Petitioner, Kent County, Delaware Levy Court, challenges a decision by the EPA to add the Houston landfill, located in Kent County, Delaware, to the National Priorities List. Kent County contends that the EPA improperly calculated the landfill's "waste characteristics" score by using only an unfiltered sample in testing the groundwater at the site. Petitioner also argues that the EPA failed to consider undisputed data concerning the size of the population potentially at risk in determining the site's "distance to nearest well/population served" score. We vacate the EPA's listing decision and remand for further consideration.

The EPA based the landfill's waste characteristics score on a single unfiltered groundwater sample despite several EPA documents suggesting that both filtered and unfiltered tests are needed to evaluate the metals content of a groundwater sample accurately. The agency does not plausibly contend that performing both tests would be too burdensome, economically or otherwise. And since we find nothing in the record that could justify failing to use filtered samples as well as unfiltered samples in this case, we conclude that the EPA acted arbitrarily in failing to conduct both tests. Furthermore, although we reject Kent County's arguments concerning the distance to nearest well/population served score, we suggest that the EPA reconsider the site's score on remand by excluding irrigation wells that do not draw from the aquifer of concern.

Section 105 of the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA) requires the establishment of a National Priorities List (NPL) of known or threatened releases of hazardous substances throughout the United States. See 42 U.S.C. § 9605(a)(8). Through an informal notice and comment rulemaking process, the EPA places hazardous waste sites on the NPL according to a score generated by the Hazard Ranking System (HRS). A hazardous waste site's HRS score reflects "the relative potential of uncontrolled hazardous substance facilities to cause health or safety problems, or ecological or environmental damage." Appendix A to Part 300--Uncontrolled Hazardous Waste Site Ranking System; A Users Manual, 40 C.F.R. Pt. 300, App. A § 1.0 (amended at 40 C.F.R. Pt. 300, App. A (1991)) (Hereinafter HRS Manual ) (unless otherwise noted, all references to the HRS Manual are to the prior version). Sites with an HRS score of 28.5 or higher are placed on the NPL. See Tex Tin Corp. v. EPA, 935 F.2d 1321, 1322 (D.C.Cir.1991) (For previous explanations of the Hazard Ranking System, see Linemaster Switch Corp. v. EPA, 938 F.2d 1299 (D.C.Cir.1991); City of Stoughton v. EPA, 858 F.2d 747 (D.C.Cir.1988); Eagle-Picher Indus. v. EPA, 759 F.2d 905 (D.C.Cir.1985) (Eagle-Picher I )).

Kent County operated its Houston Landfill between 1969 and 1980. In 1988, the EPA proposed to place the landfill on the NPL after data from tests conducted in 1984 indicated the presence of contaminants including arsenic, chromium, manganese, as well as certain organic compounds, in a monitoring well at the site. On the basis of this data, the EPA's Region III calculated the landfill's HRS score to be 38.11, well in excess of the 28.5 necessary to place the site on the NPL. The Houston Landfill was placed on the list in August of 1990.

Kent County now challenges the EPA's decision, claiming that the EPA improperly scored the Houston site. Specifically, Kent County challenges the EPA's calculation of the site's "waste characteristics" and "distance to nearest well/population served" scores. Kent County contends that the EPA improperly based the site's waste characteristics score solely on the unfiltered groundwater sample taken in 1984. The County claims that because the sample was not filtered, it may have contained soil particles with naturally occurring metals and thereby overstated the danger to groundwater posed by the site. Petitioner also argues that the EPA arbitrarily and capriciously calculated the site's "distance to nearest well/population served" score by considering wells located beyond a discontinuity in the aquifer of concern and by counting the population served by irrigation wells that did not draw from the aquifer of concern.

The EPA's decision to place a hazardous waste site on the NPL is the product of informal notice and comment rulemaking, reviewable under the arbitrary and capricious standard. Eagle-Picher Indus., Inc. v. EPA, 822 F.2d 132, 137 n. 7 (D.C.Cir.1987) (Eagle-Picher III ). We will uphold the EPA's decision if it is "consistent with the Act and the regulations promulgated thereunder, and is not arbitrary." City of Stoughton, 858 F.2d at 749 (internal quotation omitted).

