OPINION

The present case is the latest round of environmental litigation in the 33-year history of the Pacific Northwest Electric Power Planning and Conservation Act (the Power Act), 16 U.S.C. §§ 839 -839h. That statute established the Northwest Power and Conservation Council (the Council), an interstate agency composed of state-appointed representatives from Idaho, Montana, Oregon, and Washington that Congress tasked with promulgating both “a regional conservation and electric power plan” and “a program to protect, mitigate, and enhance fish and wildlife.” 16 U.S.C. § 839b(d)(l), 839b(h)(l)(A).

The Power Act was designed to resolve the conflict between the Columbia River Basin’s two great natural resources: hy-dropower and salmon. Nw. Res. Info. Ctr. v. Niv. Power Planning Council, 35 F.3d 1371 , 1375, 1377 (9th Cir.1994). Over the years, the Council’s efforts to fulfill its duties have been challenged in federal court by various regional stakeholders, including environmental groups, power companies, state governments, Indian nations, and power-consuming industrial interests. See id.; Seattle Master Builders Ass’n v. Pac. Nw. Elec. Power & Conservation Planning Council, 786 F.2d 1359 (9th Cir. 1986).

This case presents a challenge by an environmental group, the Northwest Resource Information Center (NRIC), to the Sixth Northwest Power Plan (the Plan) that the Council adopted in May 2010. NRIC’s key complaint is that the Council failed to give due consideration to the accommodation of fish and wildlife interests when it adopted the Plan. For the reasons set forth below, we AFFIRM the Plan with respect to NRIC’s “due-consideration” challenge, but REMAND the Plan to the Council for the limited purposes of (1) allowing public notice and comment on the proposed methodology for determining quantifiable environmental costs and benefits, and (2) reconsidering the inclusion in the Plan of a market-price-based estimate of the cost of accommodating fish and wildlife interests.

I. BACKGROUND

A. Statutory background

Prior decisions of this court have discussed the history, purpose, and operation of the Power Act. See, e.g., Nw. Envtl. Def. Ctr. v. Bonneville Power Admin., 117 F.3d 1520 ,1525-26, 1530-31 (9th Cir.1997); Nw. Res. Info. Ctr., 35 F.3d at 1377-79. The key source of conflict is that the extensive system of hydroelectric dams in the Columbia River Basin has been “a major factor in the decline of some salmon and steelhead runs to a point of near extinction.” Nw. Res. Info. Ctr., 35 F.3d at 1376 (quoting 126 Cong. Rec. H10687 (1980)). Hydroelectric dams have a destructive cumulative effect on salmon and steelhead fish (collectively referred to as “anadro-mous fish” because they spawn in freshwater, reach maturity in saltwater, and then return to freshwater to reproduce) mostly because they impede the path of juvenile fish to the ocean. Id. at 1376 & n. 5.

The “devastating losses of salmon and steelhead in the mid-1970s” prompted *1012 Congress to enact the Fish and Wildlife Coordination Act, 16 U.S.C. §§ 661 -666c, in 1976. Nw. Res. Info. Ctr., 35 F.3d at 1377. But that statute’s mandate of giving “equal consideration” to fish and wildlife on the one hand and hydropower projects on the other proved inadequate. Id. Congress also began considering revisions to the electric-power policies in the Northwest region because of “forecasts predicting serious power shortages during critical water years expected in the 1980’s.” Id. The Power Act, enacted in 1980, was a legislative response to both issues. It “marked an important shift in federal policy” by creating a “new obligation on the region and various Federal agencies to protect, mitigate, and enhance fish and wildlife” while not jeopardizing “an adequate, efficient, economical, and reliable power supply.” Id. at 1377-78 (internal quotation marks omitted).

Under the Power Act, the Council must first develop and adopt a fish and wildlife program. See 16 U.S.C. § 839b(h)(2)-(9) (mandating the process for adopting a fish and wildlife program before finalizing a conservation and electric power plan). The Council, in doing so, must seek recommendations from federal and state fish and wildlife agencies and from the region’s Indian tribes concerning (1) measures that can be implemented to advance fish and wildlife interests; (2) objectives for the development and operation of hydroelectric projects in the Columbia River Basin that promote such interests; and (3) research and development to, among other things, protect “anadromous fish at, and between, the region’s hydroelectric dams.” Id. § 839b(h)(2)(A)-(C). Federal and regional water-management agencies, regional electric-power-producing agencies and customers, and the general public may also submit recommendations to the Council. Id. § 839b(h)(3).

