Ohio Valley Environmental Coalition v. Fola Coal Co. (120 F. Supp. 3d 509, 2015)

MEMORANDUM OPINION AND ORDER

ROBERT C. CHAMBERS, Chief Judge.

This suit concerns allegations that Defendant Fola . Coal Company, LLC,. has violated the narrative water quality standards of three separate permits for discharges from three mines into tributaries of-Leatherwood Creek. On June 1-4, 20Í5, the Court held a bench trial regarding jurisdiction and liability, and the parties timely conducted post-trial' briefing.

As explained below, the Court FINDS that Plaintiffs have established, by a preponderance of the evidence, that Defendant has committed at least one violation of its permits governing Fola Mine No. 2 and Fola Mine No. 6 by discharging into Road Fork and Cogar Hollow high levels of ionic pollution, which have caused, or materially contributed to a significant adverse impact to the chemical and biological components of the applicable streams’ aquatic ecosystem, in violation of the narrative water quality standards that are incorporated into those permits.- However, the Court further FINDS that Plaintiffs have not met their burden in establishing liability for alleged violations with respect to discharges from Fola Mine No. 4A into Right Fork, under NPDES Permit No. WV1013815.

I. Background

Plaintiffs Ohio Valley Environmental Coalition (“OVEC”), West Virginia Highlands Conservancy, and Sierra Club filed this case pursuant’ to the citizen suit provisions of the Federal Water Pollution Control Act (“Clean Water Act” or “CWA”), 33 U.S.C. § 1251 et seq., and the Surface Mining Control and Reclamation Act (“SMCRA”), 30 U.S.C. § 1201 et seq. CompL, ECF No. 1. Before proceeding to the parties’ evidence and arguments, the Court will first discuss the relevant regulatory framework and the factual background of this case.

A. Regulatory Framework

The primary goal of the CWA is “to restore and maintain the chemical, physical, and biological integrity of the Nation’s waters.” 33 U.S.C. § 1251(a). To further this goal, the Act prohibits the “discharge of .any pollutant by any person” unless a statutory exception applies; the primary exception is the procurement of a National Pollutant - Discharge Elimination System (“NPDES”) permit. 33 U.S.C. §§ 1311(a), 1342. ■ Under, the ,NPDES? the U.S. Environmental Protection Agency (“EPA”) or an authorized state agency can issue a permit for thé discharge of any pollutant, provided that the discharge complies' with the ■ conditions of the CWA. 33 ■ U.S.C. § 1342. A state may receive approval to administer a state-run NPDES program under the authority-of 33 U.S.C. § 1342(b). West Virginia received such approval, and its NPDES program is administered through the West Virginia Department of Environmental Protection (“WVDEP”). 47 Fed.Reg. 22363-01 (May 24, 1982). All West Virginia NPDES permits incorporate by reference West Virginia Code of State Rules § 47-30-5.1.f, which states that “discharges coyered by a WV/NPDES permit are to be .of such quality so as not to cause violation of applicable water quality standards promulgated by [West Virginia Code of State Rules ■§ 47-2].” This is an enforceable permit condition. See, e.g., OVEC v. Elk Run Coal Co., Inc., No. 3:12-cv-0785, 2014 WL 29562, at *3, *6 (S.D.W.Va. Jan. 3, 2014); OVEC v. Elk Run Coal Co., Inc., 24 F.Supp.3d 532 (S.D.W.Va.2014); OVEC v. Fola (Stillhouse), 82 F.Supp.3d 673 (S.D.W.Va.2015).

Coal mines are also subject to regulation under the SMCRA, which prohibits any person from engaging in or carrying out surface coal mining operations without first obtaining a permit from the Office of Surface Mining Reclamation and Enforcement (“OSMRE”) or an authorized state agency. 30 U.S.C. §§ 1211,1256,1257. A state may receive approval to administer a state-run surface mining permit program under the authority of 30 U.S.C. § 1253. In 1981, West Virginia received conditional approval of its state-run program, which is administered through the WVDEP pursuant to the West Virginia Surface Coal Mining and Reclamation Act (“WVSCMRA”). W. Va.Code §§ 22-3-1 to -33; .46 Fed. Reg. 5915-01 (Jan. 21, 1981). Regulations passed pursuant to the WVSCMRA require permittees to comply with the terms and conditions of their permits and all applicable performance standards. W. Va. Code R. § 38-2-3.33.C. One of these performance standards requires that mining discharges “shall not violate effluent limitations or cause a violation of applicable water quality standards.” Id. § 38-2-14.5.b. Another performance standard mandates that “[a]dequate facilities shall be installed, operated and maintained using the best technology currently available ... to treat any water discharged from the permit area so that it complies with the requirements of subdivision 14.5.b of this subsection.” Id. § 38-2-14.5.C.

B. Factual Background

This controversy concerns discharges from three surface mines along the southern portion of the Leatherwood Creek watershed: ,(1) Fola Surface Mine No. 2 in Clay and Nicholas Counties, West Virginia; (2) Fola Surface Mine No. 4A in Clay County, West Virginia; and (3) Fola Surface Mine No. 6 in Nicholas County, West Virginia.' Stipulátion ¶ 6, ECF No. 53.

Defendant’s mining activities at Surface Mine No. 2 are regulated under WV/ NDPES Permit WV1013840 and West Virginia Surface Mining Permit S201293, both originally issued in 1994. Id. at ¶¶ 6-7. WVDEP reissued WV/NPDES Permit No. WV1013840 in 2001, 2004, 2008, and 2014. At the time this complaint, was filed, the 2008 reissuance was in effect. Outfall 001 of Surface Mine No. 2 discharges into Road Fork and Leatherwood Creek. Id.

Defendant’s mining activities at Surface Mine No. 4A' are regulated under WV/ NPDES Permit WV1013815 and West Virginia Surface Mining Permit S200502. Id. at ¶¶ 21, 23. WV/NPDES Permit WW1013815 was originally issued in 1993, and was reissued in 1999, 2006, 2008, and 2014. At the time this complaint was filed, the 2008 reissuance was in effect.- Outfalls 22, 23, and 027 of Surface Mine No. 4A discharge into Right Fork of Leatherwood Creek and Cannal Coal Hollow. Id.

Finally, Defendant’s mining activities at Surface Mine No. 6 are regulated under WV/NPDES Permit WV1018001 and West Virginia Surface Mining Permit S2011999, both' originally issued in 2000. Id. at ¶¶ 42-44. WVDEP reissued WV/NPDES Permit WV1018001 in 2008. Id. at ¶43. At the time this complaint was filed, the 2008 reissuance was in effect. Outlets 013, 015, and 017 of Surface Miné No. 6 discharge into Cogar Hollow, a small tributary of Leatherwood Creek. Id. at ¶ 43.

