The various pollutants are assumed to be passively transported by the wind. The dispersion graphic represents the upwind locations of air parcels that were sampled at the monitoring location. Passive particles are released continuously from the sampling location and their position is integrated in the upwind direction using a 4-dimensional wind field, that varies with horizontal position, height, and time. In addition to transport by the wind, the parcels also disperse through atmospheric turbulence. The calculation is done using the ADAPT dispersion model and the resulting pattern shows the likely source regions for the pollutants measured in the previous six hour period. More detailed information is available about the ADAPT model.
The computations are made with the most current meteorological data using forecast data fields from NOAA's NAM model. The forecast data are available at 1-hour intervals at 12 km resolution over most of the U.S. on the sigma-pressure hybrid coordinate system. An alternative source region calculation using the observations from a nearby surface meteorological site is available for evaluation. The observational data from the single site and near surface level have been interpolated to a 3-dimensional time-varying field to facilitate the upwind dispersion calculation.
The backward potential source patterns are computed for each 6-h period arriving at the monitoring location. The sampled air could have come from any upwind point. The values represent dilution factors and can be used to make a more quantitative estimate of the emissions required to match a measurement at the monitoring location. For instance the dark blue region in the example calculation represents a dilution (D) value of 2x10-12 mass per cubic meter. If for example, we had a measurement (M) of 20 μg per cubic meter, then the emission rate from any one blue square would need to be (M/D) 10,000 kg. This would only be true if the emissions only came from one grid cell, such as a power plant. However, for more regional pollutants, where there are multiple sources, that emission amount would be divided amoungst all the sources. The pattern does not suggest probability, in that source regions with higher values are more likely. All squares are equally likely. Lower dilution values, for the off-centerline grid cells, only means that a higher emission rate is required to achieve the measured value.
In the case of an area source, rather than a point source, the emissions would be spread out over all non-zero grid cells. In this example, the concentration grid had a resolution of 4 km, three times finer than the meteorology grid, which resulted in approximately 500 grid cells with values greater than zero. Cells with values less than 20 are not shown in the illustration. Using the same computational approach, except now the area wide emission rate would be M/ΣD, where the sum (1x10-10) is taken over all grid cells, resulting in a value of (20 μg 10-9 kg/μg) / 1x10-10 or 200 kg; which when divided over the area of emissions (500 cells 4x4 km square) gives an emission value of 0.025 kg km-2 or a rather modest value of 25 μg m-2.