A Distributed Computing Solution For CALPUFF
Free (open access)
K.-H. Yau & J. Thé
This paper demonstrates the skills, knowledge, and resources required for completing a refined MM5-CALPUFF analysis for a regional air quality study. The dispersion over the entire country using five years of MM5 meteorological data necessitates large scale computation. To assess additional computing resources, an intranet system developed by Lakes Environmental has been utilized to take advantage of idle processing cycles of all personal computers available in an organization. The distributed computing solution requires the modelling task be divided into a series of monthly calculations for each source. As partial results from participating computers are returned, they are summed to yield the final total impact in post processing. Further reduction of computational effort is achieved through application of advanced modelling techniques. Since nonlinearity of the model is insignificant for emissions involving only sulphur dioxide, as it is the case for our project, the final solution can be scaled linearly with respect to the source strength. Grouping of identical sources may also reduce some of the computation, although it is not utilized for the project. By application of innovative distributed computing solution and advanced modelling techniques, we have reduced the project life to one month, instead of one year if it were executed in a straightforward manner. Keywords: CALPUFF, MM5, distributed computing, modelling, dispersion modelling. 1 Introduction Recently, regulators in the U.S. have been frequently requesting more permit applicants to perform Class I area impact analyses, long range transport of pollutants, visibility studies for best available retrofit technology (BART), and
CALPUFF, MM5, distributed computing, modelling, dispersion modelling.