GOULD, Circuit Judge:

This appeal is the latest chapter in a multi-decade dispute centered on Teck Metals’ liability for dumping several million tons of industrial waste into the Columbia River. Since we last heard an interlocutory appeal in this case, the district court dismissed Teck’s divisibility defense to joint and several liability on summary judgment. At Phase I of the trifurcated bench trial, the court held that Teck was a liable party under the Comprehensive Environmental Response, Compensation, and Liability Act (“CERCLA”). At Phase II, the court found Teck liable for more than $8.25 million of the Confederated Tribes of the Colville Reservation’s response costs. The district court then certified this appeal by entering partial judgment under Federal Rule of Civil Procedure 54(b). We conclude that we have jurisdiction, and we affirm.

I

The Columbia River, the fourth-largest river in North America, begins its 1,200-mile journey to the sea from its headwaters in the Canadian Rockies. The River charts a northwest course in British Columbia before bending south toward Washington. It then widens and forms the Arrow Lakes reservoir until, thirty miles before the international border, it reaches the Hugh Keenleyside Dam. After passing through the dam’s outlet, the River is free-flowing until south of the border near Northport, Washington. There it again starts to slow and pool at the uppermost reaches of Lake Roosevelt, the massive reservoir impounded behind the Grand Coulee Dam. This case concerns the more than 150-mile stretch of river between the Canadian border and

6 PAKOOTAS V. TECK COMINCO METALS the Grand Coulee Dam, known as the Upper Columbia River.

From time immemorial, the Upper Columbia River has held great significance to the Confederated Tribes of the Colville Reservation. These tribes historically depended on the River’s plentiful fish for their survival and gave the River a central role in their cultural traditions. 1 And the Colville Tribes continue to use the Upper Columbia River to this day for fishing and recreation. Under the applicable treaties, the Tribes retain fishing rights in the River up to the Canadian border. See Okanogan Highlands All. v. Williams, 236 F.3d 468, 478 (9th Cir. 2000) (citing Antoine v. Washington, 420 U.S. 194, 196 n.4 (1975)). Those treaties draw the Colville Reservation’s eastern and southern boundaries “in the middle of the channel of the Columbia River.” Act of July 1, 1892, ch. 140, § 1, 27 Stat. 62, 62–63. The Tribes claim equitable title to the riverbed on their side of the channel, and the United States has long supported this claim. See Confederated Tribes of Colville Reservation v. United States, 964 F.2d 1102, 1105 n.7 (Fed. Cir. 1992); Opinion on the Boundaries of and Status of Title to Certain Lands Within the Colville and Spokane Indian Reservations, 84 Interior Dec. 72, 75–80, 1977 WL 28859, at *3–5.

For nearly a century, however, the Upper Columbia River has been fouled by Teck Metals’ toxic waste. 2 Teck operates the world’s largest lead and zinc smelter in Trail, British Columbia, just ten miles upstream of the U.S. border. 1 See generally U.S. EPA, Upper Columbia River Expanded Site Inspection Report Northeast Washington, app. A (Petition for Assessment of Release), https://nepis.epa.gov/Exe/ZyPURL.cgi?Docke y=P100MFOQ.TXT.

2 Teck was previously named Teck Cominco Metals.

PAKOOTAS V. TECK COMINCO METALS 7

During smelting, lead or zinc ore is heated to a molten state, during which the desired metal is separated from impurities in the raw ore. These impurities cool to form glassy, granular slag. Between 1930 and 1995, Teck discharged about 400 tons of slag daily—an estimated 9.97 million tons in total—directly into the free-flowing Columbia River. Teck washed this debris into the river using untold gallons of contaminated effluent. These solid and liquid wastes contained roughly 400,000 tons (800 million pounds) of the heavy metals arsenic, cadmium, copper, lead, mercury, and zinc, in addition to lesser amounts of other hazardous substances. ^[3]

At least 8.7 million tons of the Trail smelter’s slag and nearly all of the dissolved and particulate-bound metals in its effluent made the short trip downstream into the United States. Upon reaching the calmer waters of Lake Roosevelt, Teck’s smelting byproducts came to rest on the riverbed and banks, with larger detritus settling upstream and smaller particles settling downstream near the Grand Coulee Dam. 4 Section 107(a)(4)(A) of CERCLA provides that a PRP is liable for “all costs of removal or remedial action incurred by the United States Government or a State or an Indian tribe not inconsistent with the national contingency plan.” 42 U.S.C. § 9607(a)(4)(A). At its core, a “removal” action is defined as “the cleanup or removal” of hazardous

20 PAKOOTAS V. TECK COMINCO METALS

substances from the environment. ^[9] Id. § 9601(23). No less important, however, are several associated activities described by the statutory definition. ^[10] This case concerns two defined categories of related activities: such efforts “as may be necessary to monitor, assess, and evaluate the release or threat of release of hazardous substances,” and “as may be necessary to prevent, minimize, or mitigate damage to the public health or welfare or to the environment.” Id.

Cleanup-adjacent activities face a low bar to satisfying these definitions of “removal.” See United States v. W.R. Grace & Co., 429 F.3d 1224, 1238 (9th Cir. 2005) (“The definition of ‘removal’ is written in sweeping terms.”). Section 101(23) covers all activities “as may be necessary” to advance certain threat assessment or abatement goals. This permissive language means qualifying activities need not be performed with the intent of achieving the statutory goals; need not be absolutely necessary to achieve those goals; and need not actually achieve those goals. Rather, The district court concluded that the investigations by the Tribes’ expert consultants qualify as recoverable costs of removal. To begin with, the Tribes hired an environmental consultant, Environment International, to plan and implement a study of the Upper Columbia River Site. This consultant collected multiple sediment and pore water samples and sent those samples to independent labs for testing. An environmental engineering firm, LimnoTech, then compiled the resulting data into a comprehensive database and analyzed the data. The Tribes also employed several subject-matter experts, such as a geochemist and a metallurgist, to review the data. Finally, the Tribes retained a hydrology firm, Northwest Hydraulic Consultants, to sample and analyze upstream sediment cores from the Canadian reach of the Columbia River.

We agree with the district court that the Tribes’ data collection and analysis efforts were not an unreasonable means of furthering at least three distinct purposes embraced by CERCLA.

