Opinion for the Court filed by Circuit Judge WALD.
WALD, Circuit Judge:In these consolidated cases, petitioners seek review of several aspects of the Standards for the Use or Disposal of Sewage Sludge, 58 Fed.Reg. 9387 (1993) (to be codified at 40 C.F.R. parts 257 and 403) (“Regulations”), issued on February 10, 1992 by the Environmental Protection Agency (“EPA” or “agency”). Because petitioners have raised valid challenges to (1) the use of the 99th percentile figures from the National Sewage Sludge Survey (“NSSS”) for the Table 3 “clean sludge” caps, (2) the assumed rate and duration of application underlying the risk-based data in Table 3 as applied to heat-dried sludge, (3) the assumed exposure possibilities underlying the risk-based cap on selenium as applied to public contact sites with low potential for occupancy, and (4) the lack of data to support the risk-based cap on chromium, we remand those parts of the regulations to the EPA for modification or additional justification. We reject the challenges to the classification of “dedicated uses” as “land disposal” and to the EPA’s refusal to provide for site-specific variances from the pollutant limitations for land-applied sewage sludge.
I. BACKGROUND
A. Statutory Framework
The Clean Water Act of 1972 (“CWA” or “Act”) was enacted to “restore and maintain the chemical, physical, and biological integrity of the Nation’s waters.” 33 U.S.C. § 1251(a). The Act prohibits “the discharge of any pollutant by any person” into the navigable waters of the United States, except in compliance with various provisions of the Act, 33 U.S.C. § 1311(a), and directs the EPA to regulate the discharge of wastewater into the navigable waters by various industrial, commercial, and public sources. See 33 U.S.C. § 1311(b). As amended by the Federal Water Pollution Control Act of 1977, Pub.L. No. 95-217, 91 Stat. 1566 (codified at 33 U.S.C. § 1251 et seq.), and the Water Quality Act of 1987, Pub.L. No. 100-4, 101 Stat. 7 (1987), the CWA also requires the EPA to promulgate comprehensive regulations for the management of sewage sludge— the by-product of pre-discharge sewage and wastewater treatment by publicly and privately owned treatment works (“POTWs”).
POTWs receive sewage and liquid industrial wastes. POTW treatment of these waste streams produces a liquid effluent that meets CWA discharge standards and may be expelled into surface water and a residual material, sewage sludge, which may not be discharged into the waters. POTWs dispose of sewage sludge through incineration or landfill deposits; they also apply it to land or sell it to the public for use as a fertilizer. Implementation of the Clean Water Act of 1972’s restrictions on effluent discharge has led to more pre-discharge treatment of sewage wastes and, consequently, more sewage sludge is generated as a by-product of treatment. The production of sewage sludge each year has nearly doubled since the original enactment of the Clean Water Act. See 58 Fed.Reg. 9249.
The Federal Water Pollution Control Act of 1977, an amendment to the Clean Water Act, directed the EPA in general terms to develop a regulatory program to ensure the safe use and disposal of sewage sludge. See 33 U.S.C. § 1345(d) (1982). In 1987, Congress enacted another amendment to the CWA, the Water Quality Act, to require the EPA to issue specific regulations for the use and disposal of sewage sludge. Under the amended Act, the EPA must identify and set numeric limits for toxic pollutants that “may be present in sewage sludge in concentrations which may adversely affect public *395health or the environment,” and establish management practices for the use and disposal of sludge containing these toxic pollutants. 33 U.S.C. § 1345(d)(2). Its regulations are to be issued in two phases — the first round to be promulgated “on the basis of available information,” the second to encompass pollutants unaddressed by the first round. Id. It is the Round One regulations that are now at issue.
B. Regulatory Development
At the start of the rulemaking process, the EPA made an initial assessment in the aggregate that “current use and disposal practices for sewage sludge pose little risk to public health.” 58 Fed.Reg. 9320. Sewage sludge that meets safety requirements is a “valuable resource” as “fertilizer and a soil conditioner,” 58 Fed.Reg. 9249, and the EPA “strongly support[s] the beneficial reuse of sewage sludge.” 58 Fed.Reg. 9251. The EPA identifies “land application” as one type of beneficial reuse and defines it as “the spraying or spreading of sewage sludge onto the land surface; the injection of sewage sludge below the land surface; or the incorporation of sewage sludge into the soil so that the sewage sludge can either condition the soil or fertilize crops or vegetation grown in the soil.” § 503.11(h), 58 Fed.Reg. 9391. The Round One regulations — Standards for the Use or Disposal of Sewage Sludge— regulate land application of sewage sludge as well as surface disposal and incineration.
The Round One regulations establish limits on ten pollutants in sludge destined for land application. To set these land application pollutant limits, the EPA sought first to identify “those [pollutants] most likely to pose a hazard to human health or the environment.” 1 It enlisted federal, state, academic, and private sector experts to screen a list of 200 pollutants to determine which, if any, posed a potential risk to human health or the environment if contained in sewage sludge that was applied to or disposed of on land or incinerated. These experts selected forty-eight pollutants, for which the EPA compiled environmental profiles. Based on data and information from published scientific reports, the profiles assessed the pollutants’ general toxicity and persistence, as well as the particular pathways by which they might cause harm to human health or the environment. See 58 Fed.Reg. 9263-64 (Table III-l). Using these profiles and preliminary data about the concentration and frequency of these pollutants in sewage sludge, the EPA exempted from regulation those pollutants that presented no risk to human health or the environment at the highest observed concentration and deferred consideration of those for which it had insufficient data to make this risk determination. See 58 Fed.Reg. 9264. It initially proposed limits for 25 pollutants in sludge to be applied to land, see 54 Fed.Reg. 5761 (Table III-4), and concluded by regulating ten heavy metals in sludge applied to land in the final Round One sewage sludge regulations, see 58 Fed.Reg. 9392. These portions of the regulations establish numeric limits on pollutants in sludge that is applied to agricultural land, forests, public contact sites, or reclamation sites.
II. The LAND Application Regulations
In establishing the limits for the ten regulated heavy metals pollutants — arsenic, cadmium, chromium, copper, lead, mercury, molybdenum, nickel, selenium, and zinc — the EPA generated two sets of data.
A. The Underlying Data
The first data set on the ten pollutants describes their current concentration in sewage sludge. The data is culled from the EPA’s National Sewage Sludge Survey (“NSSS”), in which the EPA sent questionnaires to 479 POTWs — out of a national total of 11,407 — and performed sampling and analysis at 208 of the 479. See 58 Fed.Reg. 9269.2 Based on this sampling and analysis, the EPA identified the pollutant concentra*396tions in current sludge output, and calculated 99th-percentile concentration numbers: the pollutant concentration not exceeded by 99% of the sludge samples in the NSSS (“99th percentile caps”).
