Aircraft Icing Algorithms Applied to U. S. Navy Numerical Model Data: AVerification Study

摘要

The results from a verification with pilot reports of four aircraft icingalgorithms applied to Navy global numerical model data are presented. Significant differences in forecast performance among algorithms are closely related to differences in temperature and moisture thresholds utilized to infer icing. Near 850 mb (-5000 ft), three of the icing routines correctly forecast over 70% of observed icing occurrences, and obtained Hanssen and Kuipers skill scores (difference between hits and false alarms) in excess of 0.5 (0, random performance; I, perfect skill). Statistical tests indicated that differences in skill scores as a function of forecast lead time (0 to 24 hr) were generally not significant. The use of higher vertical resolution data was found very important for enhanced icing prediction performance. Overall results indicate that the ability of icing routines to differentiate icing type and intensity based on temperature, moisture and stability criteria is clearly limited. In terms of forecast skill and computational efficiency, algorithm comparisons indicate that the icing routine currently used as operational guidance at the National Centers for Environmental Prediction Aviation Weather Center would be a good selection for the Naval Research Laboratory's aviation support product suite.