Refined Modeling

The basic requirements of refined modeling are similar to screening modeling, but with a more refined data approach (i.e. refined receptor grid, hourly meteorological data, source placement, etc.).

Refined modeling uses real-world hourly meteorological data records to assess a source's impacts on ambient air quality. It is generally accepted that refined modeling results will be much lower than those predicted through the use of a screening model.

Air Quality Models

The Guideline on Air Quality Models provides a complete list of available air quality refined models and modeling techniques. This document should also be consulted for further details on the models suggested in this guideline and for those modeling situations not specifically covered in this guidance. Any deviations from the discussed recommendations should be justified to the Department.

In accordance with Appendix W, effective December 9, 2006, AERMOD will be the required model for predicting near-field (< 50km) impacts for all modeling applications. ISCST3 will no longer be acceptable for any MEDEP-BAQ modeling demostrations.

Other models may be used (for special situations) as they become approved by EPA or through prior written approval from MEDEP-BAQ on a case-by-case basis.

Given that a variety of promulgated models are available for refined modeling use (as discussed in the most recent version of the Guidelines on Air Quality Models), the rest of this page focuses on AERMOD, as it is highly probable that the AERMOD model will be used for most licensing applications.


AERMOD is an all-terrain steady-state dispersion model, based upon Gaussian and planetary boundary layer concepts, designed for use in both simple and complex terrain. It requires many user-selected parameters, many which will require sound judgment and meteorological expertise. MEDEP-BAQ strongly encourages consultants receive professional training in the use of AERMOD prior to performing and submitting an air quality demonstration.

There are 3 major components to the AERMOD modeling system:

AERMOD - The dispersion model which requires various user-selected parameters, as well as incorporates the data created in AERMAP and AERMET
AERMAP - The terrain pre-processor for AERMOD
AERMET - The meteorological pre-processor for AERMOD

In addition, there are other non-regulatory components to the AERMOD modeling system:

AERSURFACE - utility designed to calculate estimates of surface characteristics based upon Land Use/Land Cover (LULC) information.
AERSCREEN - screening model for AERMOD which does not require detailed meteorological or terrain data.
AERMINUTE - utility that processes 1-minute ASOS wind data to generate hourly average winds for input to AERMET in Stage 2.

The AERMOD modeling system (including utility programs) and associated guidance are available for download at the SCRAM website.

The "regulatory default option" should be used when modeling with the current version of AERMOD.


AERMAP is the terrain pre-processor to AERMOD which processes Digital Elevation Model (DEM) data and creates an elevation and height scale (the terrain height and location that has the greatest influence on dispersion) for each receptor in the domain. AERMAP automatically selects the closest node elevation in each quadrant with respect to the receptor or source, then weights that elevation with respect to the distance from the receptor or source. The closer the node elevation, the more weight it is given. Conversely, further distances are weighted less.


AERMAP is designed to utilize two different resolutions of DEM data: 1-degree and 7.5-minute data; however, USEPA recommends that the higher resolution 7.5 minute data be used. The USGS has provided links on their website where DEM files can be downloaded for free at AERMOD requires the source DEM data to be in the Spatial Data Transfer Standard (SDTS format) format.

Once you have navigated to the correct county and map that you will need, you can then download the necessary DEM data file(s). Typically, there will be two files associated with each DEM file, one that contains the DEM-related data and one text file. As an example, for the Augusta 1:24,000 DEM, the following files are available for download:



The first file ( 1643784.DEM.SDTS.TAR.GZ ) contains the zipped DEM data files, while the second file ( 1643784.DEM.SDTS.TXT ) is a text file that contains valuable information about the corresponding DEM data file (name, datum, resolution, projection type, map boundary coordinates, etc.). Be sure to download both files to verify that the data contained in the text file is the required data for your specific application.

The zipped file can then easily be unzipped using WINZIP or similar type archiving/compression program.

It will be necessary to convert all of the individual files contained in the *.GZ (zipped) file into a single *.DEM that AERMOD will accept for processing. This can be accomplished using the SDTS2DEM utility program. Most commercially available software packages will likely have this conversion utility program built in. If the utility is not provided in the software, a small DOS-based utility program can be downloaded at any of the following sites:

After downloading, the SDTS2DEM program will need to be saved in the same directory as the unzipped DEM data. Navigate to the folder that contains the DEM data and SDTM2DEM program and double click on the SDTS2DEM executable. If done correctly, a DOS-based window will be displayed. When prompted, enter the first four numerical characters of the SDTS file names that is to be converted. Note that you should have a number of files in your directory with the same four leading digits such as 3931CATD, 3931CATS, 3931CEL0, etc. In this case, the Augusta DEM file is 3931.