In recent years, water quality measurements from the above listed discharges have routinely shown discharges of high conductivity. Stipulation ¶14, ECF No. 53 (showing discharges from Outlet 001 at Mine No. 2 with conductivity measurements consistently around 3000 ■ pS/cm); id. at ¶ 32 (showing discharges from Outlets 22, 23, 27 at Mine No. 4A consistently ranging from approximately 1500 pS/cm to above 3000 pS/cm); id. at ¶47 (showing discharges from Outlets 013, 015, 017 with conductivity measurements consistently ranging from approximately 2500 pS/cm to 4000 pS/cm). Water quality measurements have also revealed elevated conductivity in Leatherwood Creek and its tributaries. Id at ¶ 13 (showing conductivity levels ranging from 3000 pS/cm to 4000 pS/cm in Road Fork); id. at ¶ 33 (showing conductivity levels consistently above 1000 pS/em below Mine No. 4A); id. at ¶ 46 (showing conductivity levels ranging from 3000 pS/cm to 5000 pS/cm in Cogar hollow).

On June 1-4, 2015, the Court conducted a bench trial on liability issues. At the close of the evidence, the Court entered an oral finding on general causation, but reserved judgment on issues of specific causation. Tr. 4 at 259-60, June 4, 2015, ECF No. 107. Since that time, the parties have provided timely post-trial briefing. In Section II, the Court will review the evidence and arguments concerning general causation and elaborate on its general causation finding. ■ In Section III, the- Court will move on to review the evidence and arguments concerning specific causation for each of the ■ three- mine permits at issue.

II. General Causation

Generally speaking, Plaintiffs are faced with the dual burden of establishing both general and specific causation. At the close of trial, the Court announced its finding that Plaintiffs met their burden with respect to general causation. . Tr. 4 at 259-60, ECF No. 107. Specifically, Plaintiffs proved by a preponderance of the evidence that conductivity, as a measure of a consistent mix of ions typical of alkaline' mine drainage in the Appalachian region, may cause or materially contribute to biological impairment to aquatic life as measured by the West Virginia Stream Condition Index (“WVSCI”), thereby constituting a violation of the narrative water quality standards incorporated into Defendant’s permits. Id.; accord OVEC v. Elk Run Coal Co., 24 F.Supp.3d 532 (S.D.W.Va.2014);, OVEC v. Fola Coal Co. (Stillhouse), 82 F.Supp.3d 673 (S.D.W.Va.2015) The bases for the Court’s finding on general causation are explained in further detail below, with the majority of discussion focused on Defendant’s critiques of EPA’s Benchmark, followed by a brief review of other scholarly publications on the question..

A. Introducing the EPA’s Benchmark

Yet again, the Court begins its analysis of general causation with arguments concerning the import and reliability of the EPA’s Benchmark. See Elk Run, 24 F.Supp.3d at 558-59; Fola (Stillhouse), 82 F.Supp.3d 673. In March 2011, the EPA released “A Field-Based Aquatic Life Benchmark for Conductivity in Central Appalachian Streams” (“EPA’s Benchmark” or “Benchmark”). Joint Ex. 17. The EPA’s Benchmark is the studied result of qualified authors and reviewers. Id. at ix-xiii (listing authors, contributors, and reviewers, including Defendant’s expert, Dr. Charles Menzie).

In the nearly three hundred page Benchmark, the EPA reached the conclusion that “salts, as measured by conductivity, are a common cause of impairment of aquatic macroinvertebrates” in central Appalachian streams only after considering and' then ruling out the potential confounding effects of habitat, organic enrichment, nutrients, deposited sediments, pH, selenium, temperature, lack of headwaters, catchment areas, settling ponds, dissolved oxygen, and metals. EPA’s Benchmark at. A-l, B — 1; see also id. at A-40 (“This causal assessment presents clear evidence that the deleterious effects to benthic invertebrates are caused, by, not just associated with, the ionic strength [, i.e., conductivity,] of the water.-... When [other potential] eauses are absent or removed, a relationship between conductivity and ephemeropteran [, i.e. mayfly,] richness is still evident.” (emphasis added)); id. at A-37 (“As conductivity increases, the occurrence and capture probability decreases for many genera in West Virginia ... at the conductivity levels predicted to cause effects. The loss of these genera is a severe and clear effect.”). The Benchmark also found that “of the [nine] land uses ... analyzed, only mining especially associated with valley fills[, i.e,, mountaintop mining with valley fills,] is a substantial source' of the salts that are measured as conductivity.” Id. at A-18.

The EPA ultimately concluded that the “chronic aquatic life benchmark value for conductivity” in West Virginia streams is 30Ó |xS/cm. Id. at xv. To derive this recommended high-end threshold value, the EPA used the 5th percentile of a species sensitivity distribution, based on the standard methodology for deriving water-quality-criteria, meaning that this 300 gS/cm benchmark value for conductivity is “expected to avoid the local extirpation [due to the salts measured as conductivity] of 95% of native species.” Id. at xiv.

In support of' both the specific 300 gS/cm benchmark value and the general causal linkage between conductivity and impairment to aquatic macroinvertebrates, the Benchmark contains a graph which charts, for 163 genera, the level of ionic exposure above which a genus is effectively, absent from water bodies in a region, with conductivity readings on the x axis and proportion of genera extirpated on the y axis. Id. at xiv, 18 fig. 8. A fairly consistent line is formed as conductivity and extirpation both increase, illustrating the causal connection between conductivity and significant biological impairment which Plaintiffs seek to prove. See id. at 18 fig. 8. Relatedly, the EPA reported its finding that “the probability of impairment’ at 500 jxS/cm is 0.72 and at 300 gS/cm, is 0.59.” Tr. 2 at 111, June 2, 2015, EOF No; 100; Joint Ex. 17 at A-36. Stated differently, when conductivity reaches 300 |xS/cm, it is more likely than not that the streams will suffer impairment. Moreover, the likelihood of impairment continues to increase as conductivity further exceeds that threshold. Joint Ex. 17 at A-36; Tr. 2 at 110-12, ECF No. 100.

Upon reviewing the EPA’s findings, the Scientific Advisory Board (“SAB”) made the follow comments:

Mountaintop mining and valley fills are important sources of stress to aquatic systems in the Central Appalachian region, both from the perspective of localized and cumulative regional impacts. In a companion report, the Panel provides a review .of EPA’s assessment of the impacts associated with mountaintop .mining and valley fills. There is clear ..evidence that valley fills are associated with increased levels of dissolved ions (measured as conductivity) in downstream .waters, and that these increased levels of conductivity are associated with changés in the composition of stream biological communities.

Pis.’ Ex. 128 at PE1418, Tr. 2 at 112-13, ECF No. 100. The SAB further concluded that the EPA had presented a “convincing case” for establishing the causal relátionship between conductivity and loss of genera. Pis.’ Ex. 128 at PE 1431, Tr. 2 at 115, ECF No. 100.