As the agency consistently reminds us, listing on the NPL does not require any action by any party, and does not determine any party's liability for the cost of cleanup at the site. See Eagle-Picher I, 759 F.2d at 920 (quoting the preamble to the NPL, 48 Fed.Reg. 40,658, 40,659 (1983)). It is intended to be a "rough list" of prioritized hazardous waste sites; a "first step in a process--nothing more, nothing less." Eagle-Picher Indus. v. EPA, 759 F.2d 922, 932 (D.C.Cir.1985) (Eagle-Picher II ). Therefore, we have recognized the EPA's interest in reconciling "the need for certainty before action with the need for inexpensive, expeditious procedures to identify potentially hazardous sites...." Eagle-Picher I, 759 F.2d at 921.

But the agency must remain aware that placement on the NPL has serious consequences for a site's owner. See B & B Tritech, Inc. F/K/A B & B Chemical Co, Inc. v. EPA, 957 F.2d 882, 885 (D.C.Cir.1992) (placement on the NPL has "considerable costs"); SCA Serv. of Indiana v. Thomas, 634 F. Supp. 1355, 1361-66 (N.D.Ind.1986) (recognizing the potential for damage to business reputation and loss of value in property, as well as other harmful consequences, when site is listed on NPL). While we do not require the EPA's decisions to be perfect, or even the best, see City of Stoughton, 858 F.2d at 756, we do require that they not be arbitrary or capricious.

B. The Houston Landfill's Waste Characteristics Score

A facility's "waste characteristics" score reflects the toxicity, persistence, and quantity of the most hazardous substance present at a facility that could migrate to groundwater. See HRS Manual, 40 C.F.R. Pt. 300, App. A § 3.4 On the strength of a groundwater sample taken from the Houston landfill, the EPA determined that the most hazardous substance present at the site was arsenic.

In its comments, Kent County raised concerns with the methodology used by the EPA in its groundwater tests, stating that

[i]n view of the acknowledged quality control problems encountered by EPA's contract laboratory as outlined in the [groundwater expert's report], as well as the uncertainties concerning sampling protocols that were utilized in the field, it is reasonable to assume that the single round of PMW-3 data relied upon in the D.R. constitutes an aberration and should be discarded.

Kent County Comments, at 5. The County also submitted the report prepared by its groundwater expert, Geraghty & Miller, which explained:

[I]t is presently unclear whether aqueous sample preparation in the field included filtering and fixing the samples prior to shipment for analysis. This represents an important consideration for metals analyses, since far different results for non-filtered, non-preserved samples can occur. Raw samples with high suspended solids concentrations when analyzed can result in higher reported values for metals. Characterization of ground water without geologic material matrix interference requires samples to be filtered and fixed to provide a truer representation of ground-water quality conditions.

Geraghty & Miller Report, at 10. Furthermore, in supplemental comments, Kent County submitted geological data indicating that there was naturally occurring arsenic in the soil surrounding the Houston site that may have affected the agency's groundwater tests. Letter from Wm. Roger Truitt, Attorney for Kent County, to Penelope Hansen, Chief, Hazardous Assessment Branch, EPA 2-3 (July 23, 1990). The supplemental comments also referred the EPA to the agency's own 1982 "Statement of Work" apparently instructing field testers to filter samples before performing metals analysis. Id. at 2.

The State of Delaware, in its comments, agreed with Kent County's conclusions on the use of an unfiltered groundwater sample. The State explained that "[f]iltration through a .45 micron pore size filter immediately after sample capture and prior to acid fixation is the method adopted by the U.S. Geological Survey and [the state] for processing ground water samples for metals analysis." State of Delaware Comments, at 2. Delaware also pointed out that none of the groundwater tests it performed using filtered samples detected the levels of arsenic that the EPA's single test using an unfiltered sample revealed. Id. at 2-3.

Despite the comments from Kent County and Delaware questioning the accuracy of the agency's groundwater tests, the EPA chose not to retest the site using filtered samples. Instead, in its response comments, the EPA stated that "filtering of samples is not a regulatory requirement." Response Comments, at 22-23. In support of this conclusion, the agency pointed to a memorandum (the so-called "Memorandum 15"), issued by the EPA's Office of Solid Waste and Emergency Response, which addresses methods for testing groundwater under the Resource Conservation and Recovery Act (RCRA). The memorandum states that use of the dissolved metals analysis (the method used for testing filtered samples) "does not account for those metals that are absorbed to the soil matrix and which may move back and forth in equilibrium with the ground water." Memorandum 15, at 1. The EPA also reasoned that

[w]hether or not the groundwater samples were filtered, the data may be adequate to show an observed release, e.g., if both a background well and a monitoring well were sampled without filtering, but levels of contaminants were significantly higher in the monitoring well, then it may indicate that an observed release has occurred.