After these recommendations undergo a public notice-and-comment process, see 16 U.S.C. § 839b(h)(4), the Council is required to develop a fish and wildlife program based on the recommendations, public commentary, and consultations with the federal and state fish and wildlife agencies, the “appropriate Indian tribes,” the federal agencies responsible for operating or regulating hydroelectric facilities in the region, and “any customer or other electric utility which owns or operates any such facility.” See id. § 839b(h)(5) (referring back to the “agencies, tribes, and customers” listed in § 839b(h)(4)(A)). One of the key federal agencies involved in this process is the Bonneville Power Administration (BPA), a power-marketing agency within the United States Department of Energy that is charged with implementing conservation measures and acquiring resources in accordance with the Council’s power plan. See Aluminum Co. of Am. v. Bonneville Power Admin., 903 F.2d 585 , 588 (9th Cir.1989) (describing the BPA); 16 U.S.C. § 839d(a) (setting forth the BPA’s role in implementing the Council’s power plan).

In the event that stakeholder recommendations conflict, the Council is charged with resolving any inconsistency by “giving due weight to the recommendations, expertise, and legal rights and responsibilities of the Federal and the region’s State fish and wildlife agencies and appropriate Indian tribes.” 16 U.S.C. § 839b(h)(7). The Council may reject any recommendations made by such agencies and tribes, but it must justify the rejection with a written explanation of why the recommendation does not comport with the objectives of the fish and wildlife program as outlined in the Power Act or why the recommendation would be less effective than the adopted recommendations. Id.

Congress also set forth in the Power Act a substantive mandate for the fish and *1013 wildlife program: “The program shall consist of measures to protect, mitigate, and enhance fish and wildlife affected by the development, operation, and management of [hydroelectric] facilities while assuring the Pacific Northwest an adequate, efficient, economical, and reliable power supply.” 16 U.S.C. § 839b(h)(5). The Council must include such measures that it determines will

16 U.S.C. § 839b(h)(6).

The fish and wildlife program exists apart from, but is also mandatorily incorporated into, the regional conservation and electric power plan that the Power Act requires the Council to adopt. See 16 U.S.C. § 839b(e)(3)(F) (requiring that a power plan include a fish and wildlife program). Similar to § 839b(h)’s requirements for the fish and wildlife program, the Power Act sets forth procedural and substantive requirements for the hydroelectric power plan. See id. § 839b(d)-(e).

Procedurally, the Council must review its power plan at least once every five years. 16 U.S.C. § 839b(d)(l). The power plan may also “be amended from time to time.” Id. Public hearings must precede any adoption of a power plan or any “substantial, nontechnical amendments” thereto and must conform to the section of the Administrative Procedure Act governing rulemaking. Id. § 839b(d)(l) (referencing 5 U.S.C. § 553 ). Once adopted, all conservation and resource acquisitions undertaken by the BPA must be consistent with the Council’s power plan (unless otherwise specified in the Power Act). Id. § 839b(d)(2).

The Power Act also prescribes general substantive requirements for the power plan, starting with prioritizing resources into a hierarchy. “Priority shall be given: first, to conservation; second, to renewable resources; third, to generating resources utilizing waste heat or generating resources of high fuel conversion efficiency; and fourth, to all other resources.” 16 U.S.C. § 839b(e)(l). Resources that the Council determines to be “cost-effective,” a term defined elsewhere in the Power Act, id. § 839a(4)(A), are given priority over resources that are not cost-effective, id. § 839b(e)(l).

The power plan must also establish a “general scheme” for the BPA to meet its power-providing obligations through conservation and resource acquisition. 16 U.S.C. § 839b(e)(2). This scheme must reflect the Council’s “due consideration” for the following:

Id.