The second data set on the ten pollutants is risk-based. Under its risk-based analysis, the EPA modelled 14 pathways by which pollutants in land-applied sludge could affect human health or the environment and then identified a hypothetical “highly exposed individual” (“HEI”) for each pathway and calculated a pollutant limit that would protect the HEI. The pathway model analyzes the exposure potential from the total quantity of metal in a given area of soil. The EPA proceeded on the uncontested premise that “metals persist in the soil and accumulate over time,”3 and the pathway model assesses the risk posed by the total accumulation of pollutants in a given hectare of land. The resulting pollutant limit, accordingly, “repre-sentes] the total quantity of metals that could be added to [a given area] of soil. So long as the total quantity ... for the metal is not exceeded, the exposure assessment models predict that there will be no injury to the HEI. The model is unconcerned whether the total quantity of the pollutant is received in a single load or over time.” 58 Fed.Reg. 9282 (emphasis added). The risk-based exposure model, then, is indifferent as to the concentration of a pollutant in any given load of sludge.
The EPA chose also to regulate concentration limits. “[B]y applying certain conservative assumptions” about the amount of sludge that would be applied to a given area of land, the EPA “backcalculated” from the total pollutant limits in a given area of land to a permissible sludge pollutant concentration per load. 58 Fed.Reg. 9317. The “backcal-culation” provides the EPA a means of converting the cumulative pollutant limit into a concentration cap for the pollutant in any given load of sludge. The model assumes a total amount of sludge that will be applied to a given hectare of land based on an assumed yearly application rate and assumed duration of application. The EPA assumed that ten metric tons of sludge would be applied annually to a hectare of land for 100 years. This converts into an assumption that, in total, 1000 metric tons of sludge will be applied to a given hectare of land. Based on (1) this total amount of sludge that the EPA assumed would accumulate on an area of land, and (2) the total amount of pollutant that the EPA had determined could safely accumulate on an area of land, the EPA calculated pollutant/sludge, the permissible concentration of pollutant in any application of sludge. For instance, assuming that 1000 metric tons of sludge would be applied to a hectare of land over its lifetime, and having determined that 41 kilograms of arsenic could safely accumulate in that hectare, the EPA determined that it could allow 41 kgs of arsenic in 1000 metric tons of sludge, or 41 mgs of arsenic per kilogram of sludge. This number — in mg/kg — is the EPA’s risk-based concentration cap, and derives from the “backcalculation” from the EPA’s risk analysis, which is based on the EPA’s application rate and duration assumptions.
B. The Regulatory Design
The EPA designed its final regulations of pollutants in land-applied sewage sludge on the basis of its risk-based and empirical data sets. These regulations use the following four tables in a manner described in the text below:
*397[[Image here]]
1. Ceiling Concentrations
Table 1 — Ceiling Concentrations — contains the less stringent of the two concentration limits, the risk-based concentration cap and the 99th percentile concentration cap, for each of the ten pollutants. See § 503.13(b)(1), 58 Fed.Reg. 9392. No sewage sludge may be applied to the land unless the concentration of each of the pollutants is under the Table 1 limits. Once it complies with the Table 1 limits, it must comply either with the Table 3 limits — qualifying as “clean” sludge — or with the cumulative limits in Tables 2 and 4. Each option is discussed below.4
2. “Clean” or “High Quality” Sludge
Table 3 — Pollutant Concentrations (“clean sludge caps”) — contains the more stringent of the two concentration limits, the risk-*398based concentration cap and the 99th percentile concentration cap, for each of the ten pollutants. See § 503.13(b)(3), 58 Fed.Reg. 9392. If sludge meets Table 3 — i.e., the concentration of each of the pollutants in sludge is under the Table 3 caps, as well as the Table 1 caps — it is considered “clean” sludge, and may be applied to land with no further regulation.
3. Cumulative Pollutant Limits for Bulk Sewage Sludge
If bulk sewage sludge does not meet the Table 3 caps, it must comply with Table 2— the Cumulative Pollutant Loading Rates (“CPLRs”). While Tables 1 and 3 regulate pollutant concentration in sludge, Table 2 regulates pollutant concentration in land. It contains the cumulative risk-based limit derived from the EPA’s pathway exposure model. The CPLRs represent the total amount of pollutant that can ever permissibly concentrate in a given area of land. Compliance "with the cumulative option in Table 2 requires the maintenance of centralized land application records to ensure that the total pollutant limit is not exceeded. Each time non-“elean” sludge is applied to land, the amount of sludge applied and the concentration of the pollutants in that sludge must be recorded, so that the agency can keep track of the pollutants accumulating in that piece of land. See§ 503.12(e)(2), 58 Fed.Reg. 9391 (person who applies sewage sludge in accordance with Table 2 must first contact permitting authority for prior Table 2 application records and ensure that Table 2 limits are not exceeded); § 503.12(j), 58 Fed.Reg. 9391 (must notify permitting authority of application); § 503.17(a)(5)(ii), 58 Fed.Reg. 9394 (recordkeeping requirements for person who applies sewage sludge in accordance with Table 2).5
4. Annual Pollutant Limits for Packaged Sewage Sludge
Packaged sewage sludge that does not meet the Table 3 caps must comply with Table 4 — the Annual Pollutant Loading Rates (“APLRs”). Table 4 also regulates the pollutant concentration in land. It contains the cumulative limit in “annualized form”— imposing a limit on how much pollutant can accumulate in a given area of land each year. Because packaged sludge is generally applied to home gardens, the EPA did not think it would be feasible to maintain centralized records and control of packaged sludge applications.6 Accordingly, it converted the Table 2 cumulative limits into annual limits: the total amount of pollutant that could be allowed to accumulate in one year. Assuming that packaged sewage sludge would “probably not be applied longer than 20 years,”7 the EPA determined that one-twentieth of the cumulative limit for each pollutant could be applied each year. The Table 4 APLRs are thus the Table 2 CPLRs divided by 20. Each year, non-“clean” packaged sewage sludge can only be applied in quantities such that none of the pollutant limits in Table 4 is exceeded. See § 503.13(a)(4)(ii), 58 Fed.Reg. 9392. For packaged sludge, the EPA enforces this application limit through labelling: non-“clean” packaged sewage sludge must be labeled with the maximum yearly application so as to ensure that none of the Table 4 APLRs will be exceeded. See § 503.14(e), 58 Fed.Reg. 9392 (label shall be affixed on bag or other container with: name and address of person who prepared the sludge, statement that application is prohibited expect in accordance with instructions, and the annual application rate that does not cause any of the ceilings in Table 4 to be exceeded). For instance, the Table 4 APLR for arsenic is 2 kgs of arsenic per hectare. See § 503.13(b)(4), 58 Fed.Reg. 9392. If packaged sewage sludge contained 50 mgs of arsenic per kilogram of sludge, then 40. kgs of that sludge could be applied to a hectare of land before reaching the APLR. (If there are 50 mgs of arsenic in each kilogram of sludge, then 40 kilograms of sludge contain 2 kgs of arsenic.) Such sludge must *399be labelled so that no more than 40 kgs may be applied to a hectare per year.