Plaintiffs rely . on EPA’s Benchmark here as they have elsewhere: as a scientific study which, among others, supports Plaintiffs’ general causation theory that high conductivity levels in streams impacted by alkaline mine ■ drainage causes or contributes to biological impairment. Defendant’s. many critiques of the EPA Benchmark will be considered below. Before doing so‘, however, it is necessary to briefly revisit general principles regarding the degree of deference owed EPA’s Benchmark in the analysis to follow.

“Particularly with environmental statutes such as the Clean Water Act,.the regulatory framework ... requires sophisticated evaluation of complicated data____ [A court] therefore do[es] not sit as a scientific body in such cases, meticulously reviewing all data.under a laboratory microscope.” - Crutchfield v. Cnty. of Hanover, Virginia, 325 F.3d 211, 218 (4th Cir.2003) (citation omitted) (internal quotation marks omitted). Instead, “[a] reviewing court must generally be at .its most deferential when reviewing factual determinations within an- agency’s .area of special expertise.... It is not the role of a re? viewing court, to second-guess the scientific judgments of the EPA.” Sw. Pennsylvania Growth Alliance v. Browner, 121 F.3d 106, 117 (3d Cir.1997) (citation omitted) (internal quotation marks omitted); see also Baltimore Gas & Elec. Co. v. Natural Res. Def. Council, Inc., 462 U.S. 87, 103, 103 S.Ct. 2246, 76 L.Ed.2d 437 (1983) (“[A] reviewing court must remember that the [agency] is making predictions, within its area of special expertise, at the frontiers of science. When examining this kind of scientific determination, as opposed to simple findings of fact, a reviewing court must generally be at its most deferential.”); Envtl. Def. Ctr., Inc. v. U.S. E.P.A., 344 F.3d 832, 869 (9th Cir.2003) (‘We treat EPA’s decision with great deference because we are reviewing the agency’s technical analysis and judgments, based on an evaluation of complex scientific data within the agency’s technical expertise.”); Chem. Mfrs. Ass’n v. U.S. E.P.A., 919 F.2d 158, 167 (D.C.Cir.1990) C“[W]e' give considerable latitude to the EPA in drawing conclusions from scientific and technological research, even where it is imperfect or preliminary.” (internal quotation marks omitted)).

“[technological and scientific issues ... are by their very nature difficult to resolve by traditional principles of judicial decisionmaking. For this reason, we must look at the decision not as the chemist, biologist or statistician that we are qualified neither by training nor experience to be, but as a reviewing court exercising our narrowly defined duty of holding agencies to certain minimal standards of rationality.” Reynolds Metals Co. v. U.S. E.P.A., 760 F.2d 549, 558-59 (4th Cir.1985) (internal quotation marks omitted). “[A]n agency’s data selection and choice of statistical methods are entitled to great deference, ... and its conclusions with respect to data and analysis need only fall within a zone of reasonableness.” Id. at 559 (citations omitted) (internal quotation marks omitted). In the context of agency action, “if the agency fully and ably explains its course of inquiry, its analysis, and its reasoning sufficiently enough for us to discern a rational connection between its decision-making process and its ultimate decision, [a court] will let its decision stand.” Crutchfield, 325 F.3d at 218 (brackets omitted) (internal quotation marks omitted).

In light of these precedents, and as previously analyzed by this Court, EPA’s Benchmark must be afforded deference. See Elk Run, 24 F.Supp.3d at 558-59, Fola (Stillhouse), 82 F.Supp.3d at 679-82. The EPA’s Benchmark methodically defines its inquiry, explains its reasonable analysis, and thoroughly supports its ultimate, rational conclusions. Additionally, the Benchmark underwent extensive scientific review, and it is respected as good — or even excellent — science within the relevant scientific community. Dr. Palmer, Tr. 2 at 96, ECF No. 100.

B. Critiques of the EPA Benchmark

Turning to consider newly presented evidence and argument, two recurring questions underlie the Court’s instant analysis of general causation as related to EPA’s Benchmark: (1) whether specific expertise in epidemiology is required for the development or review of EPA’s Benchmark; and, in a similar vein, (2) whether specific expertise in ecology is required for the development or review of EPA’s Benchmark. We are faced with these fundamental questions because of the apparent (and unsurprising) difficulty in finding an expert in both epidemiology and ecology. Instead, the Court heard testimony of expert epidemiologists with no formal ecological training and expert ecologists with no formal epidemiological training. From this mix of incomplete expertise, we are left with the task of sorting competing expert opinions.

As argued by Defendant, because the EPA incorporated principles of epidemiology into its causal analysis, assessing the reliability of the EPA’s findings requires review by an epidemiologist. Defendant’s expert epidemiologist, Dr. David Garabrant, reviewed' the EPA’s findings and found several areas where he believed the EPA misapplied epidemiological principles. Tr. 1 at 11, June 1, 2015, ECF No. 105 (asked whether the EPA correctly applied principles of epidemiology, Dr. Gara-brant responded, “In some ways, yes; and in some ways, no.”). Dr. Garabrant thusly criticized perceived failures on the part of EPA (1) to consider effect modification, (2) to define reliable and valid criteria for assessing confounding, (3) to adequately respond to the quality of the available data, and (4) to transparently and non-manipulatively disclose all data. Each of Dr. Garabrant’s critiques will be reviewed in turn below, but first the Court observes that Plaintiffs’ responses to these critiques can be boiled down to the suggestion that •Dr. Garabrant’s critiques are fundamentally flawed insofar as Dr. Garabrant did not adequately understand the underlying subject matter, i.e., freshwater ecology.

Taking a particularly illustrative example, Dr. Garabrant’s critique of Table B-7 of the EPA Benchmark suggests an inability to correctly apply common statistical tools presumably well within his epidemiological expertise and, in doing so, to adequately interpret available ecological data. Table B-7 offers a relatively straightforward presentation of data purporting to represent two regression lines. Joint Ex. 17 at JE0773. ‘ Dr. Garabrant demonstrated the perceived failings of the table by attempting to recreate the graph. See Tr. l.at 22-24, EOF No. 105¡ Using the data in the table, Dr. Garabrant’s recreated graph offered only the nonsensical result of predicting a total absence of mayflies at background conductivity levels. Id. Basic knowledge of ecology and observed conditions tell us that, as graphed by Dr. Gara-brant, the numbers in Table B-7 cannot be correct. Thus, Dr. Garabrant offered his expert opinion that the table was nonsense. Tr. 1 at 22 (“Something’s seriously wrong. It is not a valid result.”); id. at 24 (“We know that the .maximum number of ephemeropteran genera is 14. Of course, the minimum has to be zero.. All this [table] generates is negative numbers. It’s nonsense.”).

