Cal. Dep't of Toxic Subs. Control v. NL Indus.

Full title: CALIFORNIA DEPARTMENT OF TOXIC SUBSTANCES CONTROL et al. v. NL INDUSTRIES…

Court: United States District Court, C.D. California

Date published: Oct 20, 2022

636 F. Supp. 3d 1092 (C.D. Cal. 2022)

Opinion

Case No. 2:20-cv-11293-SVW-JPR

10-20-2022

CALIFORNIA DEPARTMENT OF TOXIC SUBSTANCES CONTROL et al. v. NL INDUSTRIES, INC. et al.

David Zaft, Jon Stephen Tangonan, Aarti Sujan Kewalramani, Dennis A. Ragen, Elizabeth B. Rumsey, CAAG - Office of the Attorney General California Department of Justice, San Diego, CA, Adam M. Shapiro, Matthew K. Edling, Timothy R. Sloane, Yumehiko Hoshijima, Sher Edling LLP, San Francisco, CA, Adrianna L. Lobato, Scott Justin Lichtig, CAAG - Office of Attorney General California Department of Justice, Sacramento, CA, Christopher T. Nidel, Pro Hac Vice, Nidel and Nace PLLC, Rockville, MD, David P. Page, Pro Hac Vice, Environmental Energy and Natural Resources Advocates PLLC, Tulsa, OK, Donald A. Robinson, Kate Marie Hammond, Sarah E. Morrison, CAAG - Office of Attorney General of California California Department of Justice, Los Angeles, CA, Thomas M. Sims, Nidel and Nace PLLC, Washington, DC, Timothy Eugene Sullivan, California Department of Justice, Oakland, CA, for California Department of Toxic Substances Control. Sarah E. Morrison, Los Angeles, CA, Donald A. Robinson, Kate Marie Hammond, CAAG - Office of Attorney General of California California Department of Justice, Los Angeles, CA, Aarti Sujan Kewalramani, Dennis A. Ragen, Elizabeth B. Rumsey, CAAG - Office of the Attorney General California Department of Justice, San Diego, CA, Adam M. Shapiro, Matthew K. Edling, Timothy R. Sloane, Yumehiko Hoshijima, Sher Edling LLP, San Francisco, CA, Adrianna L. Lobato, Scott Justin Lichtig, CAAG - Office of Attorney General California Department of Justice, Sacramento, CA, Christopher T. Nidel, Pro Hac Vice, Nidel and Nace PLLC, Rockville, MD, David P. Page, Pro Hac Vice, Environmental Energy and Natural Resources Advocates PLLC, Tulsa, OK, Thomas M. Sims, Nidel and Nace PLLC, Washington, DC, Timothy Eugene Sullivan, California Department of Justice, Oakland, CA, for The Toxic Substances Control Account. Joel L. Herz, Pro Hac Vice, Law Office of Joel L. Herz, Tucson, AZ, Kenneth A. Ehrlich, Sean Andrew McCormick, Elkins Kalt Weintraub Reuben Gartside LLP, Los Angeles, CA, for NL Industries, Inc. David C. Allen, Joel R. Meyer, Joseph M. Wahl, Barnes and Thornburg LLP, Los Angeles, CA, Dan Terzian, Erick Kees Kuylman, Thomas D. Warren, Warren Terzian LLP, Los Angeles, CA, Andrew J. Jarzyna, Pro Hac Vice, Nicholas J. Secco, Pro Hac Vice, Benesch Friedlander Coplan and Aronoff LLP, Chicago, IL, Gregory T. Frohman, Pro Hac Vice, Benesch Friedlander Coplan and Aronoff LLP, Cleveland, OH, Krista M. Enns, Lily A. North, Benesch Friedlander Coplan and Aronoff LLP, San Francisco, CA, for Gould Electronics Inc. Richard Alan Dongell, Scott L. Hengesbach, Murchison and Cumming LLP, Los Angeles, CA, Elizabeth Burnett, Clark Hill, Los Angeles, CA, Eric P. Weiss, Scali Rasmussen PC, Los Angeles, CA, for Kinsbursky Bros. Supply, Inc. Jeffrey J. Parker, Zachary Michael Norris, Stephen J. O'Neil, Sheppard Mullin Richter and Hampton LLP, Los Angeles, CA, for Trojan Battery Company, LLC. James C. MacDonald, Law Offices of James MacDonald, Irvine, CA, for Ramcar Batteries Inc. Bina R. Reddy, Pro Hac Vice, Beveridge and Diamond PC, Austin, TX, Eric L. Klein, Pro Hac Vice, Beveridge and Diamond PC, Boston, MA, Gary J. Smith, Beveridge and Diamond PC, San Francisco, CA, Jessica L. Kyle, Pro Hac Vice, Beveridge and Diamond, Baltimore, MD, Steven J. Elie, Musick Peeler and Garrett LLP, Los Angeles, CA, for Clarios, LLC. Patrick Ward Dennis, Thomas Frasca Cochrane, Gibson Dunn and Crutcher LLP, Los Angeles, CA, Alexander P. Swanson, Rutan and Tucker LLP, Irvine, CA, for Quemetco, Inc. Anna Louise Le May, Ruben A. Castellon, Raf Law Group, Los Angeles, CA, for International Metals Ekco, Ltd. Donald E. Sobelman, John M. Ugai, Linda Teresa Sobczynski, Christopher I. Rendall-Jackson, Farella Braun and Martel LLP, San Francisco, CA, for Oregon Tool Inc.

STEPHEN V. WILSON, UNITED STATES DISTRICT JUDGE

David Zaft, Jon Stephen Tangonan, Aarti Sujan Kewalramani, Dennis A. Ragen, Elizabeth B. Rumsey, CAAG - Office of the Attorney General California Department of Justice, San Diego, CA, Adam M. Shapiro, Matthew K. Edling, Timothy R. Sloane, Yumehiko Hoshijima, Sher Edling LLP, San Francisco, CA, Adrianna L. Lobato, Scott Justin Lichtig, CAAG - Office of Attorney General California Department of Justice, Sacramento, CA, Christopher T. Nidel, Pro Hac Vice, Nidel and Nace PLLC, Rockville, MD, David P. Page, Pro Hac Vice, Environmental Energy and Natural Resources Advocates PLLC, Tulsa, OK, Donald A. Robinson, Kate Marie Hammond, Sarah E. Morrison, CAAG - Office of Attorney General of California California Department of Justice, Los Angeles, CA, Thomas M. Sims, Nidel and Nace PLLC, Washington, DC, Timothy Eugene Sullivan, California Department of Justice, Oakland, CA, for California Department of Toxic Substances Control.

Sarah E. Morrison, Los Angeles, CA, Donald A. Robinson, Kate Marie Hammond,

CAAG - Office of Attorney General of California California Department of Justice, Los Angeles, CA, Aarti Sujan Kewalramani, Dennis A. Ragen, Elizabeth B. Rumsey, CAAG - Office of the Attorney General California Department of Justice, San Diego, CA, Adam M. Shapiro, Matthew K. Edling, Timothy R. Sloane, Yumehiko Hoshijima, Sher Edling LLP, San Francisco, CA, Adrianna L. Lobato, Scott Justin Lichtig, CAAG - Office of Attorney General California Department of Justice, Sacramento, CA, Christopher T. Nidel, Pro Hac Vice, Nidel and Nace PLLC, Rockville, MD, David P. Page, Pro Hac Vice, Environmental Energy and Natural Resources Advocates PLLC, Tulsa, OK, Thomas M. Sims, Nidel and Nace PLLC, Washington, DC, Timothy Eugene Sullivan, California Department of Justice, Oakland, CA, for The Toxic Substances Control Account.

Joel L. Herz, Pro Hac Vice, Law Office of Joel L. Herz, Tucson, AZ, Kenneth A. Ehrlich, Sean Andrew McCormick, Elkins Kalt Weintraub Reuben Gartside LLP, Los Angeles, CA, for NL Industries, Inc.

David C. Allen, Joel R. Meyer, Joseph M. Wahl, Barnes and Thornburg LLP, Los Angeles, CA, Dan Terzian, Erick Kees Kuylman, Thomas D. Warren, Warren Terzian LLP, Los Angeles, CA, Andrew J. Jarzyna, Pro Hac Vice, Nicholas J. Secco, Pro Hac Vice, Benesch Friedlander Coplan and Aronoff LLP, Chicago, IL, Gregory T. Frohman, Pro Hac Vice, Benesch Friedlander Coplan and Aronoff LLP, Cleveland, OH, Krista M. Enns, Lily A. North, Benesch Friedlander Coplan and Aronoff LLP, San Francisco, CA, for Gould Electronics Inc.

Richard Alan Dongell, Scott L. Hengesbach, Murchison and Cumming LLP, Los Angeles, CA, Elizabeth Burnett, Clark Hill, Los Angeles, CA, Eric P. Weiss, Scali Rasmussen PC, Los Angeles, CA, for Kinsbursky Bros. Supply, Inc.

Jeffrey J. Parker, Zachary Michael Norris, Stephen J. O'Neil, Sheppard Mullin Richter and Hampton LLP, Los Angeles, CA, for Trojan Battery Company, LLC.

James C. MacDonald, Law Offices of James MacDonald, Irvine, CA, for Ramcar Batteries Inc.

Bina R. Reddy, Pro Hac Vice, Beveridge and Diamond PC, Austin, TX, Eric L. Klein, Pro Hac Vice, Beveridge and Diamond PC, Boston, MA, Gary J. Smith, Beveridge and Diamond PC, San Francisco, CA, Jessica L. Kyle, Pro Hac Vice, Beveridge and Diamond, Baltimore, MD, Steven J. Elie, Musick Peeler and Garrett LLP, Los Angeles, CA, for Clarios, LLC.

Patrick Ward Dennis, Thomas Frasca Cochrane, Gibson Dunn and Crutcher LLP, Los Angeles, CA, Alexander P. Swanson, Rutan and Tucker LLP, Irvine, CA, for Quemetco, Inc.

Anna Louise Le May, Ruben A. Castellon, Raf Law Group, Los Angeles, CA, for International Metals Ekco, Ltd.

Donald E. Sobelman, John M. Ugai, Linda Teresa Sobczynski, Christopher I. Rendall-Jackson, Farella Braun and Martel LLP, San Francisco, CA, for Oregon Tool Inc.

Proceedings: DECISION AND VERDICT IN DEFENDANTS' FAVOR

STEPHEN V. WILSON, UNITED STATES DISTRICT JUDGE

I. INTRODUCTION

Before the Court is an action brought by the California Department of Toxic Substances Control and the Toxic Substances Control Account ("DTSC" or "Plaintiffs") under the Comprehensive Environmental Response, Compensation and Liability Act of 1980 ("CERCLA"), 42 U.S.C. §§ 9601 et seq. Plaintiffs seek recovery of environmental cleanup and response costs from a

number of owners and operators ("Defendants") of a former lead battery recycling plant located in Vernon, California (the "Vernon Plant").

The Vernon Plant sits in the center of the Preliminary Investigation Area (the "PIA"), which is a roughly circular area that extends in an approximately-1.7 mile radius around the Vernon Plant. The PIA includes an industrial area immediately surrounding the Vernon Plant within an approximately half-mile radius (the "Industrial Area"), and residential areas that extend up to 1.7 miles to the north and south of the Vernon Plant (the "Residential Areas"). Plaintiffs alleged that lead contamination spread throughout the entire 1.7-mile radius, though their cleanup efforts largely centered on the Residential Areas.

The purpose of the three-day bench trial (the "Scope Trial") was to resolve a question as to CERCLA's third prong: whether releases from the Vernon Plant caused Plaintiffs to incur response costs. See United States v. Sterling Centrecorp Inc., 977 F.3d 750, 756 (9th Cir. 2020); 42 U.S.C. § 9607(a). Certain facts were undisputed, including that (1) lead was released from the Vernon Plant over the many decades it operated; (2) the Preliminary Investigation Area in Vernon (the "PIA") was contaminated with lead; (3) Plaintiffs had incurred cleanup costs in responding to lead contamination in the PIA; and (4) lead released from the Vernon Plant had contaminated the PIA's Industrial Areas. The narrower question, therefore, was whether the costs Plaintiffs incurred in responding to contamination in the Residential Areas were attributable to the Vernon Plant.

At trial, the parties framed the question as requiring a binary answer: either the released lead contaminated only the half-mile Industrial Area surrounding the Vernon Plant, or it contaminated the Residential Areas by traveling as far as 1.7 miles away from the Vernon Plant. As the Court explains below, however, CERCLA does not require a showing that Plaintiffs' releases actually caused contamination. To the contrary. Plaintiffs needed to show only that Defendants' releases plausibly could have caused contamination throughout the PLA's Residential Areas to satisfy the statute.

Accordingly, Plaintiffs were required to present evidence that Defendants' releases plausibly could have caused contamination throughout the Residential Areas. To do so, Plaintiffs had to establish a plausible basis for selecting the areas that required remediation. This presented a difficult causation question. Lead is a ubiquitous contaminant in urban areas, and it is undisputed that lead levels in the PIA reflected contribution from various sources, including lead paint, leaded gasoline, and other industrial sites in the PIA. Because of the difficulty of tracing a contaminant like lead, Plaintiffs rely on evidence including (1) "background," which estimates the theoretical amount of lead in the PIA if the Vernon Plant had never existed; (2) wind modeling and statistical analyses of contamination; and (3) soil chemistry.

Having carefully reviewed and considered the evidence presented at trial in light of Plaintiff's burden, the Court holds that Plaintiffs did not establish that releases from the Vernon Plant plausibly could have caused contamination throughout the PIA's Residential Areas. Plaintiffs' largely speculative evidence establishes, at most, that an undetectable amount of lead landed in some undefined parts of an approximately 6,400 acre PIA. This speculation is insufficient to establish the required nexus between the Vernon Plant's releases and the cleanup costs Plaintiffs incurred throughout the Residential Areas.

More specifically, the Court concludes that Plaintiffs (1) did not meet their burden

in showing that the entirety of the 1.7-mile radius could have been contaminated, and certainly not in amounts that would have caused Plaintiffs to incur response costs in the Residential Areas; and (2) did not meet their burden in showing that discrete portions of the Residential Areas were contaminated.

In their remediation and litigation efforts. Plaintiffs targeted the Vernon Plant as the greatest contributor to the contamination in the Residential Areas, seeking tens of millions of dollar's in cleaning up the PIA's Residential Areas with little evidence that the Vernon Plant's lead actually spread throughout those areas. Tire geographic dispersion of lead was critical because, once airborne lead particles landed in the soil, there was no evidence that lead continued to propagate throughout the Residential Areas. In contrast to CERCLA cases addressing contaminated bodies of water, where contaminants could plausibly spread, requiring remediation of the entire site, there was insufficient evidence that lead here could have plausibly caused contamination throughout the Residential Areas.

Plaintiffs' evidence speaks to the possibility of contamination in some undefined portions of the PIA. But that was not the question in this case. Plaintiffs have not established plausibility' as to particular places within the PIA, as required by the courts that have spoken to this issue. Accordingly, the Court issues the following verdict in Defendants' favor, which includes its findings of fact and conclusions of law, pursuant to Federal Rule of Civil Procedure 52(a).¹

1 For all findings of fact set forth below, in making any credibility determinations regarding witness testimony, the Court has considered, among other things, the manner in which the witnesses testified, their interest in the outcome of the case, and the reasonableness of their testimony in light of all of the evidence. The Court has also considered the relevant factors in Section 1.14 of the Manual of Model Civil Jury Instructions for the District Courts of the Ninth Circuit (2017 Edition), located at http:www3.ce9.uscourts.gov/juryinrtructions/sites/default/files/WPD/Civil_Instructions_2018_9_0.pdf.

II. BACKGROUND

The Vernon Plant is a former secondary lead smelter that operated from the mid-1920s until 2015. Secondary lead smelting involves extracting lead and other materials from products such as used batteries. That process entails separating acid, plastic, and lead; treating and disposing of the acid; and melting down the lead in large furnaces. See Complaint ("Compl."), Dkt. 1 ¶¶ 41-42, 48. Though Defendants in this matter include a number of owners, operators, and other potentially responsible parties, Exide Corporation was the Vernon Plant's most recent owner.²

2 The Court's prior order at Dkt. 198, granting JX Nippon Mining & Metals Corporation's motion to dismiss for lack of personal jurisdiction, included a robust factual background of the site's history.

Over the many years it operated, the Vernon Plant released airborne lead from its stacks, which Plaintiffs allege traveled throughout the PIA and its Residential Areas. Plaintiffs conducted extensive soil sampling throughout the PIA, which is the area depicted in the below map: the black, roughly circular perimeter around the City of Vernon. The Vernon Plant and the Industrial Areas are located at the center of the map. The PIA's Residential Areas are located at the north and south ends of the map. Image materials not available for display.

At trial, the parties presented various experts, each of whom performed various analyses based on differing scientific disciplines. Specifically, Plaintiffs offered expert testimony of experts Dr. Allen Medine, see Declaration of Allen J. Medine ("Medine Decl."), ECF No. 351-2; Dr. Anita Singh, see Declaration of Dr. Anita Singh ("Singh Decl."), ECF No. 351-4; Mr. Lyle Chinkin, see Declaration of Lyle R. Chinkin ("Chinkin Decl."), ECF No. 351-5; and Dr. John Drexler, see Declaration of John W. Drexler, PhD ("Drexler Decl."), ECF No. 351-6. Plaintiffs also offered fact testimony from DTSC Employee Peter Ruttan. See Declaration of Peter Vincent Ruttan ("Ruttan Decl."), ECF No. 351-1.

Defendants offered the expert testimony of Dr. Shahrokh Rouhani, see Declaration of Shahrokh Rouhani ("Rouhani Decl."), ECF No. 352; Mr. Ranjit Machado, see Declaration of Ranjit J. Machado. P. E. ("Machado Decl."), ECF No. 353; and Dr. Walter Shields, see Declaration of Walter J. Shields, PH.D ("Shields Decl."), ECF No. 354. Defendants also called fact witnesses Dr. Kathleen Burger, a former DTSC employee, see Trial Transcript, Day 2 ("Tr. 2"), at 111, and Philip Chandler, see Trial Transcript, Day 3 ("Tr. 3"), at 5.

Following the trial, the parties submitted simultaneous Post-Trial Briefs and Post-Trial Replies. See Pls.' Post-Trial Br., ECF No. 453; Pls.' Post-Trial Reply, ECF No. 456; Defs.' Post-Trial Br., ECF No. 454; Defs.' Post-Trial Reply, ECF No. 455.³ Having considered the evidence and argument presented at trial and the subsequent briefs, the Court hereby issues the following verdict. III. LEGAL STANDARD

3 The Court adopts the parties' naming convention for referring to exhibits: PX refers to Plaintiffs' Exhibits. DX refers to Defendants' Exhibits, and JX refers to Joint Exhibits.

A. CERCLA

Congress enacted CERCLA "to provide for liability, compensation, cleanup, and emergency response for hazardous substances release into the environment and the cleanup of inactive hazardous waste disposal sites." 3550 Stevens Creek Associates v. Barclays Bank, 915 F.2d 1355, 1357 (9th Cir. 1990) (citing Pub.L. No. 96-510, 94 Stat. 2767 (1980)). "CERCLA generally must be construed liberally to accomplish its dual goals of promptly cleaning up hazardous waste sites and making polluters, rather than society as a whole, pay." Pakootas v. Teck Cominco Metals, Ltd., 905 F.3d 565, 584 (9th Cir. 2018).

"To establish a prima facie right to recovery under § 9607(a), a plaintiff must demonstrate that: (1) the site on which the hazardous substances are contained is a 'facility' as defined in CERCLA; (2) a 'release' or 'threatened release' of a 'hazardous substance' from the facility has occurred; (3) the 'release' or 'threatened release' has caused the plaintiff to incur response costs that were 'necessary' and 'consistent with the national contingency plan;' and (4) defendants are within one of four classes of persons subject to liability under § 9607(a). Carson Harbor Vill., Ltd. v. Unocal Corp., 287 F. Supp. 2d 1118, 1152 (C.D. Cal. 2003), aff'd sub nom. Carson Harbor Vill. v. Cnty. of Los Angeles, 433 F.3d 1260 (9th Cir. 2006) (citation omitted).⁴

4 The first, second, and fourth prongs were not at issue for purposes of this trial, nor was the question of whether Plaintiffs' response costs were "necessary" and "consistent with the national contingency plan" was not at issue for purposes of the Scope Trial.