5. Summary
In brief, all sludge must meet the Table 1 caps as a threshold requirement to land application. Then, there is a choice between meeting the Table 3 “clean sludge” caps — in which case there is no further regulatory control of land application — or the cumulative limits of Tables 2 and 4, in which case there are continuing recordkeeping obligations in the case of bulk sewage sludge (Table 2) or labeling requirements in the case of packaged sewage sludge (Table 4).
III. Challenges to the Table 3 “Clean Sludge” Caps
Petitioners challenge various aspects of the Table 3 “clean sludge” caps. As explained above, Table 3 contains the more stringent of the risk-based and 99th percentile concentration caps for each of the ten pollutants. For chromium and selenium, this more stringent cap is the 99th percentile number; for the other regulated pollutants, it is the risk-based cap.8 Sludge that meets both the Table 1 caps and the Table 3 caps is considered “high quality” sludge, and may be applied to the land without further regulation. Sludge that does not meet the Table 3 caps must meet the cumulative limits in Table 2 or Table 4, which involve more complicated regulatory oversight.
All of the petitioners — the Leather Industries of America (“Leather Industries”), the City of Pueblo (“Pueblo”), and, jointly, the Association of Metropolitan Sewerage Agencies and the Milwaukee Metropolitan Sewerage District (collectively “AMSA”) — challenge the EPA’s use of the 99th-percentile caps in Table 3 on the grounds that they are unrelated to risk. The AMSA also challenges the application rate and duration assumptions that underlie the risk-based caps in Table 3 on the grounds that these assumptions cannot rationally be applied to heat-dried sludge, which is applied at lower rates and for shorter durations.
A. Safe Harbor Defense
In defending the Table 3 caps, the EPA suggests that because the Table 3 caps do no more than offer land appliers an additional option, rather than impose a mandatory requirement, they should withstand review. Land appliers need not comply with Table 3 — if sludge does not meet Table 3, it can nonetheless be applied to land under the recordkeeping or labelling schemes of Tables 2 and 4. Table 3 only offers a safe harbor from the more involved regulatory controls of Tables 2 and 4. Although “[cjomplying with the ‘clean sludge’ pollutant concentrations in Table 3 may be advantageous” because it relieves the land applier from the record-keeping and management practice requirements controlling non-“clean” sludge, Table 3 is not a prerequisite to land application. EPA Brief at 24.
Because it is not requiring compliance with Table 3, the EPA suggests, it should have greater leeway in designing Table 3. As the EPA acknowledges, however, the Table 3 safe harbor “provides significant relief from the [otherwise controlling] regulatory safeguards.” EPA Brief at 24. Failure to meet the Table 3 caps subjects the would-be applier of sludge to not insignificant burdens, and undoubtedly makes non-“clean” sludge less attractive to the applier. Because the Table 3 clean sludge safe harbor provides “significant relief’ to complying sludge, the design of that safe harbor is subject to the same rational basis review as the rest of the regulatory scheme.
B. NSSS 99th Percentile Caps
Petitioners argue that the 99th percentile caps are not risk-based and therefore exceed the EPA’s statutory authority under the enabling legislation.9 The statute direct*400ing the EPA to issue the Round One pollutant limits for sewage sludge provides:
[T]he Administrator shall identify those toxic pollutants which, on the basis of available information on their toxicity, persistence, concentration, mobility, or potential for exposure, may be present in sewage sludge in concentrations which may adversely affect public health or the environment, and propose regulations specifying acceptable management practices for sewage sludge containing each such toxic pollutant and establishing numerical pollutant for each use identified under paragraph (1)(A).
33 U.S.C. § 1345(d)(2)(A)(i) (emphasis added). It further instructs that the management practices and numerical criteria so established “shall be adequate to protect public health and the environment from any reasonably anticipated adverse effects of each pollutant.” 33 U.S.C. § 1345(d)(2)(D) (emphasis added).
In determining whether the 99th percentile limits in Table 3 are a permissible interpretation of the statute, we turn to Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984), and its progeny. We must first determine whether Congress’ intent is clear as to the permissibility of the agency’s interpretation. If it is not, “the question for the court is whether the agency’s answer is based on a permissible construction of the statute.” Chevron, 467 U.S. at 843, 104 S.Ct. at 2782.
The EPA does not contest that its statutory authority is limited to promulgating regulations “adequate to protect public health and the environment from any reasonably anticipated adverse effects.” 33 U.S.C. § 1345(d)(2)(D). As a matter of Chevron's first step, then, there is no dispute that the statute clearly mandates regulations based on “reasonably anticipated adverse effects,” and, thus, bearing some relation to risk. The EPA argues, however, that the 99th percentile caps fulfill this mandate of adequate protection from reasonably anticipated adverse effects and bear a relation to risk because they provide “an additional safety mechanism.” EPA Brief at 26. The EPA suggests two ways in which the 99th percentile caps function as a safety mechanism: (1) they provide a “margin-of-safety” “necessary to ensure ‘adequate’ protection from these pollutants,” EPA Brief at 28; and (2) they prevent current sewage sludge practices — found to be safe in the aggregate — from deteriorating. EPA Brief at 27. We conclude, however, that the EPA has failed to show that the 99th percentile caps are risk-related, and thus that they accord with the express mandate of the statute.
First, the EPA states that the 99th percentile caps are based on “a margin-of-safety analysis [that] is consistent with the legislative intent underlying section 405 of the Act.” EPA Brief at 28. Whatever the underlying legislative intent, we do not view the 99th percentile caps as merely a “margin-of-safety” device. The fact that one cap is more restrictive than another does not automatically make it a “margin of safety.” Rather, a margin of safety must be rooted in an analysis of risk. “[T]he Administrator [must] base[ ] his conclusion as to an adequate margin of safety on a reasoned analysis and evidence of risk.” American Petroleum Institute v. Costle, 665 F.2d 1176, 1187 (D.C.Cir.1981) (emphasis added), cert. denied, 455 U.S. 1034, 102 S.Ct. 1737, 72 L.Ed.2d 152 (1982).
