Development of algorithms to diagnose and predict in-flight icing conditions has progressed steadily over the last five years. Evidence of this progress is provided by a verification study comparing the performance characteristics of numerous icing algorithms that have been developed during this period. These algorithms include the RAP, Stovepipe, Satellite, and NNICE algorithms, as well as a new diagnostic algorithm known as the Integrated Icing Diagnosis Algorithm (IIDA), which incorporates many features of the other algorithms. AIRMETs, the operational icing forecasts issued by the National Weather Service's (NWS) Aviation Weather Center (AWC), also are included in the comparison.
Data for the comparison were collected during winter 1997-98, from December through March. The icing algorithms utilized output from the RUC-1 model. Data for a total of 110 forecasts (including three issue times on 36-37 days) are included for each algorithm. Pilot reports (PIREPs) of icing are used to evaluate the forecasts. Both the cities-based and national-scale algorithm verification approaches are applied to evaluate the algorithms and AIRMETs. In addition to overall verification measures, day-to-day variability in the statistics is evaluated. Finally, the algorithms which produce continuous ouptut values (e.g., ranging from 0 to 1 rather than being dichotomous, Yes/No values) are verified and compared over a wide range of thresholds used to identify a Yes forecast.
Results of the comparison indicate that the algorithms have improved steadily with time, so that in many respects the performance of current algorithms is similar to the performance of the operational forecasts. In particular, the newer algorithms have similar Probability of Detection and slightly smaller Impacted Volume compared to the AIRMETs.
The 8th Conference on Aviation, Range, and Aerospace Meteorology