Multi-Model Ensemble Wake Vortex Prediction

Abstract

Wake vortices generated by aircraft as a consequence of lift constitute a potential danger to following aircraft. Real-time predictions of wake vortex behavior can contribute to avoiding dangerous situations posed by vortices hovering in the glide path, drifting to parallel runways or living longer than expected. Employing wake vortex models developed by DLR (D2P) and NASA (APA 3.2, APA 3.4, TDAWP 2.1) the capability of ensemble approaches to outperform individual wake vortex predictions is evaluated. In particular a Multi-Model Ensemble approach, called Reliability Ensemble Averaging, is examined which uses individual model weights that depend on the prediction quality of each model. Field measurement data collected during campaigns at the airports of Munich, Frankfurt and special airport Oberpfaffenhofen serve to train and to evaluate the ensemble predictions. It is shown that a well balanced Multi-Model Ensemble may outperform individual model predictions.