Sensitivity Analysis of Meteorological Inputs in a Gaussian Long-Term Dispersion Model

摘要

Meteorological data are a critical input for ISC-based Gaussian air quality models. We analyzed the effects of using 1990 Hourly United States Weather Observations (HUSWO) and 1996 International Surface Weather Observations (INSWO) databases on the annual average concentrations predicted by an ISC-based dispersion model. The 1990 and 1996 meteorological inputs differ in three respects: the methods in which the data were obtained, the number of stations with sufficient data, and, differences in meteorological conditions between 1990 and 1996. The model requires three types of data inputs: emissions, receptor locations, and meteorology. We reproduced the results of the original ISCLT-based study but with 1996 meteorological data. Because the same emissions input and fixed receptor locations were used for both model simulations, differences in annual average concentration estimates between the model simulations are entirely attributable to the differences between the meteorological inputs for these two years. Due to differences in reported variables, number and location of valid stations, and meteorology itself, statewide annual-averaged concentration estimates were found to vary from –16.9% to 84.4% in 1996 as compared to 1990. Even greater variability in modeled concentrations is found at the receptor level, thereby demonstrating the importance of meteorology in ISC-based dispersion models.