Forecast verification: A dispersion modeling perspective

Abstract

The Environmental Protection Agency currently uses AERMOD, an air quality dispersion model to aid in the forecasting of transport and dispersion of air pollution for the U.S. Typically, NWS-ASOS observations (post-processed by EPA-AERMET model) are used as input to the AERMOD model. This traditional framework of running a dispersion model based on point observations is quite problematic from a variety of theoretical standpoints (e.g., lack of representativeness of meteorological data). An alternative viable framework would be to use prognostic meteorological models in conjunction with AERMOD. Indeed, contemporary research shows that the use of prognostic models as a substitute for NWS-ASOS observations alleviates some of the longstanding dispersion modeling problems, but at the same time creates new concerns. I will elaborate on several questions that need to be adequately addressed before prognostic models can be reliably utilized in operational dispersion applications. Most of these questions are rooted in prognostic models’ (in) ability to accurately represent the boundary layer variables of interest to the dispersion modeling community (e.g., wind speed, wind direction, temperature). I will compare the potential of a new generation prognostic meteorological model called the Weather Research and Forecasting (WRF) model in capturing wind speed variable versus data from the West Texas Mesonet by statistical analysis for verification. One year of ARW WRF output is analyzed. The WRF is a 36/12 km two-way nested run using the YSU PBL scheme. With use of innovative strategies for verification of complex spatio-temporal forecast fields and novel verification measures will make this study distinct.