The reviewing authority, the protocol & model approval

A modeling demonstration is not finished when the model runs — it is finished when a reviewing authority accepts it. This page covers the three review structures Appendix W builds around every regulatory analysis: who the reviewer is, the modeling protocol agreed up front, and the rules for which model may be used at all.

Who is "the appropriate reviewing authority"?

Appendix W uses the phrase deliberately, because the answer varies:

Many States and some local agencies administer NSR permitting under programs approved into SIPs. In some EPA regions, Federal authority to administer NSR permitting and related activities has been delegated to State or local agencies. In these cases, such agencies "stand in the shoes" of the respective EPA Region. Therefore, depending on the circumstances, the appropriate reviewing authority may be an EPA Regional office, a State, local, or Tribal agency, or perhaps the Federal Land Manager (FLM).

Source · 40 CFR 51 App. W §3.0(b) — captured verbatim in the source library (live eCFR, issue date 2026-08-01).

One constant sits behind the variety: "For all approvals of alternative models or techniques, the EPA Regional office will coordinate and seek concurrence with the EPA's Model Clearinghouse" (§3.0(b)) — the body EPA created so that "fairness, consistency, and transparency in modeling decisions are fostered" across regions and agencies (§3.3(a)).

The modeling protocol: agree before you model

The Guideline's strongest procedural recommendation is to settle the approach before any work starts:

During this meeting, a protocol should be established between the preparing and reviewing parties to define the procedures to be followed, the data to be collected, the model to be used, and the analysis of the source and concentration data to be performed. … The protocol should be written and agreed upon by the parties concerned, although it is not intended that this protocol be a binding, formal legal document. Changes in such a protocol or deviations from the protocol are often necessary as the data collection and analysis progresses. However, the protocol establishes a common understanding of how the demonstration required to meet regulatory requirements will be made.

Source · 40 CFR 51 App. W §9.2.1(a). EPA's Air Quality Analysis Checklist on SCRAM is the model for its contents.

Everything this guide's earlier pages treat as a judgment call — the representative met station, the background monitor, the receptor extent, the nearby-source inventory — is exactly what the protocol pins down. A protocol conversation is cheap; re-running a five-year cumulative analysis because the reviewer rejects the met data is not.

Preferred models: the Addendum A bargain

Appendix W's model-selection system is a bargain with two sides. Use a preferred model — one listed in Addendum A to the Guideline, of which AERMOD is entry A.1 — exactly as its summary directs, and the applicability question is settled:

Preferred models may be used without a formal demonstration of applicability as long as they are used as indicated in each model summary in Addendum A.

Source · 40 CFR 51 App. W §3.1.2(a); §3.1.1(e): "No further evaluation of a preferred model is required for a particular application if the EPA requirements for regulatory use specified for the model in the Guideline are followed."

The other side of the bargain is strict:

A preferred model must be operated with the options listed in Addendum A for its intended regulatory application. If the regulatory options are not applied, the model is no longer "preferred." Any other modification to a preferred model that would result in a change in the concentration estimates likewise alters its status so that it is no longer a preferred model.

Source · 40 CFR 51 App. W §3.1.2(c).

This is the regulatory root of the DFAULT discipline running through this guide: turning off the regulatory defaults doesn't just change numbers — it changes the model's legal status, converting the analysis into an alternative-model application needing case-by-case approval. (See Rule — Regulatory Default (DFAULT).)

Alternative models: three doors, all guarded

When a preferred model won't do, §3.2 defines exactly three conditions under which an alternative may be approved:

i. If a demonstration can be made that the model produces concentration estimates equivalent to the estimates obtained using a preferred model; ii. If a statistical performance evaluation has been conducted using measured air quality data and the results of that evaluation indicate the alternative model performs better for the given application than a comparable model in Addendum A; or iii. If there is no preferred model.

Source · 40 CFR 51 App. W §3.2.2(b).

Each door has its own test. Equivalency is quantified — "the appropriate regulatory metric(s) are within ± 2 percent of the estimates obtained from the preferred model" (§3.2.2(c)). Better performance requires a statistical evaluation against measured air quality data under an agreed protocol (§3.2.2(d)). And no-preferred-model brings a five-part test: scientific peer review, theoretical applicability, adequate databases, performance evaluations showing the model "is not inappropriately biased for regulatory application," and an established protocol (§3.2.2(e)) — with a pointed footnote that for PSD, where results are used in an absolute sense, "the model should not be biased toward underestimates."

Whichever door: the decision belongs to the EPA Regional office, formally documented "in a concurrence memorandum from the EPA's Model Clearinghouse" (§3.2.1(b)) — which is also how the PSD rule's own model-substitution provision (40 CFR §51.166(l)(2)) is satisfied.

In PlumeSmart

The platform runs the preferred model as Addendum A directs — current regulatory AERMOD, regulatory options on by default, with any non-default option flagged against its regulatory basis — and its run records document model version and options so a protocol or review response can cite exactly what was run.