First, the expert consultants investigated the presence and movement of toxic wastes at the Site. We have held that section 101(23) encompasses such studies into the location and migration of materials containing hazardous substances. See Wickland Oil Terminals v. Asarco, Inc., 792 F.2d 887,

22 PAKOOTAS V. TECK COMINCO METALS

889, 892 (9th Cir. 1986) (allowing cost recovery for “testing . . . of the migration of slag particles” as an action that “may be necessary to monitor, assess, and evaluate the release or threat of release of hazardous substances”).

Second, the Tribes’ experts tested whether the slag and effluent-contaminated sediment found at the Site leach contaminants into the environment. Section 101(23) on its face covers “asses[ing] . . . [the] threat of release of hazardous substances.” 42 U.S.C. § 9601(23); see also Wickland, 793 F.2d at 889, 892 (allowing cost recovery for “conduct[ing] tests to evaluate the hazard posed by the slag”); Cadillac Fairview/California, Inc. v. Dow Chem. Co., 840 F.2d 691, 692–93, 695 (9th Cir. 1988) (same).

And third, the experts traced the origins of the slag and sediment metals found at the Site. Teck has maintained before and throughout this litigation that many other sources, including other smelters, are to blame for the Upper Columbia River’s pollution. The Tribes commissioned a study investigating this claim, but the results show that the wastes match the Trail smelter’s isotopic and geochemical “fingerprint.”

Efforts to identify the parties responsible for the disposal of toxic wastes at a site are likewise recoverable costs of removal. In Key Tronic Corp. v. United States, 511 U.S. 809 (1994), the Supreme Court considered whether a PRP could recover fees for work performed by an attorney in searching for other parties that had used a site for hazardous waste disposal. Id. at 820. The Court held that “[t]hese kinds of activities are recoverable costs of response clearly` distinguishable from litigation expenses.” Id. Indeed, searches for pollution sources are often conducted by non- lawyers, such as “engineers, chemists, private investigators,

PAKOOTAS V. TECK COMINCO METALS 23 or other professionals”—much like the Tribes’ experts here. Id.

Key Tronic appears to have rested its holding on yet another statutory definition, section 101(25). See id. at 813, 816–20. That provision defines removal and remedial actions collectively as “response” actions, and then defines all “response” actions to “include enforcement activities related thereto.” 42 U.S.C. § 9601(25). The Court in Key Tronic noted that the search in that case had prompted EPA to initiate an administrative enforcement action against another party that had been identified as disposing of wastes at the site. Id. at 820. The Court also found it significant that “[t]racking down other responsible solvent polluters increases the probability that a cleanup will be effective and get paid for.” Id. Although Key Tronic did not discuss section 101(23)’s definition of “removal,” the benefit of making an effective cleanup more likely also falls within the scope of actions identified by the district court that “may be necessary to prevent, minimize, or mitigate damage to the public health or welfare or to the environment.” Similarly, uncovering evidence that a party is responsible for hazardous waste puts pressure on that party voluntarily to clean up its pollution, which would also advance the goals of that provision. Cf. E.I. DuPont de Nemours & Co. v. United States, 508 F.3d 126, 135 (3d Cir. 2007) (“Voluntary cleanups are vital to fulfilling CERCLA’s purpose.”). And under both provisions, CERCLA’s broad remedial purpose “supports a liberal interpretation of recoverable costs” to ensure that polluters pay for the messes they create— including the difficulties of identifying them in the first place. United States v. R.W. Meyer, Inc., 889 F.2d 1497, 1503 (6th Cir. 1989) (quoting United States v. Northernaire Plating Co., 685 F. Supp. 1410, 1419 (W.D. Mich. 1988)).

24 PAKOOTAS V. TECK COMINCO METALS Teck opposes the district court’s conclusion, arguing that the Tribes’ studies implicitly fall out of the statutory definitions of “removal” because they are all “litigation- related.” To be sure, the studies were commissioned after the Tribes joined this litigation; they were undertaken to help prove Teck’s liability; and many of them were presented to the district court in Phase I of trial.

Teck’s argument relies on a pair of decisions from the Third Circuit. In Redland Soccer Club, Inc. v. Dep’t of Army of U.S., 55 F.3d 827 (3d Cir. 1995), the court held that when evaluating the “necessary” costs of response under section 107(a)(4)(B), it looks to “[t]he heart of the[] definitions of removal and remedy” and considers whether the costs are “necessary to the containment and cleanup of hazardous releases.” Id. at 850 (quoting United States v. Hardage, 982 F.2d 1436, 1448 (10th Cir. 1992)). The court then applied this rule in Black Horse Lane Assoc., L.P. v. Dow Chemical Corp., 228 F.3d 275 (3d Cir. 2000), where it held that “private parties may not recoup litigation-related expenses in an action to recover response costs pursuant to section 107(a)(4)(B).” Id. at 294. As Teck points out, the court noted that the work at issue did not “play[] any role in the containment and cleanup of the Property,” which meant it was not “necessary.” Id. at 297.

We conclude that those out-of-circuit cases are not persuasive here. The Colville Tribes bring their cost recovery action as a sovereign under section 107(a)(4)(A), so they are entitled to “all costs” rather than merely the “necessary” costs of response. Compare 42 U.S.C.

PAKOOTAS V. TECK COMINCO METALS 25

§ 9607(a)(4)(A), with id. § 9607(a)(4)(B). 11 And even if the latter standard were applicable, we have never interpreted the term “necessary” as requiring a nexus solely between recoverable costs and on-site cleanup activities. See Carson Harbor Vill., Ltd. v. Unocal Corp., 270 F.3d 863, 871 (9th Cir. 2001) (en banc) (holding that a response action is necessary if it responds to “an actual and real threat to human health or the environment”). We instead read CERCLA’s cost recovery provisions as making no distinction between cleanup and investigatory costs. Wickland, 792 F.2d at 892. Neither case cited by Teck speaks to the issue presented— whether an activity that would otherwise qualify as removal is disqualified by virtue of having a connection to litigation. See Black Horse Lane, 228 F.3d at 298 & n.13 (concluding that “the removal definition . . . exclud[es] the sort of ‘oversight’ costs” sought by plaintiff); Redland Soccer Club, 55 F.3d at 850 (concluding that plaintiffs’ health risk assessment costs are not “‘response costs’ under any of the[] definitions” of “removal” and “remedial”).

