14.4 Dust Storms: Emission Factors


Previous

HOME


Next

The use of climatological dust emission factors did not provide realistic results for the Salt Lake City example case. There are several approaches that can be taken to address this problem. If monitoring data are available, it is possible to compute the emissions factors required to match the measurements, as was done in an earlier section. For this section, we will examine the climatological data and select values that are more representative of maximum emissions that were observed in the measurements.

  1. The emission files for the continental U.S are based upon monthly averages over a four year period. Extracting all the emission factors for March and April and sorting them in ascending order, the cumulative distribution shows that we can expect a maximum value of 2.0E+06 g/m. Assume this represents the largest possible rate over the U.S. and then we set this value to each emission point in the Sevier Lake region. Each KA (dust density times emission area) value in the previous CONTROL file is replaced with the maximum value. This modified CONTROL_dust4.txt should be retrieved from the \Tutorial\dust directory. Save the changes and run the model.

  2. Once the run completes go right to the Convert to DATEM step. The model statistical results show a higher correlation (0.72) and the model under-prediction is now less than a factor of four. The scatter diagram shows some correspondence over a wide range of air concentrations, suggesting that adjusting the emission factors can provide a method to calibrate model performance to match current surface conditions in the dust source regions. In this case, perhaps the threshold friction velocities also need to be adjusted.

The dust storm emission calculations are very sensitive to the emission parameters and the module should not be treated as a black box. However, using the revised module with spatially varying characteristics, it may be possible to "calibrate" the model, refining the emission factors using measured data, and then using the calibrated model for predictions for a specific location..

5 m 29 s