agency’s action on the grounds actually considered by the agency.” Nw. Envt’l Defense Ctr. v. Bonneville Power Admin., 477 F.3d 668, 686 (9th Cir.2007). In any case, this argument has no merit because it misses the point. NHTSA’s chosen methodology for setting CAFE standards is a cost-benefit analysis that purports to take the relevant costs and benefits into account.
.Moreover, we note that guidance from the Office of Management and Budget provides that agencies are to monetize costs and benefits whenever possible. Office of Mgmt. & Budget, Office of the President, OMB Circular A-4, at 27 (2003).
. Since Petitioners filed three sets of opening and reply briefs, the briefs addressing EPCA issues are referred to as "EPCA Br." or "EPCA Reply Br.,” the brief addressing NEPA issues is referred to as “NEPA Br.” or "NEPA Reply Br.,” and the brief filed by the governmental entities is referred to as "States’ Br.” or "States’ Reply Br.”
. Known as the "Volpe model.''
. U.S. DOT/NHTSA-Dr. Michael Wang-Individual Report (Comments), NHTSA Docket No. 2005-22223-7 (Aug. 30, 2005).
. Since Reformed CAFE does not have preset corporate mínimums but will depend on manufacturers' fleet mix, the MY 2011 figure is based on NHTSA's estimate of the average CAFE level that will be required of manufacturers for MY 2011. See 71 Fed.Reg. at 17,-568.
. “Curb weight” generally means the unloaded vehicle weight. See 40 C.F.R. § 86.082-2 (2007).
. The adverse safety effects of downweight-ing touted by NHTSA only applies to smaller vehicles such as passenger cars, not light trucks:
If an increase in fuel economy is effected by a system that encourages either down-weighting or the production and sale of more small cars, some additional traffic fatalities would be expected. However, the actual effects would be uncertain, and any adverse safety impact could be minimized, or even reserved, if weight and size reductions were limited to heavier vehicles (particularly those over 4,000 lb). Larger vehicles would then be less damaging (aggressive) in crashes with all other vehicles and thus pose less risk to other drivers on the road.
NAS Report at 5 (emphasis added); see also id. at 27, 72. The only dispute here is whether NHTSA should have included weight reduction as a cost-effective technology for light trucks weighing between 4,000 and 5,000 lbs. curb weight.
. The NAS committee majority’s findings regarding the adverse safety effects of down-weighting was disputed by two dissenting members of the committee, who argued that "[t]he relationship between fuel economy and highway safety is complex, ambiguous, poorly understood, and not measurable by any known means at the present time.... The conclusions of the majority of the committee ... are overly simplistic and at least partially incorrect.” NAS Report at 117. The dissenters stated, “[tjhe bottom line is that if the weights of passenger cars and light trucks are reduced proportionally, Kahane’s study predicts that the net effect on highway fatalities in collisions among all highway users is approximately zero. Given the history of the debate on this subject, this is a startling result.” Id. at 120-21.
.Kahane explains that "the 'fatality increase per 100-pound reduction' does not mean the effect of literally removing 100 pounds from a specific [light truck or van]. It is the average increase in the fatality rates of 1991-99 models weighing W-100 pounds relative to other 1991-99 models weighing W pounds, given drivers of the same age/gender and equal values on the other factors.” Kahane Study at viii.
. "The required level of CAFE for a particular manufacturer for a given model year is calculated using the target-setting function for that model year in conjunction with that manufacturer’s actual total production and its production at each footprint value for that model year. The manufacturer’s required CAFE level is calculated by dividing its total production for the model year by the sum of the values obtained by dividing the manufacturer’s production of each vehicle model included in its fleet by the fuel economy target for that model.” 71 Fed.Reg. at 17,587; see also id. at 17,608-09.
. The legislative history that Petitioners cite concerning NHTSA's consideration of fleet mix in establishing CAFE standards does not help them. That history reveals that Congress considered and rejected a penalty-waiver provision that would have excused manufacturers from fuel economy standards if consumer preference caused their non-compliance. This legislative history argument was rejected by the D.C. Circuit in Center for Auto Safety, and we agree with the D.C. Circuit’s analysis on that issue. 793 F.2d at 1339-40.
. R.L. Polk & Co. is a major automotive market intelligence company.