The Power Act then lists seven specific items that the power plan must include: (1) an energy conservation program that includes model conservation standards, (2) research and development recommendations, (3) “a methodology for determining quantifiable costs and benefits,” (4) “a demand forecast of at least twenty years ... and a forecast of power resources estimated by the Council to be required to meet the [BPA’s] obligations,” (5) “an analysis of reserve and reliability requirements and cost-effective methods of providing reserves designed to insure adequate electric power at the lowest probable cost,” (6) the fish and wildlife program adopted pursuant to § 839b(h), and (7) a methodology for calculating surcharges recommended under § 839b(f), if any. 16 U.S.C. § 839 (b)(e)(3). Congress left to the Council’s discretion the level of detail that the power plan provides with respect to these seven elements. Id.

B. The Sixth Northwest Power Plan (the Plan)

In December 2007, the Council published a paper that identified the major issues for consideration in its next power plan. The Council also asked for stakeholder feedback on those issues and other relevant topics. Meanwhile, the Council’s primary focus was on the fish and wildlife program, culminating in the June 2009 adoption of a new program (the 2009 Program).

Unlike past programs, the 2009 Program did not include plans of detailed hy-drosystem operations for fish and wildlife because the federal agencies that operate and regulate the federal dams in the Columbia River Basin had already produced detailed plans for the operations of each facility intended to improve conditions for fish and wildlife affected by the hydrosys-tem. These plans, set forth and reviewed in biological opinions from the National Oceanic and Atmospheric Administration Fisheries and the U.S. Fish and Wildlife Service, focus on benefitting fish populations listed as threatened or endangered under the Endangered Species Act.

Rather than prescribe specific operations, the 2009 Program lays out biological objectives, principles, and strategies designed to benefit fish and wildlife. These strategies included transporting and providing safe bypasses for fish around dams, as well as spilling water over dams to allow for their passage. NRIC did not submit any recommendations or comments during the public process that led to the 2009 Program, nor did it seek judicial review once the 2009 Program was adopted.

In September 2009, the Council issued a draft version of the Plan. A period for public comments followed, and the Council held public hearings in Idaho, Montana, Oregon, and Washington. The Plan notes that the Council consulted “with various governments, entities and individuals in the region, and accepted and considered substantial written and oral comments.” In February 2010, the Council voted to adopt the Plan, and notice of that adoption was published on May 4, 2010.

Pursuant to the Power Act, the Plan incorporates by reference the 2009 Program and provides a power-demand forecast, an assessment of current and potential resources, a conservation program, an analysis of reserve and reliability requirements, and an appendix describing a methodology for determining quantifiable environmental costs and benefits. The methodology appendix, however, did not appear in the September 2009 draft version of the Plan, an omission first brought to light by public comments.

*1015 Although not required by the Power Act, the Plan also includes the BPA’s market-price-based estimate of the financial cost of the fish and wildlife measures adopted in the 2009 Program. The BPA’s cost estimate, which is $750-$900 million annually, appeared in the draft version of the Plan, but it was part of a lengthy explanation of (1) an alternative method for estimating the cost of the 2009 Program that was substantially lower, and (2) the several purposes served by providing such a cost estimate. Using a “replacement resource cost” methodology, the draft version of the Plan estimated the 2009 Program’s cost at $300 million annually, less than half of the cost estimate provided by the BPA. Neither that specific estimate nor any mention of other cost-estimate methodologies appear in the final version of the Plan.

C. Procedural background

NRIC timely filed its petition to challenge the Plan in this court after the Council published its notice of adoption. See 16 U.S.C. §§ 839f(e)(l)(A), 839f(e)(5) (providing that the adoption of a power plan is a final action subject to judicial review and that suits to challenge final actions of the Council are to be brought in the federal court of appeals for the region). This court allowed the BPA, Northwest River-Partners (an interest group of regional utilities, ports, businesses, and farmers), and the Public Power Council (a trade association representing regional consumer-owned utilities) to intervene and file response briefs in addition to the brief filed by the Council.

II. ANALYSIS

A. Standard of review

The Administrative Procedure Act, 5 U.S.C. §§ 701-06 , governs review of actions taken by the Council. Seattle Master Builders Ass’n v. Pac. Nw. Elec. Power & Conservation Planning Council, 786 F.2d 1359, 1366 (9th Cir.1986); 16 U.S.C. § 839f(e)(2). The Council’s adoption of the Plan will not be set aside unless the Plan is found to be “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” See 5 U.S.C. § 706 ; Seattle Master, 786 F.2d at 1366 .