Contrary to Dr. Garabrant’s opinion, Plaintiffs’ expert found the same Table B-7 to be perfectly sensible with .the addition of a single interpretive move. Though explained elsewhere in the Benchmark (see, e.g., Figures 13a, 13b, 13c, 13d, and 13e), the authors made no explanation of whether a logarithmic scale shquld.be used to interpret the data shown at Table B-7. Though not specifying that the data in the table would need to be logarithmically transformed, according to Dr. Baker, the appropriateness of using a log scale would be obvious to an ecologist. Tr. 4 at 229-30, ECF No. 107. With that background expertise in ecological data analysis, Dr. Baker produced two graphs, each using different scales, and each showing results consistent with the . data analyzed. Pis.’ Exs. 176 and 177.

Thus, what appeared to one epidemiological expert to be an incorrect and nonsensical table, was in fact a perfectly sensible table that the authors merely neglected to. adequately label for non-expert reviewers.' While there is likely no across-the-board answer to what respective degrees of epidemiological and ecological expertise are 'necessary to evaluate the EPA’s Benchmark, this - example serves as a ready reminder in-the analysis-to follow that something beyond wholly non-contextual data analysis may be needed.

1. Assessment of Effect Modification

Turning to Dr. Garabrant’s broader critiques of the EPA Benchmark, we begin with the suggestion ¡that the Benchmark is fundamentally flawed insofar as the EPA failed to account for effect modification. As explained by Dr. Garabrant, “[e]ffect modification occurs when the association between two factors is different depending on' the presence or absence of a third factor. If, for example, the association between an exposure and an outcome is different for men than for women, sex modifies the relationship between the exposure and the outcome.” ECF No. 90-1 at 3; see also Dictionary of Epidemiology (Miguel Porta ed., 6th ed.2014) (defining “effect modification” as a “[vjariation in the selected effect measure' for the factor under study across levels of another factor”); id. (defining “effect modifier” as “[a] pre-exposure factor across whose levels the value of the effect measure of interest varies; [a] factor that biologically, clinically, socially, or otherwise alters the effects of another factor under study”). When effect modification is present, “[combining the two groups to create a summary measure, of .association is meaningless: it is not true for men and it is not true for women.” ECF No. 90-1,at 3. Thus, before a causal analysis moves on to continue potential confounders, it is essential to first assess whether effect modification is present. Dr. Garabrant, Tr. 1 at 14-15, ECF No. 105.

Returning to Dr. Garabrant’s critique, indeed, the words “effect modification” cannot be found in the text of the Benchmark, suggesting to Dr. Garabrant that the EPA did no analysis of effect modification. In an effort to assess the presence or absence of effect modification, Dr. Gara-brant turned to the underlying data and produced a series of figures purportedly showing the presence of effect modification. See Def.’s Exs. 31-36. In each table, the percent of sites with ephemeropt-era present is represented along the y-axis, conductivity is represented along the x-axis, and blue, red, and green lines run across the graph as representations of low-, mid-, and upper-range values for a given potential effect modifier, respectively. Taking the example of 'pH as a potential effect modifier (Def.’s Ex. 31), Dr. Garabrant explained that the effect of pH can be gleaned from looking at “the vertical distance between the green line and the blue. line holding conductivity constant,” with greater vertical distance suggesting greater likelihood of effect modification. Tr. 1 at 36-37, ECF No. 105.

Using these tables to thusly visualize the data, Dr. Garabrant reached the conclusion that effect modification was present with respect to pH (Tr. 1 at 37, ECF No. 105 (discussing Def.’s Ex. 31: “[i]f pH is neutral to high, there is no relationship between conductivity and Ephemera. If pH is low, the insects are adversely affected. That’s what effect modification looks like”)), stream size (Tr. 1 at 39, ECF No. 105 (discussing Def.’s Ex. 32: “I think you have evidence here of effect modification ..You are getting a different answer according to stream size. ' That’s effect modification”)), dissolved oxygen (Tr. 1 at 41, ECF No. 105 (discussing Def.’s Ex. 33)), iron (Tr. 1 at 43, ECF No. 105 (discussing Defi’s Ex. 34)), and manganese (Tr. 1 at 44-45, ECF No. 105 (discussing Defi’s Ex. 36)). Thus, the Court has one trained epidemiologist, with no formal background or experience in ecology, claiming that the EPA neglected to consider effect modification, and in so doing, missed the presence of several effect modifiers, thereby undermining the entirety of its causal analysis.

A second trained epidemiologist offered testimony on the same issue, but reached starkly different conclusions. Responding to Dr. Garabrant’s analysis of effect modification and EPA’s Benchmark, Plaintiffs’ expert, Dr. Wing, cautioned that, “in order to make á. decision about interaction or effect modification, it’s first necessary to have some idea about the topic one is investigating because without that, one can make egregious mistakes about an analysis which can be done by someone who doesn’t know anything about the topic but could result in essentially meaningless conclusions or actually conclusions that are misleading.” Tr. 2 at 20, ECF No. 100. He further stated the belief “that the issue of effect modification or interaction is one that should be made based on subjective knowledge in the area, and it’s not one that’s simply a statistical requirement or rule.” Tr. 2 at 20, ECF No. 100.

Consistent with that fundamental reservation and despite his considerable epidemiological expertise, Dr. Wing was unable to agree with Dr. Garabrant’s conclusion that effect modification is present in the dataset and yet left unaddressed by EPA. Looking, for example, at Dr. Garabrant’s figure assessing dissolved oxygen as a potential effect modifier (Def.’s Ex. 33), Dr. Wing observed similar trends across low-, mid-, and upper-range dissolved oxygen levels. Tr. 2 at 45, ECF No. 100. According to Dr. Wing, the similarity of the trend, or slope, suggests an absence of effect modification. Id. Moreover, solely based on the graph relied upon by Dr. Garabrant, Dr. Wing explained that it was impossible to definitively assess effect modification because Dr. Garabrant neglected to include any information on sample size or precision (e.g., no slope estimates or standard error estimates are provided). Tr. 2 at 46, ECF No. 100. And so the opinion of a second epidemiologist without ecological training reaches not only a contrary conclusion about effect modification, but further identifies an analytical barrier to reliably interpreting the graphs relied upon by the first.

To that uncertain mix, Dr. Baker contributes his opinion on effect modification as an ecologist without formal epidemiological training. Like Dr. Wing, Dr. Baker similarly critiqued the absence of information on sample size or precision. Dr. Baker further called attention to the fact that values were binned across the conductivity gradient by Dr. Garabrant in a manner that failed to control for sample size within each bin. Most surprising, however, was Dr. Baker’s testimony that the EPA did assess effect modification. According to Dr. Baker, the term “effect modification” is not commonly used in ecology. Tr. 4 at 239, ECF No. 107. Instead, ecologists commonly refer to “covariation” as a “catchall term” used for both confounding and effect modification. Tr. 4' at 191, ECF No. 107. Though never using the term, Dr. Baker remains assured that the EPA considered effect modification through alternate means. Id. at 191, 239.