. In addition, pickup owners consider better fuel economy to be the most important improvement, valuing it nearly as much as all other potential improvements (e.g., greater horsepower, better off-road capability, carrying heavier loads) combined. Polk Study at 16.
. Trucks are subdivided into a number of classes based on GVWR. Class 1 trucks are those 6,000 lbs. GVWR or less. Generally, this includes minivans, small pickup trucks, and small to medium SUVs. Class 2a trucks are those between 6,001 lbs. and 8,500 lbs. GVWR, which include vans, full-size pickups such as the F-150 and Chevrolet Silverado 1500, and medium to large SUVs. Class 2b trucks are those 8,501 to 10,000 lbs. GVWR, which include some large vans, heavy duty pickups (e.g., F-250, Silverado 2500, Ford Excursion, GMC Yukon XL, and Hummer H2), and large SUVs. See UCS Cmt. at 33 & n. 13.
. Only one factor or the other is required, along with feasibility, for the Secretary to designate as an "automobile” a 6,000 to 10,-000 lbs. GVWR vehicle. 49 U.S.C. § 32901 (a)(3)(B)(ii).
. American Council for ah Energy-Efficient Economy (ACEEE) Cmt., NHTSA Docket No. 2003-16128-1156 (Apr. 27, 2004).
. NHTSA's original rationale for expanding the definition of "automobile” to include Class 2a trucks but continuing to exclude Class 2b trucks from CAFE regulation was the prediction that consumers would not use them for personal use. 43 Fed.Reg. 11995, 11997 (Mar. 23, 1978). The agency indicated that it would reconsider its decision if this prediction proved false: "[T]he agency does not expect any circumvention of this type to be as prevalent as the shift in GVWR’s across the previous 6000 pound dividing line.... [V]ehicles rated much above 7000 pounds are equipped with heavy duty suspensions and other components which make then unattractive for personal uses.... [I]f the agency’s projection in this regard proves to be incorrect, the light truck category could be further expanded to avoid circumvention of the fuel economy standards.” Id.
. NHTSA misdirects much of its attention to the argument that regulating MDPVs is "more feasible.” NHTSA Br. at 80 (emphasis in original).
. We also reject NHTSA’s assertion that Petitioners “invert the relevant question the agency sought to answer” since the agency was only concerned with the feasibility of regulating MDPVs, not other vehicles in the 8,500 to 10,000 lb. weight class. NHTSA Br. at 79. As discussed above, the agency specifically sought comments on whether it should regulate all vehicles between 8,500 and 10,-000 lbs. GVWR, see 68 Fed.Reg. at 74,930, and Petitioners presented evidence on precisely that question, see, e.g., Environmental Defense Cmt. at 9-11; UCS Cmt. at 33-35; Public Citizen Cmt. at 18-19; Polk Study.
. As opposed to, for example, setting total costs equal to total benefits, as suggested by some commenters. See 71 Fed.Reg. at 17,591 (rejecting the Union of Concerned Scientists' break-even approach).
. The Supreme Court's recent decision in National Association of Home Builders v. Defenders of Wildlife, - U.S. -, 127 S.Ct. 2518, 168 L.Ed.2d 467 (2007), is not relevant for several reasons. First, NEPA analysis is entirely distinct from analysis under the Endangered Species Act. See id. at 2535. Petitioners do not interpret NEPA as "add[ing] another entirely separate prerequisite to th[e] list,” id. at 2537, of statutory factors in 49 U.S.C. § 32902(f). NEPA imposes the obligation on every agency to evaluate the environmental impacts of its major actions so that there can be informed agency and public de-cisionmaking. See 40 C.F.R. § 1500.1. Second, unlike the EPA, NHTSA has not taken the position that its actions in setting CAFE standards involve no judgment or discretion. NHTSA asks this court to defer to its discretionary choices (based on its expert judgment) on every issue Petitioners raise under the EPCA. Third, there is no doubt that the fuel economy standards set by NHTSA will have a direct effect on greenhouse gas emissions from light trucks — and that NHTSA is thus a "legally relevant cause.” See Massachusetts, 127 S.Ct. at 1457-58 ("[Rjeducing domestic automobile emissions is hardly a tentative step. Even leaving aside the other greenhouse gases, the United States transportation sector emits an enormous quantity of carbon dioxide into the atmosphere ... more than 6% of worldwide carbon dioxide emissions .... To put this in perspective: Considering just emissions from the transportation sector, ... the United States would still rank as the third-largest emitter of carbon dioxide in the world.... Judged by any standard, U.S. motor-vehicle emissions make a meaningful contribution to greenhouse gas concentrations and hence, ... to global warming.”).