B. CAUSATION

The Scope Trial addressed only a portion of CERCLA's third prong: whether releases from the Vernon Plant caused Plaintiffs to incur response costs. See 42 U.S.C. § 9607; Carson Harbor Vill., 287 F. Supp. 2d at 1186 (referring to a "causal link" between the releases and the response costs) (citing State of Cal. on Behalf of State Dep't of Toxic Substances Control v. Ct. Galvanizing, Inc., 1996 WL 528510, *3 (9th Cir. 1996) ("[T]he relevant causation inquiry is whether there is a nexus between the release [or threatened release] and the response cost incurred")).

The scope of this causal inquiry is critical. What showing must the parties make to prove (or disprove) that releases from the Vernon Plant caused Plaintiffs to incur response costs? Plaintiffs advocate for a burden-shifting framework created by the court in Castaic Lake Water Agency v. Whittaker Corp., 272 F. Supp. 2d 1053 (C.D. Cal. 2003). Defendants assert that Plaintiffs bear the burden of proof, and that the correct causation standard is dictated by Boeing Co. v. Cascade Corp., 207 F.3d 1177 (9th Cir. 2000).

At the outset, two boundaries confine the Court's interpretation of CERCLA's complex statutory framework.⁵ The first boundary is CERCLA's plain language. Because the provision at issue explicitly requires that Defendants' release caused Plaintiffs to incur response costs, the Court cannot adopt a standard that, as a practical matter, would nullify the statute's text by omitting a causation requirement entirely. See Chubb Custom Ins. Co. v.

5 Interpreting CERCLA's language places the Court in a difficult position. As other courts have discussed, CERCLA was imperfectly drafted, see, e.g., Asarco LLC, 21 F. Supp. 3d at 801. n. 19; United States v. P.H. Glatfelter Co., 768 F.3d 662, 674 (7th Cir. 2014), and this section only guides the Court minimally.

Space Sys./Loral, Inc., 710 F.3d 946, 958 (9th Cir. 2013) ("In interpreting statutes, our task is to construe Congress's intent, and doing so requires us to 'begin, as always, with the language of the statute.'"); see also Boeing, 207 F.3d at 1185 (considering the "practical consequences" in "construing the statutory requirement of causation.").

The second boundary is CERCLA's purpose. See Carson Harbor Vill., Ltd. v. Unocal Corp., 270 F.3d 863, 877 (9th Cir. 2001) (Carson Harbor I) (en banc). "CERCLA was enacted to protect and preserve public health and the environment by facilitating the expeditious and efficient cleanup of hazardous waste sites." Id. at 880. It "must be construed liberally to accomplish its dual goals of promptly cleaning up hazardous waste sites and making polluters, rather than society as a whole, pay." Pakootas, 905 F.3d at 584.

With these concerns in mind, the Court also distinguishes between the inquiry now before it, and a separate inquiry Defendants refer to as "defendant-specific" causation. The idea of "defendant-specific" causation refers to a well-established CERCLA principle—that Plaintiffs need not show a specific defendant's waste caused Plaintiffs to incur cleanup costs. See, e.g., United States v. Alcan Aluminum Corp., 315 F.3d 179, 184 (2d Cir. 2003) ("The government is not required to show that a specific defendant's waste caused the incurrence of cleanup costs in order for strict liability to attach to that defendant."); see also United States v. Monsanto Co., 858 F.2d 160, 169-70 (4th Cir. 1988). The Fifth Circuit in Amoco Oil identified the distinction:

Several courts have interpreted the causation requirement in a different context—for determining the standard of proof necessary to show a [specific] defendant is responsible for the hazard that resulted in response costs. For that purpose, the causation requirement has been interpreted in a somewhat relaxed manner due to difficult proof problems inherent in toxic waste cases and CERCLA's broad liability provisions. Accordingly, in cases involving multiple sources of contamination, a plaintiff need not prove a specific causal link between costs incurred and an individual generator's waste. [Citations].

Amoco Oil Co. v. Borden, Inc., 889 F.2d 664, 670 n. 8 (5th Cir. 1989), as clarified on denial of reh'g (Jan. 23, 1990) (emphasis added).

Thus, CERCLA does not contain any language requiring a plaintiff to show that a specific defendant caused certain costs to be incurred.⁶ By contrast, the statute here explicitly requires that Plaintiffs establish a "causal link" between Defendants' releases and Plaintiffs' response costs. See Carson Harbor Vill., 287 F. Supp. 2d at 1186. Defendants call this "release causation."

6 CERCLA's legislative history makes this point clear. See State of N.Y. v. Shore Realty Corp., 759 F.2d 1032, 1044-45 (2d Cir. 1985); United States v. Wade, 577 F. Supp. 1326, 1333 (E.D. Pa. 1983). Though an early House version of the statute "imposed liability only upon 'any person who caused or contributed to the release or threatened release,'" the final version "imposed liability on classes of persons without reference to whether they caused or contributed to the release or threat of release." Shore Realty, 759 F.2d at 1044 (internal citations omitted) (emphasis added). In the defendant-specific context, many cases rely on this legislative history. However, CERCLA's legislative history is silent with respect to the more general release causation requirement at issue in the Scope Trial.

This distinction is critical because it affects the showing a plaintiff must make to establish liability. A "release causation" case, for which CERCLA requires a showing of causation, would be mistaken in relying on a "defendant-specific" case, for

which CERCLA does not require a showing of causation. But as the Court's below discussion of Castaic Lake illustrates, some cases mistakenly overlook the difference between the two causation inquiries.

The Court now considers the parties' arguments as to the causation framework. The Court holds, first, that Castaic Lake is incorrect in framing the Section 107 inquiry as a "burden-shifting" inquiry. CERCLA does not require Defendants to prove that "none" of the lead in the PIA came from the Vernon Plant. Second, the Court concludes that Boeing does not speak to the question of burden, but its principles are instructive in evaluating causation generally. Third, the Court finds that, to prove that Defendants' releases caused Plaintiffs to incur response costs, Plaintiffs must establish that Defendants' releases plausibly could have contaminated the PLA's Residential Areas.

1. Castaic Lake

Plaintiffs argue that the applicable release causation standard is a burden-shifting approach from Castaic Lake, 272 F. Supp. 2d at 1065-66. Under Castaic Lake, Plaintiffs must only "(a) identify the contaminant at a site; (b) identify the same (or perhaps a chemically similar) contaminant off-site; and (c) provide evidence of a plausible⁷ migration pathway by which the contaminant could have traveled from a site to the off-site locations." See id.; see Pls' Pretrial Br. at 3.

7 As the Court's discussion in Section III.B.3 indicates, other cases rely on concepts similar to plausibility, like reasonableness and justification. However, as far as the Court's research has shown, Castaic Lake was the first case to consider the word "plausible" in this context.

According to the Castaic Lake court, after Plaintiffs meet their prima facie burden, "the burden of proof [then] falls on the defendant to disprove causation." Tesoro Ref. & Mktg. Co. LLC v. City of Long Beach, No. 216CV06963VAPFFMX, 2018 WL 11348067, at *3 (C.D. Cal. Nov. 21, 2018) (citing Castaic Lake, 272 F. Supp. 2d at 1066). Plaintiffs resultantly contend that Defendants must show that "none of the lead in the PIA that DTSC investigated and/or cleaned up came from the Vernon Plant." See Pls. Pretrial Br. at 7 (emphasis added); see Tr. 3 at 117:7-25 ("If one particle ... lands somewhere, then the defendant loses.").

The Court disagrees. We decline to adopt such a framework requiring Defendants to bear the burden of disproving causation by showing that "none of the lead" came from the Vernon Plant. Plaintiffs' argument that only one particle of contamination needs to travel to the Residential Areas is sufficient to establish liability is unsupported. Indeed, as a practical matter, the Court concludes that one particle of lead landing in the Residential Areas would be unlikely to have caused Plaintiffs to incur response costs.

Castaic Lake was a release causation case where several plaintiff water companies contended that four of their wells were contaminated by perchlorate released from the defendants' nearby site. See Castaic Lake, 272 F. Supp. 2d at 1058. Considering a motion for summary judgment, the court determined that the plaintiffs met their prima facie burden of establishing release causation: perchlorate was detected at the defendants' site and at the water wells, and the plaintiffs' experts opined that perchlorate could travel via surface water from the site to the wells. Id. at 1067-68.

In response, the defendants' expert opined that (1) "surface water migration alone likely could not cause all of the contamination in Plaintiffs' wells:" and (2) "other nearby facilities might have released

perchlorate in the direction of Plaintiffs' wells, but it is impossible to determine sources based on available data." Id. at 1068. Because "neither of these opinions indicated that [the] defendants can disprove that the [] site was a cause of perchlorate contamination in [the] wells, [the] defendants [] failed to create a genuine issue of material fact that would preclude summary judgment for the [p]laintiffs." Id. (cleaned up) (emphasis in original). The court therefore granted summary judgment in the plaintiffs' favor.

To evaluate Castaic Lake, the Court traced the precedents upon which it relied, including Westfarm Associates Limited Partnership v. Washington Suburban Sanitary Comm'n, 66 F.3d 669 (4th Cir. 1995), United States v. Alcan Aluminum Corp . ("Alcan-Butler"), 964 F.2d 252 (3d Cir. 1992), Artesian Water Co. v. New Castle County, 659 F.Supp. 1269 (D. Del. 1987), aff'd on other grounds 851 F.2d 643 (3d Cir. 1988), and United States v. Bliss, 667 F.Supp. 1298 (E.D. Mo. 1987). See Castaic Lake, 272 F. Supp. 2d at 1065.

The Court's review of these cases brought it to the distinction previewed above: the difference between release causation and defendant-specific causation. Westfarm, Alcan, and Bliss were each defendant-specific cases, and each relied on defendant-specific precedents in holding that the burden of proof in CERCLA cases lies with the defendant. See Westfarm, 66 F.3d at 681; Alcan, 964 F.2d at 264-266; Bliss, 667 F. Supp. at 1311.⁸ In other words, these three cases merely affirmed the well-established principle that a CERCLA plaintiff need not establish a causal nexus between a release of contaminants and a specific defendant. See Amoco Oil, 889 F.2d at 670 n. 8.

8 Castaic Lake purportedly relied on Artesian Water, but Artesian Water did not adopt a burden shifting framework. See Artesian Water, 659 F. Supp. at 1283-84. Rather, Artesian Water only held that "the defendant's conduct is a cause of the event if it was a material element and a substantial factor in bringing it about." Id. at 1283. In light of the "uncontradicted" evidence "that the threat of contamination ... was initially the sole factor and remain[ed] a substantial factor in the continued [response costs,]" the court did not discuss any burden shifting theory. See id.

For example, in Westfarm, the Fourth Circuit referenced CERCLA's legislative history, see supra n. 6, in holding that the burden of disproving causation rested with a defendant. There, the plaintiff sued a dry-cleaning facility and a local sewer operator after discovering perchloroethylene in groundwater beneath its property. Westfarm, 66 F.3d at 673. The plaintiff traced the contaminant to the adjacent commercial dry-cleaning facility. Id. at 674. After discovering that employees at the dry-cleaning facility would pour the perchloroethylene down a chain connected to the county's sewers, the plaintiff conducted an inspection that revealed various flaws in the sewers. Id.

On appeal, there was no question as to whether the releases (generally) had caused the plaintiff to incur response costs. Id. at 681-82. Rather, the sewer operator defendant argued that the plaintiff had not traced the contaminants to the operator's sewers. Id. The court discussed Congress' decision to delete defendant-specific causation language from the statute, finding that CERCLA "imposed liability on classes of persons without reference to whether they caused or contributed to the release or threat of release," see id. (citing Monsanto, 858 F.2d at 170 n. 17). Ultimately, Westfarm concluded that the defendant's argument was insufficient to disprove causation as to that defendant. Id. at 682.

Alcan-Butler and Bliss are similar. See Alcan-Butler, 964 F.2d at 264-66; Bliss,

667 F.Supp. at 1309-11. But Castaic Lake mistakenly relied on these cases to create a rule for release causation, holding that the defendants bear the burden of disproving that a release caused the plaintiffs' response costs, and conflating two distinct inquiries under the statute.

Plaintiffs do little to address this argument. Plaintiffs merely cite cases like Westfarm and Alcan-Butler without recognizing the critical distinction between the two different types of causation inquiries. See Pls.' Post-Trial Br. at 5-7. Though Plaintiffs argue that other courts in this district have also adopted Castaic Lake, those cases have done so without considering this distinction or otherwise explaining then reasoning. See, e.g., Santa Clarita Valley Water Agency v. Whittaker Corp., No. 2:18-CV-06825-SB-RAOx, 2021 WL 2549066, at *2 (C.D. Cal. Apr. 5, 2021); Tesoro Ref & Mktg. Co., 2018 WL 11348067, at *3. This precedent is therefore of little help to the Court.

Plaintiffs also rely on Congress' policy in enacting CERCLA. See Post-Trial Reply Br., at 2-5. But as discussed above, Congress' intent in drafting CERCLA must comport with the plain language of the statute. Other than cases interpreting defendant-specific causation. Plaintiffs do not cite any persuasive authority indicating that Defendants should, as a policy matter, bear the burden of disproving release causation.

Even Asarco, which Plaintiffs rely on, stated, "Section 107(a) always requires a plaintiff to prove, in its prima facie case, that a release from a facility for which the defendant is a responsible person caused the incurrence of response costs[.]" See Asarco LLC v. Cemex, Inc., 21 F. Supp. 3d 784, 804 (W.D. Tex. 2014);⁹ see also Artesian Water, 659 F. Supp. at 1282 ("CERCLA's strict liability scheme does not diminish the necessity of demonstrating a causal connection between a release or threatened release and the incurrence of costs by a section 107 plaintiff."); Krygoski Const. Co. v. City of Menominee, 431 F. Supp. 2d 755, 763 (W.D. Mich. 2006) ("While causation may be elusive, a CERCLA plaintiff still shoulders the burden of alleging and proving an unbroken chain of events occurred which bridged his necessary response costs to the defendant's conduct.") (quoting Rhodes v. County of Darlington, South Carolina, 833 F.Supp. 1163, 1190 (D.S.C. 1992)). Without more, Plaintiffs' appeal to policy is unavailing.

9 Confusingly, Asarco adopted Castaic Lake's burden-shifting test, referencing cases like Westfarm, Alcan, Bliss, and Monsanto that did not consider the question of release causation. See Asarco LLC, 21 F. Supp. 3d at 807.

But even setting aside Castaic Lake, Plaintiffs' position is untenable. Considering the facts of this case—and CERCLA cases generally—no defendant could ever prove, as Plaintiffs claim Defendants must, that not one microscopic particle landed within a site. Importantly, because CERCLA includes a release causation requirement, placing such a burden on Defendants would eviscerate the statute's plain language. See Chubb Custom Ins., 710 F.3d at 958 ("In interpreting statutes, our task is to construe Congress's intent, and doing so requires us to 'begin, as always, with the language of the statute.'"). Thus, because CERCLA requires a nexus between Defendants' release and Plaintiffs' response costs, the Court cannot adopt any standard that nullifies CERCLA's statutory text. See id.

For these reasons, the Court disagrees with Plaintiffs' argument that Defendants bear the burden of proving that no contaminants traveled to the PIA's Residential Areas. 2. Boeing

Countering Plaintiffs' argument that Castaic Lake provides the correct standard for determining CERCLA causation, Defendants argue that the correct causation standard was stated in Boeing, 207 F.3d at 1185. According to Defendants, Boeing provides that Plaintiffs bear the burden of establishing either (1) "but-for" causation, meaning that that no soil remediation would have been necessary in residential areas in the absence of the Vernon Plant; or (2) "causal overdetermination," meaning that soil remediation in the residential areas would have been required as a result of Vernon Plant emissions even if those other sources themselves had not existed. See Defs. Pretrial Br. at 9; see Boeing, 207 F.3d at 1185.

A review of the Boeing decision indicates that Defendants overstate its holding. Boeing was a contribution action that addressed the appropriate allocation of cleanup expenses, where both the Boeing Company and Cascade Corporation contaminated the same aquifer with chemical solvents. See Boeing, 207 F.3d at 1180. Specifically, Boeing filed suit against Cascade for contribution to "offset Boeing's higher expenses." Id. at 1180; see 42 U.S.C. § 9613(f). Cascade argued that it should not be allocated any cleanup costs because the EPA had already ordered Boeing to investigate its own contamination—meaning that Boeing would have incurred cleanup costs even if Cascade had not contaminated the soil. Id. at 1182. Thus, Cascade argued that its release was not the "but-for" cause of the cleanup costs, and that it was not liable under CERCLA as a result. Id.

The Ninth Circuit disagreed. Id. at 1183-85. It recognized that "there is a limited class of cases in which conduct ought to be treated as a cause of damages even though the conduct was not a sine qua non." Id. at 1183-84. The court referred to this concept as "causal overdetermination"—when multiple sufficient causes are responsible for an effect, liability attaches to "each" of the causes. Id.

As an example, the court discussed an overhead light "controlled by front and rear light switches. One person flips up the front switch at precise the same time another person flips up the rear switch. The light goes on." Id. at 1184. Cascade's position that it was not the "but for" cause of the contamination was the same as a person in the light switch example arguing that "his switching the light on was not a sine qua non of its going on." Id. at 1185. The court explained:

It is true that the light would have gone on anyway because of the other person's conduct. If conduct had to be a sine qua non even for this overdetermined result, then neither person's conduct caused the light to go on. But the light went on. And it did so by human agency, not spontaneously. So the conclusion that Cascade's argument compels, that no one caused the light to go on, is false. Because the correct answer has to be the same for the two individuals, by eliminating the false answer we have left only one possible answer which must be true: Each of the two persons caused the light to go on.

Id. at 1185. The Court held, consequently, that "where either polluter's conduct would have caused the same response cost to be incurred in the same amount, and the conduct was of substantially equal blameworthiness, the proper construction of the causation requirement in the statute is that both polluters should be treated as having caused the response cost." Id.

While Boeing does discuss the "but for" and "causal overdetermination" standards in the context of the statutory language at issue, see Boeing, 207 F.3d at 1183 (whether a party's release of hazardous substances

caused the plaintiff to incur response costs), Boeing is silent as to the issue now before the Court: the showing that Plaintiffs must make to establish or disprove Defendants' liability under the third prong of CERCLA. It did not address the question of burden, or what evidence Boeing, as a plaintiff who incurred response costs, was required to adduce to prove that Cascade was the "but for" or causally overdetermined cause of Boeing's response costs. Id. at 1185-86.

Rather, the Ninth Circuit's holding was narrower: it merely rejected Cascade's argument that it was not liable because it was not a "but for" cause of Boeing's response costs. Id. at 1185. Though the court held that findings of but-for cause or causal overdetermination would be sufficient to establish liability, it did not hold that either were necessary. Id.

The Court therefore disagrees with Defendants' broad argument that Boeing announced a framework through which the Court can resolve the release causation question at issue. But as discussed below, the Court finds Boeing instructive to the extent that it addresses causation principles apart from the question of burden.

3. Framework

To determine the correct framework in the wake of this muddled precedent, the Court returns to the statutory text. Having considered the statute at length, the Court believes the decades-long confusion over its text stems from a failure to recognize a key concept embedded within the statute: that establishing liability requires resolving a dispute over a causal chain of several steps.

As dictated by the statute, courts ask only whether a "release caused the incurrence of response costs." See 42 U.S.C. § 9607(a). However, to actually incur response costs, an independent actor like DTSC must affirmatively decide to respond to the release: a release does not automatically cause a plaintiff to incur response costs, the way that turning on a light switch causes a light to go on. Rather, the causal connection between a release and subsequent response costs necessarily includes some intermediate cause.

