Machine Learning and Numerical Weather Prediction Prediction of Aircraft Turbulence and Icing: Operations and Verification

Over the last year, research into the application of machine learning to create aircraft icing and turbulence linear regression algorithms has led to an operational product now put into use. The output of this product is continually being observed and verified to see how well it is working and, just as importantly, where improvements need to be made. A constant influx of aircraft observational data and numerical weather model output allows periodic retraining of the machine learning model. Real world turbulence case studies also provide valuable insight into where gaps in model and forecast data need to be studied and addressed. Icing and especially turbulence cases are newsworthy and highlight the challenge and necessity of continual improvement to ensure flight safety.