Seeing no supportive authorities on point, we decline to adopt Teck’s reading of “removal” as implicitly excluding activities that have a connection to litigation. By its terms, the statute gives no weight to the timing, purpose, or ultimate use of covered activities. See 42 U.S.C. § 9601(23), (25). A plaintiff’s ongoing response action may complicate recovery, but those costs remain recoverable at trial. See Johnson v. James Langley Operating Co., 226 F.3d 957, 963

11 For this reason, we need not decide whether the Tribes’ cost of fingerprinting wastes at the Site was “necessary” in light of the study yielding a “duplicative identification” of Teck as a polluter. Syms v. Olin Corp., 408 F.3d 95, 104 (2d Cir. 2005). But in any case, we cannot fault the Tribes for paying to learn that Teck disposed of these wastes when Teck disputed that the wastes could be traced back to the company rather than to a number of other potential pollution sources.

26 PAKOOTAS V. TECK COMINCO METALS

(8th Cir. 2000) (“[P]laintiffs’ response costs in this case are not transformed into litigation costs merely by their timing with respect to their initiation of this action.”); Matter of Bell Petroleum Servs., Inc., 3 F.3d 889, 908 (5th Cir. 1993) (“With respect to costs, if any, incurred after the complaint was filed, prejudgment interest should be assessed on those costs from the date of the expenditures.”). Further, a plaintiff’s intent to use the fruits of an investigation in litigation does not excise that activity from the statutory definitions of removal. See Johnson, 226 F.3d at 963 (“[T]he motives of the . . . party attempting to recoup response costs . . . are irrelevant.” (quoting Gen. Elec. Co. v. Litton Indus. Automation Sys., Inc, 920 F.2d 1415, 1418 (8th Cir. 1990), abrogated on other grounds by Key Tronic Corp., 511 U.S. 809); cf. Carson Harbor, 270 F.3d at 872 (holding that self-serving “ulterior motive[s]” should be disregarded when determining whether response costs are necessary because “[t]o hold otherwise would result in a disincentive for cleanup”). Many, if not most, CERCLA plaintiffs study the contamination at a site with an eye to potential litigation, and it would make little sense to provide these costs only to parties that are disinclined to file suit. Finally, recoverable investigation costs do not transform into unrecoverable costs if the information obtained is later used to help prove a PRP’s liability. See Vill. of Milford v. K-H Holding Corp., 390 F.3d 926, 935–36 (6th Cir. 2004) (holding that the plaintiff could recover from the defendant the costs of identifying it as a PRP). Indeed, we would turn Key Tronic’s reasoning on its head if we read that opinion as making a defendant liable for all PRP search costs except the cost of identifying that defendant once that evidence is used in the plaintiff’s case in chief. See 511 U.S. at 820 (lauding the plaintiff’s investigation for “uncovering the [defendant’s] disposal of wastes at the site”).

PAKOOTAS V. TECK COMINCO METALS 27

We instead determine whether an activity amounts to “removal” by comparing the actions taken to the categories defined by statute. See, e.g., W.R. Grace & Co., 429 F.3d at 1246–47; Hanford Downwinders Coal., Inc. v. Dowdle, 71 F.3d 1469, 1477–79 (9th Cir. 1995); Durfey v. E.I. DuPont De Nemours & Co., 59 F.3d 121, 124–26 (9th Cir. 1995). The statutory language—not extra-textual factors— is controlling.

We conclude that the district court properly awarded the Colville Tribes all investigation expenses as costs of removal, even though many of these activities played double duty supporting both cleanup and litigation efforts. ^[12]

B

We next consider the district court’s award of the Colville Tribes’ attorney’s fees.

Shortly after CERCLA was enacted, several district courts interpreted section 107(a)(4)(A) to mean that the United States could recover its attorney’s fees for successfully bringing a response costs action. See, e.g., United States v. Ne. Pharm. & Chem. Co. (NEPACCO), 579 F. Supp. 823, 851 (W.D. Mo. 1984), aff’d in part and Teck contends that Chapman does not apply here because its holding is tied to the specific facts of that case. In Chapman, EPA ordered the defendant to remove

PAKOOTAS V. TECK COMINCO METALS 31

hazardous substances from the site, and when the defendant failed to comply, EPA itself initiated a response action. 146 F.3d at 1168–69. EPA then requested repayment for its response costs, and only after the defendant refused to pay did the United States bring a response costs action. Id. at 1169. Teck maintains that the Tribes’ response costs action is distinguishable because it is “not premised on a refused order or a refusal to fund response costs.”

We disagree. Neither background fact identified by Teck was material to the outcome in Chapman. See id. at 1173–76. Litigation may not be necessary if a defendant is cooperative, but CERCLA does not limit a government’s recovery of attorney’s fees just to those response costs actions that are absolutely unavoidable. And we follow the other circuits that have considered this issue, which have held that a government’s response costs action amounts to an “enforcement activit[y]” without so much as mentioning a requirement that there first be a disobeyed cleanup order or an unsuccessful repayment negotiation. See United States v. Dico, Inc., 266 F.3d 864, 878 (8th Cir. 2001); B.F. Goodrich v. Betkoski, 99 F.3d 505, 528, 530 (2d Cir. 1996), overruled on other grounds by United States v. Bestfoods, 524 U.S. 51 (1998); see also Reardon, 947 F.2d at 1514 (“[I]f ‘enforcement activities’ in § 9601(25) is interpreted to exclude the expenses of cost recovery actions, this would have the effect of denying the government significant amounts of attorney’s fees—which was certainly not the intent of Congress.”).

Because this case is squarely governed by Chapman, we conclude that the Colville Tribes are entitled to collect their reasonable attorney’s fees for prevailing in their response costs action against Teck. See 146 F.3d at 1176; see also Fireman’s Fund, 302 F.3d at 953.

32 PAKOOTAS V. TECK COMINCO METALS Teck also tries to evade the significance of Chapman by raising several novel challenges to the district court’s award of attorney’s fees.