Theoretically, there may be any number of intermediate causes that lead to the incurrence of response costs. For example, if the Vernon Plant released a small amount of lead for only ten minutes and caused a significant public outcry, which in turn caused DTSC to clean up a large area around the smelter, the release may have still "caused the incurrence of response costs." See id. In that case, the release would still be the "but for" of the response costs.

At trial, however, the parties directed their efforts to establishing a specific intermediate cause: contamination. Plaintiffs argue that they "cleaned up homes in the PIA because [they] determined that those homes were contaminated by lead released from the Vernon Plant." Pls.' Post-Trial Br., at 29. Or, framed differently, Plaintiffs' alleged chain of cause and effect proceeded as follows:

(1) The Vernon Plant released lead, a hazardous substance.

(2) Defendants' release caused lead contamination in the PIA.

(3) Lead contamination caused Plaintiffs to incur response costs.

Importantly, there is no dispute as to the first and third steps of this chain. The parties agree that the Vernon Plant released lead. And the parties agree that Plaintiffs were responding to lead contamination when they incurred cleanup costs in remediating the PIA. Only the second prong is at issue here: whether Defendants' releases caused lead contamination throughout the 1.7 mile radius of the PIA. For several reasons, however, the Court concludes that CERCLA does not require a showing of actual contamination. First, CERCLA's plain language does not require a showing of actual contamination. For one, the word "contamination" is conspicuously absent from this portion of the statute. See 42 U.S.C. § 9607(a).¹⁰

10 Though not cited by either side, subsection (q) of Section 107 creates an exception to liability for "contiguous property owners" who own property that, among other things, "is or may be contaminated by a release or threatened release of a hazardous substance[.]" See 42 U.S.C. § 9607(q). While subsection (q), enacted in 2002, is of limited relevance here, it suggests that, at the least, Congress is aware of the distinction between releases that cause contamination and releases that do not—and that Congress chose not to include a similar requirement by amending § 9607(a).

Second, this reasoning is consistent with the case law; most prominently, the Dedham Water Co., Inc. v. Cumberland Farms Dairy, Inc. line of cases, which are cited throughout many of the cases that address this issue. Specifically, in Dedham II 889 F.2d 1146 (1st Cir. 1989), the First Circuit explained that a defendant's releases might have caused a plaintiff to incur response costs even where that defendant's releases "did not, in fact, contaminate the plaintiff's wells." Id. at 1157.

The district court had concluded its inquiry when it found that the releases did not cause contamination, but the First Circuit held that it erroneously failed to consider

whether defendant's releases (or threatened releases) might nonetheless have caused the plaintiff to incur 'response costs' even though those releases did not in fact contaminate the wells. (A plaintiff, for example, under certain circumstances might reasonably think that a particular release would prove likely to contaminate his wells and spend money to avoid the contamination even though no actual contamination occurs.)

Dedham II, 889 F.2d at 1157. The court explained that "every court that has addressed this issue, with the exception of the district court[...], has held that it is not necessary to prove actual contamination of plaintiff's property by defendant's waste in order to establish liability under CERCLA." Id. at 1154.

Third, the Court is also persuaded that no showing of contamination is necessary because a defendant may be held liable merely for a "threatened" release.¹¹ See 42 U.S.C. § 9607(a). Generally, threatened releases merely gives rise to a risk of possible future contamination. But the statute does not make any distinction between how a plaintiff may recover based on a release or a threatened releases. So long as either a release or a threatened release caused the incurrence of response costs, a plaintiff may recover equally. Thus, because a plaintiff may recover without a showing of contamination in the context of a threatened release, it would be inconsistent to interpret Section 107 to require a showing of contamination when a release actually occurs, because there is no such distinction in the statute.

11 Plaintiffs do not seek to recover response costs for threatened releases here (nor could they, because Plaintiffs began to remediate the PIA after the Vernon Plant was no longer operating).

Congress therefore must have contemplated liability simply based on the potential for contamination from a threatened release: nothing in the statute suggests that Congress intended to create a different standard for actual releases, because the statute treats both actual and threatened releases as identical. It would be unreasonable to conclude that Congress intended to create a different standard for actual releases. So, because there is no

basis to conclude that Congress intended to require a different nexus between a release and response costs versus a threatened release and response costs, the two must require the same showing of plausible contamination.

In accordance with Congress's probable wishes, the Court is persuaded that "plausibility" captures the best approach to the release causation inquiry. The Dedham cases treated the causation question as requiring a reasonableness inquiry—i.e., whether a plaintiff "under circumstances might reasonably think" that a particular release could cause contamination, despite no contamination occurring. See Dedham II, 889 F.2d at 1157. A later case in the Dedham series, Dedham IV, similarly explained that to recover for a threatened release, "a plaintiff would first have to prove that it possessed a good-faith belief that some action was desirable in order to address a particular environmental threat. The plaintiff would then have to demonstrate that its response to the perceived threat was objectively reasonable." Dedham IV, 972 F.2d 453, 458 n. 2 (1st Cir. 1992).

Other courts, including in Amoco Oil, Castaic Lake, and KRSG I, have addressed the question from a similar perspective (though using different language). For example, in Amoco Oil, the court stated that "the question of whether a release has caused the incurrence of response costs should rest upon a factual inquiry into the circumstances of a case and the relevant factual inquiry should focus on whether the particular hazard justified any response actions." Amoco Oil, 889 F.2d at 670 (emphasis added).

The court in Castaic Lake also performed a similar analysis, requiring evidence of a "plausible migration pathway." Like Dedham IV and Amoco Oil, Castaic Lake's inquiry asks whether the plaintiff, in incurring cleanup costs, correctly assessed the likelihood that released contaminants plausibly could have traveled from one site to another. If it was not "plausible" that the contaminants could have traveled to the area that was cleaned up, a plaintiff's response costs are not compensable.

KRSG I likewise evaluated the plausibility of a migration pathway. See Kalamazoo River Study Group ("KRSG I"), 171 F.3d 1065, 1068-70 (6th Cir. 1999). The defendant there argued that "[polychlorinated biphenyls] from [the defendant's] facility have not and could not have migrated all the way down [a drainage] ditch to Morrow Lake and into the Kalamazoo River[.]" Id. at 1068. Though the plaintiff's expert opined that there was "sufficient waterflow" to carry the contaminants through the entire length of the ditch—this theory based on nothing more than the expert's "speculation" as to the "possibility" that the contaminants could have traveled through the drainage ditch. Id. at 1072. Without any "affirmative evidence" that the waterflow could have carried contaminants into the lake and move to the Kalamazoo River, it was merely possible, not plausible, that Defendants releases caused Plaintiffs to incur response costs. Id.; see also Asarco, 21 F. Supp. 3d at 807 (quoting KRSG I, 171 F.3d at 1072).¹² Finally, in Thomas v. FAG Bearings Corp., 846 F. Supp. 1382 (W.D. Mo.), on reconsideration in part, 860 F. Supp. 663 (W.D. Mo. 1994), the court held that a different release causation standard applied depending on which cause of response costs—a release, or contamination itself—was more proximate or immediate. See id. at 1390.¹³ This too asks whether a migration pathway was plausible: if there was little dispute that a release caused contamination, Thomas reasoned that a lesser showing was appropriate, because the nexus between the release and response costs would be readily apparent. Id. However, if a plaintiff responded to contamination alone, and the nexus between the release and contamination was tenuous, Thomas held that a greater showing was needed. Id.

12 Though Plaintiffs argue that KRSG I imposed an "actual contamination" requirement, the Sixth Circuit's holding does not rest on such a requirement. Rather, the plaintiff's evidence was too speculative to show that contaminants either could have, or did travel, to the Kalamazoo River. See KRSG I, 171 F.3d at 1072. Moreover, though the Sixth Circuit later recognized that applying a stricter, defendant-specific causation requirement in a contribution action was erroneous, it did not disturb its previous holding or assert that the showing required in KRSG I was somehow incorrect. See Kalamazoo River Study Grp. v. Menasha Corp. ("KRSG II"), 228 F.3d 648, 656 (6th Cir. 2000).

13 This is where the distinction between "one site" and "two site" cases arises from: if a release, contamination, and cleanup occur at one site, establishing a chain of liability is relatively simple. But if a release occurs at one site and is alleged to have traveled to a second site that a plaintiff cleaned up, it is logical to require some showing that the plaintiff was responding to the release at the first site. Though this is what Thomas attempted to explain. Thomas framed the legal standard in an unworkable and confusing way: it stated that a lesser showing is required "where die plaintiff can show that the release or threatened release by the defendant, and not the actual contamination, caused the plaintiff to incur response costs. However, where the response costs are incurred solely as a result of and in response to the actual contamination, the plaintiff must prove that the release by the defendant actually caused the contamination at plaintiffs site[.]" Id. at 1391.

As a legal principle, this plausibility question is part and parcel of causation inquiries generally: it simply reflects the distinction between "but-for" cause and something resembling a "scope of liability" or traditional tort "proximate cause" question. See, e.g., 6 Witkin, Summary of Cal. Law (11th), Torts §§ 1335-36 (2022).¹⁴ If certain response costs were implausible, because contamination itself was implausible—such costs would fall outside the scope of liability, and a court could not find that Defendants' releases caused the incurrence of response costs. See Dedham IV, 972 F.2d at 458 n. 2; Amoco Oil, 889 F.2d at 670.

14 Bliss and Wade are often cited for a principle that "die structure of CERCLA and its legislative history make it clear that traditional tort notions, such as proximate cause, do not apply." See Bliss, 667 F. Supp. at 1309 (citing United States v. Wade, 577 F.Supp. 1326, 1332-33 (E.D. Pa. 1983)). Again, however, these broad pronouncements are made in the context of establishing a specific Defendant's connection to the contaminants—defendant-specific causation. That is not the issue here.

This is also consistent with the principle that Congress intended to place some limit on liability by including a causation requirement in Section 107(a). For example, Thomas described, "A party who discovers TCE groundwater contamination in Missouri could successfully sue every party who released TCE in the entire country ... [T]here is no indication that Congress intended this absurd result." Thomas, 846 F. Supp. at 1387; see also 2 Law of Hazardous Waste: Management, Cleanup, Liability, and Litigation § 14.01[4][d] (S. Cooke ed. 1988) ("There are potential statutory and equitable problems with holding defendants associated with one site liable for an independent set of environmental harms that by coincidence occurred nearby.").

Or, returning to the Court's earlier example, if the Vernon Plant had released lead for only ten minutes, caused significant public outcry, and caused Plaintiffs to incur $150 million in cleanup costs, the "release" would still be a but-for cause of the cleanup. In that hypothetical, the nexus between the release and response costs

would be insufficient: lead is ubiquitous in urban areas like Vernon, and there would be little reason to conclude that a ten-minute release could cause the contamination in the PIA. A finding of liability would run afoul of Congress' intent in including a causation requirement in the language of Section 107(a). See Artesian Water, 659 F. Supp. at 1282 ("CERCLA's strict liability scheme does not diminish the necessity of demonstrating a causal connection between a release or threatened release and the incurrence of costs by a section 107 plaintiff.").¹⁵

15 Of course, the Court must consider Section 107(a)'s explicit textual requirement in tandem with CERCLA's general policy goal of "provid[ing] the federal government with the means to effectively control the spread of hazardous materials from inactive and abandoned waste disposal sites." Kaiser Aluminum & Chem. Corp. v. Catellus Dev. Corp., 976 F.2d 1338, 1340 (9th Cir. 1992) (citation omitted). CERCLA "intended to affix the ultimate cost of cleaning up these disposal sites to the parties responsible for the contamination." Id. Along these lines, courts have recognized that "[d]ifficult proof problems are inherent in hazardous-waste cases: the co-mingling and migration of wastes at a disposal site make identification of sources scientifically difficult and economically infeasible." Asarco LLC, 21 F. Supp. 3d at 807 (cleaned up); see also Artesian Water, 659 F. Supp. at 1282 ("From a technological standpoint, [the plaintiff's ability] to 'fingerprint' the [contaminant] in the groundwater ... is exceedingly doubtful. To impose such a requirement might permit the owners and operators of both facilities to avoid financial responsibility for the cleanup, and would thus eviscerate section 107.").

Thus, Plaintiffs cannot satisfy their causation burden by selecting a massive area of land and showing that one particle may have landed in a 6400-acre area. Plaintiffs, alternatively, have not established that lead traveled throughout or to any particular part of the PIA, including the Residential Areas. Rather, Plaintiffs' showing must turn on the likelihood of contamination throughout the Residential Areas. This is especially given that this case deals with soil contamination. In other cases, like those with contamination of water, even a trace amount of contaminants may propagate throughout an entire body of water, like a well or an aquifer.

In such a situation. Plaintiffs' proposed "one particle" approach might be more tenable. In the context of soil contamination, however, the presence of lead at one parcel does not necessarily mean that lead will also be present at an adjoining parcel.¹⁶ Yet that is what Plaintiffs' approach suggests: Plaintiffs contend that one particle of contamination over 6,400 acres of soil allows them to recover response costs for all 6,400 acres. While the size of a given site presents a difficult line-drawing question that the Court need not explicitly resolve, Plaintiffs' approach would lead to an absurd and unworkable result: that one speck of lead in a 6,400 acre area triggered Plaintiffs' expenditure of millions in remediation costs. See Thomas, 846 F. Supp. at 1387 ("A party who discovers TCE groundwater contamination in Missouri could successfully sue every party who released TCE in the entire country. Although CERCLA's liability provisions are far-reaching, there is no indication that Congress intended this absurd result.").

16 While evidence at trial showing the possibility of continued lead migration based on fugitive emissions—for example, if lead were lifted from the ground by a vehicle driving over the contaminated land—Plaintiffs' evidence explicitly suggested that such fugitive emissions were limited to the Industrial Area. [Tr. 2, at 142:14-20].

To summarize, the various ways that courts have framed CERCLA's causation standard—whether a "causal nexus," "justification," or "plausible migration pathway"—all attempt to answer a question about the proximity of the relationship

between the release and the response costs. Though the clearest connection between a release and response costs is actual contamination, CERCLA does not require a showing of actual contamination. See Dedham II, 889 F.2d at 1157. The above case law nevertheless makes clear that to satisfy CERCLA's causation requirement, Plaintiffs must establish a "plausible," see Castaic Lake, 272 F. Supp. 2d at 1068, "objectively reasonable," see Dedham IV, 972 F.2d at 458, n. 2, or "justified," see Amoco Oil, 889 F.2d at 670, theory of contamination to satisfy CERCLA's causation requirement.¹⁷

17 Plaintiffs reference A & W Smelter & Refiners, Inc. v. Clinton, 146 F.3d 1107, 1111 (9th Cir. 1998), where the Ninth Circuit stated in dicta that "[w]here a party is responsible for a particular release, that party is a cause of any response to that release, although on occasion EPA overzealousness may be another cause as well. The Fifth Circuit rather explicitly adopted [a] strained definition of the causation requirement as a way of getting around the lack of a minimum level requirement. [See Amoco Oil, 889 F.2d at 670]. We sympathize, but we believe it is not our function to read into the statute a limitation that Congress did not put there." Id. The Court is unaware of any other case that has interpreted the release causation requirement so expansively or gone so far as to say that "[w]here a party is responsible for a particular release, that party is a cause of any response to that release." See id. To the contrary, every one of the above cases discussed by the Court has interpreted the statute to require a greater showing.

* * *

This leads the Court to a related question: the extent to which Plaintiffs' self-imposed cleanup threshold of 80 parts per million ("ppm") bears on the causation standard. The threshold imposes a minimum level of contamination at which Plaintiffs would clean up the PLA. At trial, DTSC employee Peter Ruttan testified that Plaintiffs' official policy imposed a standard of cleaning up properties only where soil lead levels exceeded 80 ppm. See Tr. 2, at 62:13-63:4.

On redirect examination, Ruttan explained that "DTSC entered into an agreement with Exide ... whereby if Exide could establish that the background was higher than 80 ppm, then Exide would be allowed to only clean up properties that are above that higher background level[.]" Id. at 65:4-13; see also Tr. 1, at 16:20-22 ("[Cleanup was] caused not by the cleanup standard but because the level of lead in the [PIA] is above background."). However. Ruttan then testified that because the Long Beach background study established background of less than 80 ppm, the cleanup goal was set at 80 ppm. Tr. 2, at 65:19-66:1.

Plaintiffs argue that the 80-ppm threshold is not relevant to the causation inquiry because CERCLA's causation requirement does not specify a minimum threshold of contamination for releases to "cause" response costs. Plaintiffs rely on several cases that declined to impose or apply a "minimum quantity" threshold in a different context—determining whether a substance is "hazardous" under CERCLA. See A&W Smelter, 146 F.3d at 1110-11; Mid Valley Bank v. N. Valley Bank, 764 F. Supp. 1377, 1386-87 (E.D. Cal. 1991); Citizens Dev. Corp., Inc. v. Cty. of San Diego, 2017 WL 1089549, at *7 (S.D. Cal. Mar. 23, 2017).

In A & W Smelters, for example, the Ninth Circuit held that because CERCLA did not specify some threshold quantity that would render a substance "hazardous," even trace amounts of a contaminant like lead would satisfy CERCLA's "hazardous substance" requirement. See A & W Smelter, 146 F.3d at 1111. In other words, the court declined to read into the statute some minimum threshold requirement because "[n]othing in the law suggests that quantities of a hazardous substance

below its reportable level render it no longer hazardous." Id. at 1110. Mid Valley Bank and Citizens Dev. Corp. held similarly.

In response. Defendants argue that Plaintiffs' cleanup threshold "defines" causation in this case, because Plaintiffs' threshold defines the amount of contamination necessary for Defendants' releases to "cause" Plaintiffs to incur response costs. As put by Defendants, for example, "[T]he agency has already determined that a lead contribution of—for example—1 ppm does not cause' the agency to incur response costs." See Post-Trial Rep. at 11. Defendants cite NCR Corp., which, in discussing apportionment, noted that the defendant was an independently sufficient cause of contamination where its contamination exceeded the EPA's "maximum safety threshold." See United States v. NCR Corp., 688 F.3d 833, 839 (7th Cir. 2012).

Though this presents a close question, Plaintiffs have the better of the argument: requiring proof of a minimum causation threshold is inconsistent with CERCLA's text, Congress's intent in enacting the statute, and subsequent case law. As an initial matter, the statute itself does not require any sort of minimum showing—nor even a showing of actual contamination, as discussed above—for a finding of liability. Thus, adopting Defendants' view would require reading into the statute a minimum showing requirement, counter to its plain text.

Second, requiring proof of 80 ppm would present an unworkable causation inquiry. For example, if Defendants' releases caused an amount of contamination below 80 ppm—e.g., 50 ppm, 20 ppm, or 10 ppm—Defendants' releases alone would not have caused Plaintiffs to incur response costs. But even in the case where Defendants' releases caused only 10 ppm of contamination. Plaintiffs would still be responding, at least in part, to the 10 ppm of Defendants' contamination. Defendants ask the Court to conclude that this is insufficient. But this is inherently unworkable on the facts of this case, where it is undisputed that Plaintiffs were responding to multiple causes of lead contamination in the PIA, like lead-based paint, leaded gasoline, and smelter-related lead from sources other than the Vernon Plant.

Third, as the Court discussed above, Defendants' argument rests largely on Boeing. But Boeing did not hold that a plaintiff is required to show either but-for causation or causal overdetermination to prevail on a CERCLA claim. Indeed, such a showing would also be impossible on the facts of this case and stand in tension with CERCLA's general policy goal of "promptly cleaning up hazardous waste sites and making polluters, rather than society as a whole, pay." Pakootas, 905 F.3d at 584.

Finally, "[t]he liability standard for contribution claims is the same as the standard for cost recovery claims." KRSG II, 228 F.3d at 656. In other words, though Defendants argue that they are left "holding the bag" for all of the contamination in the PIA, CERCLA's lower standard at this stage enables them to implead third parties and resolve the case more easily via questions of divisibility, allocation, and contribution. This is what Congress intended in enacting CERCLA's statutory scheme. The Court must interpret CERCLA liberally to effectuate Congress' intent.