In contrast, this court generally reviews de novo any legal questions arising from an agency decision, such as the proper interpretation of a statutory provision. Seattle Master, 786 F.2d at 1366 . “[W]e look first to the statute’s language,” which “must ordinarily be regarded as conclusive.” Nw. Res. Info. Ctr. v. Nw. Power Planning Council, 35 F.3d 1371 , 1383 (9th Cir.1994) (internal quotation marks omitted). But substantial deference is accorded “to the interpretation given statutes by the officers or agency charged with their administration.” Nw. Envtl. Def. Ctr. v. Bonneville Power Admin., 117 F.3d 1520 , 1530 (9th Cir.1997) (internal quotation marks omitted).

B. Due consideration

NRIC first argues that the Plan fails to give “due consideration ... for ... protection, mitigation, and enhancement of fish and wildlife” as the Power Act requires. See 16 U.S.C. § 839b(e)(2). “Due consideration,” according to NRIC, obligates the Council to independently consider and give significant weight to the needs of anadromous fish when formulating a power plan. NRIC contends that this obligation stands separate from the requirement that a power plan incorporate a fish and wildlife program. See id. § 839b(e)(3)(F). The upshot of NRIC’s interpretation is that if the Council learns through developing a power plan that existing or future power resources, including the implementation of conservation measures, create the capacity for further fish *1016 and wildlife enhancements, then the Council must consider whether such enhancements would serve the Power Act’s goal of furthering fish and wildlife interests while maintaining an adequate, efficient, economical, and reliable power supply.

The Council agrees with NRIC “that the consideration due under the [Power] Act is a serious substantive obligation” and in the past has recognized that such consideration is in addition to the Power Act’s mandate of adopting a fish and wildlife program. But it rejects NRIC’s contention that the Power Act requires, or even allows, the Council to revisit the fish and wildlife measures adopted in the 2009 Program. The Council instead argues that it gave fish and wildlife interests due consideration in three ways: (1) through the evaluation of two of the many alternative power-resource scenarios analyzed for inclusion in the Plan, (2) by developing a resource plan that accommodates the 2009 Program’s fish and wildlife measures, and (8) by considering the impact of potential new power resources both on environmental quality and on fish and wildlife.

But the Council’s first and second suggested methods for providing due consideration to fish and wildlife interests miss the mark. The first method is easily dismissed because the Council fails to show how its evaluation of two particular power-resource scenarios was at all relevant to its due-consideration obligation, particularly with respect to providing “sufficient quantities and qualities of [river] flows for successful migration, survival, and propagation of anadromous fish.” See 16 U.S.C. § 839b(e)(2). One of these resource scenarios posited removing the four federal dams in the lower Snake River (a major tributary in the Columbia River Basin), while the other considered the hypothetical retirement of a number of coal-fired plants throughout the region.

The Council contends that its due-consideration obligation did not require an analysis of either scenario. It further argues that it is not authorized to include in a power plan’s resource scheme a recommendation to shut down an existing power resource. But an analysis of resource scenarios that the Council claims is beyond its power to adopt has only a tenuous bearing on the Council’s duty to set forth a resource scheme that gives due consideration to fish and wildlife interests. The Council itself notes that “[t]he resource scheme actually adopted by the Council in the power plan did not call for or assume that the events would happen; these alternative scenarios (and others) were developed to provide useful information to the region to inform further thinking and planning.” In other words, the alternative scenarios were not developed to give due consideration to fish and wildlife interests in the resource scheme actually adopted. A post-hoc reliance on these alternative-scenario analyses is thus unavailing.

The Council’s second argument—that it provided due consideration to fish and wildlife enhancements through its adoption of the 2009 Program—is likewise inadequate. Developing a resource plan that accommodates the fish and wildlife program satisfies the two statutory mandates listed below and thus cannot constitute complying with the due-consideration mandate if that mandate is properly understood as an independent obligation. See Kenaitze Indian Tribe v. Alaska, 860 F.2d 312, 317 (9th Cir.1988) (declining to adopt a statutory interpretation that would render one portion of the statute redundant when “another interpretation ... avoids such redundancy”). Those other statutory mandates are (1) the requirement that the power plan include a fish and wildlife program, 16 U.S.C. § 839b(e)(3)(F), and (2) the requirement that the plan’s power-demand forecast account for the effects of *1017 the fish and wildlife program on the availability of power resources, id. § 839(b)(e)(3)(D).

