Statistical Analysis of a Radar-Based Icing Hazard Algorithm

In-flight icing is a contributing factor to many aviation accidents, and the reliable detection of this hazard is a major concern to aviation safety. The National Weather Service is in the process of upgrading the NEXRAD S-band radar network to have dual-polarimetric capability. Several recent studies have shown that the aggregate characteristics of dual-polarimetric radar measurements are useful in distinguishing the regions of clouds containing super-cooled liquid and mixed phase from those with ice only. This has led to the development of a prototype polarized S-band Icing Hazard Level algorithm (IHLA). Using the output from the IHLA for a statistically significant number of icing and non-icing cases, the authors will compare the output to several other independent sources of in-flight icing information. These independent sources include:

• The NASA Icing Remote Sensing System depends on a Radiometrics Corporation 23-channel radiometer to derive temperature profiles and total integrated liquid water amounts. In addition, the system utilizes a Vaisala Laser ceilometer and a Metek K_a-band radar to define cloud base and top heights. An algorithm based on expected cloud structure, as inferred by these measurements, is used to determine the distribution of liquid water content within the cloud.

• Icing-related pilot reports are voluntary accounts and are our only means of in-situ diagnoses of actual atmospheric conditions encountered by aircraft in the absence of expensive icing research flights or specially instrumented fleet aircraft.

• The Current Icing Product combines visible and infrared satellite imagery, radar reflectivity, lightning observations, pilot reports and standard ground-based weather observations with numerical weather prediction model output to produce a gridded, hourly, three-dimensional representation of icing probability and severity.

The authors will explore how the S-band freezing drizzle, high differential reflectivity and mixed-phase icing modules of the IHLA behave under varying icing conditions. Quantitative analysis including probability of detection, percent of radar volume warned on and strengths and weaknesses of the algorithm will be discussed. Qualitatively, the input and output fields will be examined to determine how they match the conceptual models of in-flight icing environments.