Citizens Against the Refinery's Effects (CARE) appeals from a final ruling of the Administrator of the Environmental Protection Agency (EPA) approving a permit for the prevention of significant deterioration (PSD) of air quality issued to the Hampton Roads Energy Company (HREC). After reviewing the application, analyzing the modeling data submitted by the HREC, and digesting numerous public comments, EPA determined that the refinery would not cause significant deterioration of air quality in the Portsmouth area. We affirm the action of the Administrator in issuing the permit.The PSD Permit Program

The PSD permit program, designed to protect areas where the air is relatively clean, was added to the Clean Air Act by Congress in 1977. 42 U.S.C. § 7470. The PSD provisions require that new "major emitting facilities"1 obtain permits before locating in areas where the air is cleaner than required by the National Ambient Air Quality Standards (NAAQS).2 These are designated "attainment areas" and are divided into three classes: Class I for areas such as national parks that have very clean air where little or no deterioration is permitted; Class II for areas with clean air where some deterioration resulting from moderate economic growth may occur; and Class III for areas where more economic growth and greater air deterioration is allowed.3 42 U.S.C. § 7472. The PSD permit allows the major emitting facility to contribute to air pollution only up to specified incremental amounts permitted by the statute. 42 U.S.C. 7473(b)(1). Thus areas with air cleaner than the national standards are protected from factories, refineries, and other pollution sources that locate in those areas.

To obtain a PSD permit, the applicant must submit to EPA an "air quality dispersion" model which predicts whether the facility (1) violates the national standards, or (2) will contribute to air pollution in excess of that allowed by the PSD regulations for the attainment area involved. 42 U.S.C. § 7475. Guidelines for the modeling format have been developed and published by EPA in conjunction with private industry. After public comments were solicited, this modeling guideline was incorporated by reference into the regulations. 40 CFR 52.21(m). Included in the guidelines are various modeling devices designed to limit application of the theoretical models. Mathematical threshold calculations known as "significance levels" were approved by EPA and included in the guideline after notice and public comments were received.4 See 40 CFR 52.21(m).

HREC has been attempting for 7 years to build a 184,000 barrel refinery in Portsmouth, Virginia, the first refinery in over 20 years to be built on the east coast. For purposes of the PSD permit, Portsmouth is a Class II attainment area. HREC submitted an application for a PSD permit to EPA on June 28, 1978.5 On August 4, 1978, HREC submitted the requisite air quality modeling analysis, thus fulfilling the PSD permit requirements.

On August 17, 1978, EPA notified HREC that its application was considered complete as of August 4, 1978. In analyzing the application, EPA (1) reviewed the refinery's impact upon national standards; (2) contracted with an independent consultant to assess the refinery pollution abatement devices to make sure these complied with applicable regulations; and (3) conducted an analysis of air quality impact which resulted in additional permit conditions designed to control SO 2 emissions. EPA issued a proposed permit on October 16, 1979 and public hearings were held in November of 1979. As a result of comments received at those hearings, EPA requested additional carbon monoxide information from HREC. EPA evaluated the additional information and issued the final permit on January 25, 1980.

CARE makes three major attacks on EPA's issuance of the permit to construct the refinery. First, they argue that the modeling analysis submitted by HREC was inaccurate and therefore underpredicted the effect of the refinery on air quality in Portsmouth. Second, CARE contends that EPA should have considered the HREC application complete when the carbon monoxide data was submitted in January, 1979, rather than in August of 1978. Such a change in the completion date would bring HREC under the new monitoring requirements of the 1977 Clean Air Amendments and require a full year of air quality monitoring data from the refinery. 42 U.S.C. § 7475(e)(2). Third, CARE argues that the significance levels approved by EPA in the modeling devices caused the agency to disregard data from the model which should have been considered.

The parties agree upon the standard this Court should follow in reviewing the EPA action. The decision of an administrative agency is to be upheld unless it is found to be "arbitrary and capricious, an abuse of discretion, or otherwise not in accordance with law." Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 91 S. Ct. 814, 28 L. Ed. 2d 136 (1971); Appalachian Power Company v. EPA, 477 F.2d 495 (4th Cir. 1973). In addition the courts will accord great deference to an agency's interpretation of its own regulations. Udall v. Tallman, 380 U.S. 1, 85 S. Ct. 792, 13 L. Ed. 2d 616 (1965).

Analysis of modeling results required for PSD applications is a highly technical area particularly within the expertise of the EPA, and thus the agency interpretation should be given great weight by the court. FPC v. Florida Power & Light Co., 404 U.S. 453, 92 S. Ct. 637, 30 L. Ed. 2d 600 (1972), reh. den. 405 U.S. 948, 92 S. Ct. 929, 30 L. Ed. 2d 819 (1972). Yet CARE argues that this Court must examine in great detail the technical modeling data submitted to the agency and, in considering this data as well as materials outside the record, determine that EPA approval of the HREC permit was unlawful. This we cannot do.