First, Teck asserts that the Tribes do not have the requisite “enforcement authority” to recover the costs of any enforcement activities connected with the Upper Columbia River Site. Teck reasons that the Tribes lack the response authority bestowed on the federal government by section 104, 42 U.S.C. § 9604, which Teck claims that EPA can— but here did not—“delegate” to a state, political subdivision, or Indian tribe under section 104(d)(1)(A), id. § 9604(d)(1)(A). But this provision is irrelevant. Section 104(d)(1)(A) does not address delegation at all; it simply “authorizes EPA to enter into cooperative agreements or contracts with a state, political subdivision, or a federally recognized Indian tribe to carry out [Superfund]-financed response actions.” 40 C.F.R. § 300.515(a)(1). EPA’s regulations explain that the agency “use[s] a cooperative agreement to transfer funds”—not federal authority—“to those entities to undertake Fund-financed response activities.” Id. And in any event, the enforcement authority at issue is whether the Tribes can bring a lawsuit to recover their response costs. As Teck conceded at oral argument, the Tribes “clearly can bring a claim for recovery of response costs” under section 107(a)(4)(A), so they have all the authority needed to “enforce [this] liability provision.” Reardon, 947 F.2d at 1512–13; see also Washington State Dep’t of Transp. v. Washington Nat. Gas Co., Pacificorp, 59 F.3d 793, 801 (9th Cir. 1995) (“States [and tribes] need not obtain EPA authorization to clean up hazardous waste sites and recover costs from potentially responsible parties.”).

PAKOOTAS V. TECK COMINCO METALS 33

Teck next contends that the Tribes cannot recover their attorney’s fees because this case is not “related to” any response action at the Site, as required by section 101(25). In another statutory context, the Supreme Court has explained that the “ordinary meaning of [the] words ‘related to’ is a broad one,” meaning “having a connection with or reference to,” though that breadth “does not mean the sky is the limit.” Dan’s City Used Cars, Inc. v. Pelkey, 569 U.S. 251, 260 (2013) (alterations omitted) (quoting Rowe v. New Hampshire Motor Transp. Ass’n, 552 U.S. 364, 370 (2008)). Adopting that standard here, we conclude that an enforcement activity falls outside of section 101(25) only if it has an inadequate connection with an existing or potential response action at a given site. Although some enforcement activities can be conducted only after a response action has begun, some can be conducted beforehand. For instance, a cash-strapped property owner may wish to locate solvent polluters to split the tab before incurring response costs, and EPA may well review and approve a party’s cleanup plans before any response activities are conducted. See, e.g., Key Tronic, 511 U.S. at 820 (covering PRP searches); United States v. E.I. Dupont De Nemours & Co. Inc., 432 F.3d 161, 163, 173 (3d Cir. 2005) (en banc) (covering EPA’s review, approval, and monitoring of proposed cleanup activities). Nothing in section 101(25)’s text or the case law interpreting it requires one activity to come before the other for them to be related. The Tribes have conducted investigative activities during the course of this litigation, so the district court correctly held that this response costs suit is “related to” a response action at the Site.

Last, Teck takes issue with the attorney’s fees associated with the Tribes’ declaratory judgment claim. CERCLA provides that any court awarding response costs in a section 107(a) action “shall enter a declaratory judgment on liability

34 PAKOOTAS V. TECK COMINCO METALS

for response costs . . . that will be binding on any subsequent action or actions to recover further response costs.” 42 U.S.C. § 9613(g)(2). As a result, the declaration of Teck’s liability for future response costs is simply an additional form of relief that the Tribes obtained through the same efforts underlying their successful response costs action. See City of Colton v. Am. Promotional Events, Inc.- W., 614 F.3d 998, 1007 (9th Cir. 2010). Teck responds that declaratory relief did not need to be granted to compel Teck to fund a response action, but this mandatory relief does not require a showing of necessity. Regardless of whether future response costs are speculative—or even, as Teck insists, affirmatively unlikely—CERCLA requires that a successful plaintiff in a section 107(a) action be awarded both response costs and declaratory relief. See 42 U.S.C. § 9613(g)(2).

Teck also challenges the reasonableness of the attorney’s fees award under the standard set forth in Hensley v. Eckerhart, 461 U.S. 424 (1983). Teck contends that if we agree that the Tribes were not entitled to any costs of removal, then we should conclude that the district court misjudged the degree of the Tribes’ success. But we do not agree with Teck’s premise, so we reject its conclusion. The district court did not abuse its discretion in finding the $4.86 million attorney’s fees award to be reasonably proportionate to the properly awarded $3.39 million for investigation expenses. See Webb v. Ada Cty., 285 F.3d 829, 837 (9th Cir. 2002). The ratio between attorney’s fees and the degree of success obtained is also reasonable when one considers that the Tribes earned a valuable declaratory judgment, which “confer[s] substantial benefits not measured by the amount of damages awarded.” Hyde v. Small, 123 F.3d 583, 584 (7th Cir. 1997); see also In re Dant

PAKOOTAS V. TECK COMINCO METALS 35

& Russell, Inc., 951 F.2d 246, 249–50 (9th Cir. 1991) (noting that CERCLA plaintiffs often “spend some money responding to an environmental hazard” and then bring a response cost action to recover their “initial outlays” and to obtain “a declaration that the responsible party will have continuing liability for the cost of finishing the job”).

In sum, we conclude that the district court properly awarded the Colville Tribes their attorney’s fees, and we do not disturb the finding that approximately $4.86 million is a reasonable award in this case.

IV

The final question presented is whether the district court erred in granting summary judgment on Teck’s divisibility defense to joint and several liability. ^[13]

We review the district court’s grant of summary judgment de novo, and we may affirm on any basis supported by the record. Kohler v. Bed Bath & Beyond of California, LLC, 780 F.3d 1260, 1263 (9th Cir. 2015). Viewing the evidence in the light most favorable to the nonmoving party, we must determine whether there is “no genuine dispute as to any material fact,” Fed. R. Civ. P. 56(a), and whether the district court correctly applied the relevant substantive law, see Devereaux v. Abbey, 263 F.3d 1070, 1074 (9th Cir. 2001) (en banc).

CERCLA liability is ordinarily joint and several, except in the rare cases where the environmental harm to a site is shown to be divisible. United States v. Coeur d’Alenes Co., 767 F.3d 873, 875 (9th Cir. 2014); see also Martha L. Judy, Coming Full CERCLA: Why Burlington Northern Is Not the Sword of Damocles for Joint and Several Liability, 44 New Eng. L. Rev. 249, 283 (2010) (counting only four decisions finding divisibility out of 160 cases).

