Reading AERMOD's results: ranks, tables & the design value

AERMOD computes a concentration at every receptor for every averaging period, for every hour of the meteorological record. A five-year run with a ten-thousand-receptor grid produces hundreds of millions of numbers. Almost none of them are the answer. The answer is a handful of ranked values, selected in the exact form the standard defines — and the output pathway is where you tell the model which ranks to report.

This page explains the ranking vocabulary that runs through every AERMOD output file and every regulatory conversation: what a rank is, what "high-eighth-high" means, why the rank you request must match the standard's form, and how to read the tables that come back.

What a "rank" actually is

Take one receptor and one averaging period — say 24-hour averages at receptor R-427, over one year. That receptor has 365 twenty-four-hour averages. Sort them from highest to lowest. The first-highest is the largest of those 365 values; the eighth-highest is the eighth one down the list.

Two things trip people up here:

Ranking happens per receptor, not across the map. Each receptor keeps its own sorted list. The eighth-highest at R-427 usually occurred on a different day than the eighth-highest at R-428.
"Highest" across receptors is a second, separate step. After each receptor has its eighth-highest value, the analysis takes the highest of those across the whole receptor network. That is the number compared with the standard. The shorthand H8H — "high eighth high" — names both steps at once: the highest, across receptors, of the eighth-highest at each receptor.

The same two-step shape gives us H1H (highest first-highest — the peak anywhere, the single largest number the run produced), H2H, H4H, H6H, and H8H. When a reviewer asks "what's your H1H?", they are asking for the peak concentration anywhere in the domain; when they ask for the H8H, they are asking for a specific percentile of a specific distribution.

H8H in two steps: each receptor ranks its own values and keeps the 8th-highest; the design value is the highest of those across all receptors

Why the rank is not a preference

Each standard names its own rank, because each standard is written with a different tolerance for rare events. From the AERMOD User's Guide:

For the 1-hour NO2 standard, the modeled design value is based on the 98th-percentile of the daily maximum 1-hour values, which is represented by the eighth highest of the daily maximum 1-hour values across the year. The 1-hour SO2 modeled design value is based on the 99th-percentile, or fourth highest, of the daily maximum 1-hour values across the year.

Source · AERMOD User's Guide (EPA-454/B-26-001) §3.2.17, p. 3-59.

For particulates, the same idea with different numbers:

Modeling demonstrations of compliance with the PM-10 NAAQS are based on the High-N+1-High value over N years, or in the case of five years of NWS meteorological data, the High-6th-High (H6H) value over five years.

Source · AERMOD User's Guide §3.2.16.2, p. 3-59.

And for fine particulate, the User's Guide spells out the full two-step, multi-year recipe in one sentence — worth reading slowly, because it is the clearest statement of how ranks, receptors, and years combine:

the model calculates the eighth-highest 24-hour concentration at each receptor for each year modeled, averages those eighth-highest concentrations at each receptor across the number of years of meteorological data, and then selects the highest, across all receptors, of the N-year averaged eighth-highest values.

Source · AERMOD User's Guide §3.2.16.1, p. 3-57.

Note the order in that recipe: rank within each year, average the ranks across years, and only then take the maximum across receptors. Averaging first, or taking the maximum first, produces a different — and wrong — number. The Part I pages on design-value forms cover why each standard is built this way; the point here is that the rank is dictated by the regulation, and requesting the wrong one produces a technically valid run that answers the wrong question.

Two important refinements of "highest"

The daily-maximum filter. For 1-hour nitrogen dioxide and sulfur dioxide, the ranking is not over all 8,760 hours. It is over the 365 daily maximum 1-hour values — each day contributes only its worst hour, and then the year's 365 daily maxima are ranked. AERMOD applies this special processing automatically when the pollutant is NO₂ or SO₂ and 1-hour averages are requested; adding other short-term averaging periods to the same run turns it off, with a warning, and reverts to ranking all hours (§3.2.17, p. 3-60).

Multi-year averaging. With five years of meteorology in one run, AERMOD averages the per-year ranked values, matching the standards' three-year averaged form with an unbiased five-year estimate. Annual standards for NO₂ and SO₂ work differently — they are based on the highest individual year rather than an average — which is why those runs are handled with the multi-year option rather than a single concatenated file. The rule pages in this guide cover that distinction in detail.

The three printed tables

The main output file (aermod.out) contains up to three kinds of summary table, each controlled by its own keyword:

Ranked values by receptor (RECTABLE) — the workhorse. You name the averaging period and which ranks you want, and AERMOD prints, for every receptor, the value at each requested rank. This is where design values come from. Modern AERMOD accepts any rank up to the 999th highest:

the RECTABLE keyword had been modified to allow user-specified ranks of short-term averages (for all pollutants) up to the 999th highest value.

Source · AERMOD User's Guide §3.7.1, p. 3-165.

Overall maximum values (MAXTABLE) — instead of per-receptor ranks, the top N values anywhere in the domain, with the receptor and date for each. Ask for the top 50 and you get a table of the fifty largest averages the run produced, each tagged with where and when it happened. This is the fastest way to see whether the peaks cluster at one location (a single dominant source or a terrain feature) or scatter across the map.

Concurrent values by day (DAYTABLE) — every receptor's value for every day processed. Occasionally invaluable for debugging a specific episode, and enormous:

Note: This option can produce very large output files, especially when used with a full year of data and very short period averages, such 1-hour and 3-hour. It should therefore be used with CAUTION.

Source · AERMOD User's Guide §3.7.1, p. 3-167.

Source groups: the other axis of every table

Every table above is produced per source group. A source group is a named set of sources you define — often "ALL" for everything, plus groups such as the applicant facility alone, or a set of neighboring sources.

This is not bookkeeping; it is what makes the regulatory comparisons possible. The significant-impact test needs the project's own sources by themselves; the cumulative NAAQS test needs everything together; and a source-contribution analysis needs each contributor separately. Defining thoughtful groups before the run is far cheaper than re-running to get them, and it is the input that the MAXDCONT analysis (covered later in this Part) depends on entirely.

Reading a results table without being misled

Three habits save real time:

Check the rank in the header before reading the number. Every table states its averaging period, source group, and rank. A number copied from the wrong table is the most common reporting error in the field.
Look at where the maximum is, not just what it is. A peak sitting on the fence line next to the tallest stack tells a different story than a peak two kilometers away on a ridge — and the second one points at terrain, which is a different refinement path.
Compare against the standard in the standard's own units. AERMOD reports in micrograms per cubic meter; some standards are written in parts per million or parts per billion. The conversion depends on the pollutant's molecular weight and reference conditions, and it is a conversion worth doing carefully once and documenting.

In PlumeSmart

Results are presented with the rank and averaging period named on every number, the design value computed in the standard's own form, and the comparison against the applicable standard shown directly — with the underlying ranked tables available in full for the record.