Advancements in Combining Datasets for In-Flight Icing Diagnoses

Advancements in numerical weather prediction (NWP) models continue to enhance the quality of in-flight icing forecasts and diagnoses. When diagnosing current in-flight icing conditions, observational datasets are combined with NWP model output to form a more accurate representation of those conditions. Surface observations are heavily relied upon to identify cloud coverage and cloud base height above observing stations. One of the major challenges of using these point-based or otherwise limited observations of cloud properties is extending the influence of the observation to nearby points on the model grid. An alternate solution to the current method for incorporating these point-based observations into the in-flight icing diagnoses was developed.

The basis for the new method is rooted in a concept borrowed from signal and image processing known as dithering. We leverage the confidence associated with a truth dataset (METAR observations) to decide how much to rely on NWP-derived METAR fields. Initial results show that incorporating NWP-derived data to assist in areas with sparse METAR observations creates an icing diagnosis that is visually more realistic. Circles of icing surrounding the point- based METAR observations are nearly eliminated in favor of a smoother cloud and resultant in-flight icing diagnosis.