Statistical Methods for Diagnosing Aviation Hazards and Their Likelihood from Numerical Weather Prediction Models: Past, Present, and Future (Invited Presentation)

While weather-related aviation hazards have existed since the beginning of aviation travel, the ability to forecast aviation weather to diagnose and therefore mitigate these hazards has greatly improved but challenges still exist today. Advances in Numerical Weather Prediction (NWP) models have in general outpaced advances in aviation forecasting, partially due to the fact that early NWP models did not provide direct model output of aviation hazards such as ceiling heights, visibility, convection, and lightning. More recently, modelers have developed algorithms to provide NWP output of these aviation hazards, and while this new type of model output is useful, the techniques are still in flux and opportunities for improvement still exist.

Throughout the years, a variety of methods have been employed to statistically post-process the NWP direct model output to assist with model interpretation and forecasting. The National Weather Service (NWS) Meteorological Development Laboratory (MDL) has worked for 50 years on techniques to statistically post-process the NWP direct model output to provide added value to the NWP output. MDL’s Model Output Statistics (MOS) products were developed to aid in statistically interpreting NWP output to produce guidance of sensible weather to aid forecasters, and the Localized Aviation MOS Program (LAMP) was designed with a focus on providing statistical guidance to aid in aviation meteorology forecasting. The NWS evolution to gridded forecasts via the National Digital Forecast Database, the transition to Digital Aviation Services, and the inception of the National Blend of Models to provide a starting point for the official NWS gridded forecasts are the latest NWS initiatives to provide gridded aviation forecasts to help mitigate aviation hazards at any point in the area of interest.

This talk will describe the history of statistical methods for diagnosing aviation hazards and their likelihood from NWP models. A summary of the current state of the science of the statistical post-processing methods will be provided and current challenges and plans for mitigating them will be discussed. A discussion of the future of statistical post-processing for aviation meteorology, including the expected future emphasis on probabilistic forecasting for better aviation decision support, will be provided.