Ohio v. United States Environmental Protection Agency , 798 F.2d 880 ( 1986 )

MERRITT, Circuit Judge.

This opinion follows our earlier opinion in Ohio v. E.P.A., 784 F.2d 224 (6th Cir.1986). There we held that the EPA’s use of CRSTER, a computerized atmospheric model, was arbitrary and capricious in the absence of site-specific validation studies. We then ordered the parties to propose a program for validating the CRSTER model, and a set of interim emissions limits for use until the validation process has been completed. The parties have now responded, and the Court has before it and has read and considered (1) the Group III petition for rehearing and rehearing en banc, (2) EPA’s petition for rehearing, (3) each party’s response to those two petitions, and (4) each party’s response to the two questions presented for comment in the Court’s opinion and order of February 26,1986, 784 F.2d 224.

As the full Court has not exercised its discretion to review this case en banc, the case is back before the panel. We hold as follows:

First, the panel reaffirms its decision. On the present record the use of the CRSTER model without validation to forecast SO2 diffusion at the two plants adjacent to Lake Erie is arbitrary and capricious. Instead of simply deferring to EPA’s decisions regarding scientific and computer models, as EPA and the Group III petitioners urge, the legislative history of the 1977 Clean Air Act indicates that the Courts are to conduct a “searching review” of the basis of EPA modeling and test procedures like the one at issue here. See No. 294, 95th Cong. 1st Sess., 1977 U.S. Code Cong. & Ad.News 1077, 1397, 1402. See also H.Conf.Rep. No. 95-564,1977 U.S. Code Cong. & Ad.News 1502, 1559 (calling for “thorough, comprehensive review”). Our inquiry indicates that the specific meteorological and geographic problems present at these sites were not adequately resolved in using the CRSTER model without validation. Based on the CRSTER model, EPA has reversed its field and has allowed 400% more SO2 than its original order permitted — all without evaluation, validation, or empirical testing of the model at the site, a site where the parties agree that the lakeshore effects on the dispersion of SO2 are significant but as yet untested and unknown (at least on the basis of the present record). See Motor Vehicle Mfg. Assn. v. State Farm Mutual, 463 U.S. 29, 41-42, 103 S.Ct. 2856, 2865-66, 77 L.Ed.2d 443 (1983) (agency “changing of course by rescinding rule” requires closer judicial scrutiny). By no means does the Court insist that all models be validated at all sites. We find only that the accuracy of CRSTER at the site has not been sufficiently demonstrated to meet the arbitrary and capricious standard of review.

Second, although the Court originally considered saving the time and expense of starting over at the agency level by ordering the parties to propose an appropriate evaluation procedure for the CRSTER model, both EPA and the Group I petitioners have requested that we remand the case to EPA for de novo resolution of the issue.

Accordingly, the case will be remanded to EPA for further proceedings consistent with this opinion and our opinion of February 26, 1986. EPA will submit a proposed order of remand within 20 days from the date of filing of this opinion. The other parties will submit their comments and alternative proposals for an order of remand within 10 days following the date of filing of EPA’s proposed order of remand.

It is so ordered.