A Case Study Comparison of the Performance of Operational Prediction Models Used in the United States

Abstract

The performance of the National Weather Service, Air Force and Navy large-scale numerical prediction models was studied for the case of a relatively small-scale but important weather-producing frontal system in the Midwest over the period 1200 GMT 5 October 1974 to 0000 GMT 7 October 1974. Forecasts were analyzed both for the operationally important parameters of precipitation, surface pressure and 500 mb heights and for such key diagnostic parameters as vertical motion and thermal and vorticity advection. Results showed the importance of resolving small synoptic-scale features in the initial conditions as well as the role of model resolution, basic dynamics formulation, and planetary boundary layer representation in the forecasts. There was a wide range of performance among the four models. The National Weather Service (NWS) Limited Fine Mesh Model clearly gave the best 24 h forecasts, compared to all the other models including the NWS Primitive Equation Model.