Accordingly, the Court holds that Plaintiffs bear the burden of establishing that lead plausibly caused contamination throughout the PIA's Residential Areas. Critically, plausibility is more than possibility. See Asarco, 21 F. Supp. 3d at 807 (quoting KRSG I, 171 F.3d at 1072 (finding assumption based "solely on speculation and possibility" insufficient. "Rather, the Court must conduct an inquiry into the

facts as a whole—considering factors such as the location of the cleanup site and defendant's facility, the geology and hydrology of the area, and the nature and quantity of the contamination" to determine whether Defendants' releases plausibly could have caused contamination throughout the Residential Areas. See id.

The Court next applies this framework to the facts of this case.

IV. DISCUSSION

Causation presents significant questions for courts tasked with the interpretation of scientific evidence. See REFERENCE MANUAL ON SCIENTIFIC EVIDENCE (THIRD), at xiii-xiv (2011). "Scientific inference typically requires consideration of numerous findings, which, when considered alone, may not individually prove the contention." Id. at 19-20. "It appears that many of the most well-respected and prestigious scientific bodies ... consider all the relevant available scientific evidence, taken as a whole, to determine which conclusion or hypothesis regarding a causal claim is best supported by the body of evidence." Id. at 20 (citing Milward v. Acuity Specialty Prod. Grp., Inc., 639 F.3d 11, 18 (1st Cir. 2011) ("No matter what [scientific] methodology is used, 'an evaluation of data and scientific evidence to determine whether an inference of causation is appropriate requires judgment and interpretation.'")).

The Court considers the various lines of evidence presented by the parties and their experts. "Since this was a bench trial involving primarily battling experts, it is up to the Court to weigh the evidence presented and attempt to ascertain the truth." Coeur D Alene Tribe v. Asarco Inc., 280 F. Supp. 2d 1094, 1103 (D. Idaho 2003).¹⁸ As mentioned, there are three primary lines of evidence in dispute: (1) background, (2) statistical analyses—which are related to wind data; and (3) soil chemistry.

18 As discussed in detail below, the Court finds all expert testimony discussed here admissible under the standard articulated in Daubert v. Merrell Dow Pharmaceuticals, 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993). The Court, rather, does not find Plaintiff's experts persuasive aiding its inquiry into the cause of lead contamination in the PIA and concludes that Plaintiffs are unable to satisfy their burden to establish release causation.

In its subsequent discussion, the Court addresses each line of evidence individually. The Court then weighs the various lines of evidence and concludes: (1) the background evidence weighs in Defendants' favor; (2) the statistical analyses weigh in Defendants' favor—though the wind modeling weighs slightly in Plaintiffs' favor; and (3) the soil chemistry evidence weighs slightly in Defendants' favor. After considering the totality of the evidence, the Court concludes that Plaintiffs have not met their burden to establish that Defendants' releases plausibly could have caused contamination in the PIA's Residential Areas.

A. "BACKGROUND" LEAD LEVEL

To determine the extent of possible contamination in the PLA, the parties rely on a concept known as "background," which presents a theoretical baseline of "the amount of lead that would be in the soil around the Vernon Plant if the Vernon Plant had never existed." See Trial Tr. Day 1, 25:12-15; see Medine Decl. ¶ 50. Comparing contaminant levels to a background reference area is an EPA-approved method for determining the extent of contamination at a CERCLA site. At trial, the parties disputed the extent to which three

different background studies—the Long Beach Study, the Waterstone Study, and the Wu Study—can establish contamination in the PLA.

For the below reasons, the Court finds that the evidence of background weighs in Defendants' favor.

1. Summary of Studies

Plaintiff summarizes the results of these studies as follows:

Reference Background PIA Sample Area Average Average Location Long Beach 63 153 Open Yard Waterstone 214 333 Dripline Wu 181 256 All

See PX054.¹⁹

19 The Wu Study is only of limited importance here: the parties' Post-Trial Briefs, Dr. Singh, and Dr. Shields did not address it. Dr. Medine briefly mentioned the study in his declaration, but concluded that it was not as appropriate as the Long Beach and Waterstone Studies for "as direct a comparison" with the PIA. See Medine Decl. ¶¶ 78-79. At trial. Plaintiffs only mentioned the study in passing to suggest that the 75 ppm difference between the Wu PIA Average and the Wu Background Average indicated that contamination from other sources was greater than the contribution from the Vernon Plant. Given its minimal evidentiary value, the Court need not address the Wu Study. The Court also notes, for sake of clarity, that the Wu Study is distinct from a different study Defendants rely on called the Wu & Johnston Study. See DX080.

Conducted by Advanced GeoServices, a contractor retained by Exide, the Long Beach Study was a study of a site located approximately eleven miles from the southern border of the PIA. Medine Decl. ¶¶ 61-64. Advanced GeoServices selected the site because of "its proximity to major freeways, a historically industrial area absent the Exide Facility or other secondary lead smelter, and a sizable rail yard with intermodal facility and switching yard." See JX003.00010. "The housing stock [was] similar in age, size, and density" to the PIA. Id.; see also JX008.00071-72; Singh Decl. ¶ 45.

The study relied on soil samples taken at several depths from the yards of 19 Long Beach homes. See Singh Decl. ¶ 47. The samples at each depth for the home were then combined to form a single composite sample. The study then sampled 19 homes in the Northern PIA and 20 homes in the Southern PIA using a similar technique. The Long Beach data is summarized at PX088 and in Dr. Singh's declaration. See Singh Decl. ¶ 18. Dr. Singh estimated that, on average, the PIA's soil lead level was 256 ppm. See Singh Decl. 4. The average soil lead level for the 39 PIA homes referenced in the Long Beach Study was "175 ppm for the Northern PIA and 131 ppm for the Southern PIA." Id. ¶ 6. By comparison, the average background level as based on the Long Beach study was approximately 77-82 ppm. Tr. 2, at 30:10-12; see Singh Decl. ¶ 48.²⁰ Thus, this amounts to an approximately 100 ppm difference. Second is the Waterstone Study, which plays a more limited role in the parties' dispute than the Long Beach Study. The Waterstone Study, also performed by a contractor retained by Exide, was based on dripline data from soil lead investigations taken from 30 school districts throughout Los Angeles. See PX053.00010-11; see also PX091 (sample locations). The Waterstone Study authors wrote that the thirty sites were selected out of nearly a hundred options because they had "conditions similar to those found at the Vernon Facility in urban areas with roadways, freeways, and industrial uses." Tr. 3, at 32:8-13.

20 Dr. Singh's testimony at trial was that she performed her own statistical analysis of the Long Beach Study samples and calculated her own background number. See Tr. 1, at 206:16-21; Tr. 2, at 27:24-28:1 ("I just computed the background for Waterstone. I computed background for Long Beach."). However, Dr. Singh's declaration appeals to quote from the Long Beach study itself. See Singh Decl. ¶ 48 (citing PX092). Regardless, if Dr. Singh did perform the calculation herself, as she stated. Defendants do not take issue with her methodology in performing the calculation—so any difference there is trivial. Defendants also cross-examined Dr. Singh about a third calculation she performed: a "site-extracted background" level as calculated in Dr. Singh's export rebuttal report. See Tr. 1, at 213:4-16; 218:18-19. To perform this calculation. Dr. Singh used a software called ProUCL and evaluated the approximately 146,232 samples of soil in the PIA and determined that "mean of the site-extracted background population is 147 milligrams per kilogram and UCL95 is 148.2 milligrams per kilogram." See Tr. 1, at 213:12-16; 226:3-228:4. Dr. Singh did not testify to this calculation in her declaration, and stated that it did not "become part of [her] conclusion in this case." Tr. 2., at 33:4-6. Rather, Dr. Singh only performed the calculation to rebut the position of a defense expert who was initially to be a pail of the trial. Dr. Henderson. See Tr. 2. at 31:8-14.

The Study's average dripline background lead level was 214 ppm. See PX05.00019; see also Medine Decl. ¶ 74; Singh Decl. ¶ 7. Comparatively, the dripline lead data in the PIA shows lead contamination of 333 pm. Id. This amounts to a 119-ppm difference between the PIA's Residential Areas and the Waterstone Study.

2. Analysis

Whether the Long Beach and Waterstone studies provide appropriate background measures presents a close question. The average lead level in the PLA's Residential Areas is 90 ppm greater than the Long Beach reference area; the average lead dripline level in the PIA's Residential Areas is 119 ppm greater than the Waterstone reference area. According to Plaintiffs, these differences show that the Vernon Plant's releases contributed to contamination in the PIA. See PX054; see Singh Decl. ¶¶ 4-7.²¹

21 Dr. Medine testified that he did not hold himself out as an expert in background in this case, see Tr. Day 1 at 92:11-22, and that he would defer to Dr. Singh's testimony on the question, see id. at 93:3-94:17. However, he also stated that his "knowledge of background analyses" generally allows him to "examine the results" and "use [them]" if needed. Id. at 94:2-5. Given that Dr. Mediae's analysis of the Long Beach Study is largely duplicative of Dr. Singh's as to background, and because Plaintiffs rely on Dr. Singh's analysis of the study, the Court finds it unnecessary to discuss Dr. Medine's testimony separately.

But for the above studies to provide a suitable baseline from which to evaluate the PIA's lead contamination, each reference area and its soil samples must be sufficiently similar to the PIA and the PIA soil samples. See, e.g., Singh Decl. ¶ 20 (evaluating background requires an "apples to apples" comparison); JX011.00025 ("Background samples should be collected from areas near the site that are not influenced by site contamination, but that have the same basic characteristics (e.g., soil type, land use).").

Here. Plaintiffs have not satisfied their burden to produce evidence that the Long Beach and Waterstone studies relied on sample areas that were sufficiently comparable to the PIA. Though the studies each included brief explanations of their choices of reference areas, noting the areas' similar "industrial" histories, there is insufficient evidence that the studies considered a type of industrial history more relevant to the PIA: the historical presence of lead-emitting

industries. The PIA included "roughly 355 fixed sources of industrial lead in the area, including at least half a dozen other major lead smelters, foundries, and comparable lead sources that were closer to the residential areas than the Vernon Plant." See Shields Decl. ¶¶ 55-58. But there is no evidence that either the Long Beach Study area or the Waterstone Study reference area included the lead-emitting industries present in the PIA. Tr. 3, at 78:23-79:13; see Shields Decl. ¶ 63.

Dr. Kate Burger, a former DTSC employee, affirmed this view of the Long Beach Study. In 2015, Dr. Burger prepared a preliminary report estimating the Vernon Plant's impact in the PIA (the "2015 Report"). Tr. 2, at 112:10-15; see JX010, Report. However, Dr. Burger did not use the Long Beach Study to estimate background in the 2015 Report. Tr. 2, at 115:7-15.

At her deposition, Dr. Burger testified that she did not do so because "the [Long Beach] lead levels came back relatively low. Didn't seem representative of the City of Vernon." Id., at 116:16-117:4. Defense counsel then asked Dr. Burger, "It was your understanding that the Long Beach study had a different industrial history than the area around Vernon; collect?" Tr. 2, at 118:8-10. Dr. Burger answered, "Yes. It would have been impossible to match the industrial histories at both locations." Id. 118:11-12. She also stated that the Long Beach Study results were too low to be an accurate proxy for Vernon when she assessed the initial 20,000 soil samples taken from the PIA. See Tr. 2, at 116:11-117:7.²²

22 On cross examination. Plaintiffs' counsel suggested that Dr. Burger did not rely on the Long Beach Study for some other reason. First, Plaintiffs' counsel adduced evidence that Dr. Burger relied on approximately 20,000 PIA samples in drafting her report, though the current universe of PIA samples includes approximately 146,000 samples. See id. at 119:11-22.
Second, Plaintiffs' counsel attempted to suggest that Dr. Burger did not rely on the Long Beach Study because the Long Beach Study was based only on yard samples, rather than dripline and roadway data, which the PIA did include. Id., at 122:12-25. Dr. Burger's testimony on this point was initially unclear, so the Court asked her to explain how the different sampling methods affected Dr. Burger's approach to the Long Beach study. See id., at 123:6-125:25. Dr. Burger's testimony remained unclear: at most, Dr. Burger's testimony suggested that the different sampling data may have been a different reason why she did not to use the Long Beach study. Ultimately, Dr. Burger stated that she did not rely on the Long Beach Study because she "was trying to estimate what the urban lead levels appeared to be closer to the Vernon Plant." Id., at 125:22-25.

Defendants contend that the Waterstone Study "suffers from the same flaw as Long Beach: its data were collected in 30 school districts across Los Angeles and there is no indication that the industrial profile of those 30 districts bears any relation to the Vernon area." See Def. Post-Trial Br., at 22 (citing Tr. 2, at 43:23-44:3); see Shields Decl. ¶ 63. Dr. Singh, by contrast, testified that the Waterstone Study was a reliable point of comparison for the dripline-²³ average in the PIA, see Singh Decl. ¶¶ 7, 20. Dr. Shields testified that the Waterstone Study was generally a "good background reference area" for the PIA, however, he had reservations about the study as source of data because it also considered "a much less industrial area on average than the

23 The dripline "accumulates lead from lead paint and other aerial sources of lead deposited on the roof and walls of homes," and is "typically the highest concentration of lead on a parcel[.]" Id.; see also Tr. 1, at 133:9-19 ("The Waterstone study was really a worst case scenario for looking at a drip line lead. In addition to focusing on eroding paint or areas of the building that were in disrepair, the Waterstone study also included industrial areas, the effects of leaded gasoline, and other types of sources.").

PIA." See Tr. 3, at 79:2-6; see also id., at 30:10-31:5.

On this question, Dr. Shields' testimony was significantly more credible than Dr. Singh's, who had initially testified that the Waterstone data points "were comparable in every which way to the residences in PIA like industrial area, closer to the highways and, you know, all that," Tr. 2, at 25:6-11, but then later testified that she did not know the sample sites' proximity to industry—and that she simply assumed they were sufficiently similar to the PIA's Residential Areas because "[t]hose areas were chosen by the defendants[.]" Tr. 2, at 43:23-44:5. And when Dr. Singh performed a "site-extracted background calculation," she concluded the PIA's background number was 147 ppm, a different number from the Long Beach and the Waterstone background averages.

The age of homes at the background reference areas also raised uncertainty in the use of the Long Beach study as an appropriate measure of background. Dr. Shields opined that the houses in Long Beach were too new compared to the older housing stock in the PIA. See Tr. 3, at 90:15-91:3 ("The median age of the homes in the PIA is 1927 for the date of original construction, and the comparable statistic for the Long Beach is 1950 or 1951."). This is contrary to Dr. Singh's conclusion that the "Long Beach Study area has comparable housing stock with a median age of home construction of 1950." See Singh Decl. ¶ 50. The date of construction is important because it is considered a "surrogate" for the effects of various sources of lead over time—like lead-based paint, gasoline, and atmospheric deposition. Tr. 3, at 61:23-62:6.

On cross examination, Dr. Shields agreed that the rough difference in lead concentrations in the soil between homes built in 1950-1960 and homes built in 1920-1930 was approximately 40 ppm. See Tr. 3, at 95:2-9; see JX009 (depicting statistical evidence of home age). Plaintiffs argue that the difference between the Long Beach and Waterstone studies and the PIA average is nearly 100 pm—significantly greater than 40 ppm—meaning that home age could not account for the entire difference in these background figures.

But this evidence does not resolve the Court's uncertainty. It would require the Court to make a speculative and unsupported inference about the PIA's lead levels. There is also an absence of evidence as to the accuracy with which the studies were performed. Though the average lead values in the Residential Areas were approximately 100 ppm greater than the background values, there is no analysis before the Court about the contractors' decision-making process or the ways they conducted the studies. Plaintiffs only presented the studies, which refer briefly to the reasons the reference areas were chosen, and argue that the studies were performed correctly because they were performed by contractors retained by Exide. However, Plaintiffs could have chosen to present testimony of one or both of these contractors, or could have tasked Dr. Medine, Dr. Singh, or another expert with evaluating the reference areas' industrial histories and the integrity of the studies themselves. Even though Plaintiffs bear the burden of presenting this evidence to establish release causation, they did not do so. Plaintiffs also did not present any evidence as to the contractors' qualifications, educational and experiential pedigrees, or methodology.

Plaintiffs argue, in response, that the Vernon Plant's emissions accounted for more than 98% of the total lead releases reported in the PIA. See Medine Decl. ¶ 33. Plaintiffs would have the Court disregard any uncertainty because they argue that, regardless of the background number, this release evidence is sufficient to

conclude that the PIA's average contamination had to have exceeded background.

But the uncertainty about the fit of each background area is fatal to Plaintiffs' argument. Without a plausible measure of the amount of lead in the PIA's Residential Areas if the Vernon Plant had not existed, the Court cannot compare the average lead levels to the respective background measures, regardless of the percentage of releases caused by the Vernon Plant. In essence, Plaintiffs contend that the Court must trust the studies, even though there is little evidence that allows the Court to do more than speculate as to whether the background areas were chosen correctly or are sufficiently similar to the PIA's Residential Areas.

To the contrary, the background areas are unlike the PIA's Residential Areas given the PIA's 355 fixed sources of industrial lead in the area, including at least half a dozen other major lead smelters, foundries, and comparable lead sources that were closer to the residential areas than the Vernon Plant. Though both studies reference the historical presence of industry in their respective reference areas, they do so without mention of whether these were lead-emitting industries, a critical point for evaluating the fit of the background studies to this case.

Moreover, Plaintiffs have not established that the approximately nineteen homes sampled in the Long Beach Study or the thirty school districts used in the Waterstone Study provide a sufficiently large sample size to generate an accurate background figure. The sizes of these studies differ greatly from the hundreds of thousands of samples taken in the PIA.

The evidence of background here does support Plaintiffs' theory of contamination throughout the PIA in general. But though Plaintiffs concluded that the PIA extended 1.7 miles away from the Vernon Plant, the background evidence does not support this determination of distance. As with Plaintiffs' other experts. Dr. Singh did not determine whether lead released from the Vernon Plant migrated throughout the PIA, or to any one portion of the PIA. See Tr. 1, at 209:17-20.

Thus, even if the Vernon Plant accounted for 98% of the releases in the PLA,²⁴ the background evidence does not show where those releases actually landed. The analysis of the scope of the contamination is particularly relevant here because differences in the scope could account for millions of dollars of remediation. Without more, the evidence is insufficient to conclude that Plaintiffs have met their burden to show that contaminants plausibly migrated throughout the PIA. See, e.g., Licciardi v. Murphy Oil USA, Inc., 111 F.3d 396, 398-99 (5th Cir. 1997) (reversing the district court's finding of liability as to lead contamination where the district court relied on a background study, the U.S. Geological Survey, which not only was not "establishe[] ... as a standard, requirement, or criterion," but also because it was "not clear that the Survey figure used by the district court was relevant; according to an expert who testified at trial, the 'background level' relied upon by the district court was established '30-50 miles away from the [site]'").

24 As discussed elsewhere, we do not accept this finding. Briefly, the Court does not accept this statement because the Court could not trace the basis for this opinion given by Plaintiffs' expert, Dr. Medine. Plaintiff also did not satisfactorily account for this industrial history. The statement that 98% of the releases in the PIA came from the Vernon Plant is thus simply a conclusion, inconsistent with the historical and scientific evidence presented in this case. See discussion infra, Part IV.D.

Therefore, the background studies weigh against a finding that the Vernon Plant plausibly could have caused contamination

in the entirety of the PIA's Residential Areas.

B. ANALYSIS OF CONTAMINATION

Next, the parties dispute the extent to which soil contamination in the PIA is consistent with a pattern of contamination expected from a point source like the Vernon Plant. The parties agree that, generally, a "decreasing gradient" pattern would best represent contamination from a point source—which means that as distance from the source increases, concentrations of a contaminant decrease.

The parties also agree that contamination in the PIA does not "clearly" show a decreasing gradient—but each side offers its own explanation as to why. Plaintiffs argue that the lack of a decreasing gradient is a result of various confounding factors that, over the years, have masked patterns throughout the PIA. Defendants argue that there is no decreasing gradient of contaminants because the Vernon Plant did not cause a detectable amount of contamination in the PIA.