Response Comments, at 23.

Shortly after the EPA added the Houston site to the NPL, Kent County discovered several documents, prepared by the EPA's own experts, that addressed the filtration issue. Essentially, the documents indicate that, under CERCLA, the proper method for analyzing the metals content of a site's groundwater is to use both filtered and unfiltered samples. Kent County now moves to supplement the administrative record with these documents, arguing that they illustrate the arbitrary nature of the EPA's decision to rely on a single unfiltered groundwater sample in scoring the site. The EPA insists, however, that the documents discovered after the Houston site was added to the NPL were not part of the administrative record and may not be considered by this court on petition for review.

On November 12, 1991, this court granted Kent County's motion to enlarge the record to include the internal EPA documents. We have previously held, however, that such a summary order should be read only as allowing petitioner to supplement the judicial record and not as resolving the question regarding the scope of the administrative record. See Linemaster Switch, 938 F.2d at 1305. Therefore, we must now decide whether Kent County may supplement the administrative record.

The EPA argues that our decision in Linemaster Switch precludes Kent County's request to supplement the administrative record. In Linemaster Switch, we refused to allow a petitioner to supplement the administrative record with material not before the agency when it made its decision where the petitioner failed to timely submit the material to the proper division of the EPA. Linemaster Switch, 938 F.2d at 1305. We held that allowing such a result would "effectively require EPA to comb all regional files for potentially relevant data before listing a site on the NPL," and would be inconsistent with prior decisions emphasizing the necessarily abbreviated nature of the NPL listing process. Id.

But, this is not a case where petitioner failed to submit relevant documents to the proper division of the EPA. In fact, Kent County never knew the documents existed until after the EPA issued its decision.

In San Luis Obispo Mothers for Peace v. NRC, 751 F.2d 1287, 1327 (D.C.Cir.1984), vacated in another part, 760 F.2d 1320 (D.C.Cir.1985) (en banc), and aff'd, 789 F.2d 26 (D.C.Cir.) (en banc), cert. denied, 479 U.S. 923, 107 S. Ct. 330, 93 L. Ed. 2d 302 (1986), this court recognized that supplementing the administrative record might be proper "if petitioners made a prima facie showing that the agency excluded from the record evidence adverse to its position...." While we do not suggest that the agency purposefully excluded the documents, it appears that the EPA was at least negligent in failing to discover them when it searched for documents to support its position that only unfiltered samples were required. We are not at all convinced by the EPA's argument that, because it cannot be expected to find all documents that address a particular issue before making its decision, its failure here to find any of the CERCLA documents discussing filtration is necessarily excusable.

We do not think that it was the EPA's responsibility to find all documents discussing filtration located in any office of the EPA. Here, however, because the EPA itself looked outside its national CERCLA files and relied on a single memorandum from another program (RCRA), we think it was arbitrary and capricious for the agency not to examine the Region III CERCLA files for documents discussing filtration before relying on the single RCRA memorandum. This is true particularly in a case like this one where the EPA's policy regarding filtration varies region to region. Had the EPA simply checked the files at the Region III office, it would have found the documents that discuss what appears to be a well-aired debate between using filtered and unfiltered samples. The documents relate to the position of the agency's own experts on the question central to this case. To deny their relevance would be inconsistent with rational decisionmaking by an administrative agency.

Accordingly, in reviewing the EPA's decision to place the Houston site on the NPL, we will consider the following EPA documents submitted by Kent County: (1) Memorandum from Patricia J. Krantz, Chief, QA Section, Subject: Filtered vs. Unfiltered Groundwater (August 24, 1987) (hereinafter Filtered vs. Unfiltered Groundwater ); (2) EPA, Superfund Ground Water Issue, "Ground Water Sampling for Metals Analyses," (March 1989) (hereinafter Groundwater Sampling for Metals Analyses ); and (3) EPA Region III QA Directives, "Field Filtration Policy for Monitoring Well Groundwater Samples Requiring Metals Analyses," (April 23, 1990) (hereinafter Region III Directives ). We will not, however, consider the "Record of Decision for the M & T DeLisa landfill," because the EPA's final decision in that matter was not issued until September 20, 1990, after the EPA issued its decision in Kent County's case.