The 99th percentile caps are not related to risk. In its initial version of the regulations, the EPA proposed to cap all pollutants at the higher of the pathway-generated numbers and the 98th percentile level. This proposal came under heavy attack from “a specially convened group of sewage sludge experts,” *401the Land Practices Peer Review Committee (“PRC”). 58 Fed.Reg. 9267.10 The PRC, composed of “experts from EPA, academia, environmental groups, and units of state and local government agencies,”11 concluded that the 98th percentile approach was “arbitrary,” would “either over- or under-regulate,” and “ha[d] no technical merit.”12 It pointed to the absence of any relation between the percentile numbers and risk, noting that, because the percentile concentrations are purely descriptive, they “may be insignificant from a risk standpoint” just as easily as they “may pose significant risks.”13 We can discern no reason — and the EPA has provided none — why the 99th percentile numbers are less “arbitrary” or have more “technical merit” than the 98th percentile figures. We find no support for the 99th percentile caps as a risk-based margin of safety for chromium and selenium, the two pollutants capped by the 99th percentile in Table 3.14
Second, the EPA argues that the 99th percentile caps reflect a legitimate “antibaek-sliding” approach. EPA Brief at 27. The EPA notes that its risk assessment suggests that current practices pose little risk to human health, and concludes that it should ensure that there is no deterioration from current practices. See 58 Fed.Reg. 9283. The antibaeksliding rubric, however, cannot evade the requirement that the numeric limits on land application be risk-related — based on reasonably anticipated adverse effects. The conclusion that current sludge composition is safe absent a showing that alternative sludge composition would not be safe does not justify a mandate to freeze current sludge quality. In Natural Resources Defense Council v. EPA, 859 F.2d 156 (D.C.Cir.1988), we upheld an antibaeksliding system that prevented a retreat by industrial facilities from standards regulating pollutant discharge that had been established in individual permits once the individual permitting process was replaced by national standards. In that case, the EPA had been authorized in the first place to issue the standards from which backsliding was prohibited, and those standards were based on the same statutory factors as the subsequent national standards. See 859 F.2d at 201. In sum, when the statute mandates risk-based regulation, standards from which facilities may not retreat must also be risk-based.15
We also note another significant distinction between the 99th percentile caps and the antibaeksliding regime approved in Natural Resources Defense Council v. EPA. The Natural Resources Defense Council system was site-specific and required only that each individual facility maintain its existing standards. See 859 F.2d at 201. The 99th percentile caps, by contrast, are not tailored to *402individual facilities. There will, by statistical necessity, be a certain number of POTWs that currently produce sludge with chromium and selenium concentrations above the 99th percentile, and these facilities must do more than avoid sliding back in order to meet the clean sludge caps. Without a finding of risk, the EPA is without a basis for imposing the antibacksliding mandate.
C. The Application Rate and Duration Assumptions
For all other pollutants except chromium and selenium, the risk-based concentration cap is more stringent than the 99th percentile cap, and is thus the Table 3 “clean” sludge cap. The AMSA challenges the risk-based caps in Table 3. It argues that the assumptions about the rate and duration of sludge application underlying the risk-based concentration caps in Table 3 are irrational with respect to heat-dried sludge, which is applied at lower rates for shorter durations. For whatever reason, the EPA chose not to respond to this particular claim, and the AMSA has been less than totally clear about what parts of the regulations are allegedly infected by the use of these assumptions. We are, accordingly, somewhat handicapped in evaluating the challenge. Nonetheless, on the record, we conclude that the EPA has not adequately justified its use of the assumed rate and duration of application to apply the risk-based caps in Table 3 to heat-dried sludge.
The EPA’s primary risk-based data is in Table 2. As explained above, supra pp. 396-97, the EPA used the assumed application rate of 10 metric tons/hectare and duration of 100 years to convert the Table 2 cumulative risk-based limits into the risk-based concentration caps used in Tables 1 and 3.16 The AMSA challenges the use of these assumptions in Table 3. It argues that these assumptions are irrational as applied to heat-dried sludge because it “is inconceivable that heat dried sludge could ever be applied to land at a rate of 1,000 metric tons/ha.” AMSA Brief at 12. The AMSA points to undisputed evidence in the record that the recommended application for Milorganite — a principal heat-dried sludge product — is no more than 3.5 metric tons/ha per year17; and that heat-dried sludges “are applied at low annual rates, usually around 2 to 3 metric tons per hectare, but rarely over 5 metric tons per hectare.”18
Under the Administrative Procedure Act, the assumed application rate and duration must bear some rational relationship to the actual application rates and durations for land application of sludge. See Edison Elec. Institute v. EPA, 2 F.3d 438, 446 (D.C.Cir.1993). The EPA has provided no response to the AMSA’s claim that the assumed rate and duration are irrational as applied to heat-dried sludge. The EPA’s explanation of the application rate and duration assumptions on the record is minimal. As explained above, the EPA used these assumptions to “backealculate” from the total limit on pollutant accumulation in land to a concentration cap on pollutant in any given application of sludge. The agency’s only stated basis for its assumed annual sludge application rate (“AWSAR”) of 10 metric tons per hectare and duration of 100 years is that it “believes that the pollutant concentrations derived from [these assumptions] are conservative because it is unlikely that any one site will receive 10 metric tons of sewage sludge per hectare per year for 100 consecutive years.” 58 Fed.Reg. 9317.
From the NSSS data, the EPA had information on AWSARs. It reported that typical AWSARs were 7 metric tons per hectare for agricultural land and 18, 26, and 74 metric tons per hectare, respectively, for a public contact site, forest, and a reclamation site. See 58 Fed.Reg. 9317. The EPA does not suggest that its assumptions are tailored to this data or to the information it had about *403heat-dried sludge; rather it suggests only that the assumptions are conservative enough to provide protection under each of these AWSARs. Its reasoning is as follows. The assumed application rate of ten metric tons for a duration of 100 years results in a total assumed sludge application of 1000 metric tons per hectare. This would be conservative enough to encompass the 74 metric ton rate, the EPA explained, because the 74 metric ton rate would not likely continue for more than the 13 years it would take to approach the assumed total of 1000 metric tons per hectare. (Seventy-four metric tons of sludge applied yearly for 13 years equals a net application of 962 metric tons.) Making the same calculation for the other observed AWSARs of 7, 18 and 26 metric tons per hectare, the EPA concluded that these types of applications would not likely continue for longer than the 142, 55, and 38 years it would take to reach the assumed total of 1000 metric tons per hectare. See 58 Fed.Reg. 9317. Besides indicating that the combination of the assumed AWSAR and duration can safely accommodate much higher actual application rates, the EPA offers no reason for selecting them. Nor does it offer any reason for using these assumptions to derive caps for heat-dried sludge which it knows will be applied at application rates and durations well below the assumed numbers.