There are several lines of relevant evidence to this question, including wind modeling and statistical analysis. The Court discusses each in turn.

1. Wind Modeling

The Court begins by discussing the PIA's prevailing winds. The Vernon Plant is located "at the confluence of winds that come in from the Santa Monica Bay and from Long Beach." Tr. 2., at 106:13-15. Wind carries small particles, so a point source generally deposits contaminants in a pattern consistent with that area's prevailing winds. Dr. Medine, Mr. Chinkin, and Dr. Shields agree that they would expect to see higher lead concentrations in the predominant wind directions. See Tr. 1, at 41:11-16; Chinkin Decl. ¶ 14; Shields Decl. ¶ 21.

The parties agree, accordingly, that the Court must consider the multi-directional impact of winds on the Vernon Plant's emissions to assess evidence of contamination.

i. Predominant Wind Direction

The primary line of wind evidence is based on a Health Risk Assessment ("HRA") prepared by the Exide consultant Environ in 2013. See Revised AB2588 Health Risk Assessment, PX007. As part of the HRA, Environ conducted air modeling using an Environmental Protection Agency model called AERMOD. AERMOD is a mathematical model that predicts air dispersion of contaminants from a variety of emission sources. See Machado Decl. ¶ 17; see Tr. 2, at 90:18-21. The results of the model are reproduced below. Image materials not available for display.

Each of the colored lines in the image is a "contour line" or an "isopleth," which depicts average ambient lead concentrations based on the amount of airborne lead in an area as predicted by the AERMOD model. The parties rely on this figure—and its underlying data—for two reasons.

First, Defendants rely on the figure as evidence of the predominant winds in the PIA: the exhibit shows that the largest lobes of the isopleths are, in order of decreasing magnitude, to the southwest, east, and then north. See id.; see also Tr. 1, at 77:17-79:21. Thus, Defendants argue that southwest, east, and north were the predominant wind directions in the PIA. See Shields Decl. ¶ 21. On cross-examination, Plaintiffs' expert Dr. Medine agreed—after some back-and-forth—that this model showed that the southwest, east, and north wind directions were the predominant wind directions in the PIA. See Tr. 1, at 79:6-21.

Plaintiffs indicate, however, that wind data AERMOD relied on was not collected at the Vernon Plant, but offsite. For example, the Resurrection Catholic High School was the closest location to the Vernon Plant where air monitoring has been performed.²⁵ See Medine Decl. ¶ 92. Dr. Medine explained that he "observed significant variability in prevailing directions from various wind stations and relative to the high school[,]" creating uncertainty as to the predominant wind directions.

25 The Court is uncertain if evidence was presented as to the station's distance from the Vernon Plant. Accordingly, the Court takes judicial notice of the results of a Google Maps search, which indicates that the Resurrection Catholic School is 1.9 miles away from the Vernon Plant. See Crandall v. Starbucks Corp., 249 F.Supp. 3d 1087, 1099 (N.D. Cal. 2017) ("Courts have generally taken judicial notice of facts gleaned from internet mapping tools such as Google Maps or Mapquest.").

Id. Thus, Plaintiffs contend that the AERMOD results only show predominant wind direction to a limited extent. See Post-Trial Reply, at 11; see Tr. 1, at 170:8-11.

According to Dr. Medine, the wind data collected at the Vernon Plant showed that the primary wind direction was to the northeast, and that winds also blew to the east, north, northwest and southwest. Tr. 1, at 170:12-171:8. Dr. Medine testified, however, that when he performed his directional analyses, he relied on the off-site data and the AERMOD model, because he did not have the wind data collected at the Vernon Plant. Tr. 1, at 171:9-13; see also id. at 168:4-170:2; DX295.

Plaintiffs asked defense expert Dr. Shields about the onsite and offsite wind data on cross examination. See Tr. 3, at 69:25-74:9. Dr. Shields agreed with Plaintiffs—that there was an inconsistency between the AERMOD wind directions and the site-specific wind data. Id., at 72:11-13. However, he explained that he deferred to Environ meteorologists who decided to use the wind data from the Resurrection Catholic School to perform the AERMOD modeling. Id., at 72:17-24, 74:2-9.

Indeed, the AERMOD study explained why Environ decided to use offsite wind data. See PX007.00025. Even though DTSC recommended the use of onsite meteorological data in a 2011 comment letter, the onsite data did not include as complete of a dataset as the offsite data. Id. Moreover, the offsite wind station was located closest to the site, the terrain around the station was similar to that of the Vernon Plant, and there were no intervening terrain features between the sites. Id. at PX007.00025-26.

Plaintiffs also rely on the AERMOD results as evidence of contamination in the PIA: they argue that AERMOD "demonstrates that particulate emissions, including lead, from the Vernon Plant would have been transported by the wind several miles away from the Plant." See Post-Trial Br., at 13 (citing Chinkin Decl. ¶ 24). Consistent with Defendants' argument as to the predominant wind direction, Plaintiffs' expert Mr. Chinkin opined that the HRA modeling results "demonstrate that airborne lead concentrations would have been transported at least 2.7 miles to the southwest, 1.9 miles to the east-northeast, and 1.7 miles to the north[.]" Chinkin Decl. ¶ 24.

But Plaintiffs cannot have it both ways. If the AERMOD model is inadequate because it relied on wind measurements not taken at the Vernon Plant, see Post-Trial Reply, at 11, the Plaintiffs cannot rely on the AERMOD results as evidence of the distance and direction in which contaminants traveled. Still, because both sides relied on the offsite data in performing their analyses, see e.g., Medine Decl. ¶ 132; PX058, the Court does not entirely reject AERMOD as evidence of the PIA's prevailing winds. But given the experts' agreement that there is inconsistency between the onsite and offsite wind data, the Court cannot overlook this uncertainty as to the predominant winds in its discussion of the experts' statistical analyses of contamination—particularly where, as discussed above, Plaintiffs bear the burden of establishing release causation here.

ii. Contamination

Next, Plaintiffs rely on the AERMOD results as evidence of contamination, requiring the Court to determine whether AERMOD can reliably estimate contamination in the PIA. This endeavor reveals several questions about AERMOD's ability to estimate the amount, distance, and directionality of contamination.

Defendants first challenge the reliability of AERMOD's inputs. To use AERMOD, a "modeler inputs several pieces of information into the model, including

weather data, terrain features, emissions factors, building information, and stack parameters." Chinkin Decl. ¶ 20. The model, however, is not infallible. Plaintiffs' expert Dr. Chinkin explained on cross examination that estimating model inputs is "intrinsically uncertain" and model results can "vary widely" based on inputs. Id. at 94:6-22, 98:11-13. Accordingly, Defendants argue that the AERMOD results are uncertain because no expert in this case reviewed or verified any of the model's inputs. Tr. 2, at 97:22-98:1.

This uncertainty alone is not necessarily fatal to the AERMOD results. Defendants themselves did not evaluate the inputs, and nor do they adduce any evidence to suggest that Environ's work or the South Coast Air Quality Management District's review of the modeling was somehow faulty. Like Plaintiffs' attempt to create uncertainty about the offsite wind data above, Defendants rely on AERMOD's wind inputs, but now claim that other similar inputs might be unreliable.

But, more significantly, Plaintiffs rely more broadly on AERMOD to show that lead could have traveled to some portion of the PIA's Residential Areas. Defendants contend that AERMOD's results are both unreliable and unpersuasive because AERMOD was run in "concentration" rather than "deposition" mode. Machado Decl. ¶ 17. Plaintiffs have not met their burden in answering this contention.

As a result, Defendants have the better of the argument: as the Court explains below, because AERMOD was run in concentration mode, the AERMOD results do not plausibly show contamination at the PIA's outer bounds or the amount of airborne lead that could have been deposited in particular areas.²⁶

26 Defendants move to exclude Mr. Chinkin and the HRA evidence under Daubert. See Mot. to Exclude Lyle R. Chinkin, ECF No. 386. Under Federal Rule of Evidence 702, which governs the admissibility of such testimony, "[a] witness who is qualified as an expert by knowledge, skill, experience, training, or education may testify in the form of an opinion or otherwise if: (a) the expert's scientific, technical, or other specialized knowledge will help the trier of fact to understand the evidence or to determine a fact in issue; (b) the testimony is based on sufficient facts or data; (c) the testimony is the product of reliable principles and methods; and (d) the expert has reliably applied the principles and methods to the facts of the case." See Fed. R. Evid. 702. While the way in which "reliability is evaluated may vary from case to case," United States v. Frazier, 387 F.3d 1244, 1262 (11th Cir. 2004) (en banc), in all cases, "[t]he trial judge ... must find that [the proffered testimony] is properly grounded, well-reasoned, and not speculative before it can be admitted." Fed.R.Evid. 702 Advisory Committee's note (2000 amends.). Here, to the extent the Court's subsequent discussion about AERMOD's deficiencies is consistent with Defendants' motion, the motion is granted. AERMOD cannot reliably quantify the outer bounds of contamination or the amount of contamination in the PIA's Residential Areas because it was run in concentration mode. Dr. Chinkin's failure to cite any authority for this level of analysis supports Defendants' argument. Simultaneously, however, AERMOD remains useful to support the more general proposition that some airborne lead traveled to certain undefined portions of the Residential Areas. Defendants' motion does not meaningfully challenge this claim, which, as Dr. Chinkin testified, is premised on foundational scientific principles of gravity causing particles to fall. Accordingly, the Court does not exclude the model in its entirety.

Some preliminary scientific background about a particle's airborne migration is illustrative here. When airborne particles are released from a source, they travel in a plume. See id. As the plume travels with the wind, particles fall out of the plume, a process called deposition. Id. The plume gradually depletes in size until, eventually, no plume remains. Tr. 2, at 100:5-6. When run in "concentration mode," AERMOD calculates airborne particle concentrations at varying distances from the emission source. Machado Decl. ¶ 17. Unrealistically, however, running AERMOD in concentration mode assumes that no deposition occurs. Thus, concentration mode "overstate[s] how far particles will travel, because it knowingly falsely assumes that none of those particles will ever fall to the ground." Id. ²⁷ AERMOD can be run in deposition mode—where it performs an additional calculation "to determine how much airborne material is dropping to the ground at specified distances from the emission source," see id.—but it was not run in deposition mode in this case. The Court therefore finds this evidentiary gap troublesome, as it shows that Plaintiff has not met its burden in showing exactly how much, and where, the lead particles may have landed.

27 Given the proposes of the HRA, running AERMOD in concentration mode is reasonable: an ERA takes AERMOD's output of airborne lead concentration and multiplies it by a risk factor "to determine what risk the public is being exposed to." Tr. 2, at 92:7-12. In other words, the HRA is designed to estimate a worst-case scenario for health risk. It was not designed to estimate an accurate amount of contamination in a given area.

There are additional problems with the AERMOD use in this case. Running AERMOD in concentration mode will find a non-zero plume for whatever geographic area a modeler sets, known as a domain. See Tr. 2, at 103:17-104:15. In other words, the way AERMOD was designed, the model will generate a plume for the entire domain chosen by the modeler. See id., at 104:22-105:5. Though EPA guidance states that a modeler should not set the domain to exceed 50 kilometers, a modeler can set the domain to whatever the modeler wants. See id., at 103:23-104:4. Therefore, "[i]f the modeler sets a ten-mile domain, the model will generate a plume for the entire ten-mile area. If the modeler sets a thirty-mile domain, the model will generate a plume for a thirty-mile area." Post-Trial Br., at 24 (citing Tr. 2, at 104:13-105:5; Machado Decl. ¶ 18).

Nevertheless, Plaintiffs' expert Dr. Chinkin stated that "[t]he HRA Modeling results demonstrate that airborne lead concentrations would have been transported at least 2.7 miles to the southwest, 1.9 miles to the east-northeast, and 1.7 miles to the north, well beyond the Vernon Plant's fenceline." See Chinkin Decl. ¶ 24. Dr. Chinkin explained that that, even if AERMOD shows airborne lead in a certain area based on "concentration mode," at least "some of that lead" is necessarily being deposited because "basic physics tells us that particles are heavier than air." Tr. 2, at 100:23-101:3; see Chinkin Decl. 18.

Contrarily, Dr. Chinkin testified at his deposition, "Don't put a lot of weight in the pattern of the AERMOD. That is not what it has been validated for especially in an area where it is complex winds like this." See Tr. 2, at 106:6-9. At the same time, he explained that even if the "exact pattern can be off," the "relative understanding of the shape of the pattern is still useful information for determining how far particles were traveling from that facility." Id., at 106:22-25.

The Court is unable to reconcile these statements. Though defense expert Mr. Machado "agree[d] that the shape of the isopleths ... can provide an indication of the directionality of particulate transport from the Vernon plant," see Machado Decl. ¶ 24; see also Tr. 2, 147:9-11, directionality and distance present different issues.

Indeed, defense experts Dr. Shields and Mr. Machado generally agreed as to the direction in which airborne lead traveled

they testified that the wind carried lead particles away from the Vernon Plant. And when asked if the Vernon Plant contributed to lead found in the residential soils in the PIA, Dr. Shields testified that the Vernon Plant did contribute some lead to the residential areas—just not in an observable amount. See Tr. 3, at 16:22-17:7; 20:15-19; 24:1-13 ("I have said all along that the Vernon plant contributed some lead, not observable amount, but there had to be some lead deposited in the residential areas.").

In describing an "observable amount," Dr. Shields explained that his statistical analysis could only detect contamination from the Vernon Plant above 50 ppm. Id., at 75:19-76:21. Though Dr. Shields did not calculate how much lead was actually emitted from the Vernon Plant or determine how much traveled to the PIA's residential areas, he stated that, "based on [his] experience, [] most of the lead released would have been deposited very close to the plant." Id., at 18:8-9.

Mr. Machado explained that a number of factors were relevant to determining how far particles traveled, depending on the particles' size. For larger particles—those approximately 25 to 30 microns in size—Mr. Machado agreed that gravity would be the primary force that would pull these particles to the ground. See Tr. 2, at 127:14-128:12. For smaller particles, those "closer to 5 microns," see id., at 128:15-17, the primary mechanism pulling particles toward the ground was turbulence. Id., at 127:20-22. Particles emitted from a stack, like at the Vernon Plant, would generally be smaller. Id., at 133:13-18.

Defendants argue that "[t]he point of his testimony was that forces such as turbulence could cause fine particles to fall to the earth sooner than they would have under gravity's influence alone; that is, closer to the Vernon Plant." Def. Post-Trial Reply, at 8. However, Mr. Machado's testimony does not support this claim:

Q: And those particles that are [] 5 microns or less [], where would those settle out?

A: It will really depend on the air currents, how the turbulence in the air takes those particles. That is a science that is, I will have to say not very well understood. Nevertheless, those particles will hit the ground at some point downwind.

Tr. 2, 135:2-8. As Plaintiffs argue, "Mr. Machado's nod to turbulence does not contest what both sides' experts agree on: that lead transported in the air will deposit in soil as it travels." Pls.' Post-Trial Br., at 17.

Mr. Machado explained that a variety of factors are responsible for how far a lead particle might travel, including wind direction, wind speed, gravity, turbulence, particle size, particle density, interception or impaction. Mr. Machado stated, "the smaller particles will travel, I would say—it is difficult to say how far, but miles away, it could easily travel miles away." Tr. 2, at 142:24-143:1. Thus, Mr. Machado does not concretely opine as to the areas in which contaminants may have landed.

Having considered the above evidence and argument, the Court holds that even if the AERMOD results show that contamination from the Vernon Plant reached some undefined portion of the PIA's Residential Areas, the AERMOD results cannot estimate the outer bounds of contamination or the precise amount of in the PIA. Even if Plaintiffs establish that lead traveled could have traveled somewhere beyond the PIA's half-mile Industrial Areas, taking Plaintiffs' AERMOD argument to its logical conclusion—that lead could have traveled throughout the entire PIA—again requires rank speculation. Because it was run in concentration mode, AERMOD cannot

calculate the outer bounds of contamination, and like background, does not show that contamination actually went any specific distance beyond the PIA's Industrial Areas.

This is insufficient for Plaintiffs to meet their burden: Plaintiffs seek to recover response costs throughout the PIA's Residential Areas by broadly defining the entire 6,400 acre site as a single site, but the AERMOD model does not show that lead plausibly could have traveled throughout the PIA—only that it might have traveled beyond the Industrial Area to some limited, undefined portions of the Residential Areas. Though Defendants' experts agreed with the general notion that some lead traveled to the Residential Areas generally, the lack of specificity is insufficient to establish contamination throughout the PIA's residential areas. See Asarco LLC, 21 F. Supp. 3d at 808 (holding that there was a lack of causation for purposes of CERCLA liability as to one of two sites where the plaintiff "presented only minimal, anecdotal evidence that surface-water runoff that traveled offsite would migrate" to another location).

AERMOD may be useful in some circumstances. But here, it does so only for the limited purpose described above. In light of Plaintiffs' burden to establish release causation, the AERMOD evidence does not show that Defendants' releases plausibly could have caused contamination in the PIA's Residential Areas.

2. Statistical Analyses

Next, the parties rely heavily on their respective statistical analyses. Plaintiffs' expert Dr. Medine, who performed directional analyses and an inverse-weighted-distance contouring analysis of soil samples throughout the PIA, ultimately concluded that the Vernon Plant was the predominant source of contamination throughout the PIA. Defendants' expert Dr. Rouhani, who performed geostatistical analyses and a kriging analysis, reached the opposite conclusion: that the Vernon Plant's impacts were confined to the industrial areas surrounding the Vernon Plant.

The Court begins by discussing the experts' respective directional and geostatistical analyses. First, the Court addresses and denies Defendants' Daubert motion to exclude Dr. Medine's testimony as to his directional analyses. Second, the Court summarizes Dr. Rouhani's testimony. Third, the Court compares the two and concludes that Dr. Rouhani's geostatistical methods are more persuasive than Dr. Medine's directional analyses. After weighing the two methods, this hue of evidence weighs in Defendants' favor.

The Court then turns to the experts' contouring and kriging analyses. The Court concludes that both analyses suffer from deficiencies, but that the kriging analysis weighs slightly in Defendants' favor.

i. Directional Analyses and Geostatistics

a) Dr. Medine

We first discuss Dr. Medine's testimony. To determine whether soil lead data in the PIA indicated a decreasing gradient moving away from the Vernon Plant, Dr. Medine conducted several directional analyses of the data. Medine Decl. ¶¶ 38, 42. Based on these analyses. Dr. Medine ultimately concluded that the "level of soil lead [in the PIA] does decrease as one moves farther away from the Vernon Plant." Id. ¶ 41.

Dr. Medine recognized that the decreasing gradient pattern was not "visible to the naked eye," because the pattern of contamination from a point source "can often be interrupted or even completely masked" by "other sources of lead" in a "complex urban environment like metropolitan Los Angeles." Id. 38. Other "human disturbances" can also distort a decreasing

gradient, like the "redevelopment of homes and the addition of landscaping." Id. ¶ 39.

Therefore, Dr. Medine relied on statistical "tools" that enabled him to "identify statistical patterns in the PIA," despite the potential for disturbances in the soil lead data. Id. ¶¶ 40-41. Dr. Medine concluded, accordingly, that the Vernon Plant's releases contributed to lead contamination in the PIA.

Defendants challenge Dr. Medine's statistical analysis under Daubert. Though Defendants' motion presents a close question, the Court denies the motion: Plaintiffs satisfy then burden to establish Dr. Medine's reliability and helpfulness by preponderance of the evidence. However, as the below discussion illustrates, the Court is skeptical of Dr. Medine's conclusions, and considers his methodology and opinions in its discussion of the merits of his testimony and Plaintiffs' case.

1) Methodology and Daubert

Defendants focus their attack on Dr. Medine's methodology and move to exclude his testimony as unreliable under Daubert. See ECF No. 387. Under Daubert, the Court, in its role as a gatekeeper, must "ensure that all admitted expert testimony is both relevant and reliable." Grodzitsky v. Am. Honda Motor Co., 957 F.3d 979, 984 (9th Cir. 2020) (quoting Wendell v. GlaxoSmithKline LLC, 858 F.3d 1227, 1232 (9th Cir. 2017)). "The focus of the district court's analysis must be solely on principles and methodology, not on the conclusions that they generate,' and 'the court's task is to analyze not what the experts say, but what basis they have for saying it.'" Id.