In Burlington Northern, the Supreme Court confirmed that “‘[t]he universal starting point for divisibility of harm analyses in CERCLA cases’ is § 433A of the Restatement (Second) of Torts.” Burlington N. & Santa Fe Ry. Co. v. United States, 556 U.S. 599, 614 (2009) (Burlington Northern II) (quoting United States v. Hercules, Inc., 247 F.3d 706, 717 (8th Cir. 2001)). Under the Restatement,

PAKOOTAS V. TECK COMINCO METALS 39

“when two or more persons acting independently cause a distinct or single harm for which there is a reasonable basis for division according to the contribution of each, each is subject to liability only for the portion of the total harm that he has himself caused.” Id. (quoting United States v. Chem- Dyne Corp., 572 F. Supp. 802, 810 (S.D. Ohio 1983)) (alteration omitted). “But where two or more persons cause a single and indivisible harm, each is subject to liability for the entire harm.” Id. (quoting Chem-Dyne, 572 F. Supp. at 810).

The divisibility analysis involves two steps. First, the court considers whether the environmental harm is theoretically capable of apportionment. See Restatement (Second) of Torts § 434 cmt. d. This is primarily a question of law. See United States v. Burlington N. & Santa Fe Ry. Co., 520 F.3d 918, 942 (9th Cir. 2008) (Burlington Northern I), rev’d on other grounds, 556 U.S. 599 (2009); United States v. NCR Corp., 688 F.3d 833, 838 (7th Cir. 2012); Hercules, 247 F.3d at 718; Bell Petroleum, 3 F.3d at 896. Underlying this question, however, are certain embedded factual questions that must necessarily be answered, such as “what type of pollution is at issue, who contributed to that pollution, how the pollutant presents itself in the environment after discharge, and similar questions.” NCR, 688 F.3d at 838. Second, if the harm is theoretically capable of apportionment, the fact-finder determines whether the record provides a “reasonable basis” on which to apportion liability, which is purely a question of fact. Restatement (Second) of Torts §§ 433A(1)(b), 434 cmt. d; see also Burlington Northern II, 566 U.S. at 615; NCR, 688 F.3d at 838; Hercules, 247 F.3d at 718; Bell Petroleum, 3 F.3d at 896.

40 PAKOOTAS V. TECK COMINCO METALS

At both steps, the defendant asserting the divisibility defense bears the burden of proof. See Restatement (Second) of Torts § 433B(2); see also Burlington Northern II, 556 U.S. at 614; NCR, 688 F.3d at 838. This burden is “substantial” because the divisibility analysis is “intensely factual.” United States v. Alcan Aluminum Corp., 964 F.2d 252, 269 (3d Cir. 1992) (Alcan-Butler). The necessary showing requires a “fact-intensive, site-specific” assessment, PCS Nitrogen Inc. v. Ashley II of Charleston LLC, 714 F.3d 161, 182 (4th Cir. 2013), generating “concrete and specific” evidence, Hercules, 247 F.3d at 718. But that is not to say that the defendant’s proof must rise to the level of absolute certainty. See Burlington Northern II, 556 U.S. at 618. Rather, the defendant must show by a preponderance of the evidence—including all logical inferences, assumptions, and approximations—that there is a reasonable basis on which to apportion the liability for a divisible harm. See Restatement (Second) of Torts § 433A cmt. d; see also, e.g., Hercules, 247 F.3d at 719; Bell Petroleum, 3 F.3d at 904 n.19.

In the context of a motion for summary judgment, however, the burdens operate somewhat differently. Teck’s answer pleaded divisibility as an affirmative defense for which Teck would bear the burden of proof at trial. ^[14] To defeat this affirmative defense on summary judgment, the Colville Tribes and the State of Washington took on both the Teck counters that the first question on the motions for summary judgment is whether the alleged harm could be divided “under any set of facts,” which would mean Teck had no burden of production on the overall harm.

We disagree. Even on a Rule 12(b)(6) motion to dismiss—that is, before discovery—a non-moving party is

42 PAKOOTAS V. TECK COMINCO METALS

held to more than an “any set of facts” standard. See Bell Atl. Corp. v. Twombly, 550 U.S. 544, 562–63 (2007). It is not the court’s job to envision hypothetical scenarios in which a mix of pollution from multiple sources could potentially be divisible. Rather than relying on judicial imagination, Teck was required to “make a showing sufficient to establish the existence of an element essential to” its divisibility defense: that the harm is theoretically capable of division. Celotex, 477 U.S. at 322.

Teck then argues that, at most, its burden of production extended only to addressing the harm from the specific pollutants that Teck is alleged to have contributed to the Site. In the operative complaints, the Tribes and the State sought “the costs of remedial or removal actions, natural resource damage assessment costs, and natural resource damages that [plaintiffs] have incurred and will continue to incur at the Upper Columbia River and Lake Roosevelt where hazardous substances have come to be located.” The district court read these pleadings as alleging a harm caused by “all of the hazardous substances released or threatened to be released from the Site, from whatever source.” But in Teck’s view, the harm pleaded is impliedly limited to the six hazardous substances alleged to have originated from the Trail smelter, so Teck contends that it can disregard all other types of pollution found with its wastes at the Site.

The environmental harm in this case is not so limited. Section 107(a) imposes strict liability on all PRPs, even if those persons are in fact not responsible for any pollution at all. United States v. Atl. Research Corp., 551 U.S. 128, 136 (2007). That is because “Congress has . . . allocated the burden of disproving causation to the defendant who profited from the generation and inexpensive disposal of hazardous

PAKOOTAS V. TECK COMINCO METALS 43

waste.” Monsanto, 858 F.2d at 170. It certainly is not always an easy task to determine the entire extent of contamination at a site. See NCR, 688 F.3d at 841. The Restatement makes clear, however, that “[a]s between the proved tortfeasor who has clearly caused some harm, and the entirely innocent plaintiff, any hardship due to lack of evidence as to the extent of the harm should fall upon the former.” Restatement (Second) of Torts § 433B cmt. d.

