Nearby sources vs. the monitor: avoiding double-counting

The background page established the split: nearby sources are modeled explicitly, everything else is represented by the monitor. This page covers the trap between those two rules — a nearby source whose impact the monitor already measures — and the tools for handling it.

The problem, in one example

Your project (call it plant A) needs a cumulative Prevention of Significant Deterioration (PSD) demonstration. Two kilometers away sits plant B, a large existing source. The area's ambient monitor is a few kilometers downwind of B, and on most days B's plume passes right over it.

If you now model B explicitly and add the monitor's background to every hour, B is counted twice: once through its modeled concentrations, and again inside the monitor data its plume helped create. The double-counted demonstration can fail on paper while the real air is fine — the classic inflated cumulative analysis.

The Guideline on Air Quality Models (40 CFR part 51, Appendix W) names this connection directly:

A key point here is the interconnectedness of each component in that the question of which nearby sources to include in the cumulative modeling is inextricably linked to the question of what the ambient monitoring data represents within the project area.

Source · 40 CFR 51 App. W §8.3.3(a).

The deciding test: is the source "adequately represented"?

The rule from §8.3.3(b): "All sources in the vicinity of the source(s) under consideration for emissions limits that are not adequately represented by ambient monitoring data should be explicitly modeled." Flip it around and you get the double-counting rule: a source the monitor does adequately represent belongs in the background — and only there.

How do you decide? The Guideline prescribes a visual assessment of the modeling domain, weighing three things (§8.3.3(b)(i)):

• Geometry — "the proximity of nearby sources to the project source and the ambient monitor." A source between the monitor and the prevailing wind is probably in the data; a source on the far side of the domain probably is not.
• Magnitude — "the nearby source's level of emissions with respect to the ambient data." If plant B emits enough to matter and the monitor reads barely above regional levels, the monitor is not capturing B.
• The dispersion environment — "meteorological patterns, terrain, etc." A ridge between B and the monitor, or a wind rose that rarely blows B toward it, changes the answer.

Sources that fail the visual test get "further qualitative and quantitative analysis before being explicitly modeled" (§8.3.3(b)(ii)) — and the Guideline sets expectations for how many survive:

The number of nearby sources to be explicitly modeled in the air quality analysis is expected to be few except in unusual situations.

Source · 40 CFR 51 App. W §8.3.3(b)(iii) — the call is made with "professional judgment by the appropriate reviewing authority."

That expectation is Appendix W's framework — not a cap. The same paragraph is explicit that the determination "calls for the exercise of professional judgment by the appropriate reviewing authority" and that the guidance is not intended "to comprehensively prescribe which sources should be included." In practice the split varies widely by agency: some reviewing authorities apply the representativeness framework narrowly, while others direct applicants to explicitly model extensive inventories — cumulative analyses with hundreds of inventory sources are routine in some state programs. Whatever the split, every explicitly modeled nearby source uses the Table 8-2 emission inputs (§8.3.3(c); see Emission inputs), and the allocation is agreed with the reviewing authority in the modeling protocol.

The toolbox when the geography isn't clean

Real domains are messy: the monitor sees B in some wind directions and not others, or B's mornings but not its afternoons. AERMOD and the Guideline give you graduated tools — all of them about choosing what the background term means, never about subtracting anything from it:

1. Choose a different monitor. The §8.3.2(b) preference — "closest to and upwind of the project area," with similar characteristics — often dissolves the problem: an upwind monitor never saw B in the first place.
2. Sector-varying background. AERMOD's BGSECTOR capability assigns different background values by wind direction (up to six sectors, each at least 30° wide — AERMOD User's Guide §3.3.8). Example: for winds from the north — the directions where the project's peak impacts occur and B's plume misses the monitor — use the monitor's north-wind values; B is modeled explicitly and the background is clean. The double-count disappears sector by sector.
3. Temporal pairing. Where the mismatch is by season or hour rather than direction, App. W §8.3.2(c)(iii) allows pairing background "by season and/or hour of day" (see Background concentrations & representative monitors).
4. Accept the conservatism, knowingly. For a screen that passes anyway, leaving B in both places is merely conservative. The tools above matter when a demonstration is failing — removing a double-count is one of the legitimate refinements to reach for before anything drastic (see the refinement discussion elsewhere in this guide).

What is not in the toolbox: arithmetic subtraction of a modeled source from monitor data. The Guideline's framework allocates each source to one side of the ledger by judgment and documentation, agreed with the reviewing authority — not by post-processing the monitor record.

How this shows up in a failing run

The signature of a double-count is specific enough to diagnose: the cumulative design value fails only at receptors in the sector toward plant B, the controlling hours are the hours the wind carries B toward the monitor, and the project-alone impact at those receptors is comfortably small. When you see that pattern, re-examine what the monitor represents before touching anything else in the model.

In PlumeSmart

Background can be entered uniformly or as sector-varying values (the platform generates the corresponding BGSECTOR/BACKGRND records), and monitor candidates are listed with their Air Quality System (AQS) identities and locations relative to the facility — the geometry the §8.3.3 visual assessment is made from. Inventory sources carry their own roles so the modeled-versus-monitored split stays explicit and reviewable.