Performance evaluation of AERMOD, CALPUFF, and legacy air dispersion models using the Winter Validation Tracer Study dataset

Abstract

The performance of the steady-state air dispersion models AERMOD and Industrial Source Complex 2 (ISC2), and Lagrangian puff models CALPUFF and RATCHET were evaluated using the Winter Validation Tracer Study dataset. The Winter Validation Tracer Study was performed in February 1991 at the former Rocky Flats Environmental Technology Site near Denver, Colorado. Twelve, 11-h tests were conducted where a conservative tracer was released and measured hourly at 140 samplers in concentric rings 8 km and 16 km from the release point. Performance objectives were unpaired maximum one- and nine-hour average concentration, location of plume maximum, plume impact area, arc-integrated concentration, unpaired nine-hour average concentration, and paired ensemble means. Performance objectives were aimed at addressing regulatory compliance, and dose reconstruction assessment questions. The objective of regulatory compliance is not to underestimate maximum concentrations whereas for dose reconstruction, the objective is an unbiased estimate of concentration in space and time. Performance measures included the fractional bias, normalized mean square error, geometric mean, geometric mean variance, correlation coefficient, and fraction of observations within a factor of two. The Lagrangian puff models tended to exhibit the smallest variance, highest correlation, and highest number of predictions within a factor of two compared to the steady-state models at both the 8-km and 16-km distance. Maximum one- and nine-hour average concentrations were less likely to be under-predicted by the steady-state models compared to the Lagrangian puff models. The characteristic of the steady-state models not to under-predict maximum concentrations make them well suited for regulatory compliance demonstration, whereas the Lagrangian puff models are better suited for dose reconstruction and long range transport.