Toward the Improvement of Aircraft-Icing Forecasts for the Continental United States

Abstract An automated procedure is developed for detecting and forecasting atmospheric conditions conductive to aircraft icing over the continental United States. The procedure uses gridded output from the Nested-Grid Model, and is based on the manual techniques currently in use at the National Aviation Weather Advisory Unit in Kansas City, Missouri. Verification of the procedure suggests forecasting performance on par with the human forecasters. Unfortunately, efforts at more-rigorous performance analysis are hindered by the inadequacies of the verification database, which consists of pilots’ subjective reports of airframe ice buildup. In general, no-ice conditions are not reported. The physics of aircraft icing are reviewed, and the current manual techniques are discussed. The automated procedure provides an infrastructure for implementing incremental improvements in the algorithm as observations and numerical models improve.