Even standing independently, the rebuttal arguments offered by Dr. Wing and Dr. Baker arguably do enough to dispose of Dr. Garabrant’s critique of the EPA Benchmark related to analysis of effect modification.' Dr. Wing’s testimony effectively draws the adequacy of Dr. Gara-brant’s expertise into question, and Dr. Baker’s testimony demonstrates the importance of ecological expertise in reading and evaluating the EPA’s work. Furthermore, Plaintiffs’ expert testimony does not stand alone; it is acéompanied by and consistent with the expertise and analysis of the EPA — an expert federal agency acting in its area of expertise.

2. Analysis of Confounding

Dr. Garabrant further argued that the EPA performed an unreliable analysis of confounding, thereby rendering EPA’s causal conclusions invalid. The Dictionary of Epidemiology defines “confounding” as

[T]he ■ distortion of a ‘-measure of the effect of an exposure on an outcome due to the association of the exposure with other; factors that influence the occurs rence of the outcome.- Confounding occurs when all or part of the apparent association between the exposure and the outcome is in fact accounted for by other variables that affect the outcome and are not themselves affected by exposure,”

As argued by Dr. Garabrant; the EPA failed to adequately and reliably assess confounding, instead relying on an unverified and subjective methodology. Tr. 1 at 16-18, ECF No. 105.

Asked to comment on the validity of EPA’s approach to analyzing confounding, Dr. Garabrant hesitated to say whether the approach was valid or not. Tr. 1 at 16, ECF No. 105 (“It’s hard to say whether it is valid. I have never seen it used. I haven’t seen any-validation of it. I-have never seen any test of this method to show that it works. So I would say it’s not known whether it’s reliable or not. EPA created it.”). Beginning broadly, Dr. Gar-abrant called attention to the following paragraph from the Benchmark explaining the EPA’s approach to confounding in its causal analysis:

Weighing evidence for confounding factors differs from weighing evidence for causation. The causal assessment in Appendix A determines whether dissolved salts are an important cause of biological impairment ■ in the region. This assessment of confounding accepts the result of the causal assessment and attempts to determine whether any of the known potential confounders interfere with estimating effects of conductivity to a significant degree.

EPA Benchmark at B-3. As explained by Dr. Garabrant, this paragraph reflects an analytical error on the part of the EPA akin to “putting the cart before the horse.” Tr. 1 at 17,‘ ECF No. 105 (“[T]he idea that you accept the result of causal assessment and then look at confounding is simply putting the cart before the horse. It’s backwards.”). In addition to this analytical error, Dr. Garabrant suggested that EPA’s approach was relatively arbitrary and subjective. Tr. 1 at 16, ECF No. 105.

Dr. Garabrant then went on to explain that epidemiologists commonly rely on a relatively straightforward way to identify the presence of confounding effects: compare the results of a crude analysis testing the association between conductivity and extirpation against the results of an adjusted analysis that introduces a potential confounding factor. Tr. 1 at 18-19, ECF No. 105. Using the same dataset relied upon by the EPA, Dr. Garabrant performed precisely that analysis. See Def.’s Ex. 40. The results of that analysis are reproduced in the following table:

Referring to this table, Dr. Garabrant offered testimony that any change in parameter estimate after adjusting for a given variable greater than 10% signals the presence of confounding. Tr. at 71-72. On cross-examination, however, Dr. Garabrant readily acknowledged that while “there is widespread agreement -that more than fifty percent change is important,” in the range of ten to twenty percent, judgments about confounding would depend on the analyst’s background knowledge in the subject matter at issue. Tr. 1 at 72, ECF No. 105. As already mentioned, Dr. Garabrant is without precisely that background knowledge. Accordingly, the Court is left to conclude that Dr. Garabrant’s'ten percent threshold is itself arbitrary and unreliable.

Beyond analytical differences of opinion, the Court is further unmoved by Dr. Gara-brant’s analysis of confounding given the quality of underlying data and the nature of certain variables. First, as will be discussed at length in the next section, some of the variables analyzed by Dr. Garabrant for confounding are known to ecologists to have little to no relevance in the context of West Virginia streams impacted by alkaline mine drainage. (e.g., orthophosphates, see infra Section II.B.3.a). Second, and also discussed below, the database lacks a significant number of data points for some of these variables unless total and dissolved values are considered together. (e.g., magnesium, calcium, selenium, and manganese, see infra Section IÍ.B.3.a).

Like Dr. Garabrant, this Court would be unable to set anything but an arbitrary threshold for recognizing potential confounding variables. Instead, -the Court continues to. rely on the expertise of ecologists. and testimony assuring the Court that the EPA engaged a reasonable and verified analysis of confounding.

3. Adequacy of the underlying data

Dr. Garabrant offered two critiques of the data relied upon by EPA in developing the Benchmark. First, Dr. Garabrant highlighted missing data points (e.g., limited number of data points for dissolved calcium). Second, Dr. Garabrant criticized EPA’s presentation of data and its analysis thereof as misleading.

a. “Missing ” Data

With respect to allegations of fatally missing data, Dr. Garabrant prepared a table reporting the number and percent of missing data points for each variable missing greater than 50% of the possible data points. Def. Ex. 38 (reproduced below).

According to the table, greater than 98% of the data points are missing for dissolved magnesium, dissolved calcium, dissolved manganese, dissolved orthophosphates, and total orthophosphates. Def.’s Ex. 38. Additionally, between 67% and 78% of the data was missing for various land cover variables, 85% of the data was missing for dissolved selenium and 77% missing for total selenium. According to Dr. Gara-brant, these deficiencies in the dataset prevented the EPA from meaningfully analyzing potential effects of these variables. Tr. 1 at 24-25, ECF No. 105.

However, as an ecologist, Dr. Baker was not similarly troubled by the missing data. First, he explained that deficiencies found in dissolved magnesium, were well compensated for by data on total magnesium. Tr. 4 at 199-201, ECF No. 107. As is the case with magnesium, greater than 50% of the sites in the WVDEP database had available data -points for total calcium and total manganese. Id. Similarly, when the data points for both dissolved and total selenium are combined, greater than 50% of the sites had data on selenium levels. Id.