An agency has discretion to design rules that can be broadly applied, sacrificing some measure of “fit” for administrability. At the same time, however, “[a]n agency must justify its failure to take account of circumstances that appear to warrant different treatment for different parties.” Petroleum Communications v. FCC, 22 F.3d 1164, 1172 (D.C.Cir.1994). In this case, the EPA did not provide any justification for its assumptions of 10 metric tons/heetare for 100 years in the case of heat-dried sludge use, when it had information in the record (1) regarding the actual rate and duration of use of heat-dried sludge, and (2) data showing that heat-dried sludge was not an anomalous type of land application. Given that the EPA had at hand the information necessary accurately to prevent the known risks, it must provide some explanation for ignoring it in favor of blanket, highly conservative assumptions. Absent any further justification, the EPA has not supplied a rational basis for its assumed application duration and rate in the case of heat-dried sludge.19
We therefore require that the EPA reconsider the Table 3 risk-based caps with regard to heat-dried sludge in order either to justify its general assumptions on rate and duration or to provide more tailored caps that fit the data on heat-dried sludge.
IV. City of Pueblo’s Additional Challenges
In addition to its challenge to the 99th percentile caps, Pueblo challenges the risk assessment that underlies the pollutant limits for selenium, found in Tables 1 and 2. It challenges both the EPA’s decision to regulate selenium at all and the exposure assumptions used to derive the risk-based limits. It also challenges the EPA’s refusal to provide a variance procedure. We agree that the EPA has not justified the use of high-occupancy exposure assumptions to regulate the selenium content of sludge applied to low-occupancy sites. In declining to provide var*404iances, however, the EPA acted within its discretion.
A. Risk-Based Cap on Selenium,
The EPA’s method of risk assessment for each identified pollutant was to model 14 possible exposure pathways through which sewage sludge applied to the land could pose a threat to human health or the environment,20 and a prototype “Highly Exposed Individual” (“HEI”) for each pathway.21 The EPA analyzed each pathway for each pollutant, and the pathway producing the most restrictive limit on the pollutant became the basis for the cumulative pollutant loading rates (CPLRs) in Table 2, and then the risked-based caps in Tables 1 and 3.
For selenium, the pathway leading to the most stringent cap was Pathway 3, “Sewage Sludge -> Human,” which “assesses the hazard to a child of ingesting undiluted sewage sludge.” 1 Technical Support Document at 5-104, reprinted in J.A. at 632. The highly exposed individual for this path is a l-to-6year-old child who ingests sewage sludge daily for a maximum of 5 years. See id. Accordingly, the final risk-based concentration cap for selenium, which governs all land application of sewage sludge, is based on the Pathway 3 scenario.
Pueblo first challenges the inclusion of selenium in the regulations at all, on the grounds that it conflicts with an earlier EPA profile of selenium. In 1985, the EPA issued a preliminary analysis of selenium as part of a series assessing a total of 32 chemicals of potential concern in sewage sludge.22 It concluded then that “[n]o human health hazard due to Se[lenium] is expected when either sludge-amended soil or pure sludge is ingested.” 23 For some of the other pathways, it did identify risks when sludge containing typical amounts of selenium was applied at high rates or when sludge containing high amounts of selenium was applied at lower rates.24 The profile explained that it was a “rapid screening tool” and if a significant hazard was indicated, a more detailed assessment would be undertaken.25 The final regulations cap selenium for precisely the pathway that the profile found to present no threat: human digestion of sewage sludge. As the EPA explained, however, the 1985 profile was a preliminary assessment. Our task today is to evaluate the final rule based on its underlying, more detailed, assessment. If the final regulation of selenium is adequately supported on its own terms, any conflict with the earlier, preliminary assessment is not significant. Accordingly, we now turn to whether that condition — adequate support for the regulation — is satisfied.
With respect to the final regulation, Pueblo challenges the risk-assessment underlying the risk-based limits on selenium in Tables 1 and 2 as applied to Pueblo’s land application practices. As explained above, the risk-based limit for selenium is derived from Pathway 3 — human ingestion of sludge. Pueblo objects to the use of the Pathway 3 HEI — a child who ingests sewage sludge daily for up to 5 years — to regulate the application of sewage sludge to “public contact sites” to which children will not have access — “including all of the various land application sites utilized by Pueblo,” such as highway medians, roadside cemeteries, golf courses, and industrial parks. Pueblo Brief at 14.
The EPA recognizes that public contact sites “include both those with ‘a high potential for occupancy,’ such as parks, and those with ‘a low potential for occupancy,’ including highway medians and roadside cemeteries.” *405EPA Brief at 48-49. Nevertheless, it chose “to be conservative” and define the HEI based on sites with a high potential for occupancy. Id. at 49. See also 1 Technical Support Document at 5-360, reprinted in J.A. at 686.
The EPA has failed to demonstrate a rational relationship between its highly conservative exposure assumptions and the actual usage regulated by those assumptions. See Edison Elec. Institute v. EPA, 2 F.3d at 446. Indeed, the EPA has acknowledged the misfit, but argues that in a “rulemaking of staggering complexity, the Agency was not required to refine its analysis so precisely as to devise a separate exposure analysis for children who ingest sludge on highway medians or in cemeteries.” EPA Brief at 49. Although the EPA is not held to a standard of precise refinement, it is held to one of rationality and it must supply a reasoned basis for its regulatory choices. If, as Pueblo’s practices suggest, a significant proportion of sewage sludge application involves sites with low potential for public and child contact, then it is irrational, at least without further explanation, to sweep these applications willy-nilly into a category based on a high-child-exposure model. Accordingly, we remand the selenium limits in Tables 1 and 2 for further justification or modification.
B. Site-Specific Variances
Pueblo also argues that the EPA’s decision not to allow site-specific variances “effectively negated Pueblo’s right [under the Administrative Procedure Act] to petition the Agency to amend or repeal the land application limitations for selenium.” Pueblo Brief at 19. The absence of a variance procedure by which Pueblo could request an exemption from the requirements, however, has no bearing on Pueblo’s right to petition the agency to amend or repeal its caps. The EPA was within its discretion in rejecting a variance procedure because “site-specific pollutant limits would have to be developed on a site-by-site basis for possibly thousands of land application sites.” 58 Fed.Reg. 9309. Cf. Edison Elec. Institute v. EPA, 2 F.3d at 446 (upholding EPA’s rejection of a variance mechanism). While a variance procedure might save a marginally overbroad general rule, the agency was under no general duty to establish such a procedure, and its reconsideration of the general rule on remand will presumably remove any need for an escape hatch.