First, Defendants argue that Dr. Medine is not qualified to use statistical tools in his assessment of the evidence because he is not a statistician. "The issue of qualifications is governed by Federal Rule of Evidence 702 which contemplates a broad conception of expert qualifications. As the terms of the rule state, an expert may be qualified either by 'knowledge, skill, experience, training, or education.' Moreover, the advisory committee notes emphasize that Rule 702 is broadly phrased and intended to embrace more than a narrow definition of qualified expert." See Thomas v. Newton Int'l Enterprises, 42 F.3d 1266, 1269 (9th Cir. 1994).

Dr. Medine testified that, in his 40-year career in environmental science, see Medine Decl. ¶¶ 11-16, he has "used statistics ... on many projects." See Opp'n to Daubert Mot., Declaration of Lily A. North in Support of Mot. ("North Decl.") Ex. A, at 27:23-28:13. While he admits that he did not "consider himself a statistician" because he "not have significant training in statistics for [his] college degrees[,]" he explained that he "used statistics throughout [his] career and [has] worked with other individuals that perform statistical analyses for the projects that [he's] been involved with." Id. At trial, Dr. Medine testified that he has had "several classes in statistical methods," including in "advanced numerical analysis," and that "statistics were often used in many other classes that [he] took." Id., at 73:8-11. He also stated that, "when he was teaching at the University of Connecticut and the University of Colorado," he "would often use various kinds of methods to describe data to graduate-level engineers and scientists." Id., at 73:12-15.

Based on his testimony, the Court is satisfied that Dr. Medine's years-long career in environmental science renders him sufficiently qualified to offer a statistical opinion in this case. The authorities cited by Defendants are inapposite. For example, in In re Live Concert Antitrust Litigation, 863 F. Supp. 2d 966, 994 (C.D. Cal.

2012), the Court found an economist unqualified to classify various concert genres because he had no qualification or experience to do so. In Grodzitsky, the Court excluded an expert who failed, as a general matter, to reliably analyze the issue in the case. See Grodzitsky v. American Honda Motor Company, Inc., No. 2:12-CV-1142-SVW-PLA, 2015 WL 2208184 at *4-6 (C.D. Cal. Apr. 22, 2015). Here, in contrast, Dr. Medine's expertise in environmental science is relevant to his expert opinion on how statistics may show the effects of lead contamination in a certain area.

Next, Defendants argue that Dr. Rouhani's geostatistical analysis is the recommended method for evaluating the geographic extent of contamination, rather than Dr. Medine's directional analyses. Defendants cite guidance from the EPA stating that "[g]eostatistics is widely recognized for offering graphical methods that are ideally suited for delineating contaminant zones." See JX011, at 18, Superfund Lead-Contaminated Residential Sites Handbook. However, Defendants' motion overstates the import of the EPA guidance, which does not say that geostatistical methods are the only methods one should use to evaluate contamination.

Plaintiffs, on the other hand, cite several examples of studies and regulatory guidance, which suggest that Dr. Medine's general approach—evaluating contamination based on distance from the Vernon Plant—may be appropriate depending on the circumstances. See Opp'n to Daubert Mot., at 9-11. Having reviewed Plaintiffs' cited authorities, the Court concludes that there is nothing per se unreliable about performing a directional analysis to evaluate soil contamination; the studies Plaintiffs cite are thorough in describing their evaluation of contamination with distance. See, e.g., Ex. C to Opp'n, Tacoma Smelter Plume Site, Final Report, ECF No. 418-2. The differences between Dr. Rouhani's geostatistical approach and Dr. Medine's directional analyses go to the weight of the parties' evidence, rather than its admissibility.

This leads the Court to the heart of Defendants' motion: the methods Dr. Medine used to conduct his directional analyses and interpret the 217, 316 soil lead measurements taken in the PIA.

Dr. Medine began his directional analyses by limiting the data he used. Medine Decl. ¶ 118. In his opinion, he stated that this was "the best way to reduce localized impacts from other sources, and therefore represent[ed] the best chance to identify any impacts from the Vernon Plant." Medine Decl. 43. Specifically, Dr. Medine decided not to rely on data collected from a residential property's side yards, gardens, parkway, and "driplines," or data from public parks and schools. See Tr. 1, at 85:6-86:22. Thus, Dr. Medine only relied on samples identified as coming from residential front yards or back yards because these "would be more reflective of atmospheric emissions[.]" Tr. 1, at 90:3-8.

As Defendants argue, this excluded a significant amount of data points from the analysis: for example, in Dr. Medine's regression analysis of the North Sector, Medine Decl. ¶ 42; Ex. PX050, limiting the analysis to front yard and back yard measurements excluded 8,540 out of 20,810 of the available data in that area. And Dr. Medine only performed his analyses for a limited area of the PIA: rather than performing directional analyses for each direction, Dr. Medine limited his analysis to only 113° of the 360° circle that DTSC drew around the Vernon Plant when it created the PIA.

Next, Dr. Medine averaged the data by creating a parcel average for the front yard and back yard of each home in the PIA; Dr. Medine grouped those averages together based on their distance from the PIA and averaged the data again. Tr. 1, at 95:8-96:5. This process was called binning,

referring to Dr. Medine's process of placing the parcel values into "bins" of different distance intervals. For example, Dr. Medine conducted an analysis with 1000 ft bins, creating an average value of lead from data points taken between 0-1000 ft. from the Vernon Plant, 1000-2000 ft., 2000-3000 ft., and so on.

However, somewhat confusingly, the bins themselves were not all equal sizes: for example, in Dr. Medine's North Sector analysis, the bin from 3000-5000 ft. was actually 2000 ft. wide, while the other bins measured 1000 ft. in width. And the last bin, based on the step out samples, spanned a distance of 11,000-24,000 ft. from the Vernon Plant.

Adding to the confusion, though Dr. Medine's overall analysis focused on certain directions for each of his identified sector, like north and northeast, the first three bins included soil samples from every direction around the Vernon Plant. Given the importance of predominant winds in evaluating contamination, the omnidirectional component of this analysis fails to account for wind impact in the first 3000 ft. from the Vernon Plant. Further, Dr. Medine's last bin, which was based on step out samples, did not filter out the other types of samples that Dr. Medine had previously filtered; the step out samples were based on parkway data. See Medine Decl. ¶ 44.

Also somewhat inconsistently, Dr. Medine used only front yard data in performing his IDW Contouring analysis, described below. Medine Decl. ¶ 123. He stated that, "[f]rom a review of aerial photographs of residential properties in the PIA, it was apparent that residents would use the back yards of activities that might affect lead concentrations in the soils, including concrete patios, vehicle parking, vehicle maintenance, workshops/sheds, and other residential activity, and, as such, the front yard samples would better reflect the contribution of airborne emissions of lead on soils." Id. The Court is left puzzled as to why Dr. Medine included these samples in conducting his directional analyses.

After this, Dr. Medine then used Microsoft Excel to create a regression trendline. However, Dr. Medine's trendline used an unweighted form of regression, which assumes that all data points have the same variance. Here, however, Dr. Medine's data points do not have the same variance—because they are the average of a number of different samples. For example, the average assigned to the distance of 8,500 ft. from the Vernon Plant is based data from 425 parcels, while the average assigned to the distance of 10,500 ft. is based on data from only 17 parcels. In statistical terms, the 10,500 ft. bin average is 25 times more variable than the 8,500 bin average. According to Defendants, "it is a basic statistical principle that when data have highly variable variances, weighted regression should be performed." See Mot. at 22. Plaintiffs did not account for why Dr. Medine did not do so.

The results of Dr. Medine's analysis of the PIA's North Sector are excerpted below. Image materials not available for display.

See PX092. Dr. Medine relies on this chart to opine that there is a strong trend of decreasing contamination as distance increases from the Vernon Plant.

Defendants contend that Dr. Medine's analysis is unreliable because Dr. Medine did not sufficiently explain the bases for his various choices. The Court disagrees, however, because Dr. Medine's choices were consistent with choices made by Environ, which described reasons for excluding dripline samples, and defense expert Dr. Shields, who also relied on the Environ report in using front and back yard samples. Indeed. Dr. Shields agreed with Dr. Medine that soils closest to the painted structure in older homes typically have the highest lead concentrations.²⁸ See Opp'n to Daubert Mot., at 15. Moreover, the Long Beach Study, performed by Advanced Geoservices, also excluded dripline samples.

28 Plaintiffs also contend that Dr. Rouhani excluded large segments of the PIA by using "only the maximum lead concentration at each sample location, which meant that he excluded more than 2/3 of the sampling data." See Opp'n to Daubert Mot., at 2. However, Plaintiffs do not articulate any problem with such an exclusion; Dr. Rouhani explained that using the maximum lead concentration at each location would, if anything, overstate the impact of emissions from the Vernon Plant.

Similarly, Plaintiff's opposition to the Daubert motion, and the evidence at trial, sufficiently support Dr. Medine's choice to exclude school and park sampling data, as such data was consistent with disturbances in certain of these areas. See Opp'n to Daubert Mot., at 17. While Defendants contend that Dr. Medine did not do enough to verify such disturbances throughout, this is an argument that goes to the weight of Dr. Medine's testimony, rather than its admissibility. Next, discussing Dr. Medine's averaging of various data points, Plaintiffs cite several studies that used averages within distance bins to evaluate contamination. On one hand, Plaintiffs do not address Defendants' critique that this averaging method masks variability. And Plaintiffs do not contest the merits of Defendants' argument that using an unweighted regression was inappropriate, deeming this argument "the ipse dixit of [Defendants'] attorneys." However, Defendants do not meaningfully explain why or how a weighted regression would lead to different results here, merely citing a National Institute of Standards and Technology website with an "Engineering Statistics Handbook" in a footnote of their motion. See Mot., at 22. Indeed, Plaintiffs cite various studies suggesting that regression analyses are appropriate tools for delineating site contamination. While geostatistics may be a better method in this case, the Court cannot conclude that this renders Dr. Medine's testimony unreliable.

Ultimately, the "general rule [is that] the factual basis of an expert opinion goes to the credibility of the testimony, not the admissibility, and it is up to the opposing party to examine the factual basis for the opinion in cross-examination." Loudermill v. Dow Chemical Co., 863 F.2d 566, 570 (8th Cir. 1998); see, e.g., United States v. Hicks, 103 F.3d 837, 846 (9th Cir. 1996), overruled on other grounds by United States v. W.R. Grace, 526 F.3d 499 (9th Cir. 2008) ("[T]he impact of imperfectly conducted laboratory procedures is better approached as an issue going not to the admissibility, but to the weight of the DNA profiling evidence." (internal quotations and citation omitted)). While the line between Dr. Medine's methodology and the "factual basis" for his opinion is somewhat blurry, the Court is satisfied that Plaintiffs' burden to establish reliability and helpfulness under Daubert is satisfied.

b) Dr. Rouhani

Unlike Dr. Medine, Dr. Rouhani used geostatistical methods to perform his analysis. Geostatistics assess the "spatial relationship among all of the data points located along a chosen direction." See Rouhani Decl. ¶¶ 14, 19. According to Dr. Rouhani, "geostatistics is better able to evaluate datasets that contain a mixture of background and impacted data." Id. 15. To support his opinion, Dr. Rouhani cites, first, various studies that have applied geostatistical methods in analyzing contaminated sites, and second, EPA guidance explaining that geostatistics is "ideally suited for delineating contaminant zones." Id. 16-18.

Dr. Rouhani's geostatistical analyses led to two different outputs: variograms and a kriging map. Variograms evaluate the differences between pairs of data points lying along the chosen direction based on the pairs' spatial correlation. Id. ¶¶ 19-20. "Because variograms are based on the difference between paired sample values, where each member of the pair is affected by the same localized source of lead, the method automatically accounts for it." Id. ¶ 21.

Ultimately, Dr. Rouhani's variogram analysis indicated that "the effects of emissions from the former Vernon smelter on measured soil lead concentrations are confined to the industrial areas surrounding the Vernon smelter." Id. ¶ 38. The variograms showed no "statistical evidence of any relationship between emissions from the former Vernon smelter and the measured lead concentrations within the PIA residential areas." Id. ¶ 43. Dr. Rouhani's analysis would have detected contamination on a level of 1-9 ppm, consistent with the sensitivity of Plaintiffs' methods.

Dr. Rouhani then performed several sensitivity analyses for the PIA's residential areas by using three alternative sets of data. Id. 47. First, Dr. Rouhani excluded

all samples that were collected within 10 feet of buildings, where the lead-based paint contamination would be greatest. Id. ¶ 48. Second, he ran his calculations based exclusively on the soil samples DTSC collected from the "parkways" (i.e., the strips of grass located between the street and the sidewalk), because the parkway sample locations would be relatively far away from the corresponding residence. Id. ¶ 49. Third, he re-ran his calculations using Dr. Medine's dataset from his directional analysis. Id. ¶ 50.

After performing these sensitivity analyses, Dr. Rouhani concluded that the different datasets would not alter his opinion: "lead emitted from the former smelter [did] not [have] a statistically detectable effect on DTSC's surface soil lead measurements within the PIA residential areas." Id. ¶¶ 51-52.²⁹

29 Plaintiffs move to exclude Dr. Rouhani's sensitivity analysis of front yard and backyard samples and several of his conclusions as untimely disclosed pursuant to Fed. R. Civ. P. 26 and 37. See Mot. to Exclude Undisclosed Expert Opinions, ECF No. 382, Rule 37(c)(1) provides that a party who "fails to provide information ... as required by Rule 26(a) or (e) ... is not allowed to use that information ... to supply evidence on a motion, at a hearing, or at a trial, unless the failure was substantially justified or is harmless." Fed. R. Civ. P. 37(c)(1). "The determination of whether a failure to disclose is justified or harmless is entrusted to the broad discretion of the district court." San Francisco Bay Area Rapid Transit Dist. v. Spencer, No. 04-04632-SI, 2007 WL 421336, at *4 (N.D. Cal. Feb. 5, 2007). "Nondisclosure is harmless if it does not prejudice the other party." Auto. Indus. Pension Tr. Fund v. Tractor Equip. Sales, Inc., 73 F. Supp. 3d 1173, 1181-82 (N.D. Cal. 2014) (citing Yeti by Molly, Ltd. v. Deckers Outdoor Corp., 259 F.3d 1101, 1106 (9th Cir. 2001)). Defendants disclosed this analysis to Plaintiffs approximately a week before Dr. Rouhani's deposition, almost two months before trial. At Dr. Rouhani's deposition. Plaintiffs' counsel introduced the supplemental analysis as an exhibit and asked what it was and whether he would be using it. See North Decl. to Opp'n, Ex. G (Rouhani Tr.) at 238:19-242:19. After Dr. Rouhani discussed the analysis. Plaintiffs' counsel asked no further questions of Dr. Rouhani. Given this sequence of events, Plaintiffs' claims of "sandbagging" and harm are unavailing—Plaintiffs were able to seek clarification of the basis for Dr. Rouhani's testimony at his deposition and had sufficient time (nearly two months) to prepare for further cross examination at trial. See Gallegos v. Seeley, No. 3:18-CV-01322-JAH-MSB, 2021 WL 4295290, at *5 (S.D. Cal. Sept. 21, 2021).

c) Analysis

Thus, unsurprisingly, Dr. Medine and Dr. Rouhani come to opposite conclusions. Of the two, the Court finds Dr. Rouhani's testimony more compelling: Dr. Medine's methodology and the factual bases for his opinion are plagued with uncertainty. Plaintiffs' arguments are largely speculative and fail to identify any plausible reason why Dr. Rouhani's geostatistical approach and sensitivity analyses are inferior to Dr. Medine's.

First, considering the experts' methodologies, Dr. Rouhani's geostatistical method presents a more persuasive approach. Geostatistics is explicitly endorsed by the EPA. JX011, at 18. In the 1980s, the EPA first applied geostatistics to delineate a contaminant zone at a lead smelter, see DX272 (recommending geostatistical techniques to "contour map the lead plume around an urban secondary lead smelter" to manage remediation), and has done so on a variety of occasions. While Plaintiffs correctly point out that the EPA guidance does not identify geostatistics as the only method for delineating contamination, the guidance does not suggest that an analysis like Dr. Medine's is appropriate. See Chem. Mfrs. Ass'n v. U.S. E.P.A., 919 F. 2d 158, 167 (D.C. Cir. 1990) ("It is not the role of courts to 'second-guess the scientific judgments of the EPA,' and we give considerable latitude to the EPA in drawing

conclusions from scientific and technological research, even where it is 'imperfect' or 'preliminary.'" (citations omitted)).

Turning to Dr. Medine's analysis, however, Plaintiffs' evidence is not as compelling or persuasive as the EPA's guidance. Though Plaintiffs cite several studies that relied on the use of directional techniques like Dr. Medine, the Court is unable to determine from the evidence before it whether these studies themselves have been reviewed or affirmed, either in courts or in the scientific literature. Though the studies appear sufficient to establish Dr. Medine's reliability under Daubert, the Court finds the EPA's guidance more convincing here.

Moreover, Plaintiffs offer little in the way of rebutting Dr. Rouhani's geostatistical analyses. For example, at trial, Plaintiffs attempted to suggest that geostatistics is improper where data is not spatially correlated. See Tr. 2, at 154:5-7. Plaintiffs cited draft EPA guidance written by Dr. Rouhani in 2004, where he stated, "simpler classical statistical approaches may be applicable in some situations where contamina[nt] concentrations are not spatially structured." Id., at 156:12-14. This general statement that simpler approaches may be applicable does little to establish Plaintiffs' point—that geostatistics is inappropriate in this case because the contaminant concentrations were not spatially structured.

Defendants, on the other hand, offer a strong criticism of Dr. Medine's choices in this case. As explained above, Dr. Medine compared averages from (1) omni-directional industrial zone samples to (2) unidirectional residential front-yard and backyard samples to (3) step-out parkway samples to determine that a trend existed in the PIA. While Dr. Medine presented his graph as establishing a trend based on these very specific comparisons, he also included a graph of some of the raw data, which does not show such a trend "to the naked eye." See Medine Decl. ¶ 102.

Image materials not available for display.

Dr. Medine essentially asks the Court to trust that his decisions to filter certain kinds of data in certain zones, create varying bin sizes, and create a graph using averages of thousands of data points reveals the true trend of contamination in the PIA. But given the uncertainties in this evidence, the Court cannot conclude whether Dr. Medine's opinions in fact depict the underlying trend of contamination in the PIA, or whether Dr. Medine's graph is the result of his choices in filtering data. Even if certain of his choices made sense in isolation—like excluding dripline samples—Plaintiffs do not present sufficient evidence for the Court to conclude that all of Dr. Medine's choices were correct, or why he made such choices inconsistently (like in the case of step-out samples). Plaintiffs also do not indicate how his opinions would vary if he had made slightly

different choices. As Defendants argue, Dr. Medine had never made a graph with all of the PIA data.³⁰ And Dr. Medine admitted that changing data points would change the trend line—even if a trend line might look similar. Without a point of comparison, there is no way for the Court to evaluate these opinions. See Tr. 1, at 121:5-122:2.

30 Though Dr. Medine also included several half-mile bin analyses in his expert report and declaration, Plaintiffs did not present these analyses at trial or in their post-trial briefing. Though the Court does not know why Plaintiffs chose not to do so, one reason may be because Dr. Rouhani concluded in his expert report that Dr. Medine's half-mile analyses were not statistically significant. Whatever the reason in Plaintiffs' strategy, this is yet another sign of Dr. Medine's one-track mind in approaching his analysis. For example, where Dr. Medine observed an increase in lead concentrations with distance, he simply did not include that directional analysis in his declaration. See Medine Decl. ¶ 143.

