Every output file AERMOD can write, and when you'd want it

Beyond the printed tables, AERMOD writes special-purpose output files — ten of them — each answering a different question. Most runs use two or three. This page is the catalog: what each file contains, the question it answers, and the traps worth knowing before you request it.

The User's Guide introduces the set this way:

The AERMOD model provides options for seven types of output files for specialized purposes.

Source · AERMOD User's Guide (EPA-454/B-26-001) §3.7.2, p. 3-168. (The list that follows in that section now names ten keywords — the sentence predates the additions made for the 1-hour standards.)

AERMOD's special-purpose output files, grouped by the question each one answers

For plotting concentration maps

PLOTFILE — the one nearly every project uses. It writes, for a chosen averaging period, source group, and rank, a simple table of receptor coordinates and the value at each:

The data are written to the file in the order of x-coord, y-coord, concentration so that the file can easily be imported into a graphics package designed to generate contour plots.

Source · §3.7.2.3, p. 3-174.

This is the file behind every isopleth map in a permit application. Two rules attach to it. First, the rank must already be requested in the ranked-value table — the model will not plot a rank it did not tabulate:

Note: The averaging period and high value for which a PLOTFILE is generated must also be included on the RECTABLE keyword… The RECTABLE keyword entry must be specified on the OU pathway prior to the PLOTFILE entry.

Source · §3.7.2.3, p. 3-173.

Second, annual and period averages do not take a rank at all — there is only one such average per receptor, so the rank parameter is omitted.

For post-processing raw results

POSTFILE — every calculated average at every receptor, in time order, either as a compact binary file or as plottable text. This is the file to request when you intend to do arithmetic AERMOD does not do for you: combining runs, applying a time-varying background, or building a custom statistic.

The size warning is real:

Note: The POSTFILE option can produce very large files and should be used with some caution. For a file of hourly values for a full year (8760 records) and 400 receptors, the resulting file will use about 14 megabytes of disk space.

Source · §3.7.2.2, p. 3-172.

Scale that mentally before requesting hourly output on a large grid: the size is roughly the number of averaging periods times the number of receptors times four bytes. Ten thousand receptors, hourly, five years is hundreds of gigabytes — a request that will succeed and then be unusable.

For finding exceedances

MAXIFILE — a log of every time and place a value equaled or exceeded a threshold you set. Each record carries the date, the receptor's coordinates and elevation, and the value:

a listing of every occurrence where the result for that averaging period/source group equals or exceeds the threshold value.

Source · §3.7.2.1, p. 3-169.

Set the threshold at the standard and you have an exceedance register. Its second use is as the feeder for event processing: each violation becomes a named event that can be re-run for source-by-source contributions, with names like TH240019 — the nineteenth 24-hour threshold violation (§3.7.2.1, p. 3-170).

MAXDAILY — for 1-hour nitrogen dioxide and sulfur dioxide, the daily maximum 1-hour concentration for each day of the record, for one source group. It is the intermediate step of the design-value calculation, exposed as a file: useful when you want to see the seasonal or operational pattern behind the ranked values rather than just the ranks (§3.7.2.9, p. 3-182).

MXDYBYYR — the same daily maxima summarized by year for each requested rank, so you can see how much each individual year contributes to a multi-year averaged design value. When one year dominates a five-year average, this file is where that shows up (§3.7.2.10, p. 3-182).

For model evaluation and diagnostics

RANKFILE — values by rank, formatted for quantile-quantile plots (a chart comparing modeled and observed distributions). It requires the overall-maximum table to be requested first, and it differs from that table in one important way:

the RANKFILE output differs from the results in the MAXTABLE output in that duplicate date/hour occurrences are removed.

Source · §3.7.2.5, p. 3-176.

EVALFILE — built for formal model-evaluation studies: arc-maximum normalized concentrations hour by hour, requiring receptors organized into arcs. Outside an evaluation study you will not use it (§3.7.2.6, p. 3-177).

SEASONHR — average concentration by season and hour of day. This is a quietly useful diagnostic file: it answers "when does this facility's impact actually happen?" A pattern peaking in winter pre-dawn hours points at stable, low-mixing conditions; a summer afternoon peak points at convective looping or terrain impingement. The season index is 1 for winter through 4 for fall, and the record count column reports how many non-calm, non-missing hours went into each average (§3.7.2.7, p. 3-179).

TOXXFILE — unformatted raw results above a threshold, structured for the TOXX component of the TOXST model. A specialized toxics-assessment interface, not part of a criteria-pollutant demonstration (§3.7.2, p. 3-168).

For source contributions

MAXDCONT — which source group caused each high value, paired in time and space. It is the most consequential output file in the regulatory workflow and has its own page next.

EVENT processing — a companion path for detailed source contributions at specific times. Events come from threshold violations or are user-specified, and the output level is chosen between a contribution summary and full detail:

The SOCONT option provides the average concentration (or total deposition) value (i.e., the contribution) from each source for the period corresponding to the event for the source group. … the DETAIL option provides the hourly average concentration (or total deposition) values for each source for every hour in the averaging period, and a summary of the hourly meteorological data for the event period.

Source · §3.7.3, p. 3-183.

The detail option is the closest thing AERMOD offers to a full explanation of a single bad hour: every source's contribution, hour by hour, alongside the weather that produced it.

Three formatting options worth knowing

SUMMFILE writes the summary of high ranked values — the part of the main output file most people actually read — to its own file, together with the model setup options summary (§3.7.4, p. 3-183). Convenient when the main output file is hundreds of megabytes.

FILEFORM switches result files from fixed-decimal to exponential notation. It exists for two opposite problems: very small values losing precision, and very large values overflowing the fixed format. There is a regulatory catch:

However, FILEFORM EXP cannot be used with the DFAULT option with the MODELOPT keyword when the pollutant specified with the POLLUTID is a criteria pollutant … If FILEFORM EXP is specified when modeling any one of the criteria pollutants, the DFAULT option must be omitted from the MODELOPT list of options specified or AERMOD will automatically reset FILFORM to FIXed format.

Source · §3.7.4, p. 3-184. Since version 26135 the chosen format applies to all result files, including the main output file and the summary file.

Since dropping the regulatory-default option makes a run non-regulatory, exponential formatting is effectively a tool for non-criteria work such as air toxics — not something to turn on casually in a criteria-pollutant permit demonstration.

NOHEADER suppresses the header records in formatted output files, which some plotting and post-processing tools prefer (§3.7.4, p. 3-185).

Choosing a sensible output set

For a typical permit demonstration:

Ranked tables at the ranks the standard requires — always.
A contour file at the design-value rank for each pollutant and averaging period you must map.
A source-contribution analysis if the cumulative result approaches or exceeds a standard.
A threshold file if you need an exceedance register.
Raw concurrent results only when you have a specific post-processing plan for them.

The discipline is to request what you will read. Every extra file costs disk, run time, and — when a reviewer receives a folder of files nobody can explain — credibility.

In PlumeSmart

The platform requests the output set each analysis needs, generates contour maps and ranked tables directly from the run, and keeps the raw model files available for download, so the record a reviewer receives matches exactly what the model produced.