V. LeatheR Industries’ Challenges to the Chromium Caps
Leather Industries challenges the Table 2 risk-based pollutant limit of 3,000 kg/ha for chromium.26 This risk-based figure was derived from Pathway 8 — Sewage Sludge -> Soil -> Plant — the most stringent pathway for chromium. Pathway 8 addresses the problem of phytotoxicity: when plants absorb certain quantities of certain metals their yield can be sharply reduced. The EPA determined that a pollutant would be capped at that concentration which evidence showed would create phytotoxicity effects leading to a more than 50% drop in plant growth.27 Leather Industries argues that the EPA does not have authority to regulate on the basis of phytotoxicity, and that, even assuming it has such authority, the resulting pollutant limit has insufficient evidentiary support. We conclude that the EPA has authority to protect against phytotoxicity, but that it lacks adequate support for its final limit.
A. Phytotoxicity as a Regulatory Criterion
Leather Industries first argues that the EPA does not have statutory authority to regulate on the basis of phytotoxicity because reduced crop yield is a purely economic concern, not a “human health” or “environmental” concern. Leather Industries Brief at 10-11. Leather Industries did not pursue this claim in its reply brief or at oral argument, and we find the claim untenable. The EPA’s mandate to establish standards for pollutants that “may be present in sewage sludge in concentrations which may adversely affect ... the environment,” 33 U.S.C. *406§ 1345(d)(2)(D), surely encompasses the authority to protect crop yield, an indisputable aspect of the “environment.”
B. The Chromium Limit
Leather Industries offers a more compelling challenge to the quality of the data underlying the final chromium limit. Modell-ing the phytotoxicity pathway involves two steps: (1) determining the phytotoxicity thresholds — “the concentration of each metal in the tissue of each plant group ... associated with 50 percent reduction in biomass,” (“Step 1”), 1 Technical Support Document 5-200, reprinted in J.A. at 657; and (2) determining the quantity of pollutant in the soil that would cause the plant to absorb that amount of pollutant (“Step 2”).
1. Step 1: Phytotoxicity Threshold
In the first step — determining the phyto-toxicity threshold for chromium — the EPA relied on short-term laboratory studies to determine the concentration of pollutant in plant tissue that would reduce plant growth by 50%. As far as we can tell, the EPA had only one study relating to this step of the analysis for chromium, a 1975 study of corn grown in pots containing sludge-amended soil (“Mortvedt & Giordano Study”).28 Leather Industries argues that the EPA improperly relied on this study because (1) some of its conclusions about the phytotoxic effects of chromium were based on experiments using hexavalent rather than trivalent chromium, and (2) the study assessed phytotoxicity using pot rather than field studies.
a. Trivalent vs. Hexavalent Chromium
Chromium can exist in two states, hexava-lent and trivalent. With regard to risk to human health, only hexavalent chromium is toxic, and the EPA has “delisted” chromium in tanning industry waste because this chromium is in the trivalent form.29 Chromium in sewage sludge is also in the trivalent state, and the agency so assumed in this regulatory process. See 58 Fed.Reg. 9297. There are two complications, however, which may nevertheless make sewage sludge chromium an appropriate target of regulation. First, there is some evidence that trivalent chromium can oxidize to hexavalent chromium. Second, there is some evidence — which may be related to the oxidation possibility — that trivalent chromium can cause phytotoxicity in plants.
As to the first complication, there are several studies cited in the record showing that trivalent chromium can oxidize to hexavalent chromium, though the levels of conversion appear to be low. A survey of these studies led one scientist to conclude that a “[trivalent chromium]-containing sludge could release low levels of [hexavalent chromium] over a long-period of time.”30 Because it is not now possible to estimate the extent of hexavalent chromium formation, that scientist recommended regulating all chromium — hexavalent and trivalent — alike.31 As to the second complication, the Mortvedt & Giordano study on which the EPA relied to determine the chromium phytotoxicity threshold found that trivalent chromium is less toxic to corn, but nevertheless causes some phytotoxicity.32 There is, then, enough genuine scientific debate regarding trivalent chromium’s potential harm and its potential conversion to the hex-avalent state to warrant significant discretion on the part of the agency in its choice to use data on the harm posed by hexavalent chromium to regulate trivalent chromium.
More important, however, from Our reading of the underlying reports, it appears that to the extent that trivalent chromium is less toxic to plants than hexavalent chromium, it is not because plants can absorb triva*407lent chromium with no harm, but because trivalent chromium is less available for plant uptake. Thus, the hexavalenVtrivalent distinction appears to be critical to Step 2, discussed below, but not to Step 1. The Mortvedt & Giordano study, for instance, explains that the lower toxicity of trivalent chromium “suggests that [trivalent chromium] may have been fixed by the soil in forms which were less available than [hexavalent chromium] to plants.”33 The Peer Review Committee likewise explains that “[h]exava-lent chromium is more soluble and more bioavailable for plant uptake than the trivalent chromium usually found in sludges and field soils.”34
The relevant distinction between hexava-lent and trivalent chromium is that trivalent chromium is less available for plant uptake. This characteristic — availability for plant uptake — is not relevant to Step 1. And in Step 2, discussed below, the EPA’s protocol requires that the uptake-based cap be based on field studies of actual sewage sludge, so any observed uptake would necessarily be of chromium found in sludge, in whatever form. In other words, the apparent reason that trivalent chromium is less toxic to plants is that it cannot get into the plant tissue, not that it is harmless once it is in plant tissue. Therefore, we see no problem with using trivalent and hexavalent chromium data to determine that amount of chromium, once in, that is harmful, so long as there is data on trivalent or sewage sludge chromium getting in to the plant. As of now, there appears not to be, as we discuss below.
b. Pot v. Field Studies
Leather Industries also argues that the EPA’s reliance on the Mortvedt & Giordano study was flawed because that study uses pot, rather than field, studies. This distinction too, however, is relevant to uptake, and thus to the second step, not the first. Pot studies test the growth of plants in pots containing actual sludge or soil with salt spikes. Many eommenters urged, to the agency that salt or pot studies “would drastically over-estimate plant uptake.” 58 Fed. Reg. 9294. This is the case for salt studies because metal-containing salt spikes “are not bound to an organic matrix and are, therefore, more freely taken up by plant roots,” and for pot studies because “pots tend to restrict the area of root growth and the small amount of contained soil tends to concentrate and retain the sewage sludge pollutants around the roots, thus accelerating uptake.” Id. The EPA has acknowledged that salt and pot studies are inadequate to model plant uptake and has disclaimed their use for uptake analysis.35 The EPA did rely on pot studies for the first step, unrelated to uptake. In this context, we see no infirmity with pot studies, and Leather Industries has identified none.