On the other hand, Plaintiffs criticize Dr. Rouhani's initial decision to use of all of the data from the PIA in his geostatistical analysis, rather than accounting for human disturbance, construction activities, and redevelopment. But like Dr. Rouhani, Dr. Medine did not conduct a systematic analysis of disturbances throughout the PIA. See Tr. 1, at 124:19-23, 125:18-25, 179:22-180:14. To the extent Plaintiffs fault Dr. Rouhani for failing for doing so, they simultaneously undermine then own case.

Further, because disturbances are common in contaminated residential sites, see Trial Tr. 2, at 227:5-7, it is unlikely that the EPA would recommend geostatistics for use at complex lead remediation sites if it were unable to account for such disturbances. And critically, Plaintiffs do not challenge Dr. Rouhani's sensitivity analyses, where Dr. Rouhani excluded the same data that Dr. Medine excluded and concluded that the new variogram results did not change his conclusions.

For these reasons, the Court finds Dr. Rouhani's geostatistical analysis more persuasive than Dr. Medine's directional analyses. This evidence weighs in Defendants' favor, suggesting that Defendants' releases did not plausibly cause contamination throughout the PIA's Residential Areas. See Oakland Bulk & Oversized Terminal, LCC v. City of Oakland, 960 F.3d 603, 607, 613 (9th Cir. 2020) (holding that a district court's determination that the plaintiffs did not establish substantial evidence of a substantial danger of a fire risk in a breach of contract case was not clearly erroneous after finding the plaintiff's experts' testimony to be inconsistent and "riddled with inaccuracies, major evidentiary aps, and erroneous assumptions, to the point that no reliable conclusion about health or safety dangers could be drawn from it"); see also Asarco LLC, 21 F. Supp. 3d at 807-08, 810-11 (weighing the credibility and persuasiveness of the parties' experts as to causation of contamination at two different sites and finding causation at one of the sites even when one of the parties' experts "call[ed] into question the impact of arsenic from the Plant on the soil at the [site].").

ii. Interpolation

Next, Dr. Medine and Dr. Rouhani each utilized a statistical technique called "interpolation," which, at a general level, is a method of estimating unknown data points from existing known data. Dr. Medine performed an inverse-distance-weighed contouring analysis ("IDW Contouring"), which allowed him to create a model of contamination throughout the entire PIA, even in areas where no soil samples were taken. Medine Decl. ¶ 46. Dr. Rouhani performed a similar technique known as "kriging," which is a geostatistical form of interpolation that "also predicts the level of soil lead that would occur at every

location between the [known soil values]." Rouhani Decl ¶ 54; see Tr. 1, at 243:6-10.³¹

31 At the outset, Plaintiffs cite a study to suggest that, as a general matter. IDW Contouring is a more effective method than Kriging. See Post-Trial Reply, at 15 n. 8 (citing Sims Decl., Opp'n to Def. Mot. to Exclude Ops., Ex. T, at 132, ECF No. 418-4). However, the Court is not persuaded that Plaintiffs' citation to a single study requires the Court to conclude at this point that one method is more effective than the other. Substantively, the study itself is uncertain, and notes that "[s]ome studies have shown that kriging interpolation is the best [citation], and some other studies consider that the IDW interpolation is similar with or even better than [kriging]." See id. Ultimately, because Plaintiffs' argument was not raised at trial and is confined to a footnote in their reply to Defendants' trial brief, there is insufficient evidence to conclude that one method of interpolation is more effective than another.

a) IDW Contouring

Dr. Medine explains that his IDW Contouring results show "an area of high lead concentration in the immediate vicinity of the Vernon Plant, with decreasing concentrations in an outward direction in 360-degrees from the Vernon Plant." Id. ¶¶ 48, 124; see PX052. Below, the Court includes an excerpt of Dr. Medine's IDW Contouring results overlaid with vectors showing the PIA's prevailing wind directions. See DX299; see Tr. 1, at 52:16-24 (Dr. Medine's testimony explaining that he superimposed the four predominant wind directions—southwest, north, east, and northeast—and one non-predominant wind direction—northwest).

Image materials not available for display.

See DX299; PX052.

As discussed earlier, the offsite wind data in this case indicated that the PIA's prevailing winds blew, first, to the southwest, second, to the east, and third, to the north. Inconsistent with the evidence of predominant winds, Dr. Medine's IDW Contouring results show that the lead levels were highest from the northwest to the northeast. Dr. Medine agreed, according to his map, that "a nonpredominant wind direction has higher concentrations of lead than a predominant wind direction." Id., at 54:18-21. Defendants argue that this defies Plaintiffs' model of contamination. However, Dr. Medine offered two explanations for this inconsistency—first, that there was less data collected to the southwest and to the northwest, and second, that onsite wind data showed that the primary wind direction was to the northeast, and that winds also blew to the east, north, northwest and southwest. Tr. 1, at 170:12-171:8; see supra Section IV.B.1.i. (discussing prevailing winds). But Dr. Medine testified that when he performed his directional analyses, he relied on the off-site data and the AERMOD model, because he did not have the onsite wind data. Ultimately, however, Dr. Medine explained that he had not "determined why the northwest vector shows what it shows from the containing." 56:25-57:1.

Defendants also contend that a study known as the Wu & Johnston Study—prepared by independent researchers—shows the same "backwards relationship" between wind direction and soil lead concentrations in the PIA. See Shields Decl. ¶ 30 (citing Wu & Johnston, "Assessing Spatial Characteristics of Soil Lead Contamination in the Residential Neighborhoods Near the Exide Battery Smelter," Case Studies in the Environment (2019)).

The Wu & Johnston study mapped the average parcel soil lead concentrations in the residential areas, and designated clusters of higher lead concentrations as "hot spots" and lower lead concentrations as "cold spots." Id. The study shows certain "hot spots," or areas with higher average lead concentrations, in neighborhoods to the north, northeast, and northwest of the Vernon Plant, even though these directions "only receive modest wind." Id. ¶ 31. The other largest "hot spot" is to the southeast of the Vernon Plant, which is a non-predominant wind direction. Id. On the other hand, the study's "cool spots" are located southwest and east of the Vernon Plant, the PIA's two primary wind directions. Id.

A figure from the study is excerpted below.

Image materials not available for display.

See DX109.

Plaintiffs' expert Dr. Medine largely agreed that the Wu & Johnston Study was

inconsistent with his IDW Contouring map and with the "hallmark" of a point source like the Vernon Plant. Tr. 1, at 48:15-50:17. However, he also stated that it was inconsistent with Dr. Rouhani's kriging analysis. Tr. 1, at 50:17-21; 144:18-23. For example, neither Dr. Medine nor Dr. Rouhani's analyses showed an area of high lead concentration in the PIA's northeast comer—even though the Wu & Johnston study does. Id., at 143:23-144:2. Similarly, though the Wu & Johnston Study shows a hot spot on the southeast side of the PIA, Dr. Rouhani's map does not show a hot spot there. Id., at 144:5-13.

Plaintiffs also refer to home age in the PIA, and conclude that the Wu & Johnston Study "does little more than show that PIA neighborhoods with the oldest homes generally have the highest lead levels, while those with newer development have the lowest lead levels." Post-Trial Reply, at 11-12 (citing Tr. 1, at 145:16-149:1. According to Dr. Medine's testimony, the PIA's newer homes were located toward the south of the PIA—where more of the "cold spot" areas were located in the Wu & Johnston Study. Dr. Medine testified that a complex relationship exists between the age of homes and impacts from airborne emissions, lead paint, and leaded gasoline. See Medine Decl. ¶ 172. Defendants do not disagree that home age is an important factor in evaluating contamination.

Nevertheless, Defendants argue the presence of so many cold spots downwind of the Vernon Plant—regardless of home age—is inconsistent with the theory that the Vernon Plant was a source of lead to the Residential Areas. Even considering Dr. Medine's later claim that the primary wind direction at the Vernon Plant may have been to the east-northeast, the Wu & Johnston Study shows no "hot spots" in that direction either. See Tr. 1, at 170:12-171:8; see DX109.

Rather, as Defendants argue—and the Court agrees—"the graphs show that the highest and most widespread lead levels are due north of the Vernon Plant." These "hot spot" areas are far from the Vernon Plant, across more than a mile of less-contaminated areas closer to the facility. Ultimately, based on the above evidence and argument, the Court disagrees with Plaintiffs' claim that the Wu & Johnston Study "does little more than show that PIA neighborhoods with the oldest homes generally have the highest lead levels, while those with newer development have the lowest lead levels." Post-Trial Reply, at 11. The Wu & Johnston Study is inconsistent with both sides' interpolation evidence, leaving the Court with additional uncertainty, and further underscoring Plaintiffs' inability to fill the evidentiary gaps needed to meet their burden under Section 107.

b) Kriging

Defendants argue that the Vernon Plant is "immediately surrounded by an inner band of heavy lead contamination, then by a second band of much lower contamination levels throughout the industrial area of Vernon." See Defs.' Post-Trial Br., at 8. In Dr. Rouhani's kriging map, this second band of lower contamination takes the form of "a blue halo in the intermediate zone of the industrial area, between the more heavily contaminated core around the Vernon Plant and the residential areas[.]" See id.; see DX261 (below). Defendants argue that this shows a lack of PIA contamination attributable to the Vernon Plant, because airborne lead from the Vernon Plant could not have skipped over the zone of lower contamination in the Industrial Areas and then contributed heavier contamination to the Residential Areas. Id. ³² Image materials not available for display.

32 Defendants also rely on another exhibit, DX084, which is a map depicting soil lead sample concentrations by color. This too shows that the samples taken in the "intermediate industrial zone" generally had lower amounts of lead. See DX084.

See DX261.

Plaintiffs argue, however, that these lower lead levels are a result of human disturbance, like construction and redevelopment in that area. Plaintiffs cite to an EPA study to support their argument, which stated:

The date of construction, or building age, measures the amount of time since the construction of the building and, in many cases, is the last major disturbance of soil. Thus, the building age likely measures the length of time lead—from the housing unit and/or neighboring activity sources—has been accumulating on the soil.

See JX13.0019. Defendants do not dispute the general proposition that that soil disturbances can cause lower lead levels.

Accordingly, at trial, Plaintiffs presented several satellite photos of certain properties, which indicated redevelopment of a few of the parcels that produced low soil samples. See Tr. 2, at 209:14-210:5; Pls.' Post-Trial Br., at 21. Figure 1 in Plaintiffs' Post-Trial Brief shows two of the disputed areas, and Shaded Areas A and B. Image materials not available for display.

Two of these photos are reproduced below, which undisputedly show that certain samples were taken from land that had been redeveloped. See PX153.00010, PX153.004 (showing Shaded Area B).

Image materials not available for display.

Defense expert Dr. Rouhani admitted that he did not investigate the extent of any soil disturbances as to the samples in this intermediate industrial area. Tr. 2, at 208:1-7.³³ Defense expert Dr. Shields agreed that, when shown several images of a redeveloped parcel, at least as to that example, development would be responsible for the low set of samples taken from that parcel. See Tr. 3, at 69:2-7.

33 Defendants also suggested at trial that kriging, as a method, somehow accounted for these disturbances, but Dr. Rouhani explained that the software merely output the blue halo based on his inputs. Defendants do not explain how the kriging map accounted for disturbances.

Defendants therefore primarily challenge the anecdotal nature of Plaintiffs' evidence. As to the industrial area, Plaintiffs' photos only showed the re-development at the Shaded Areas A and B in the above figure. See Figure 1. Plaintiffs also did not evaluate the other history of development in the industrial area, nor did they perform any statistical analysis of disturbances in the area. See Tr. 1, at 124:19-23, 125:18-25, 179:22-180:14. The Court agrees that the evidence in support of Plaintiffs' argument is limited. However, Shaded Area B, to the north of the Vernon Plant, accounts for almost all of the samples in the northern intermediate zone. While the areas south and east of the Vernon Plant included a few additional sample areas that Plaintiffs did not show photographs of, the samples undisputedly contain low amounts of lead—some that Defendants cannot explain.

Accordingly, given the small handful of samples taken from this intermediate zone, the Court believes that the disturbance evidence at least partly undermines Dr. Rouhani's kriging map. DX084, below, illustrates that Dr. Rouhani's kriging map and the "blue halo" was only based on a fraction of the total samples around the Vernon Plant.

Image materials not available for display.

See DX084 (with circles added to indicate where the shaded areas A and B are in Plaintiffs' Figure 1). Dr. Rouhani agreed that disturbances would alter any impacts from the Vernon Plant at the disturbed sites, and Plaintiffs' evidence shows that Dr. Rouhani's kriging method did not account for these areas. Tr. 2, at 185:19-21.

Accordingly, because the data pattern in the industrial area is itself based only on a limited number of samples, the Court disagrees that Plaintiffs' evidence is "not nearly enough to call into question the clear and widespread data pattern in the industrial area." See Def. Post-Trial Br., at 9. The Court is not persuaded that Dr. Rouhani's kriging map accurately portrays all of the contamination in the redeveloped areas in the industrial zone of the PIA. But because the Court cannot speculate as to the possibility of redevelopment for samples of which Plaintiffs failed to present evidence, the Court cannot wholly rely on Dr. Rouhani's kriging analysis.

Thus, in comparing Dr. Medine's IDW Contouring and Dr. Rouhani's kriging, the Court cannot determine that one model more accurately depicts contamination in the PIA than the other. Both models are flawed in the ways discussed above. The Court therefore cannot rely on these models in reaching its verdict.

C. SOIL CHEMISTRY

Next, the parties' dispute focuses on the chemical makeup of soil in the PIA. Dr. Shields, Dr. Medine, and Dr. Drexler discussed soil chemistry. Dr. Shields and Dr. Medine focused on a ratio referred to as the lead-to-antimony ratio, while Dr. Drexler analyzed certain soil samples using a technique called "microscopy." 1. Lead-to-Antimony Ratio

Antimony is a metal often found at secondary lead smelters because of its use in lead batteries and as an "alloy for certain lead products." Shields Decl. ¶ 65; see also Tr. 1, at 117:19-21. "Antimony is also found in typical urban soils, but at much lower levels, from things like wheel weights or friction material on brake pads." Shields Decl. ¶ 65.

Dr. Shields and Dr. Medine agree that "[t]he statistical tendency of antimony and lead to be found together in soil samples will indicate whether the lead came from a smelter." Id. ¶ 66: see Tr. 1, at 118:5-8 (Medine's testimony) ("[S]cientists look for antimony in the soil to determine if lead came from [] a secondary smelter.").³⁴

34 Though Dr. Drexler did not explicitly testify about the lead-to-antimony ratio at trial, one of the forms of smelter-related lead he evaluated also included antimony, suggesting its usefulness as a measure of smelter-related contamination. See Tr. 2., 82:14-18.

i. Dr. Shields

To evaluate the correlation between lead and antimony in the PIA, Dr. Shields calculated two R2 values: one for the PIA's industrial area, and another for the PIA's residential areas. Shields Decl. ¶¶ 67-69. The R2 value in the industrial area was 0.81, indicating a strong relationship between lead and antimony. Id. 69. The R2 value in the residential areas was 0.18, indicating a weak relationship between lead and antimony. Id.

Dr. Shields also analyzed antimony to determine if there was a decreasing gradient from the Vernon Plant to the residential areas. Id. ¶ 71. He determined there was a sharp decreasing gradient of the lead-to-antimony ratio within the 0.5-mile radius surrounding the Vernon Plant, but that there was no decreasing gradient within the Residential Areas of the PIA. Id. The lack of a decreasing gradient in the Residential Areas suggests that the lead present in those areas did not come from the Vernon Plant. Thus, Dr. Shields concluded that the Vernon Plant's releases did not cause contamination in the Residential Areas surrounding the Vernon Plant. Id. Plaintiffs did not cross-examine Dr. Shields as to these calculations or as to the lead-to-antimony ratio generally.

ii. Dr. Medine

Countering Dr. Shields' testimony, Dr. Medine opined that the lead-to-antimony ratio in the PIA is consistent with Plaintiffs' theory that the Vernon Plant's releases caused contamination in the PIA's residential areas. See Medine Decl. ¶¶ 148-154. Dr. Medine agreed with Dr. Shields' basic premise—that "scientists look for antimony in the soil to determine if lead came from [] a secondary smelter." Tr. 1, at 118:5-8. Second, Dr. Medine did not dispute the R that Dr. Shields calculated. Tr. 1, at 119:1517.

In terms of methodology, the relevant portion of Dr. Medine's declaration is unwieldy and difficult for the Court to parse. See Medine Decl. ¶¶ 148-154. Dr. Medine begins by discussing two studies as the basis for his conclusion. Id. 148. One study found that the lead-to-antimony ratio "remained very constant in a storm drain from the site downstream for over 1,000 meters." Id. A second study found that, after calculating ratios at two different plants, ratios with a "median of 12" and a "range of 1.6-34" were characteristic of secondary lead smelters." Id.

Dr. Medine then calculated the lead-to-antimony ratios within the industrial area and the north sector of the residential areas. Id. ¶ 149. But rather than calculating an R2 value like Dr. Shields, Dr. Medine simply compared the ratios in the industrial area and the north sector, and concluded that both ratios were "characteristic of secondary lead smelters." Id.

¶¶ 152-154. In other words, Dr. Medine concluded that, because lead-to-antimony ratios found in the residential areas were "remarkably similar to the ratios found near the Vernon Plant," emissions from the Vernon Plant were "transported to soils in the Vernon Plant." Id. ¶ 49. However, Dr. Medine also opined that "[t]he ability to explore [the lead-to-antimony] relationship in further detail was limited because the level of antimony in residential soils was often too low to be detected by the laboratory." Id.

iii. Analysis

The crux of the parties' dispute here is which expert's analysis of lead-to-antimony is more accurate and persuasive: Dr. Shields' analysis of lead-to-antimony correlation in the industrial and residential areas, or Dr. Medine's more general comparison of lead-to-antimony ratios in these same areas.

Compared to the other evidence in this case, the evidence on this question is relatively limited. Plaintiffs claim that Dr. Shields ignored Dr. Medine's cited studies and "focused instead on a statistical analysis of the association between lead and antimony concentrations at the same sampling location." See Pls.' Post-Trial Reply, at 12. Plaintiffs also argue that "Defendants cite [the Eckel Study, which is] not in evidence[,] to suggest Dr. Medine's ratio analysis is flawed. This study does not support Defendants' argument, as the study used the same ratio analysis that Dr. Medine utilized." Id. (citation omitted).³⁵

35 The study at issue is W. P. Eckel, et al., Investigation of unrecognized former secondary lead smelting sites: confirmation by historical sources and elemental ratios in soil, 117 Env. Pollution 273 (2002).

Defendants argue, based on the Eckel study, that evaluating correlation is more appropriate than comparing ratios because "[t]he antimony/lead ratio is meaningful only where the data show a high R2 relationship between antimony and lead, and a low standard of deviation." Defs.' Post-Trial Br., at 17. But even if the Court could consider the Eckel study—which is not in evidence—it would be unable to because it could not access the study online. The Court is therefore ill-equipped to conclude that either a correlation or ratio analysis more capably determines whether the lead-to-antimony ratio in the PIA's residential areas is consistent with emissions from the Vernon Plant.

Nevertheless, we return to the fact that Plaintiffs have failed to satisfy their burden. In this instance, they have not met their burden with respect to the lead-to-antimony evidence: Plaintiffs did not cross-examine Dr. Shields about his calculations, and Dr. Medine did not rebut any of Dr. Shields' testimony at trial, undermining Plaintiffs' belated attempt to challenge Dr. Shields' analysis. Plaintiffs adduce no evidence to suggest that Dr. Shields' regression analysis was somehow flawed. And finally, Dr. Medine does not challenge Dr. Shields' testimony that there is no decreasing gradient of antimony in the Residential Areas relative to distance from the Vernon Plant. See Shields Decl. 71. Accordingly, the Court concludes that the evidence about the lead-to-antimony ratio suggests that the Vernon Plant's emissions could not have plausibly contaminated the Residential Areas of the PIA.