Enter the output file name (the name must be 8 characters or less) of the DEM file that you wish to create (you do not need to provide an extension to the filename, as the *.DEM extension will automatically be assigned to the filename). The SDTS2DEM utility will then execute for several seconds. Once completed, a *.DEM file with the provided name should be successfully created in your directory.

At this point, it is best to verify that the converted *.DEM file just created is compatible with the modeling software being used.

TROUBLESHOOTING TIP: Certain modeling software will not recognize the *.DEM file, as *.DEM files created using the SDTS2DEM utility are designed to work in a UNIX (i.e., non-PC based) format. This is because the *.DEM file does not have record delimiters, which some DOS-based modeling software needs to correctly read the file. It may be necessary to run the CRLF (Carriage Return Line Feed) utility which adds a carriage return/line feed to each record of the *.DEM file to make it DOS compatible. Most commercially available software packages will likely have this conversion utility program built in.


Seamless elevation data can also be used in AERMAP in the same manner as individual DEM files. Using seamless data has several major advantages over using individual DEM files; seamless data eliminates any continuity problems at edges of individual DEM files, seamless data files can encompass a much larger area in one continuous GEOTIFF file, seamless data eliminates handling of multiple files, etc.

Seamless DEM data can be extracted for a user-defined domain at:

You can then choose the resolution of DEM data (1, 1/3 and 1/9 arc-second data) on the left side of the screen. The user can then use the on-line tool to select the DEM domain and download the data in the GEOTIFF (*.TIF) format for inclusion into AERMAP.


The AERMAP terrain preprocessor requires the user to define a modeling domain, which is defined as the area that contains all the receptors being modeled and is the geographic extent in which the effects of terrain will be considered. The domain must be a quadrilateral in the same orientation as the DEM data. It is important to note that the size of the domain can significantly affect the receptor height scales calculated because AERMAP determines the largest terrain height difference in the domain as a part of the process of determining height scales.

Anchor locations are used to anchor a local coordinate system (e.g., user location of 0, 0) to a UTM coordinate system. If you are using UTMs for all data (stacks, receptors, etc.), make the anchor location equal to the primary source location.

The modeling domain must be large enough so that the AERMOD receptor network will be both sufficiently detailed and extensive enough so as to fully represent the immediate surrounding terrain and the entire domain being modeled.

While the domain (which can span multiple DEM files) can be specified either in the UTM or the latitude/longitude coordinate system, MEDEP prefers that the domain be in UTM coordinates.


As an initial starting point, construct a grid centered on the source with 50 or 100 meter spacing out to a distance of 2 kilometers (km), aligned with the Universal Transverse Mercator (UTM) grid marks. From 2 km to 5 km, construct a grid with 500 meter spacing, and beyond 5 km, use a 1 km spacing. For calculating impacts in the near-wake and cavity regions of structures, a receptor spacing of no more than 10 meters is recommended in the immediate vicinity of the stacks and nearby buildings.

In certain cases, additional special-purpose receptors may be required by MEDEP-BAQ for inclusion in the analysis. Some examples of special-purpose receptors are:

Fenceline receptors at the closest occurrence of "ambient air", as defined by Chapter 116;
Receptors within the wake region of the controlling building;
Receptors at state/international borders;
Receptors in Class I areas; and
Receptors at any additional designated sensitive locations.

The above approach for selecting a receptor network must completely represent the surrounding terrain in the area of expected maximum impact of the proposed source, the area of maximum impact of nearby sources and the area where all sources will combine to cause maximum impact. It is possible that the applicant may need to identify these locations through a trial and error process. If in doubt about receptor spacing, consult with the project meteorologist as to what receptor spacing would be appropriate for the proposed modeling analysis.

During setup processing, AERMAP will check all of the sources and receptors coordinates to ensure that they lie within the specified domain, and therefore, within the area covered by the DEM file(s). If a receptor is found to lie outside the domain, or if the domain extends beyond the area covered by the DEM data, AERMAP generates a fatal error message, and further processing of the data is aborted.

Further guidance on the use of AERMAP can be found in the User's Guide For The AERMOD Terrain Preprocessor (AERMAP) (EPA, 1998).