In their first attack, CARE raises questions about four aspects of the air dispersion model submitted by HREC. First, CARE argues that the model should have incorporated five years of weather data instead of the one year of data actually used. The EPA modeling guidelines recommend that five years of data be used. See 40 CFR 51.21(m) incorporating by reference Guideline on Air Quality Models, EPA-450/2-78-027 (1978). EPA permitted the use of only one year of data because only one year was locally obtainable and compatible with the computer model used. CARE, using evidence outside the agency record, seeks to prove that data was available for the recommended five-year period, if not at Portsmouth then at other weather stations. The record demonstrates that EPA approved what it believed to be the best evidence available according to its interpretation of its own guidelines. As such, these decisions do not appear to be irrational, arbitrary, or capricious. This is particularly true where, as here, the agency makes only modeling recommendations. To hold EPA rigidly subject to the suggestions in the modeling guideline would elevate mere agency recommendations to the stature of a regulation, and thus tie the hands of any agency giving guidance to the public in following its regulations.

Second, CARE argues that EPA erred in deciding that HREC had filed its application before one of the other pollution sources, since both applications were filed on the same day. This decision is critical; the second applicant must include potential pollution from the first applicant in its modeling analysis. See 40 CFR 52.21(o)(2)(ii). Absent a clear error of judgment, the agency will generally be upheld on a discretionary decision such as who filed first. See Bowman Transport Inc. v. Arkansas-Best Freight System, 419 U.S. 281, 95 S. Ct. 438, 42 L. Ed. 2d 447 (1974). Without a tie-breaking mechanism in the regulations, the agency had to exercise its discretion in deciding which application to consider first. The primary concern is clean air; as long as EPA considers both applications and the polluting effect of both sources, the agency determination of who files first is a matter of judgment left within the province of the administrator. Where, as here, the regulation specifying who files first is ambiguous and open to interpretation, the agency interpretation will be given great weight by the reviewing court. See Udall v. Tallman, 380 U.S. at 16, 85 S.Ct. at 801.

Third, CARE contends that the model was inaccurate because sulfur dioxide "removal rates"6 were originally used and later were improperly removed from the model. EPA contends that a "worst case" analysis was used, and that the removal rates were taken out from the days where the effect was most intense. This is a disagreement of experts, and as such the discretion of the agency is to be given great weight, absent evidence of arbitrary and capricious actions. See Smithkline Corp. v. Food and Drug Administration, 587 F.2d 1107 (4th Cir. 1978).

Finally, CARE argues that the nitrous oxide input into the model was inaccurate by a factor of ten, thereby creating inaccurate base readings. EPA argues that the error was not raised in public comments before the agency and therefore may not be considered by this Court. See United States v. L. A. Tucker Truck Line, 344 U.S. 33, 73 S. Ct. 67, 97 L. Ed. 54 (1952). The facts demonstrate here that the error which was raised during the public comment period was of a lesser degree than the error presented on appeal. EPA did consider the lesser error in the model and decided it wasn't significant. Given CARE's demonstrated ability to accurately assess information and given the need for specific, accurate objections during the technical administrative process, EPA was justified in considering only the lesser modeling error as presented by CARE and in ruling accordingly. The agency reasoning is properly explicated in the record. See E. I. du Pont de Nemours & Co. v. Train, 541 F.2d 1018 at 1026.

Beyond the errors in the dispersion model, CARE contends that the agency also erred in concluding that the application was complete as of August 4, 1978. CARE argues that EPA should have known that additional carbon monoxide data would be needed and that EPA should have considered the application complete when that data was received in January, 1979, despite the agency letter to HREC on August 17, 1978 indicating that the application was considered complete as of August 4, 1978. Such a completion date would bring the permit under the provisions of the 1977 Clean Air Act which requires one year's monitoring data prior to submission of the PSD application. 42 U.S.C. 7475(a). Thus, to accept the CARE argument that the application was not complete is to require an additional year of data from HREC. 42 U.S.C. 7475(a).

The regulations establish the date of a "complete" application as the date all materials necessary for approval have been filed. 40 CFR 52.21(r)(1). Just because additional information is found necessary after the agency begins to analyze the submission during the public comment stage does not mean that the application is incomplete. If this were the case no application would ever be complete until EPA received the last morsel of information. To approve such a result would be disruptive of the administrative process and would serve no useful purpose.

Lastly, CARE argues that the significance levels used in the models to prevent the theoretical from becoming the absurd contravene the meaning and purpose of the statute. The significance levels, designed to key the modeler into only "significant" levels of incremental pollution, are included in the modeling guidelines and apply only to modeling. See EPA, Guidelines on Air Quality Models 10. Thus these levels are not designed for the absolute NAAQS requirements of the Act. Appellant's reliance upon statutory limits on air pollution is misplaced because the levels adopted by EPA were to be used in measuring the application of the theoretical model instead of actual air pollution.

The deference normally given administrative agencies in interpreting their own regulations as well as the highly technical nature of the modeling techniques make the EPA particularly well suited to make determinations as to whether to issue a PSD permit or not. Where, as here, the agency did nothing which may be construed as arbitrary, capricious, an abuse of discretion, or outside the statute, the only course possible is to affirm the final action of the administrator.