2. Microscopy - Dr. Drexler

Plaintiffs also called Dr. John Drexler as a rebuttal expert. Dr. Drexler is a geochemist who specializes in the use of microscopic technology, also known as microscopy. See Declaration of John W. Drexler ("Drexler Decl.") ¶¶1-5, ECF No. 351-6.³⁶ Dr. Drexler analyzed 22 soil samples

36 Plaintiffs designated Dr. Drexler as a rebuttal expert. Defendants sought to exclude Dr. Drexler's testimony as outside the scope of proper "rebuttal" testimony because no defense expert had been designated to offer "microscopy" testimony. See Mot. to Exclude, ECF No. 339. Under Federal Rule of Civil Procedure 26(a)(2)(D)(ii), rebuttal expert testimony is permitted after the parties' initial disclosures only where it "is intended solely to contradict or rebut evidence on the same subject matter identified by another party." See Fed. R. Civ. P. 26(a)(2)(D)(ii). Such testimony "cannot be used to advance new arguments or new evidence." Blake v. Securitas Sec. Servs., Inc., 292 F.R.D. 15, 17 (D.D.C. 2013); see also Peals v. Terre Haute Police Dep't, 535 F.3d 621, 630 (7th Cir. 2008) ("The proper function of rebuttal evidence is to contradict, impeach, or defuse the impact of evidence offered by an adverse party.") (internal quotation omitted); TCL Commnc'ns Tech. Holdings Ltd. v. Telefonaktenbologet LM Ericsson, 2016 WL 7042085 at *4 (C.D. Cal. Aug. 17, 2016) ("[O]ffering a different, purportedly better methodology is a proper way to rebut the methodology of someone else."). Here, Dr. Drexler's testimony is properly within the confines of rebuttal testimony. Dr. Drexler offers a method of assessing the physical characteristics of PIA lead to determine the lead's source—microscopy—countering Dr. Shields' method of evaluating the lead-to-antimony ratio. Though Defendants argue that this requires an overly broad interpretation of Rule 26, Defendants' cited authorities are inapposite. The Court declines to exclude Dr. Drexler's testimony on this basis.

from 12 distinct residential or commercial addresses located either within or near the PIA. See Tr. 2, at 68:769:5. Based on his microscopic analysis of these soil samples, Dr. Drexler concluded that 17 of 22 samples included "pyrometallurgical" or "smelter-related" forms of lead. See Drexler Decl. ¶¶ 3, 2931.³⁷

37 Defendants also contend that the soil samples Dr. Drexler evaluated were spoliated. As to this argument, Court agrees with Defendants. Dr. Drexler received the samples in "late" 2016, see Tr. 2, at 69:7-13, and stored them in a drawer in his lab, but he was unable to find them when he looked for them in 2022, see Tr. 2, at 69:19-70:2. The duty to preserve evidence "arises when the party has some notice that the evidence may be relevant to litigation, or potential litigation." Morris v. California Dep't of Corr. & Rehab., Div. of Juv. Just., No. CV 07-5954-JFW (CWX), 2008 WL 11340255, at *2 (C.D. Cal. July 31, 2008) (citing United States v. Kitsap Physicians Serv., 314 F.3d 995, 1001 (9th Cir. 2002)). Parties "engage in spoliation of documents as a matter of law only if they had some notice that the documents were potentially relevant to the litigation before they were destroyed." Kitsap Physicians Serv., 314 F.3d at 1001 (internal quotation marks omitted). "The duty to preserve relevant evidence can arise before the filing of the Complaint, when litigation is reasonably foreseeable." Morris, 2008 WL 11340255, at *2 (internal quotation marks omitted). Here, Plaintiffs contend that litigation was not reasonably foreseeable when they retained Dr. Drexler in 2016 and sent him samples. Despite this four-year gap, however, the Court finds that litigation was "reasonably foreseeable" in 2016. DTSC and Exide had a lengthy ongoing relationship for decades. The AB2588 Heath Report was created in 2013; DTSC ordered Exide to perform cleanup of some properties as early as August 2014; and the Plant was shut down in 2014. California allocated hundreds of millions of funds for cleanup in April 2016. See AB 118 (Santiago) SB 93 (De Leon). Plaintiffs implicitly admit the foreseeability of litigation in a different portion of their brief by contending that Exide "had no incentive to conduct a skewed background analysis that would point to its own liability," see Pls. Post-Trial Br., at 24 n. 10. In other words. Plaintiffs' argument implicitly acknowledges that Exide knew that the background studies it conducted in 2014 could lead to its liability. For these reasons, contending that litigation was not reasonably foreseeable by October 2016 is unreasonable, and the Court concludes that exclusion of Dr. Drexler's testimony is warranted on the basis of spoliation. Nevertheless, the Court evaluates the merits of Dr. Drexler's testimony.

Dr. Drexler's testimony is unavailing because it is unpersuasive on the merits. Dr. Drexler initially opined that 17 soil samples had pyrometallurgical lead, suggesting that the Vernon Plant caused contamination in each of those samples. On cross-examination, Defendants revealed that this testimony was misleading. Defendants elicited testimony from Dr. Drexler that soil samples at five of the twelve properties Dr. Drexler evaluated were taken at the site of or immediately adjacent to other

smelters. Tr. 2 at 75:5-21. Yet Dr. Drexler's declaration did not mention this key fact.

To the contrary—and perhaps disingenuously—Plaintiffs attempted make a geographic connection between the Vernon Plant and these samples. For example, Dr. Drexler was told by Plaintiffs to note that a sample collected 2.6 miles away from the Vernon Plant contained pyrometallurgical lead—even though that sample was actually collected at the site of a different smelter. See Drexler Decl. ¶ 29 ("Pyrometallurgical forms of lead were found in the sample collected furthest from the Exide Site."); see also Tr. 2, at 75:5-9 ("Q. And this sample [], would you agree with me that it was more likely if it was from a smelter at all to have come from the Western Lead smelter than the Exide smelter 2.6 miles away? A. Yes."). This undermines Dr. Drexler's credibility and Plaintiffs' claims.

In response, Plaintiffs argue that six³⁸ of Dr. Drexler's samples, taken between 0.8 and 1.3 miles to the north and south of the Vernon Plant, were associated with a battery recycling facility. The two closest battery recycling facilities were the Vernon Plant and the adjacent Federated Metals smelter. Because the Vernon Plant operated for about 90 years, Plaintiffs argue, "it would be unreasonable to conclude" that at least one of the six samples didn't come from the Vernon Plant. See Post-Trial Br., at 23-24.

38 Plaintiff's Post-Trial Reply Brief refers to seven samples, not six. See Post-Trial Reply Br., at 17.

It is unclear to the Court how Dr. Drexler's definition of "pyrometallurgical" corresponds to Plaintiffs' conclusion that the six samples were "associated with a battery recycling facility." Dr. Drexler defined "pyrometallurgical" as a

metallurgical process that uses heat and various chemicals or compounds to melt concentrate or some other stock, in this case sometimes we are talking about batteries, battery waste but could be concentrate from a mine. It could be just recycled lead products in this case because it was a lead facility that we are talking about here. So it is using high heat to melt the material down and then using the heat and gravity to extract a more pure compound.

See Tr. 2, at 77:1-10. A "pyrometallurgical" process is identical to a smelting process. Id., at 78:4-6.

Dr. Drexler then said that he was able to link more specific forms of lead to either a primary or secondary smelting facility, with the caveat that "primary or secondary" are "general term[s]" because "some primaries can become secondary in their history, and some secondaries can act as a primary in their history." Id., at 81:10-15.

Though he had seen "almost all of [these pyrometallurgical phases] in both secondary and primary," id., Dr. Drexler explained that "some, in particular, would have an indication of a specific secondary smelter." Id., at 81:17-18. Specifically, phases with elements like arsenic, antimony, and tin, which are "very characteristic of a lead battery recycler." Id. at 82:23-83:1.

Dr. Drexler then explained that because a battery recycling facility is "generally classified as a secondary," the six samples at issue were "associated with a battery recycler." Id., at 84:4; see Drexler Decl., Ex. 19. The Court asked, and Dr. Drexler clarified, however, that his expertise did not extend to making a geographic determination

as to lead dispersion from a specific facility. Tr. 2, at 83:24-84:7.

The mere possibility that a few samples out of the tens of thousands collected throughout the PIA contained lead that was "generally associated" with a battery recycler does not empower the Court to find Dr. Drexler's testimony persuasive. Even though the Vernon Plant and Federated Metals were located between 0.8 and 1.3 mile aware from where these six samples were taken, the Court can only speculate as to the possibility that the lead in the samples was linked to the Vernon Plant. And as discussed above, Defendants' cross-examination significantly undermined Dr. Drexler's credibility.

Plaintiffs' evidence as to Dr. Drexler's microscopy analysis is thus inadequate to establish that the Vernon Plant plausibly could have caused contamination throughout the PLA's Residential Areas.

D. CONCLUDING ANALYSIS

With the above summary and analysis of the evidence and argument complete, the Court returns to the causation inquiry at the heart of this case: whether Defendants' releases caused the incurrence of Plaintiffs' response costs. See United States v. Sterling Centrecorp Inc., 977 F.3d 750, 756 (9th Cir. 2020); 42 U.S.C. § 9607(a). Plaintiffs must establish that lead plausibly could have migrated from the Vernon Plant to the PIA's Residential Areas.

Weighing the sum of the evidence dictates a verdict in Defendants' favor. Construing CERCLA liberally, the Court concludes that Plaintiffs are unable to meet their burden to establish release causation. First, considering background, Plaintiffs fail to establish the most critical component of a viable background measure: the history of lead-emitting industry in the background study reference areas, as opposed to the PIA. The Court agrees that Plaintiffs did not need to make a showing of "perfect" fit. But Plaintiffs' showing here depends exclusively on the determination of the contractors retained by Exide and any possible inference that this determination may support.

Specifically, on the record before the Court, the history of lead-emitting industry in the PIA creates too much uncertainty for the Court to attribute much weight to the background evidence. The starting point for background—the amount of lead that would be expected to be in PIA soils if the Vernon Plant had not existed—is essentially undefined, because the Court cannot determine how closely the Long Beach and Waterstone reference areas minor the PIA. Plaintiffs have not met their burden to show that the background studies are of sufficient evidentiary and persuasive value. See, e.g., Tr. 2, at 43:23-44:5 (testimony that the Waterstone Study area involved 30 neighborhoods from all over Los Angeles and that Dr. Singh did not know their proximity to industry).

Plaintiffs' theory of background essentially requires the Court to guess at what an appropriate starting point of contamination might be—yet we have no basis for doing so. Since Plaintiffs bear the burden of establishing causation, their failure to present any compelling evidence on the studies' reliability through their own experts or fact testimony from the contractors retained by Exide precludes the Court from drawing any determinative inferences about the measure of background. This remains true even considering Dr. Shields' initial testimony about the Waterstone Study, which Dr. Shields credibly qualified at trial. The background line of evidence therefore weighs in Defendants' favor.

Plaintiffs make much of the fact that their expert. Dr. Medine, testified in his declaration that the Vernon Plant accounted for 98% of the releases in the PIA. Dr. Medine stated in his declaration that "[t]he

TRI data is limited because it only goes back to 1987, and thus the TRI only addresses the more recent emissions. Nonetheless, the TRI data demonstrates that the Vernon Plant was the largest single contributor of lead among the industrial sources in the PIA, accounting for more than 98% of the total lead releases that were reported." Medine Decl. ¶ 33.

Though Plaintiffs attempt to invoke this percentage as some evidence of contamination, it is merely conclusory. Dr. Medine premises this statement on information compiled on an EPA website known as the Toxic Releases Inventory, what he refers to in his declaration as the TRI. When the Court attempted to search the TRI repository for this information, it was unable to find the basis for Dr. Medine's claim. The Court is thus uncertain as to how Dr. Medine calculated the 98% figure—or how it would have been possible to accurately do so—and therefore finds this figure unpersuasive. Relatedly, Dr. Medine also admitted that the data only stretched back to 1987 which, again, does not satisfactorily account for the area's 100-year long heavily industrial history. This conclusory statement therefore does not answer the question at issue during the trial: whether Defendants' releases could have plausibly caused contamination throughout the PIA.

Second, similarly, while the EPA-endorsed AERMOD model provides evidence of contamination traveling to some undefined parts of the PIA's Residential Areas generally, Plaintiffs have not met their burden in showing that contamination plausibly could have spread throughout the 1.7 mile radius surrounding the Vernon Plant. Because AERMOD was run in concentration mode, it cannot estimate the outer bounds of contamination in the PIA. See Section IV.B.I.ii. Thus, the model provides little basis for concluding not only that lead plausibly could have traveled to any defined area within the PIA, but also provides little basis for determining whether lead could have contaminated the entirety of the PIA.

The critical flaw in Plaintiffs' case is that, even though defense experts Dr. Shields and Mr. Machado agreed that some minute amount of lead could have traveled beyond the PIA's Industrial Areas, Plaintiffs have not offered any evidence regarding the outer boundaries of contamination in the PIA's Residential Areas.

This lack of evidence—a common flaw in Plaintiffs' arguments throughout this case—is critical because the outer bounds of the PIA are the basis for Plaintiffs' argument for recovery of cleanup costs. Dr. Shields testified that the Vernon Plant contributed a non-observable amount of lead to the PIA's Residential Areas, and Mr. Machado testified that particles could "easily" travel "miles" away, depending on a number of factors. But Plaintiffs overstate the value of this evidence because the experts' testimony does not enable Plaintiffs to meet their burden when considering the totality of the circumstances. Indeed, Dr. Shields and Machado concluded that the contamination in the Residential Areas was limited and cannot be defined the way Plaintiffs assert it can be.

Third, the statistical analyses also weigh in Defendants' favor. Dr. Rouhani's geostatistical analysis is the EPA-preferred method. Though Plaintiffs attempt to suggest that Dr. Rouhani's geostatistical method relied on incorrect data, Dr. Rouhani's sensitivity analyses showed no change in any potential pattern of contamination even when using Dr. Medine's vetted data set. Rather, Dr. Rouhani testified that the only spatial trend of contamination was detectable in the Industrial Areas. Dr. Medine's regression analysis, on the other hand, is, at best, difficult for the Court to evaluate. The Court is unable to conclude whether Dr. Medine's analysis shows an

actual trend of contamination in the PIA, or simply arrives at a trend because of his methodology in averaging data, filtering data, and mixing and matching inconsistent data sets at varying distances and directions from the Vernon Plant.

Moreover, Dr. Rouhani's conclusion that lead did not have a "statistically detectable effect" in the PIA is consistent with the testimony of Dr. Shields and Mr. Machado: some particles of lead may have still traveled to undefined portions of the Residential Areas, as Dr. Shields and Mr. Machado testified, without establishing a definite pattern of contamination throughout the PIA's Residential Areas. Neither Dr. Shields nor Mr. Machado concluded that contaminants could have affected all of the PIA, or even a specific defined portion of the Residential Areas.

Fourth, as discussed above, the parties agree that antimony is a reliable indicator of smelter-related lead, but Dr. Shields' testimony about the lead-to-antimony suggests a weak relationship between lead and antimony in the Residential Areas. Plaintiffs and Dr. Medine chose not to rebut Dr. Shields' analysis at trial. Though Dr. Medine contends that the lead-to-antimony ratio need only fall within a certain range to suggest the presence of smelter-related lead, Dr. Medine's declaration does not explain the basis for his reasoning. Moreover, Dr. Medine's opinion here is inconsistent with the viability of his own preferred methodology—use of a regression analysis—in the context of emissions from a point source.

For these reasons, Plaintiffs do not satisfy then burden to establish that lead from the Vernon Plant could have plausibly caused contamination throughout the PIA's Residential Areas—either in any defined parts or the whole of the PIA. The evidence does indeed show a possibility of lead from the Vernon Plant traveling to some undefined portion of the PIA's Residential Areas. Indeed, the Court agrees that, as a general matter, some unknown amount of contaminants likely reached some undefined portions of the PIA beyond the Industrial Areas. But this possibility does not rise to the level of plausibility. And to find that Plaintiffs have met their burden here would require a substantial leap unsupported by the evidence—or lack thereof—here.

The Court's decision does not rest on the amount of contamination at any given location in the PIA. In other words, if the evidence showed that particles could have spread throughout the PIA's Residential Areas, the fact that contamination was minute would not be an impediment to Plaintiffs' case. But here. Plaintiffs ask the Court to conclude that lead could have traveled 1.7 miles away from the Vernon Plant by simply invoking a "Residential Area" that is thousands of acres, without allowing the Court to do anything more than make a speculative guess as to the parts of the PIA to which lead may have traveled.

The Court does not compel Plaintiffs to establish plausible contamination on a granular level, on a house-by-house or block-by-block basis. To the contrary, the Court recognizes the difficulty of the line-drawing inquiry before it and liberally interprets CERCLA to effectuate Congress' intent. But the Court cannot ignore the substantial evidentiary gaps in Plaintiffs' case. For the Court to conclude that a 6000-acre area plausibly could have been contaminated, the geographic evidence must be more concrete.³⁹ As the Court explained above, see supra ns. 6-7, the PIA presents a distinct causation inquiry because the cleanup costs and boundaries of the site depended entirely on Plaintiffs' investigation and subsequent determination of which parcels they believed were contaminated. Water contamination cases provide a helpful point of comparison because they do not require a court to evaluate the scope of a contaminated area. If contaminants are introduced into a well, it is likely that the entire well will require remediation. Contaminants would spread throughout the water, contaminating the entire site and rendering Castaic Lake's plausible migration pathway standard all the more sound—because the only question is whether the contaminants could have traveled to the body of water. Here, by contrast, once lead landed in the PIA's soil, there is no evidence that it continued to spread with continued wind or vehicle turbulence. See supra, n. 16. Thus, Plaintiffs must show that the lead could have traveled throughout the Residential Areas. They have not done so here. Even a difference in a tenth of a mile can lead amount to a difference of millions of dollars in cleanup costs.

39 Part of the problem here involves California's apparent decades-long failure to adequately monitor and enforce the environmental law based on the Vernon Plant's emissions. Had California done so, it is likely that many of the evidentiary gaps in this case would not exist. For example. Plaintiffs' decision to use the AERMOD concentration mode results and failure to perform deposition mode modeling (or any other modeling) is surprising: in a case of this gravity, the Court had no model from which it could evaluate actual contamination in the Residential Areas. Or. that Plaintiffs attempted to create a theory of soil contamination based on seven samples evaluated by Dr. Drexler, when hundreds of thousands of samples were taken throughout the course of the case, is astonishing. We again echo that Plaintiffs' case is defined by its many evidentiary holes precluding the Plaintiffs from satisfying their burden.

Though the evidence shows that some undetectable amount of lead from the Vernon Plant may have reached the area beyond the Industrial Areas, Plaintiffs' burden was to establish that contaminants plausibly could have traveled throughout the PIA. Here, there is insufficient evidence from which the Court can conclude that Plaintiffs have met their burden. Even in light of the Court's liberal construction of CERCLA's statutory framework and Congress' underlying intent, CERCLA's broad remedial purpose does not empower the Court to water down causation to mere possibility. Plaintiffs' evidence may show possibility but does not rise to the level of plausibility. See Asarco LLC, 21 F. Supp. 3d at 804. "A liberal construction does not mean one that flies in the face of the structure of the statute." Newsome v. Cnty. of Santa Fe, 922 F. Supp. 519, 524 (D.N.M. 1996) (citations and quotations omitted). As discussed above, we decline to construe CERCLA's plain meaning and structure as including no causation requirement at all.

Therefore, the Court holds that Plaintiffs have not met their burden to establish that Defendants' releases caused Plaintiffs' incurrence of response costs in the PIA's Residential Areas. See United States v. Sterling Centrecorp Inc., 977 F.3d 750, 756 (9th Cir. 2020); 42 U.S.C. § 9607(a). But because the parties do not dispute that Defendants' release caused the incurrence of response costs within the PIA's Industrial Area, Plaintiffs may recover their costs incurred in responding to contamination in the Industrial Area.

V. CONCLUSION

The Court hereby enters a verdict in Defendants' favor on the Scope Trial: Defendants' releases only caused Plaintiffs to incur response costs within the half-mile radius surrounding the Vernon Plant.

IT IS SO ORDERED.