Warm-season precipitation behaves differently in intensity, coverage, lifespan, and daily cycle depending on large scale weather patterns. Without separating large-scale conditions when evaluating a model’s performance, biased conclusions could be drawn, which could negatively impact the model configuration and development. Researchers at the Pacific Northwest National Laboratory and the University of Arizona classified the major large-scale patterns associated with heavy warm-season precipitation over the Great Plains. Using these classifications, they evaluated the performance of the Weather Research and Forecasting (WRF) model in simulating precipitation under different weather patterns. Their research identified the large-scale conditions under which WRF performs the best and the worst.
Evaluating model performances under different weather patterns provides a unique way for the community to identify model problems and development focus. The developed database for six types of weather systems over the Great Plains in warm seasons serves as a reference dataset for model evaluation and a useful tool for weather forecasting with WRF. The gained an understanding of WRF biases will help future development of the model.
Researchers used a competitive neural network known as the self-organizing map (SOM) to identify the major weather patterns for warm-season rain over the Southern and Northern Great Plains from 2007 to 2014. Precipitation from Weather Research and Forecasting (WRF) simulations run by the National Severe Storms Laboratory was evaluated against National Centers for Environmental Prediction Stage IV observation from two perspectives: the dominant large-scale pattern (extratropical cyclone vs. subtropical ridge) and the associated convective activities (light vs. moderate vs. severe). In general, WRF performs the best in simulating the warm sector precipitation under mid-latitude, low-pressure areas classified as an extratropical cyclone and performs the worst in simulating intense, shorter duration convective precipitation at the periphery of a subtropical ridge. The persistent better performance is strongly tied to stratiform-dominated light convection events, whereas the missing nocturnal precipitation remains the major issue for the simulation of severe convection.