2. Filtered vs. Unfiltered Groundwater Tests

We turn now to Kent County's argument that the EPA acted arbitrarily and capriciously by using only the unfiltered groundwater sample in reaching the site's waste characteristics score. The EPA contends that there is "no regulatory requirement" that it use both filtered and unfiltered samples in evaluating a site's groundwater. Relying on Memorandum 15, the agency claims that it recommends only the "Total Recoverable Metals Method", which uses unfiltered samples, as the "standard technique" for performing metals analysis. The EPA explains that using a filtered sample would exclude from test results significant organic compounds like those found at the Houston site, as well as metals that may have been released from the site but which are temporarily attached to soil particles floating in the groundwater. See Memorandum 15, at 1-2 ("A dissolved metals analysis does not account for those metals that are absorbed to the soil matrix and which may move back and forth in equilibrium with the groundwater...."); see also Response Comments, at 23. This explanation, however, fails to tell us why the EPA cannot use both the Total Recoverable Metals Method, which uses only unfiltered samples, and the dissolved metals analysis, which uses filtered samples.

As far back as 1982, before the samples at the Houston site were taken, an EPA "Statement of Work" suggested that groundwater samples be filtered in the field before testing. Later agency documents refer to the need for both filtered and unfiltered samples. As the Region III Directives explain:

Monitoring wells sometimes produce turbid water (water containing suspended solids). The turbidity can be due to disruption of the adjacent geologic formations during well purging or poor design and initial development of the well. When particles containing metal species are suspended into the groundwater and are not removed, they dissolve when the sample is preserved to a pH [of less than] 2. High levels of aluminum, manganese, and iron in unfiltered samples often indicate the presence of these particles. Without filtration, concentrations of this mobile metal contamination in the groundwater are often over estimated.

Region III Directives, at 1. Other EPA documents issued prior to the agency's decision also recommend performing both filtered and unfiltered tests. The memorandum entitled "Filtered vs. Unfiltered Groundwater" states that "groundwater samples for metals should be split into two portions: filtered and unfiltered. Any difference in concentration between the total and dissolved fractions may be attributed to the original metallic ion content of the particles...." Filtered vs. Unfiltered Groundwater, at 2. And a report issued by a group of EPA scientists in 1989 also concludes that "[a] dual sampling approach [is] recommended with collection of both filtered and unfiltered samples." Ground Water Sampling for Metals Analyses, at 1.

The only document that the EPA has to support its use of a single unfiltered sample is Memorandum 15 which recommends the unfiltered testing method as the "standard technique" for metals analysis. Memorandum 15, at 1. But even Memorandum 15 states that using both a filtered and unfiltered test would be "worthwhile" in some circumstances. Id. at 1-2. In light of the other EPA documents that recommend both filtered and unfiltered tests, this single memorandum that is somewhat equivocal on the issue is insufficient to justify the EPA's decision to rely on the unfiltered sample in scoring the Houston site.

At oral argument, counsel for the government could not persuasively claim that performing both filtered and unfiltered tests would be infeasible, economically or otherwise. When pressed, counsel acknowledged that the EPA has never asserted that performing both tests would be too burdensome; any attempt to so argue now would be a post hoc rationalization, which we have long held cannot save an arbitrary agency decision. See Algonquin Gas Transmission Co. v. FERC, 948 F.2d 1305, 1316 (D.C.Cir.1991) (citing SEC v. Chenery Corp., 332 U.S. 194, 67 S. Ct. 1575, 91 L. Ed. 1995 (1947)). Because the EPA's own experts recommended both filtered and unfiltered tests, and because the EPA failed to offer any reason why it would have been infeasible to do both, we conclude that the EPA's decision to list the Kent County site on the basis of the single unfiltered sample was arbitrary and capricious.

We do not suggest that the EPA must use both tests in all future listing decisions. There may well be instances where the agency could justify using only unfiltered tests; for example, at a particular site where wells have been inspected and properly purged to ensure that a groundwater sample is not tainted with soil particles. In fact, the Region III Directives list a number of "exceptions" where using both filtered and unfiltered tests would not be required. See Region III Directives, at 1. However, the EPA has referred to nothing in the supplemented record, and we have found nothing, that would justify using only a single unfiltered sample in this case. In fact, the record indicates that there is naturally occurring arsenic in the soil surrounding the Houston landfill, and Delaware found lower levels of arsenic in the groundwater when it tested the site using filtered samples.