The Council further argues that the 2009 Program places “hard nonpower constraints on the hydrosystem,” thus demonstrating the Council’s due consideration. But the two separate Power Act provisions identified above already require building the power plan around those measures. Moreover, the fish and wildlife measures in the 2009 Program account for the Power Act’s command that fish and wildlife protection not jeopardize an efficient and reliable power supply. See 16 U.S.C. § 839b(h)(5). The Power Act does not authorize the Council to adopt a fish and wildlife program that maximizes fish protection in an absolute sense, but rather requires the Council to adopt the best measures possible in light of the Power Act’s energy goals. See id. § 839b(h)(5).

So while the Council equates accounting for the measures in the 2009 Program with providing due consideration for fish and wildlife interests, those measures in fact are already limited by the Power Act’s countervailing interest in assuring that the region has appropriate power resources. Characterizing those measures as “hard nonpower constraints on the hydrosystem” thus belies the fact that the measures themselves already reflect both fish and wildlife interests and power interests.

Finally, the Council suggests that it provided due consideration to fish and wildlife interests and to environmental quality by assessing the impact that potential new power resources might have on those interests. The Council’s “primary example” of this method of due consideration is the “Protected Areas” policy, which guards 44,000 miles of regional streams against new hydroelectric development. But that policy, as the citation in the Council’s response brief indicates, is part of the 2009 Program. The Council does not explain how one aspect of the 2009 Program can satisfy an obligation in the power-plan process that the Council has recognized is separate from its duty to adopt and incorporate a fish and wildlife program into the power plan.

Putting the Protected Areas policy aside, however, the Council’s basic point holds true: The Power Act’s due-consideration requirement is aimed specifically at new power-resource acquisitions, not at existing resources. This is reflected in the statute’s requirement that “[t]he plan shall set forth a general scheme for implementing conservation measures and developing resources pursuant to section 839d ... with due consideration by the Council for ... protection, mitigation, and enhancement of fish and wildlife.” 16 U.S.C. § 839b(e)(2) (emphasis added). Section 839d, in turn, governs the BPA’s conservation measures and resource acquisitions. Id. § 839d(a)-(b) (ordering the BPA to implement conservation measures and acquire resources in accordance with the Council’s power plan).

The language of 16 U.S.C. §§ 839b(e)(2) and 839d thus forecloses NRIC’s argument that the Council must reconsider the fish and wildlife measures adopted in the 2009 Program in order to satisfy the due-consideration requirement. Those measures concern adjustments to the existing hydro-system, whereas the Power Act ties the due-consideration requirement specifically to future conservation and resource acquisition. To be sure, the Power Act does not appear to prohibit such reconsideration, see id. § 839b(h)(9) (providing the time frame within which “[t]he Council shall adopt such program or amendments thereto” (emphasis added)), but the Council’s duty of providing due consideration to fish and wildlife interests in setting a general agenda for future conservation and re *1018 source acquisition does not require reconsideration of the 2009 Program.

NRIC has shown why consideration of additional fish and wildlife measures in the existing hydrosystem is not an unreasonable proposition. From the Fifth Northwest Power Plan to the current Plan, the amount of power that the Council estimates that the region can conserve in a cost-effective manner has doubled from approximately 2,950 average megawatts to 5,900 average megawatts, such that conservation alone can meet 85 percent of the region’s demand growth for the next 20 years. In comparison, fish and wildlife measures in the 2009 Program “reduce hydroelectric generation by about 1,200 average megawatts relative to operation with no constraints for fish and wildlife.” The increase in estimated conservation capacity from the last power plan to the current Plan is therefore more than double the power impact that current fish and wildlife measures have on the hydrosystem.