In line with CERCLA’s pleading requirements, the complaints here identified six of Teck’s pollutants just to establish the company’s liability. The complaints cannot be fairly read as needlessly narrowing this suit to recovery for harm caused solely by those pollutants. As a result, Teck was required to produce evidence showing divisibility of the entire harm caused by Teck’s wastes combined with all other River pollution—not just the harm from sources of Teck’s six metals alone. 15

C

With the standards of review thus established, we turn to evaluating the evidence submitted on summary judgment.

The district court primarily granted summary judgment on the ground that Teck did not have enough evidence to show that the harm at issue is theoretically capable of apportionment. The court reasoned that Teck’s evidence

15 Teck does not contend, nor does the record reflect, that Teck’s heavy metals formed an area of pollution that was distinct from areas with non-metal pollutants. And that would be an argument for apportioning liability based on distinct harms, not a single divisible harm. See Restatement (Second) of Torts § 433A(1).

44 PAKOOTAS V. TECK COMINCO METALS

could not establish divisibility because it failed to account for the entire harm at the Site. Reviewing the parties’ submissions de novo, we agree that there was no genuine dispute of fact for trial on the question whether the harm to the Upper Columbia River is theoretically capable of apportionment.

At the first step of the divisibility analysis, a court cannot say whether a harm “is, by nature, too unified for apportionment” without knowing certain details about the “nature” of the harm. Burlington Northern I, 520 F.3d at 942, rev’d on other grounds, 556 U.S. 599 (2009); see also Bell Petroleum, 3 F.3d at 895 (“The nature of the harm is the key factor in determining whether apportionment is appropriate.”). As one commentator has explained: “Even if a party’s waste stream can be separately accounted for, its effect on the site and on other parties’ wastes at the site must also be taken into account.” William C. Tucker, All Is Number: Mathematics, Divisibility and Apportionment Under Burlington Northern, 22 Fordham Envtl. L. Rev. 311, 316 (2011). That is, “a defendant must take into account a number of factors relating not just to the contribution of a particular defendant to the harm, but also to the effect of that defendant’s waste on the environment.” Id. Those factors generally include when the pollution was discharged to a site, where the pollutants are found, how the pollutants are presented in the environment, and what are the substances’ chemical and physical properties. See NCR, 688 F.3d at 838. Chief among the relevant properties are “the relative toxicity, migratory potential, degree of migration, and synergistic capacities of the hazardous substances at the site.” United States v. Alcan Aluminum Corp., 990 F.2d 711, 722 (2d Cir. 1993) (Alcan-PAS).

PAKOOTAS V. TECK COMINCO METALS 45

Teck’s divisibility expert identified hundreds of heavy metal sources that may have contributed to Upper Columbia River’s pollution throughout its watershed over the course of more than a century. At Teck’s direction, however, Dr. Johns expressly curtailed his divisibility analysis to the six hazardous substances allegedly “attributable to Teck.” But Teck did not claim that these were the only pollutants found at the Site.

Both the Tribes and the State pointed out this deficiency in their motions for summary judgment. The Tribes cited evidence of the Site containing the hazardous substances antimony, beryllium, chromium, nickel, radon, selenium, thallium, 2,3,7,8-tetrachlorodibenzo-pdioxin, polycyclic aromatic hydrocarbons (“PAHs”), polychlorinated biphenyls (“PCBs”), and DDTs. And one of the State’s experts submitted a declaration stating that EPA was evaluating the Site for around 199 contaminants of concern, including PAHs, PCBs, dioxins and furans, and pesticides. This declaration further showed that sediment samples found Teck’s metals physically mixed with other hazardous substances in the northern stretches of the Site. Zinc, for example, “was detected with other metals like antimony, arsenic, cadmium, copper, mercury, and lead, and also in several instances with up to 14 reported organic PAH chemicals present, as well as less frequently with pesticides like 2,4-DDT, 4,4 DDE, and 4,4-DDT.”

Despite this evidence, Teck’s opposition to the motions for summary judgment continued to rely on Dr. Johns’s limited analysis. Teck reiterated its assumption that the Site’s harm was solely traceable to the specific metals that Teck discharged. While conceding that its slag was “co- located” with “other slag and tailings,” Teck made no mention of its pollutants being found alongside non-metal

46 PAKOOTAS V. TECK COMINCO METALS

pollutants. And Teck relied on Dr. Johns’s view that if Teck’s slag “is not leaching,” as he believed, then “the location of the slag in sediment is irrelevant to the apportionment analysis.”

On these points Teck erred. At the outset, Teck repeatedly misapprehended the harm here. For the purpose of apportioning CERCLA liability, the relevant “harm” is the entirety of contamination at a site that has caused or foreseeably could cause a party to incur response costs, suffer natural resource damages, or sustain other types of damages cognizable under section 107(a)(4). See, e.g., Burlington Northern II, 556 U.S. at 618 (suggesting that the harm is “the overall site contamination requiring remediation” in a response cost action); NCR, 688 F.3d at 840–41 (“[T]he underlying harm caused [is] the creation of a hazardous, polluted condition . . . .”); Burlington Northern I, 520 F.3d at 939 (holding that each share of liability for the harm is “the contamination traceable to each defendant”), rev’d on other grounds, 556 U.S. 599 (2009); Chem-Nuclear Sys., Inc. v. Bush, 292 F.3d 254, 259 (D.C. Cir. 2002) (“[T]he harm at issue was the release or threatened release of hazardous substances into groundwater . . . .” (internal quotation marks omitted)).

Dr. Johns instead based his apportionment methods on three inconsistent notions of the Site’s harm: (1) “the extent of sediment contamination by hazardous substances released at the Site”; (2) “harm [to] the river,” namely “the surface water”; and (3) “the placement of hazardous substances” at the Site. Dr. Johns’s first and second measures of the harm are incomplete because they look only to the actual releases of hazardous substances from toxic wastes at the Site, ignoring the fact that wastes with a “threatened release of hazardous substances” are likewise contamination that could

PAKOOTAS V. TECK COMINCO METALS 47

give rise to response costs. Chem-Nuclear Sys., 292 F.3d at 259 (emphasis added); see also 42 U.S.C. § 9607(a)(4). Further, the second measure excludes contamination deeper than five centimeters, even though remedial activities like dredging would obviously need to excavate these materials too. Only Dr. Johns’s third apportionment method—the approach that he sketched briefly in his deposition rather than outlining in his detailed report—correctly recognized that the presence of contaminants throughout the Site is the relevant harm.