The agency insists that its decision to list the site was proper whether or not the groundwater sample was filtered because the arsenic level in the sample taken from the monitoring well was significantly higher than the arsenic level in a sample taken from a background well. See Response Comments, at 23. However, as Kent County points out, this explanation ignores the possibility that the monitoring well contained turbid water (i.e. water with soil particles containing naturally occurring arsenic) while the background well did not. We find nothing in the EPA's response comments that refutes this possibility. Consequently, we reject the agency's argument that the difference in arsenic levels between the background well and the monitoring well adequately supports its listing decision.

C. The Site's Distance to Nearest Well/Population Served Score

Kent County also challenges the EPA's calculation of the landfill's distance to nearest well/population served score. Kent County argues that the EPA improperly considered wells located beyond a discontinuity in the aquifer of concern in determining the site's distance to nearest well score, and erred in failing to exclude irrigation wells that did not draw from the aquifer of concern in calculating the site's population served score.

The distance to the nearest well is measured from the hazardous substance to the nearest well that draws from the aquifer of concern. If a discontinuity exists between the hazardous substance and a number of wells, however, those wells, and the population served by them, are removed from the target population figure. HRS Manual, 40 C.F.R. Pt. 300, App. A § 3.5. Kent County's comments suggested that "Brown's Branch," an above ground stream located near the site, served as a discontinuity in the Pleistocene Aquifer, the aquifer of concern. Kent County Comments, at 7.

The EPA rejected Kent County's assertion that Brown's Branch represented a discontinuity in the aquifer, pointing out that the stream did not completely transect the aquifer of concern. Response Comments, at 31. Referring to a figure in the HRS Manual, the EPA explained that, to reduce the distance to nearest well factor, "a discontinuity must completely transect the aquifer of concern throughout the 3-mile radius and thereby totally isolate a portion of the aquifer from the site." Id. Because Kent County acknowledges that it has not demonstrated that Brown's Branch completely transects the aquifer of concern, the only question is whether the EPA erred in reading the HRS as requiring a complete discontinuity.

We agree that the HRS manual is not completely clear on the definition of a "discontinuity." A discontinuity could mean only a physical transection or it could include certain hydrogeologic mechanisms as well. Since the meaning of the term "discontinuity" is ambiguous, we will defer to the EPA's reasoned interpretation of it. Eagle-Picher III, 822 F.2d at 138 n. 23; see also United States v. Larionoff, 431 U.S. 864, 872, 97 S. Ct. 2150, 2155, 53 L. Ed. 2d 48 (1977) (agency's interpretation of its own regulations given controlling weight unless plainly erroneous or inconsistent with the regulation). Nothing in Kent County's arguments persuades us that the EPA's reading of the HRS was erroneous.

In its comments, the County noted that "a number of residential and most irrigation wells [within three miles of the landfill] do not draw from the Pleistocene Aquifer because they tap one of the two [deeper] aquifer units." Kent County Comments, at 8. Kent County then suggested that the EPA use the report prepared by the County's groundwater expert which recalculated the population served score by excluding the residential wells that drew from the deeper aquifers. Id. The report did not, however, exclude irrigation wells that drew from deeper aquifers.

After reviewing Kent County's comments, the EPA adopted the conclusions reached by the County's groundwater expert and recalculated the site's population served score by excluding the deeper residential wells. Kent County now argues that the EPA arbitrarily and capriciously failed to exclude irrigation wells as well. But the County never addressed the issue of irrigation wells in its comments, and its expert's groundwater report never recalculated the population served score excluding the irrigation wells. We have continually stressed that parties opposing NPL listing must present their claims clearly and specifically to the agency before raising them in a petition for review. Tex Tin, 935 F.2d at 1323; see also Northside Sanitary Landfill, Inc. v. Thomas, 849 F.2d 1516, 1519 (D.C.Cir.1988) (party submitting data must alert the agency to its position and contentions). Because Kent County failed to do that here, we reject its argument that the EPA acted arbitrarily and capriciously by failing to exclude the irrigation wells when recalculating the population served score. We note, however, that since the EPA must reconsider the site's waste characteristics score on remand, the agency might want to recalculate the population served score as well.

The EPA's decision to rely on a single unfiltered groundwater sample in scoring the Houston landfill was arbitrary and capricious. The agency's own experts have recommended using both filtered and unfiltered groundwater tests and the EPA has not persuasively claimed that it would be too burdensome to do so. We vacate the EPA's decision to place the Houston site on the NPL and remand for further consideration. Since the EPA must reevaluate the landfill's HRS score on remand, we also suggest that the agency recalculate the site's population served score by excluding the deeper irrigation wells.

So ordered.