2. Step 2: Plant Uptake
After arriving at a phytotoxicity threshold of 3.0 micrograms of chromium per gram of plant tissue, the EPA had to determine the soil concentration of chromium in sewage sludge that would cause plants to absorb chromium up to the phytotoxicity threshold. The EPA looked to field studies of corn grown on sludge-amended soil to make this determination. These field studies, however, provided no data on plant growth in soil with chromium levels in excess of 3,000 kg/ha. EPA Brief, at 73. Moreover, the data the studies did provide indicated no risk of phytotoxicity in soils with up to 3,000 kg/ha of chromium. Indeed, the studies showed an inverse relation between soil concentration and plant concentration of chromium: the higher the soil concentration of chromium, the lower the plant concentration. As one study explains: “the probability of exceeding the threshold is greater for plants grown in soil not receiving any sewage sludge than for those grown in sludge-treated soils.”36
*408Having no other data available on the connection between soil concentration and plant uptake, the EPA chose the 8000 kg/ha threshold. It explained that it chose this limit because it was “the upper boundary of the range for which EPA had data,” and the EPA “had no data indicating that chromium loading rates in excess of 3000 kg/ha would be safe for plants.” EPA Brief at 72. The EPA further justifies this decision by explaining that “[w]hile EPA believes that metals are bound to the sludge and thus relatively unavailable to be ‘taken up’ by plants, the understanding of this process is still developing.” EPA Brief at 73-74. In the face of this uncertainty, it points to “data suggesting that chromium may not remain bound to the sludge at high loading rates.” EPA Brief at 74. This data, however, is the Mortvedt & Giordano study, which, as explained above, used pot studies, and the EPA has disclaimed reliance on pot studies to model plant uptake.37 Under the EPA’s own protocol, the Mortvedt & Giordano study cannot support any conclusions about plant uptake of chromium from sewage sludge amended soils, and thus cannot justify an otherwise unsupported cap premised on uptake-potential.
In sum, the EPA’s relevant evidence (1) provided no data for chromium uptake at soil concentrations greater than 3000 kg/ha, and (2) showed no uptake danger at a soil concentration of 3000 kg/ha or at any other concentration. Indeed, it indicated a declining probability of plant uptake with increased soil concentration. Based on this data, the EPA chose as the ceiling chromium soil concentration 3000 kg/ha because that was the highest concentration for which it had data. While the EPA “may ‘err’ on the side of overprotection,” it “may not engage in sheer guesswork.” American Petroleum Institute, 665 F.2d at 1186-87.
VI. Additional Claims
The AMSA challenges the EPA’s classification of dedicated beneficial use sites as “surface disposal” rather than “land application,” arguing that this classification is arbitrary and “promotes a negative public perception of dedicated sites as ‘dumping grounds.’” AMSA Brief at 36. Dedicated beneficial use sites are sites “generally owned, operated, and controlled by, or are controlled under long-term leases to, the municipal sludge operator.” 58 Fed.Reg. 9259-60. Although dedicated sites may be used “to produce crops, such as corn, which are sold as animal feed or for alternative fuel production,” 58 Fed.Reg. 9260, the EPA did not classify dedicated sites as a type of land application. The EPA explains this decision on the grounds that dedicated sites involve the application of sludge “at greater than agronomic rates” and public access is generally strictly controlled, thus making it appropriate for other aspects of the regulatory regime to differ. 58 Fed.Reg. 9259. The EPA acknowledges that its “distinction is one of degree, since both land application sites and dedicated sites involve the placement of sewage sludge on the ground.” EPA Brief at 40. Indeed, land reclamation sites, which also involve application at greater than agronomic rates, are apparently classified as land application sites in the final rule. See § 503.11(n), 58 Fed.Reg. 9391. The EPA has not explained its reason for distinguishing between land reclamation sites and beneficial use sites. We note, however, that while reclamation sites generally receive only one application at greater than agronomic rates, beneficial use sites receive repeated applications.
The EPA’s reason for distinguishing beneficial use sites from other land application practices is not entirely clear, but the fact that these sites receive repeated applications of sludge at greater than agronomic rates *409seems a plausible basis for distinction. Where the agency’s line-drawing does not appear irrational and the AMSA has not shown that the consequences of the line-drawing are in any respect dire — the only harm it has claimed to identify from the classification involves unsubstantiated claims of potential public disfavor — we will leave that line-drawing to the agency’s discretion.
VII. Conclusion
In sum, we uphold the EPA’s refusal to provide site-specific variances and its decision to classify dedicated sites as a type of “land disposal.” We hold that the EPA has failed to demonstrate that the 99th percentile caps in Table 3 are based on risk, as required by the statute, and therefore remand those Table 3 caps. We also find that the EPA failed to establish a rational relationship between the assumed application rate and site life underlying the risk-based concentration caps in Table 3 and the actual usage of heat-dried sludge, which is regulated by Table 3. As applied to heat-dried sludge, we remand those caps as well. We further hold that the EPA failed to provide a rational basis 'for applying the risk-based cap on selenium based on high occupancy exposure assumptions to public contact sites with low potential for occupancy. Accordingly, we remand the Table 1 selenium cap as applied to public contact sites with low potential for occupancy. Finally, we hold that the EPA failed to provide evidentiary support for its Table 2 cumulative pollutant limit on chromium, and remand that limit as well.
So ordered.
. Office of Water Regulations and Standards, U.S. EPA, Summary of Environmental Profiles and Hazard Indices for Constituents of Municipal Sludge 1 (1985), reprinted in J.A. at 889.
. Elsewhere in the record, the EPA uses the figure of 180 for the number of POTWs at which it performed sampling and analysis. 58 Fed.Reg. 9268. We are not certain which, in fact, is correct.
. 1 U.S. EPA Technical Support Document for Land Application of Sewage Sludge 6-17 (1992), reprinted in J.A. at 728 (“Technical Support Document”).
. Bulk, not packaged, sewage sludge applied to a lawn or home garden must comply with the Table 3 "clean sludge” regulations. See § 503.13(a)(3), 58 Fed.Reg. 9391-92. This provision is not at issue.
.From our reading of the regulations, it appears that if both “clean” and non-"clean” sludge are applied to land, only the pollutant contribution from the non-"clean” sludge must be recorded and controlled.
. See 1 Technical Support Document at 6-18, reprinted in J.A. at 729.
. Id. at 6-7, reprinted in J.A. at 718.
. For nickel, the risk-based and 99th percentile caps are identical.
