Addressing air quality problems with genetic algorithms: A detailed analysis of source characterization

The purposes for modeling air contaminants have evolved, and the models themselves have co-evolved to meet the changing needs of society. The original need for air contaminant models was to track the path of pollutants emitted from known sources. Therefore, the initial purpose of the models was to track and estimate the downwind transport and dispersion (T&D). Since dispersion results from turbulent diffusion, which is best modeled as a stochastic process, most models for the dispersion portion are based on a Gaussian spread. Because many environmental problems have their sources in a region that is far from the impact, there came a need to identify remote sources of pollution. For instance, the acid rain problem that was highly studied in the 1980s was widely thought to be caused by upwind polluters. Power plant emissions in the Ohio Valley were blamed for acid rain in New York and New England. To test this conjecture, receptor models were developed. This type of model begins with monitored pollution concentrations and back calculates the sources. Some models of this type were based on a backward trajectory analysis while others separated out the mix of chemical species present in the sample and computed likely sources given knowledge of the species composition of the potential sources. These models pointed to the Ohio Valley for the source of the acid rain precursors. Receptor models are still popular for attributing pollutants to their sources.