More importantly, all of Dr. Johns’s analysis overlooked the fact that “the mixing of the wastes raises an issue as to the divisibility of the harm.” Chem-Dyne, 572 F. Supp. at 811. Mixing of pollutants “is not synonymous with indivisible harm,” Alcan-PAS, 990 F.2d at 722, but it does create a rebuttable presumption of such harm, see id.; see also Monsanto, 858 F.2d at 172; Chem-Dyne, 572 F. Supp. at 811. The State put this presumption at issue by submitting evidence of Teck’s metals being found with unrelated pollutants, yet Teck chose not to address the potential for synergistic harm from these pollution hotspots.

Teck responds that the only relevant synergistic effects are from substances that are chemically commingled, not just physically interspersed. To that end, Dr. Johns opined that Teck’s slag cannot chemically interact with other substances based on his understanding that the slag does not leach pollutants.

We are not persuaded. Even if pollutants do not chemically interact, their physical aggregation can cause disproportionate harm that is not linearly correlated with the amount of pollution attributable to each source. In Monsanto, a key case addressing chemical commingling, the Fourth Circuit explained: “Common sense counsels that a

48 PAKOOTAS V. TECK COMINCO METALS

million gallons of certain substances could be mixed together without significant consequences, whereas a few pints of others improperly mixed could result in disastrous consequences.” 858 F.2d at 172. Also common sense, however, is the old adage that sometimes dilution is the solution to pollution. See, e.g., Carol M. Browner, Environmental Protection: Meeting the Challenges of the Twenty-First Century, 25 Harv. Envtl. L. Rev. 329, 331 (2001). For example, “[i]f several defendants independently pollute a stream, the impurities traceable to each may be negligible and harmless, but all together may render the water entirely unfit for use.” W. Keeton et al., Prosser and Keeton on Law of Torts § 52, p. 354 (5th ed. 1984). The Second Circuit thus allowed a PRP to be apportioned no liability if “its pollutants did not contribute more than background contamination and also cannot concentrate,” provided that there were no EPA thresholds below those ambient contaminant levels. Alcan-PAS, 990 F.2d at 722. And the Third Circuit has held that “the fact that a single generator’s waste would not in itself justify a response is irrelevant . . . , as this would permit a generator to escape liability where the amount of harm it engendered to the environment was minimal, though it was significant when added to other generators’ waste.” Alcan-Butler, 964 F.2d at 264.

Without knowing more about the accumulation of Teck’s wastes with unrelated pollutants, with like materials, and by themselves, a court could not tell whether “their presence is harmful and the River must be cleaned.” NCR, 688 F.3d at 840. That question is particularly important here because the most likely remedy for the Site will involve cleaning up some, but not all, of the contaminants in the 150- mile long stretch of river. See 40 C.F.R. § 300.430(f)(1)(ii)(D) (requiring EPA to select a cost-

PAKOOTAS V. TECK COMINCO METALS 49

effective remedy). More intensive remediation will no doubt be prioritized where the level of contamination, and the accompanying danger, is the greatest.

In conclusion, once the State identified mixing of Teck’s metals with non-metal pollutants, Teck was required to rebut the presumption that these pollution hotspots caused greater harm than the sum of the individual pollutants, each of which may be so widely dispersed as to be harmless on its own. Teck did not carry its burden of showing that the harm is theoretically capable of apportionment by simply “considering the effects of its waste in isolation from the other contaminants at a site.” United States v. Alcan Aluminum Corp., 315 F.3d 179, 187 (2d Cir. 2003) (Alcan- Consolidated).

On a related issue concerning the significance of the buildup of slag, we again reject Teck’s contentions. Contrary to Dr. Johns’s mistaken assumption, the buildup of Teck’s slag with other metal-bearing slag or tailings and even on its own affects the extent of the harm. Disproportionate harm can occur whether or not the slag actively leaches pollutants because, as mentioned, the mere threat of leaching can prompt a response action, and the accumulation of materials that pose a potential risk makes a response action more likely. See 42 U.S.C. § 9607(a)(4); Chem-Nuclear Sys., 292 F.3d at 259. Teck responds that Dr. Johns’s declaration at least creates a disputed issue of fact on this point that precludes summary judgment, but in light of the statutory scheme, no rational trier of fact could believe this unsupported assumption that the distribution of the slag is irrelevant. See Matsushita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574, 587 (1986). And because Teck’s slag itself contains a mixture of pollutants, Teck also had to proffer evidence that the clustering of these pollutants did

50 PAKOOTAS V. TECK COMINCO METALS

not create disproportionate environmental harm. No reasonable factfinder could otherwise assume, as Dr. Johns’s apportionment methods require, that rocks and sand from landslides and erosion, for example, are candidates for remediation on par with Teck’s toxic slag. See id.

Finally, because the divisibility of the Upper Columbia River’s contamination turns on the specific facts of that contamination, Teck is also mistaken in arguing that river pollution is categorically divisible under the Restatement. See NCR, 688 F.3d at 838. Besides, the Restatement provides dueling examples of river pollution, and the types of harm for which section 107(a) provides damages—and which the Tribes seek—are more akin to the illustration of an indivisible harm than a divisible harm. Compare Restatement (Second) of Torts § 433A cmt. i, illus. 15 (river pollution poisoning animals is indivisible), with id. cmt. d, illus. 5 (river pollution depriving a riparian owner of the use of water for industrial purposes is divisible). The Seventh Circuit reached the same conclusion in NCR, writing: “The problem here is not that downstream factories were prevented from using the [river] for some period, but that wholly apart from water usage, a toxic chemical in the water causes significant and widespread health problems in both animals and in humans.” 688 F.3d at 842.

We hold that Teck did not make a sufficient showing to establish that liability for environmental harm to the Site is theoretically capable of apportionment. We fully agree with the district court that “because [Teck] has failed to account for all of the harm at the [Upper Columbia River] Site, it cannot prove that harm is divisible.” And to borrow the apt words of Alcan-Consolidated, a case involving a defendant- appellant not carrying its burden of production at trial rather than on a motion for summary judgment,

PAKOOTAS V. TECK COMINCO METALS 51

appellant did not satisfy its substantial burden with respect to divisibility because it failed to address the totality of the impact of its waste at [the Site]; it ignored the likelihood that the cumulative impact of its waste [mixture] exceeded the impact of the [mixture’s] constituents considered individually, and neglected to account for the [mixture’s] . . . physical interaction with other hazardous substances already at the site.

