A 99-Day Assessment of the Weather Research and Forecasting Model over the Southwest United States

An assessment was conducted over a 99-day period during winter over complex terrain to evaluate the accuracy of forecasts produced by the Advanced Research version of the Weather Research and Forecasting model (WRF-ARW). The Army Weather Running Estimate-Nowcast Real-Time (WREN_RT) system is a scripted system that provides forecasts by executing WRF-ARW and its preprocessors used to produce the WRF-ARW forecasts for this study. WREN_RT aims to provide forecasts for ingestion into decision aids that produce knowledge products for Warfighters. These products include the 2-D distribution of weather phenomena that can impact Army missions and systems. Two methods of spatial verification were used on the WRF output to compute skill scores for a range of neighborhood sizes and thresholds. The more advanced method, called Fuzzy or neighborhood verification, was used to compute the Fractions Skill Score to augment the scores and error statistics produced by the traditional categorical method. For ground-truth data, both methods used a new set of gridded observations, called the UnRestricted Mesoscale Analysis, which was recently evaluated for such use. The results of the assessment showed that WRF scored very well for most thresholds, but not so well for others.