AERMET processes commercially available or on-site meteorological data, representative of the modeling domain, for use in AERMOD. AERMET uses meteorological measurements of several boundary layer parameters to compute vertical profiles of: wind direction, wind speed, temperature, vertical potential temperature gradient, vertical turbulence (sigma-w) and horizontal turbulence (sigma-v).

AERMOD requires two types of meteorological data files which must be pre-processed by AERMET:

Surface met data file - (*.SFC)
Profile met data file - (*.PFL)

AERMET processes meteorological data in three stages:

  1. The first stage (Stage1) extracts meteorological data from archived data files and processes the data through various quality assessment checks.
  2. The second stage (Stage2) merges all data available for 24-hour periods (surface data, upper air data, and on-site data) and stores these data together into a single file.
  3. The third stage (Stage3) reads the merged meteorological data and estimates the necessary boundary layer parameters for use by AERMOD.
  4. AERMET then passes this data along for use in AERMOD in two files:

    A surface file (*.SFC) of hourly boundary layer parameters estimates. The surface file also contains the surface characteristics (noontime albedo, Bowen ratio and roughness length) of the area being modeled.
    A profile file (*.PFL) of multiple-level observations of wind speed, wind direction, temperature, and standard deviation of the fluctuating wind components.

At the present time, AERMET is designed to accept data from any for the following sources:

standard hourly National Weather Service ASOS data from the most representative site;
hourly on-site wind, temperature, turbulence, pressure, and radiation measurements (if available);
morning soundings of winds, temperature, and dew point from the nearest NWS upper air station.

When pre-processing the meteorological data using AERMET, users are usually faced with the lack of proper meteorological data.

AERMET minimally requires the following parameters to be present in the hourly surface met data file for succesful processing:

Year, Month, Day, Hour
Wind Speed
Wind Direction
Dry Bulb Temperature
Cloud Cover (tenths)

It is typically considered to be helpul to include any additional surface data that might be available.

Note that when one of the above parameters is missing, AERMET will still produce the *.SFC and the *.PFL files with missing indicators applied to all missing values (e.g., -99 for windspeed, etc).

AERMOD will typically run succefully with missing values in the surface file, however AERMOD will not produce concentration impacts (all concentration values will be 0.0) for missing data periods. A warning message will be displayed in the message section of the AERMOD main output file stating the number and percentage of missing hours.

For deposition calculations, the following additional meteorological parameters are required:

Station Pressure
Relative Humidity
Precipitation (required for wet deposition calculations)

AERMET will apply a default station pressure value of 1013.25mb if this parameter is missing.

Before beginning to process any meteorological data, it is very important to determine that the meteorological data is representative. Representative is defined as the extent to which a set of measurements taken at the collection site (met tower) reflects the actual conditions at the application site (source/facility). The collected meteorological data should closely mimic the conditions affecting the transport and dispersion of pollutants in the area of interest as determined by the locations of the sources/receptors being modeled. It is recommended to work in close consultation with the MEDEP project meteorologist when preparing a meteorological database for use in dispersion modeling.


In AERMET, one can specify monthly variations of three surface characteristics for up to 12 different contiguous non-overlapping sectors. These surface characteristics include are listed below. Each wind sector can have a unique noon-time albedo, Bowen ratio, and surface roughness value.

Noon-time Albedo (r) : the fraction of total incoming solar radiation reflected by the surface. Typical values range from 0.1 for thick forests to almost 1.0 for fresh snow.
Bowen Ratio (Bo) : the ratio of the sensible heat flux to the latent (evaporative) heat flux.
Surface Roughness Length (Zo) : the height at which the mean horizontal wind speed is zero . Values range from less than 0.001 meter over calm water surface to 1 meter or more over a forest or urban area.

In order to account for these parameters, the monthly surface characteristics will need to be calculated on a sector-by-sector basis for the measurement site by running the AERSURFACE program. AERSURFACE uses the 1992 USGS Land Cover Datasets to determine the land cover types for user-specified locations.

The AERSURFACE program, user's guide and other related programs can be downloaded at the SCRAM website.

More information on the 1992 Land Cover Datasets can be found at the USGS Land Cover Institute.

The NLCD92 data can be downloaded in two different forms:

complete files by state;
files for a user-specified domain of arbitrary size and location from a “seamless data server.”

AERSURFACE supports both forms of the NLCD92 data.