. The AMSA also argues that the 99th percentile caps are in conflict with the section of the statute authorizing removal credits. We find this particular tack meritless. Under the removal credit system, POTWs may issue removal credits to indirect dischargers — the industrial facilities that discharge waste to the POTWs — in those cases where the POTWs perform treatment that would otherwise be the responsibility of the indirect dischargers. These removal credits prevent dou*400ble treatment by the POTWs and the indirect dischargers. Under the statute, removal credits are only available if they do not prevent the ultimate sewage sludge from complying with the regulations at issue, the sludge use and disposal regulations. See 33 U.S.C. § 1317(b). The AMSA argues that "removal credits are consistent with the national policy that the discharge of toxic pollutants in toxic amounts be prohibited.” AMSA Brief at 23 (emphasis in original). This argument adds nothing to its position: to the extent that the sludge regulations are valid, the removal credits can be conditioned on those sludge regulations. The important question is whether the sludge regulations are valid on their own terms.
. There is no indication in the record of the regtdatoiy or statutory basis for this peer review committee.
. Memorandum of July 24, 1989, reprinted in J.A. at 30.
. Coop. State Research Serv. Technical Comm. W-170, Peer Review: Standards for the Disposal of Sewage Sludge 86-87 (1989), reprinted in J.A. at 60-61 (“Peer Review Report”).
. Id. at 86, reprinted in J.A. at 60.
. Clearly, the EPA’s mandate to establish standards "adequate to protect public health and the environment from any reasonably anticipated adverse effects of each pollutant," 33 U.S.C. § 1345(d)(2)(D), does not give the EPA blanket one-way ratchet authority to tighten standards. Cf. Contract Courier v. Research & Special Programs Admin., 924 F.2d 112, 115 (7th Cir.1991) (“Statutes do more than point in a direction, such as 'more safety.’ They achieve a particular amount of that objective, at a particular cost in other interests. An agency cannot treat a statute as authorizing an indefinite march in a single direction.”).
.Leather Industries challenges the 99th percentile cap for chromium on the grounds that the underlying data did not include any POTW that accepts significant amounts of wastewater from a leather tannery. Leather Industries Brief at 18 n. 10. Specifically, Leather Industries states that while the highest reported level of chromium concentration from the NSSS is 3,750 mg/kg, sludges from the POTWs receiving significant amounts of wastewater from tanneries have chromium concentrations in excess of 30,000 mg/kg. Leather Industries Brief at 18 n. 10. The EPA has not contested this claim.
Pueblo argues that the 99th percentile cap for selenium is arbitrary because it "has penalized communities in Western states where naturally occurring concentrations [of selenium] are high.” Pueblo Brief at 17.
Because we have remanded the 99th percentile caps on chromium and selenium on other grounds, we do not reach these claims.
.These assumptions were used to generate all the risk-based concentration caps — those in Table 1 as well as those in Table 3. The AMSA has only challenged Table 3, so we do not address the use of these assumptions in Table 1.
. Comments of Milwaukee Metropolitan Sewerage District at 2 (Aug. 3, 1989), reprinted in J.A. at 160.
. Comments of Milwaukee Metropolitan Sewerage District at 19 (July 18, 1989), reprinted in J.A. at 151.
. We note that for packaged sewage sludge, failure to meet the Table 3 caps sends the sludge to Table 4. The only additional burden imposed by Table 4 is that the producer of the sludge must provide a label with the actual application rate that will ensure that the annual cumulative limits are not exceeded, see supra part II.B. Thus, with respect to packaged sewage sludge, the excessively conservative application assumptions may be quite easily “cured” with a simple label as to actual safe application rates.
Where the assumed rates underlying Table 3 are inaccurate with respect to packaged heat-dried sludge, then, the producer need only supply its own label with the rates that are accurately calculated to protect safety. No other burden is imposed on the producer. Tables 3 and 4 may well be a rational way to accommodate the actual application rates of packaged heat-dried sludge. We cannot tell, however, if this is the purpose of the system. Moreover, to the extent that POTWs produce one type of sludge product for packaged and bulk use, they would not be greatly helped by relief only for the packaged use. If they produced sludge that was not "clean,” they could relatively easily label it for packaged sale, but bulk use would still be considerably more burdensome. We invite the agency to elaborate on these issues as it justifies or reconsiders the risk-based caps in Table 3.
. Pathway 1, for instance, is “Sewage Sludge - > Soil -> Plant -> Human.” The EPA describes all the pathways at 58 Fed.Reg. 9284-88.
. The HEI for Pathway 1 on nonagricultural land, for instance, is "a person who regularly harvests edible wild plants ... from forests or range lands that have been amended with sewage sludge.” 1 Technical Support Document at 5-4, reprinted in J.A. at 613.
. See Office of Water Regulations and Standards, U.S. EPA, Environmental Profiles and Hazard Indices for Constituents of Municipal Sludge: Selenium (1985), partially reprinted in J.A. at 258-267.
. Id. at 2-2, reprinted in J.A. at 264.
. See id. at 2-1 to 2-2, reprinted in J.A. at 263-64.
. Id. at i, reprinted in J.A. at 259.
. The Table 2 CPLR is the basis for the risk-based concentration limit as well, which is used in Table 1.
. The EPA used plant growth as a proxy for crop yield. This has not been contested.
. See J.J. Morvedt & P.M. Giordano, Response of Com to Zinc and Chromium in Municipal Wastes Applied to Soil, 4 J. Envtl. Quality 170 (1975), reprinted in J.A. at 906.
. See R.J. Bartlett, “Chromium,” in Penn. Agric. Experiment Station, Penn. State Univ., Criteria and Recommendations for Land Application of Sludges in the Northeast 49 (1985), reprinted in J.A. at 930 ("Bartlett”).
. Id. at 50, reprinted in J.A. at 931.
. See id. at 50-51, reprinted in J.A. at 931-32.
. See'Mortvedt' & Giordano at 173, reprinted in J.A. at 909.
. Mortvedt & Giordano at 173-74, reprinted in J.A. at 909-10.
. Peer Review Report at 46, reprinted in J.A. at 56.
. See 1 Technical Support Document at 5-200, reprinted in J.A. at 657.
. A.C. Chang et al., A Methodology for Establishing Phytotoxicity Criteria for Chromium, Copper, Nickel, and Zinc in Agricultural Land Application *408of Municipal Sewage Sludges, 21 J. Envtl. Quality 521, 529 (1992), reprinted in J.A. at 1039.
. The EPA also mentions two data points showing a yield reduction of plants grown in soil with chromium levels of 1,518 kg/ha and 3,036 kg/ha. See 2 Technical Support Document at F-46 to F-47, reprinted in J.A. at 818-19 (data points 285 and 294). As far as we can tell, however, this data involves plants grown in soil containing eight different heavy metals, and does not isolate the effect of chromium. Indeed, the EPA's cursory mention of these data points suggests that they could not have played any prominent role in its final regulatory choice.