On the Performance of the AFWA version of the PSU/NCAR MM5 model for short-range forecasting in Alaska, the Western Arctic and North Pacific

The Air Force Weather Agency (AFWA) has adopted the Penn State/NCAR MM5 modeling system for regional numerical weather prediction over a large number of regions, or theatres, where there are either military actions in progress or important bases where exercises are conducted on a routine basis. One of these theatres of interest includes Alaska and adjacent areas of the Arctic Ocean, the North Pacific Ocean, Canada and Siberia. Currently forecasts are conducted for this theatre four times daily on a 45 km grid encompassing all these areas, while there are also twice daily forecasts on a 15 km nested grid which focuses on mainland Alaska.

We have been receiving AFWA Alaskan theatre MM5 forecast output on a fairly regular basis since April 2000 and have using the AFWA output as input to two model-based algorithms to diagnose and predict in-flight icing over Alaska. During this period we have been examining the performance of the model forecasts through a series of standard skill scores including root mean square error, average absolute error, S1 score and average bias. In addition, we are using pilot reports in tandem with contingency table-based statistics to evaluate the performance of the in-flight icing forecast algorithms.

In our presentation, we will briefly describe the model data, icing algorithms and measures of skill. Then we will discuss in detail the model and icing algorithm performance from a domain-averaged point of view and conclude with an examination of several interesting case studies to highlight strengths and weaknesses of the AWFA version of the MM5 as applied to the Alaskan theatre.