Per USEPA Guidance:

1. The determination of the surface roughness length should be based on an inverse distance weighted geometric mean for a default upwind distance of 1 kilometer relative to
the measurement site. For all analyses, divide the 1 kilometer ring into twelve 30 degree sectors (with the sectors extending from 0 - 30 degrees, 30 - 60 degrees, etc). Sectors are defined clockwise, as the direction from which the wind is blowing with north at 0°/360°.
2. The determination of the Bowen ratio should be based on a simple unweighted
geometric mean (i.e., no direction or distance dependency) for a representative domain,
with a default domain defined by a 10x10 kilometer domain centered on the measurement site.
3. The determination of the albedo should be based on a simple unweighted arithmetic
mean (i.e., no direction or distance dependency) for the same representative domain as
defined for Bowen ratio, with a default domain defined by a 10x10 kilometer domain
centered on the measurement site.

MEDEP-BAQ recommends that the surface characteristics be developed on a month-by-moth basis in the following manner:

Midsummer with lush vegetation
June through August
Autumn with unharvested cropland
Sepetmber, October
Late autumn after frost and harvest, or winter with no snow
Winter with continuous snow on the ground
December through March
Transitional spring with partial green coverage or short annuals
April, May

Further guidance on AERMET surface characteristics can be found in the AERMET User's Guide.

Interactive Sources

Once the results of the significant impact analysis have been established, MEDEP-BAQ will provide the applicant with a list of any sources that are to be included in the final modeling analysis and the model input data for these sources. This list will contain all data required for model input including source location(s), emission rates, stack parameters, and model-ready BPIP-PRIME input files or necessary building dimensions for the applicant to determine direction-specific building parameters.

Since the goal of the interactive source analysis is to capture the maximum impact from all sources combined, it is possible that the refined receptor grid used for the applicant alone will need to be modified (enlarged, refined, etc.) to accommodate emissions from the other sources included in the analysis.

NO2 Modeling Methodologies

EPA currently allows for a 3-tiered screening approach for modeling ambient NO2 impacts from point sources:

Tier 1 is performed by obtaining the maximum average concentration and assuming a total conversion of NO to NO2. If the concentration exceeds MAAQS, NAAQS and/or increment standards, a tier 2 analysis may be explored.
Tier 2 uses the Ambient Ratio Method (ARM) where the maximum average concentration (Tier 1 estimate) by an empirically-derived NO2/NOx value of 0.75 (national annual default). A ratio differing from 0.75 may be used if the source provides sound data that is representative of the area where the maximum annual impact occurred, although this will likely involve USEPA approval on a case-by-case basis. If the concentration exceeds MAAQS, NAAQS and/or increment standards, a tier 3 analysis may be explored.
Tier 3 uses the Plume Volume Molar Ratio Method (PVMRM), a non-regulatory default option available in AERMOD, to model the NO-to-NO2 conversion chemistry in exhaust plumes. The PVMRM is recognized as a detailed Tier-3 screening approach for NO2 modeling as outlined in USEPA’s memorandum from Tyler Fox (OAQPS) to EPA Regional Air Division Directors, "Applicability of Appendix W Modeling Guidance for the 1-hour NO2 National Ambient Air Quality Standard" (June 28, 2010).

When exhaust plumes from sources are emitted, they usually contain mostly nitric oxide (NO). As the plume is transported downwind, the NO is converted to NO2, typically through the interaction with ozone (O3):

NO + O3→ NO2 + O2

This chemical transformation, know as titration, is considered to be the primary mechanism for the conversion of NO to NO2.
The PVMRM option calculates the moles of NOx (NO and NO2) using the stack emission rate and travel time of the plume. The moles of O3 are determined using hourly representative background ambient O3 concentrations and plume size. The NO2/NOx ratio is then calculated and used to scale the predicted NOx concentrations within AERMOD.

Detailed information on the application of the PVMRM method is available from MEDEP.

Use of the Tier-3 PVMRM method is considered to be a non-regulatory default option and requires written approval by MEDEP and USEPA on a case-by-case basis.

Additional Class II Analyses

Federal Guidance and Chapter 115/140 require that any new major source or sources undergoing a major modification provided additional impacts that would occur as a direct result of the general, residential, commercial, industrial and/or other growth associated with the construction and operation of the source.

In addition, an analysis of impairment of visibility, soils and vegetation that would occur as a result of the source is also required. Visible emissions from the source are typically minimized by controlling the emissions through the implementation of BACT (Best Available Control Technology) for new sources or modifications or Best Practical Treatment (BPT) for existing sources.

Further guidance relating to these analyses are provided in A Screening Procedure for the Impacts of Air Pollution Sources on Plants, Soils and Animals (1980), the draft PSD/NSR Workshop Manual (1990).