A Monte Carlo study of uncertainties in benzene and 1,3-butadiene concentrations calculated by AERMOD and ISC in the Houston ship channel area

In the interest of better understanding air quality studies of toxics in urban areas, a Monte Carlo (MC) probabilistic uncertainty study was conducted for a 15 km by 15 km domain centered on the Houston Ship Channel. This study, sponsored by the American Petroleum Institute (API), derives from a similar MC uncertainty study sponsored by the EPA and carried out over a much broader domain in the Houston metropolitan area using straight-line Gaussian plume models (i.e., ISC3ST and AERMOD). The current study follows the EPA 1996 scenario and much of their methodology, including the emphasis on average-annual concentrations at receptors defined as the centroids of population census tracts. The focus of the current study is on uncertainties in ISC3ST and AERMOD predictions of average-annual concentrations of benzene and 1,3-butadiene, due to uncertainties in emissions and meteorological inputs and dispersion model parameters. The emissions uncertainties were estimated to be about +/- a factor of three (i.e., covering the 95 % range) for 21 benzene emissions categories and for 13 1,3-butadiene emissions categories. The uncertainties in meteorological inputs (such as wind speed and stability) and dispersion model parameters (such as σz) also were estimated for the 95% range. ISC3ST (in urban mode and in mixed urban-rural mode) and AERMOD were run 100 times in MC mode for each pollutant, using random and independent perturbations of all inputs, in order to estimate 1) the total uncertainty of the annual averaged concentrations, and 2) the inputs with uncertainties that are most strongly correlated with uncertainties in predicted concentrations. The results of the MC runs with ISC3ST and AERMOD show that the 95 % range in predicted annual averaged concentrations is about +/- a factor of two or three for both benzene and 1,3-butadiene, with little variation by model. The input variables whose variations have the strongest effect on the predicted concentrations are mobile sources and some industrial sources (dependent on chemical), as well as wind speed, surface roughness, and σz. In most scenarios studied, the uncertainties of the emissions input group contributed more to the total uncertainty than the uncertainties of the meteorological/dispersion input group.