. NHTSA’s reliance on Foundation for North American Wild Sheep is misplaced because in that case—as here—the agency’s EA "failed to address certain crucial factors, consideration of which was essential to a truly informed decision whether or not to prepare an EIS.” 681 F.2d at 1178. The EA failed to consider crucial factors relating to the effect of reconstruction and use of a road that passed directly through an area occupied by Desert Bighorn Sheep, which are sensitive to encroachment on their territory. The EA’s insufficiency forced the court to “speculate,” even though "the very purpose of NEPA’s requirement that an EIS be prepared for all actions that may significantly affect the environment is to obviate the need for such speculation by insuring that available data is gathered and analyzed prior to the implementation of the proposed action.” Id. at 1179.
. There are also some inconsistencies. Petitioners point out, for example, that the EA does not explain how the lifetime emissions of MY 2011 vehicles (697 mmt) could be less than MY 2010 (700 mmt) for the baseline alternative, see Final EA at 29, given that fuel economy is held constant and vehicle miles traveled (VMT) are expected to increase each year, id. at 8, 34.
. 22.5 mpg for MY 2008, 23.1 mpg for MY 2009, and 23.5 mpg for MY 2010.
. See U.S. E.P.A., Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990-2004, at 2-1 (Apr. 15, 2006) (stating that “[i]n 2004, total U.S. greenhouse gas emissions were 7,074.4 tetragrams of carbon dioxide equivalents (Tg C02 Eq.).”).
. This document is a revision of Environmental Defense's June 2005 analysis of optimal light truck fuel economy improvement rates titled, "Cost-Effective Targets for a 2008 + Light Truck CAFE Rule.” See Environmental Defense Cmt.App. A at A-l.
. Petitioners did not waive this “tipping” argument, as NHTSA argues. Evidence concerning "tipping” was presented to the agency during the comment period and is in the administrative record.
. Wetlands Action Network v. U.S. Army Corps of Engineers, 222 F.3d 1105, 1122 (9th Cir.2000), cited by NHTSA, is not on point, because petitioners there wanted the Corps to build a saltwater marsh instead of a freshwater marsh.
. We are not persuaded by NHTSA’s argument that City of Los Angeles v. NHTSA supports its position. In that case, state and local governmental entities and environmental groups brought separate challenges to NHTSA’s decision not to prepare an EIS covering its CAFE standards for MYs 1987-1988 and 1989. 912 F.2d at 482. Pursuant to 15 U.S.C. § 2002(a)(1), (4), NHTSA lowered the passenger car standard from the statutory 27.5 mpg to 26.0 mpg for MYs 1987-88 and 26.5 mpg for MY 1989. Id. With respect to MYs 1987-88, the court held that city and state petitioners had standing to sue under NEPA on air pollution grounds, but it rejected their challenge on the merits. The court held that NRDC had standing to challenge the MY 1989 on global warming grounds, but it did not disturb NHTSA’s conclusion that an EIS was not required. Chief Judge Wald dissented on the court's merits ruling and would have held that NHTSA acted arbitrarily in concluding that the 1989 CAFE standard would not have a significant impact on global warming and would have remanded to NHTSA. Id.
Then-Judge Ruth Bader Ginsburg joined in Judge D. Ginsburg's opinion on NRDC’s NEPA challenge, and she provided two reasons for her concurrence: "(1) NRDC's apparent acceptance of NHTSA's finding that the 1.0 mpg CAFE rollback at issue would yield a 'maximum theoretical increase of less than one percent in greenhouse gases,' ... and (2) NRDC’s failure even to allege that such an increase 'would produce any marginal effect on the probability, the severity, or the imminence’ of the global warming disaster petitioners project.” Id. at 504 (citation omitted). These reasons do not apply here. Petitioners have provided substantial evidence that even a small increase in greenhouse gases could cause abrupt and severe climate changes. Cf. Massachusetts, 127 S.Ct. at 1458 (''[T]he rise in sea levels associated with global warming has already harmed and will continue to harm Massachusetts. The risk of catastrophic harm, though remote, is nevertheless real. That risk would be reduced to some extent if petitioners received the relief they seek.”).