Assuming that the new conservation estimates were unknown to the Council when formulating the 2009 Program, that program underestimated the degree to which the region could accommodate fish and wildlife measures while maintaining an adequate power supply. The Plan itself notes that fish and wildlife measures must take into account the region’s power supply, but that evaluations of the power system are “necessarily preliminary” during the consideration of the fish and wildlife program. And although total electricity demand is not the only relevant factor when evaluating the power system—the Council must account for variables like peak demand periods and fluctuating wind generation—a substantial increase in potential conservation should nonetheless be relevant to fish and wildlife planning.

But articulating a retroactive approach that the Council chose not to follow is insufficient to meet NRIC’s burden of showing that the Council acted in a manner that was “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” See 5 U.S.C. § 706 . NRIC has not pointed to any part of the Power Act that requires the Council to reconsider fish and wildlife measures in light of its evaluation of the regional power system from the subsequent power-planning process. Absent such a showing, we will not second-guess the due consideration that the Council gave to fish and wildlife interests in the adoption of the Plan.

C. Methodology

NRIC next argues that the Council’s failure to include a methodology for evaluating environmental costs and benefits in the draft version of the Plan was both contrary to the Power Act and arbitrary and capricious. It also contends that the methodology included in Appendix P of the final version of the Plan “fails to provide a rational method for calculating environmental costs and benefits of resources or measures necessary to meet the goals of the [Power] Act.”

“The choice of methodology is a highly technical question which falls within the unique expertise of the Council.” Seattle Master Builders Ass’n v. Pac. Nw. Elec. Power & Conservation Planning Council, 786 F.2d 1359,1370 (9th Cir.1986) (upholding the Council’s methodology for determining the cost-effectiveness of conservation measures). “Unless an abuse of discretion is demonstrated, this court will not substitute its judgment on particular ... methodology.” Id.

One of the required elements of a power plan is “a methodology for determining quantifiable environmental costs and benefits under section 839a(4) of this title.” 16 U.S.C. § 839b(e)(3)(C). As with the other elements listed in § 839b(e)(3), the Power Act instructs the Council to set forth the *1019 methodology “in such detail as the Council determines to be appropriate.” Section 839a(4) provides the definition of “cost-effective,” which calls for comparing resources on the basis of “incremental system cost,” where the “system cost” of a resource is an “estimate of all direct costs ..., including ... quantifiable environmental costs and benefits.” Id. § 839a(4)(A)-(B).

1. Procedural challenge

No methodology for determining quantifiable environmental costs and benefits appeared in the draft plan, and the Council did not submit for public comment the methodology it included in Appendix P to the final version of the Plan. The Council acknowledges that the omission of its methodology from the draft plan was an “unfortunate error,” but characterizes the omission as an irrelevant procedural point. Essentially, the Council asks us to find that its error was harmless. This court has held that “[a]n agency may rely on harmless error only when its mistake is one that clearly had no bearing on the procedure used or the substance of [the] decision reached.” Sagebrush Rebellion, Inc. v. Hodel, 790 F.2d 760, 765 (9th Cir. 1986) (ellipsis and internal quotation marks omitted).

Omission of the methodology from the draft plan was harmless, the Council argues, because its methodology was self-evident from the draft plan. The Council, however, has cited no part of the draft plan to support that assertion. Rather, the citations to the record in the Council’s brief reflect the Council’s recognition of the omission and its decision to include the methodology as an appendix to the final version of the Plan.

The Council likewise contends that no amount of additional public notice and comment would have led the Council to adopt the type of methodology that NRIC argues is appropriate. But that post-hoc litigation position, even if true, is insufficient to demonstrate that the error “clearly had no bearing on the procedure used or the substance of [the] decision reached.” See Sagebrush Rebellion, 790 F.2d at 765 (internal quotation marks omitted).

Including the methodology in the draft version of the Plan, which draft went through the notice-and-comment process, might not have produced any substantial differences from the methodology that appears in the final version of the Plan, but the Council has not clearly established that this is so. See Riverbend Farms, Inc. v. Madigan, 958 F.2d 1479, 1487 (9th Cir. 1992) (explaining that “harmless error analysis in administrative rulemaking must ... focus on the process as well as the result” because otherwise “an agency could always claim that it would have adopted the same rule even if it had complied with the APA procedures”). For this reason, we will remand the Plan to the Council for the limited purpose of adopting a methodology through the appropriate notice-and-comment process. See 16 U.S.C. § 839b(d)(l) (outlining the process for adopting any significant changes to the power plan).