With respect to alleged missing data on land cover categories, Dr. Baker was similarly untroubled. Dr. Baker testified that, as an ecologist, one would not be likely to consider any of these variables as potential confounding variables and would therefore likely ignore these categories as immaterial. Id. Finally, with respect to ortho-phosphates, Dr. Baker explained again that the lack of data would not trouble an ecologist in this context, because ortho-phosphates in high concentrations are associated with agricultural landscapes, which are generally not found near — much less coextensive with — mining areas in West Virginia. Id. at 201. Stated differently, the absence of land cover data or orthophosphate measurements would only trouble a reviewer to the extent that he did not have the necessary background familiarity with ecology and land use patterns to independently recognize the insignificance of the variables.

b. “Hidden” Data

In addition to purportedly fatally missing data, Dr. Garabrant also criticized the EPA Benchmark for “hiding data.” As explained by Dr. Garabrant, a series of tables provided in Appendix B (Tables B-8 (habitat), B-13 .(Embeddedness), B-15 (pH), B-23 (stream size), B-25 (dissolved oxygen), B-28 (iron), B-29 (Aluminum), B-30 (Manganese)) all share a common flaw: failure to include significant chunks of data for conductivity levels between 200 ¡xS/cm and 1500 pS/cm, instead, only showing data at extreme conductivity conditions (i.e., < 200 fxS/cm and > 1,500 pS/cm). In some instances, these tables fail to include significant mid-range data for not only conductivity, but also for the variable of interest (e.g., mid-range iron data was not included in Table B-28). Dr. Garabrant recognized and critiqued the absence of mid-range data, and was unable to provide any methodological or analytical justification for its absence.

Making good use of database access, Dr. Garabrant recreated the suspect tables to include mid-range data (see Def.’s Exs. 39, 43-49) and then continued to plot the data represented in each table (see Def.’s Exs. 30-36). Based on this information, Dr. Garabrant reached two conclusions: (1) effect modification is present with respect to each co-variate represented; and (2) there is. a consistent absence of conductivity effects regardless of .co-variate levels until conductivity reaches 1200 pS/cm to 1500 pS/cm. These conclusions suggested to Dr. Garabrant that there are serious flaws in the EPA’s analysis. Tr. 1 at 33-46, ECF No. 105. In turn, Dr. Garabrant’s only explanation for how the Benchmark reached publication despite hidden data became the suggestion that the SAB and peer reviewers would not have had access to the dataset and the ability to perform the analysis he did. Tr. 1 at 48-49, ECF No. 105.

Through the testimony of Dr. Baker and Dr. Wing, Plaintiffs supplied convincing rebuttals to each criticism related to “hidden data” raised by Dr. Garabrant. Not only are these rebuttal arguments convincing, but to some extent, they also highlight the drawbacks of data analysis performed with relatively limited understanding of the subject matter being analyzed. First, Dr. Baker explained a methodological reason EPA did not include mid-range data in the tables: given the nature of .the data, mid-range. values were irrelevant to answering the question asked. Tr. 4 at 203-04, 207-11, EOF No. 107. The EPA-datar set was qualitatively limited (or coarse) in that the data captured presence or absence of mayflies, but not information on abundance or variety, The category “mayflies” includes a variety of discrete species, each with particular sensitivity to- conductivity. Some mayflies are uniquely sensitive, while others are uniquely tolerant to conductivity. As established in the then published' literature, all mayflies — sensitive and tolerant alike — can be expected to have a négative response to conductivity levels in excess of 1,500 |xS/cm. Stated differently, until conductivity exceeds 1,500 (xS/cm, the available data would likely show some'mayfly present. If some mayfly is present, however, that does not necessarily tell us anything about abundance (i.e., only one bug could be present) or variety (i,e., only one species of mayfly could be present). Accordingly, if the question is whether or not mayflies may be present regardless of co-variate influences, the data only allow us to answer that question- if we look to the extreme conductivity ranges (i.e., lowest conductivities where we would expect even the most sensitive mayflies to be present and the highest conductivities where we would expect even the most tolerant mayflies to be absent).

Given Dr. Baker’s more nuanced explanation of the import of the data shown and the data not shown, Dr, Garabrant’s criticism of hidden data does little to impugn the work of EPA. scientists, instead illustrating Dr. Wing’s point that data analysis is not the same as the interpretation of data. While an epidemiologist may be- qualified to run data analyses on any dataset, it should not be assumed that an epidemiologist is necessarily otherwise qualified to interpret the results of that analysis.

Second, Dr. Baker tailored his interpretation of the data presented according to the function. of the tables in the overall analysis. Importantly,, these tables were not relied upon by the EPA .to identify threshold effects; these tables were introduced to explain EPA’s confounding analysis. Presence or absence of some mayfly genera without any data on abundance or variety explains very little that would help to identify a conductivity threshold at which the most sensitive macroinverteb-rates suffer extirpation.

4. Inter-state differences in species sensitivity

Dr.- Garabrant’s observations about the differences between XC95 values in West Virginia and- in Kentucky are among his most immediately compelling observations. The Benchmark’s XG95 values report genera-specific response thresholds at which you can expect 95% of freshwater macroin-veterates to tolerate conductivity levels. While -it is not surprising to expect different genera to have different response thresholds, Dr. Garabrant made the troubling observation that there are variances in the response thresholds within genera based on whether the data was sourced from West Virginia or Kentucky. For example, cross comparison of tables reveals the following differences between response thresholds for genera in West Virginia and Kentucky, among others: ■

It is easy to share Dr. Garabrant’s, shock upon noticing that different genera apparently have different response thresholds depending on whether they are observed in one state or another. Surely biological responses should not vary according to political boundaries; indeed, as aptly stated by Dr. Garabrant, “bugs don’t know where they live.” Tr. 1 at 29, ECF No. 105.

Though Dr. Garabrant could only imagine such discrepancies suggested fatal methodological and analytical flaws, Dr. Baker readily offered sound explanations rooted in the nature of data analysis and data collection. Tr. 4 at 211-14, ECF No. 107. First, whatever the differences in XC95 values of a few genera, across the entire datasets, the XC95 are very well correlated. Id. at 211-12. Second, the West Virginia dataset includes several thousand samples; the Kentucky dataset includes roughly two hundred. Id. at 213-14. All else being equal, based on sample size alone, the Kentucky,data would have a larger possibility of error and the West Virginia database would be .more reliable. Id. In addition to differences in the quantity of data, the quality of the data for each state is.unique. Id.,at 214-15. In Kentucky, sampling protocol directs that all bugs collected in the sample be counted. Id. In West Virginia, the sampling protocol directs that only a subset of the total sample be counted. Id.