315 F.3d at 187. Although Teck must only produce evidence sufficient to create a genuine issue of material fact at the summary judgment stage, for the reasons stated above, it has not done so here.

As an additional ground for summary judgment, the Tribes and the State argued that Teck did not have enough evidence to show a reasonable basis for apportioning liability. The district court briefly considered this argument and again sided with the plaintiffs on the ground that Teck did not show that the chosen proxy—volume of hazardous substances deposited in the Upper Columbia River—was proportional to the environmental harm. We agree that the lack of a reasonable factual basis for apportioning Teck’s liability provides yet another reason for upholding the district court’s grant of summary judgment on Teck’s divisibility defense.

A defendant asserting a divisibility defense must show that “there is a reasonable basis for determining the contribution of each cause to a single harm.” Burlington Northern II, 556 U.S. at 614 (quoting Restatement (Second) of Torts § 433A(1)(b)). What is reasonable in one case may

52 PAKOOTAS V. TECK COMINCO METALS

not be in another, so apportionment methods “vary tremendously depending on the facts and circumstances of each case.” Hercules, 247 F.3d at 717. Still, the basis for apportionment may rely on the “simplest of considerations,” most commonly volumetric, chronological, or geographic factors. Burlington Northern II, 556 U.S. at 617–18 (quoting Burlington Northern I, 520 F.3d at 943). The only requirement is that the record must support a “reasonable assumption that the respective harm done is proportionate to” the factor chosen to approximate a party’s responsibility. Bell Petroleum, 3 F.3d at 896, 903 (quoting Restatement (Second) of Torts § 433A cmt. d).

Here, no rational trier of fact could find that Teck has provided a reasonable basis for apportionment. All three of Dr. Johns’s apportionment methods are variants of a volumetric approach in that they are premised on an estimate of the mass of pollutants at the Site. But as the Fourth Circuit has noted, “[v]olumetric contributions provide a reasonable basis for apportioning liability only if it can be reasonably assumed, or it has been demonstrated, that independent factors had no substantial effect on the harm to the environment.” Monsanto, 858 F.2d at 172 n.27. Teck “presented no evidence, however, showing a relationship between waste volume . . . and the harm at the site.” Id. at 172. Instead, the available record undercuts the reasonableness of Teck’s assuming a proportional relationship between waste volume alone and the Site’s contamination, for two main reasons.

First, as the Tribes point out, Teck’s evidence shows that geographic factors clearly affected the river’s contamination throughout this massive site. The Trail smelter’s pollution entered the Upper Columbia River at the international border and, according to Dr. Johns, Teck’s slag deposits extend only

PAKOOTAS V. TECK COMINCO METALS 53

45 river miles south. But Dr. Johns accounted for the potential contribution of metals from sources as far as 150 miles downriver, many of which were concentrated at more than ten different confluences between the River and its tributaries. Further, conditions varied greatly throughout the Site; the River is free flowing close to the Canadian border, causing less sediment to accumulate, but it eventually slows and forms Lake Roosevelt, preserving more sediment. As discussed above, these differences in pollution hotspots will doubtless entail varying remediation needs and injuries to the natural environment. See Hercules, 247 F.3d at 717. But even if the harm from those hotspots is capable of division, the fact that contamination strongly correlates with geography means that this is an independent factor that substantially affects the environmental harm at issue. Any proxy for the harm that did not account for geography thus could not be found reasonable.

Second, Teck’s evidence also shows that the passage of time could have a substantial impact on the river’s contamination given the long time period under consideration. Dr. Johns accounted for materials deposited into the Columbia River from the late 1800s through the present. He testified in his deposition that over time, the accumulation of new sediment could bury old contaminants, and in his declaration he said that remediation is not needed if contaminants are buried beneath at least five centimeters of sediment. Further, Dr. Johns acknowledged that over time, slag may slowly release—and thus lose—hazardous substances to the surrounding environment. The upshot is that older wastes may present less of a need for cleanup than more recently disposed wastes. On this record, no reasonable fact-finder could assume that the time at which wastes entered the River is irrelevant to determining the extent of harmful contamination at the Site.

54 PAKOOTAS V. TECK COMINCO METALS

Other independent factors could also affect the environmental harm here, but were similarly ignored by Teck. To take a ready example, some pollutants in the Upper Columbia River may be more toxic than others, like lead compared to zinc. And pollutants may have different migratory potentials based on the media in which they are deposited, such as glassy slag, powdery tailings, or suspended particulates. See Monsanto, 858 F.2d at 173 n.26; see also, e.g., United States v. Manzo, 279 F. Supp. 2d 558, 572–73 (D.N.J. 2003) (rejecting a volumetric apportionment theory where the defendants did not account for relative toxicity and migratory potential).

Absent evidence of how these factors affected the contamination of the Site, any apportionment would have been arbitrary. The district court properly “refused to make an arbitrary apportionment for its own sake.” Burlington Northern II, 556 U.S. at 614–15 (quoting Restatement (Second) of Torts § 433A cmt. i). But Teck of course can always bring a contribution action under section 113(f), 42 U.S.C. § 9613(f), against other pollution sources it identified, which “mitigates any inequity arising from the unavailability of apportionment.” PCS Nitrogen, 714 F.3d at 182.

In holding that Teck did not carry its burden of production, we do not mean to suggest that Teck had to rush the ongoing RI/FS and exhaustively document every contaminant at the Site to save its divisibility defense from summary judgment. That was not required. What was required, however, was that Teck survey the Site, “comprehensively and persuasively address the effects of its waste,” and come up with an apportionment method that a rational trier of fact could find reasonable. Alcan- Consolidated, 315 F.3d at 187. Teck did not do so here.

PAKOOTAS V. TECK COMINCO METALS 55

V

For the foregoing reasons, we affirm the district court’s judgment holding Teck jointly and severally liable for the Colville Tribes’ costs of response.

AFFIRMED.