2. Substantive challenge

NRIC also challenges the substance of the methodology adopted by the Council, arguing that the methodology “fails to provide a rational method for calculating environmental costs and benefits of resources or measures necessary to meet the goals of the [Power] Act.” Because we hold that this aspect of the Plan must be remanded to remedy the procedural defect, and because the notice-and-comment process might result in a change to the substantive methodology, we decline to address NRIC’s challenge based on the current wording of Appendix P.

*1020 D. Inclusion of the BPA’s cost estimate

NRIC’s final challenge to the Plan centers on the Council’s decision to include the BPA’s estimate that the cost of the 2009 Program totals $750-$900 million per year. That estimate, based in part on market prices applied to foregone power generation, appears in the Plan at least four times. No other estimate of the financial cost of the fish and wildlife program to the BPA’s operation of the hydro-system appears in the Plan. NRIC argues that the inclusion of what it calls an “inflated” cost estimate will influence the Council’s and the public’s perception of what fish and wildlife measures are possible while maintaining a power supply that is economical.

In contrast, the draft version of the Plan included a much lower estimate of the 2009 Program’s cost ($300 million), which reflects “a long-term amortized replacement resource cost.” The draft of Appendix M then noted that assessing the costs in terms of market prices is “sometimes” important, such as when the BPA calculates the credit it receives for expenses made for nonpower-related operations. Finally, the draft noted that “[t]he traditional ‘market price’ calculation of the total effect on generation from fish and wildlife operations is essentially irrelevant to the power plan’s resource development efforts.”

The reason for removing the discussion of fish and wildlife costs from Appendix M, according to the Council, was that the discussion elicited contentious comments on a subject that the Council deemed irrelevant. Concentrating on the cost issue arguably “threatened to interfere with the Council’s focus on ... the resource strategy.” But that reasoning does not explain why the Council retained the BPA’s cost estimate in Appendix M and also inserted it elsewhere in the Plan. The Council provides no reasoned basis, either in the record or in its brief, for why it eliminated the lower, resource-replacement cost estimate but mentioned the BPA’s higher, market-rate estimate multiple times.

A decision by an agency is arbitrary when it fails to “articulate a satisfactory explanation for its action[,] including a rational connection between the facts found and the choice made.” Motor Vehicle Mfrs. Ass’n of the U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 , 103 S.Ct. 2856 , 77 L.Ed.2d 443 (1983) (internal quotation marks omitted). Rather than articulate a rationale for its decision to retain the BPA’s cost estimate, the Council argues that the estimate had no bearing on the development of the Plan. From that premise, the Council concludes that including the estimate in the Plan was not unreasonable or arbitrary. The Council’s argument, however, conflates the issue of whether the inclusion of the BPA’s estimate was arbitrary with the question of the estimate’s ultimate effect on the Plan, and no legal authority is cited to support its argument.

To be sure, “we must take ‘due account’ of the harmless error rule” when we review agency final actions under the Administrative Procedure Act. Cal. Wilderness Coal. v. U.S. Dep’t of Energy, 631 F.3d 1072, 1090 (9th Cir.2011); see 5 U.S.C. § 706 (requiring a reviewing court to take “due account ... of the rule of prejudicial error”). As noted above, this court has defined “harmless error” in the administrative-rulemaking context as an error that “clearly had no bearing on the procedure used or the substance of [the] decision reached.” Sagebrush Rebellion, Inc. v. Hodel, 790 F.2d 760, 765 (9th Cir. 1986) (internal quotation marks omitted). Although the burden is on the party attacking the agency decision to show that the error was harmful, Cal. Wilderness, 631 F.3d at 1091 , we “must exercise great *1021 caution in applying the harmless error rule in the administrative rulemaking context” because “[h]armless error is more readily abused there than in the civil or criminal context,” id. at 1090 (internal quotation marks omitted). “ ‘[T]he “burden” of showing that an error was harmful is not ... a particularly onerous requirement.’ ” Id. (quoting Shinseki v. Sanders, 556 U.S. 396, 410 , 129 S.Ct. 1696 , 173 L.Ed.2d 532 (2009)).