Given these differences in collection methods and database size, one would expect that the EPA would identify two different and loosely associated benchmarks. Yet the methodology used by-the EPA identified remarkably similar benchmark values notwithstanding species-specific differences. Dr. Garabrant correctly concludes that the datasets are imperfect— individually and relatively. So far as this Court understands ecological study, data-sets are invariably imperfect. Notwithstanding perennial deficiencies in information, it. remains the task of,the scientist to distill-reliable (and in the regulatory context, actionable) results. Based on Dr. Baker’s explanation of the distorting effect of the quality and quantity of data in West Virginia as opposed to Kentucky, the Court remains confident that the EPA Benchmark presents reliable findings based on the information available. This is particularly the case with respect to the EPA’s analysis and conclusions based on the WVDEP database. Tr. 4 at 216, ECF No. 107 (“Overall, I would expect the West Virginia dataset to be a little •■bit more precise given the nature of the data size.”).

C. Scholarly Publications and Expert Opinions

In addition to EPA’s Benchmark, Plaintiffs further relied on a seemingly ever-growing colléction of published, -peer-reviewed journal articles addressing the connection between conductivity and impairment in ' Appalachian streams. Through the testimony of experts, ■ the Court was introduced to myriad peer-reviewed articles. In revisiting that collection of articles below, note the complete absence of peer-reviewed scientific articles to the contrary. Tr. 2 at 95, ECF No. 100. Instead,- the scientific community repeatedly reaches and reports the same conclusion despite the use of multiple methodologies relying on a variety of data-sets and conducted by a range of expert scientists. Given that growing and consistent body of scientific study, it is not surprising that Dr. Palmer is of the opinion that “there is a strong relationship and evidence of causation between high conductivity and impairment” in central Appalachian streams impacted by alkaline mine drainage. Tr. 2 at 94, ECF No. 100 (“I have at this point absolutely no doubt. There are so many studies that have been done, using very different methods and very different places that have all reached the same conclusion.”); see also Tr. 3 at 125-26 (Dr. Baker explains that the relationship between elevated conductivity and biological impairment is very strongly supported, to the point that he would sooner consider it a fact of science than a theory).

The scientific literature concerning the relationship between conductivity and impairment likely began in earnest in 2003 with publication of an Environmental Impact Study (EIS) of mountaintop mining valley fills. Tr. 2 at 99, ECF No. 100. Authors of the EIS examined changes in water chemistry and biological assemblages, finding increased concentrations of sulfates and dissolved solids, increased specific conductance, and a coincident decrease in sensitive taxa in impacted streams. Tr. 2 at 99-100, ECF No. 100. Other early publications included a paper by Kennedy et al. the following year, relaying the finding that exposure to elevated conductivity levels resulted in loss of organisms (Joint Ex. 9; Tr. 2 at 100, ECF No. 100), and a 2005 publication by Hartman examining the relationship between conductivity and mayfly richness. Tr. 2 at 100, ECF No. 100.

In 2008, Gregory Pond et al. published a paper in the Journal of North American Benthological Society, titled “Downstream effects of mountaintop coal mining: comparing biological conditions using family- and genus-level macroinvertebrate bioas-sessment tools.” Joint Ex. 13. In the underlying study, the authors conducted field sampling in order to analyze differences in water chemistry and macroinver-tebrate assemblages at mined and un-mined sites. Tr. 2 at 104. Pond et al., concluded that there was strong evidence of a causal relationship between conductivity and biological impairment. Id. at 104 (“Our results indicate that [mountaintop removal mining] is strongly related to downstream biological impairment, whether raw taxonomic data, individual metrics that represent important components of the macroinvertebrate assemblage, . or [multimetric indexes] are considered. The severity of the impairment rises to the level of violation of water-quality standards (WQS) when states use biological data to interpret narrative standard's.”). Moreover, the authors particularly noted that mayflies were especially sensitive to changes in water chemistry. Id. at 104.

Furthermore, Pond et al. calculated correlation coefficients of “[GLIMPSS] and [WVSCI] and genus- and family-level non-metric multidimensional scaling (NMS) axis scores [verses] a truncated list of environmental variables,” including conductivity, embeddedness scores, sediment deposition scores, and total rapid biological protocol (“RBP”) habitat scores. See Table 5, Pis.’ Ex. 173 at JE0196. The authors found statistically significant correlations between the metrics and the total RBP habitat score (GLIMPSS 0.38; WVSCI 0.43), but not the embeddedness scores (GLIMPSS 0.23; WVSCI 0.22) or the sediment deposition scores (GLIMPSS 0; 20; WVSCI 0.28). Id. The correlation coefficient for conductivity was almost twofold the other Values (GLIMPSS — 0.91; WVSCI -0.80). Id. These findings support the conclusion that “[w]ater quality structured benthic communities more than habitat quality.” Pis.’ Ex. 173 at JE0198.

Though relatively insignificant, Pond 2008 did find some positive correlation between habitat quality and aquatic life; however, subsequent studies have made further efforts to parse the difference between habitat influenced effects and the effects of water chemistry. See e.g. Pis.’ Ex. 173 at PE Í537 (“This suggests that degradation of water quality and the resultant increases in specific conductivity, component ions, and trace metals limit aquatic life regardless of habitat quality.”); Virginia Tech, Pis.’ Ex. 173 at PE 1703 (“Nonetheless, the extensive effort undertaken to locate test. sites with abiotic conditions comparable to those of reference sites was successful in minimizing biotic influence from non-TDS [total dissolved solids] stressors, including poor habitat quality. This was an important-step toward defining TDS sensitivity ... ”); Pond 2014 (“Habitat can be a limiting factor, but by design, we removed significant habitat degradation factors by selecting sample reaches with relatively good habitat and intact riparian vegetation at reference and VF sites ...”); 'id. ■ (“Overall, biological variation was strongly correlated with water chemistry and less by reach-scale habitat and landscape conditions. Since ion concentrations explained the greatest amount of biological impacts and were the most altered (compared to reference), this suggests tjiat recovery is potentially hindered by ions, even in forested reaches long after reclamation.”); Hitt et al, Joint Ex. 8, Pis.’ Ex. 173 at JE0129 (regarding impacts to fish assemblages, the authors noted that “[o]bserved effects of [mountaintop removal mining] could not be explained by changes in physical habitat conditions”).

The following year, Pond published a second article, “Patterns of Ephemeropt-era taxa loss in Appalachian headwater streams,” in Hydrobiologia. Pls.’ Ex. 131. As was the case with his earlier work, Pond again relied on experimental fieldwork, but he conducted unique fieldwork in a different area. Id. In this second article, Pond' compared mayfly assemblages at some ninety-two sites' in Kentucky, focusing on taxa richness (i.e., “the number of different groups of mayflies,” Tr. 2 at 105, ECF No. 100) and relative abundance. In so doing, Pond discovered that both mayfly richness and relative abundance were significantly higher at reference sites and both were significantly lower at mined sites. Tr. 2 at 105-06, ECF No. 100. Furthermore, consistent with earlier analyses, Pond reported that “[Relative mayfly abundance was most strongly correlated to specific conductance (r = 0.72) compared to total habitat score (r = 0.59).” Pis.’ Ex. 173 at PE1526; id. at PE1536 (“Analyses from WV mining areas (Hartman et al., 2005; Merricks et al., 2007; Pond et al. 2008) indicated that the decline in mayflies from mountaintop mining correlates most strongly to specific conductance.”).

In the same year thát Pond 2010 was published, Science Policy Forum published an article co-authored by Dr. Palmer and titled “Mountaintop Mining Consequences.” Pis.’ Ex. 133. Here, rather than experimental fieldwork, the authors relied on data from a variety of sources, including data from the WVDEP database. Tr. 2 at 106, ECF No. 100. Through that data, the authors again examined the relationship between water chemistry and mining activities. Id. at 106. In so doing, the authors again observed that mining contributed to poor water chemistry, particularly marked by elevated conductivity levels, and that significant declines in ma-croinvertebrate taxa resulted. Id.

The following year, the Journal of the North American Benthological Society published a paper by Eric Merriam et al., titled “Additive effects of mining and residential development on stream conditions in a central Appalachian watershed.” Joint Ex. 11. As explained by Dr. Palmer, Merriam et al., examined the combined effects of streams impacted by mining as compared to streams otherwise impacted by development, finding that mining impacts do contribute to changes in macroin-vertebrate community structure. Tr. 2 at 108-09, ECF No. 100. These changes in community structure would appear to have been more closely related to changes in water chemistry as compared to changes in habitat. Pis.’ Ex. 173 at JE0173-74 (“We found significant effects of mining on in-stream conditions. Increased levels of mining resulting in poorer water quality, primarily through increases in specific conductance and associated dissolved chemical constituents_Mining had no measurable effect on habitat complexity or quality.”). Thus, relying on unique data and methodology the authors were able to conclude as follows:

Our results are similar to those of recent studies that have identified changes in water quality to be the dominant stres-sor in mined systems (Fulk et al. 2003, Freund and Petty 2007, Pond et al. 2008, Petty et al. 2010, Pond 2010). Increased specific conductance is consistently the dominant stressor in streams affected by mountaintop removal mining in southern West Virginia (Hartman et al. 2005, Merricks et al. 2007, Pond et al. 2008).... Furthermore, increased specific conductance is a consistently -imporr tant predictor of ecological condition in these systems .,. Our results corroborate those of numerous studies in which Ephemeroptera was identified as one of the most sensitive taxa to increases in ionic strength associated with large-scale surface mining in the Central Appalachian region.

Pis.’ Ex. 173 at JE0174.

This brings us to publication of EPA’s Benchmark. As evident by the foregoing discussion, by the time the EPA published the Benchmark, scientific literature on the subject was already well developed, and according to Dr. Palmer, had already established a likely relationship between conductivity and impairment in Appalachian streams impacted by alkaline mine drainage. Tr. 2 at 110, ECF No. 100. Nevertheless, studies examining the probable relationship between mining, high conductivity, and impairment continued to reach publication in peer-reviewed scientific journals.

In 2011, Dr. Bernhardt and Palmer published an article titled “The environmental costs of mountaintop mining valley fill operations for aquatic ecosystems of the Central Appalachians” in the Annals of the New York Academy of Sciences. Joint Ex. 1; Tr. 2 at 117, ECF No. 100. There, the authors concluded that there was a significant relationship between mining, activities and changes in the chemical composition of streams below mining.. Tr. 2 at 117,. ECF No,. 100. Such changes were strongly associated with biological impairment of those streams. Id. at 117. Particularly, the authors explained that “[a]ll available data show that it becomes increasingly unlikely to find an unimpaired aquatic benthic community as conductivity increases.” Pis.’ Ex. 173 at JE0010. Elaborating on the same point, the article goes on to say that:

Whether or not individual component ions within mining-derived runoff reach streamwater concentrations that are individually lethal or toxic to aquatic life, the cumulative effect of elevated concentrations of multiple contaminants is clearly associated with a substantial reduction in water quality and biological integrity in streams and rivers below mine sites. All research to date indicates that conductivity is a robust measure of the cumulative or additive impacts of the elevated concentrations of multiple chemical stressors from mine sites that lead to biological impairment of streams.

Pis.’ Ex. 173 at JE0014 (emphasis added).

Dr. Lindberg next joined Dr. Bernhardt on a paper published in the Proceedings of the National Academy of Sciences titled “Cumulative impacts of mountaintop mining on an Appalachian watershed.” Pis.’ Ex. 136. In this study, the authors “document the cumulative impact of more than 100 mining discharge outlets and approximately 28 km2 of active and reclaimed surface coal miñes on the Upper Mud River of West Virginia.” Pis.’ Ex. 173 at PE1759. In so doing, they observed that “[a]ll tributaries draining mountaintop-mining-impacted catchments were characterized by high conductivity and increased sulfate concentration.” Id. More broadly, the unique approach taken in this paper established the cumulative impacts of mining in a watershed, with conductivity, sulfates, and selenium all significantly increasing with increased mining. Tr. 2 at 118-19, ECF No. 100; Pis.’ Ex. 173 at PE1763 (“Our synoptic survey approach conclusively demonstrates that the observed increases in conductivity and [selenium] concentration can be attributed directly to the areal extent of surface coal mining occurring in the watershed.”).

Still in 2011, another study authored by Dr. Pond reached publication in Hydro-biologia: “Biodiversity loss in Appalachian headwater streams (Kentucky, USA): Ple-coptera and Trichoptera communities.” Pis.’ Ex. 137. Here, Dr. Pond again documented the effects of mining and residential land use disturbances on macroinver-tebrates, particularly stonefly (Plecoptera) and caddisfly (Trichoptera) assemblages. Tr. 2 at 119, ECF No. 100. Dr. Pond found not only extirpation of these genera associated with mining disturbances, but further remarked that habitat factors could not explain the observed impacts. Tr. 2 at 119-20, ECF No. 100; Pis.’ Ex. 173 at PE1775-76 (“no habitat factors were significantly correlated with relative abundancé metrics”).

In 2013, Pond published yet another coauthored paper on the subject; this time in Environmental Monitoring and Assessment and titled “Calibration and validation of a regionally and seasonally stratified macroinvertebrate index for West Virginia wadeable streams.” Pis.’ Ex. 138. The article describes “the development, validation, and application of a geographically- and seasonally partitioned genus-level index of most probable stream status (GLIMPSS) for West Virginia wadeable streams.” Pis.’ Ex. 173 at PE1786. Importantly, the genus-level index developed therein proved to. be a more reliable predictor of stream quality than its family-level counterpart, WVSCI. Tr. 2 at 121, ECF No. 100 (“pointing out that a genus-level index is much more appropriate to use because the family-level index is not adequately sensitive [] because it lumps genera that have very different tolerance levels.”). These findings would suggest