The Council argues that it “did not consider or rely on how [the BPA] reports the costs of the Council’s fish and wildlife program in developing or adopting the resource strategy or other required elements of the Sixth Power Plan.” But even if we assume that assertion to be true, the decision is nonetheless harmful because the final action that we are obligated to review under 5 U.S.C. § 706 is the adoption of the Plan as a whole, not merely the Plan’s “resource strategy” or its “other required elements.”

As much as the Council stresses both in the Plan and in its brief before this court that the financial cost of the 2009 Program was irrelevant to the development of the Plan’s resource decisions, assessing that cost remains important. The Council itself identified “at least four different purposes for assessing the cost of fish and wildlife operations” in the draft of the Plan that was submitted for public comment. Moreover, the overarching purpose of the Power Act is the protection of fish and wildlife while maintaining an adequate and economical power supply. See Nw. Res. Info. Ctr. v. Nw. Power Planning Council, 35 F.3d 1371 , 1378-79 (9th Cir.1994).

The Council’s implicit endorsement of a cost estimate of fish and wildlife measures that is more than double the estimate produced by an alternative methodology is directly relevant to the fundamental balance that the Power Act commands the Council to achieve. Whether those measures cost $750 million annually rather than $300 million annually will quite likely affect where that balance is struck when the Council and the region’s stakeholders develop future fish and wildlife programs and power plans. The Council’s contention that the BPA cost estimate had “no bearing” on the Plan as a whole thus rings hollow when in fact the Council incorporated the estimate multiple times into both the Plan’s body and Appendix M.

Because the Council has provided no basis for adopting the BPA’s cost estimate throughout the Plan, and because “[w]e may not supply a reasoned basis for the agency’s action that the agency itself has not given,” Motor Vehicle Mfrs. Ass’n of the U.S., Inc. v. State Farm, Mut. Auto. Ins. Co., 463 U.S. 29, 43 , 103 S.Ct. 2856 , 77 L.Ed.2d 443 (1983), the Council’s unsupported decision was arbitrary. NRIC, moreover, has borne its relatively light burden of showing that the Council’s arbitrary decision was harmful. See Shinseki, 556 U.S. at 410 , 129 S.Ct. 1696 .

This conclusion, however, does not require setting aside the entire Plan. On remand, the Council must reconsider the parts of the Plan that contain the BPA’s cost estimate. The Council is not foreclosed from including that estimate in the Plan, but it must develop a reasoned basis for doing so. Nor is the Council required to include the resource-replacement cost estimate developed in the draft of Appendix M, but the decision to include or exclude that estimate must be grounded in reasoning reflected in a record that this court may review. See, e.g., Ariz. Cattle Growers’ Ass’n v. U.S. Fish & Wildlife, Bureau of Land Mgmt., 273 F.3d 1229, 1236 (9th Cir.2001) (explaining that “[t]he reviewing court may not substitute reasons for agency action that are not in the record,” and that “[¡Judicial review is mean *1022 ingless” if the court cannot “review the record to ensure that agency decisions are founded on a reasoned evaluation of the relevant factors” (internal quotation marks omitted)).

As a final note, we respectfully disagree with the partial dissent’s characterization of our analysis on this issue as one of “passion” and “zeal” to edit the Council’s work product. We instead view the Council’s unexplained inclusion of the BPA’s cost estimate to the exclusion of the substantially lower resource-replacement cost estimate as both an implicit finding of fact under 5 U.S.C. § 706 (2) of the Administrative Procedure Act and a “fail[ure] to consider an aspect of the problem” under Greater Yellowstone Coalition v. Lewis, 628 F.3d 1143, 1148 (9th Cir.2010). We thus agree with our dissenting colleague’s closing comment that “[w]e are not yet commissioned to serve as a Judicial Editorial Review Board,” but we find the comment inapposite.

III. CONCLUSION

For all of the reasons set forth above, we AFFIRM the Plan with respect to NRIC’s due-consideration challenge, but REMAND the Plan to the Council for the limited purposes of (1) allowing public notice and comment on the proposed methodology for determining quantifiable environmental costs and benefits, and (2) reconsidering the inclusion in the Plan of the BPA’s estimate of the 2009 Program’s cost to hydrosystem operations. Each party will bear its own costs on appeal.