Comparison and validation of two high-resolution weather forecast models at Frankfurt Airport

Abstract

In recent years the 'Nowcasting Wake Vortex Impact Variables' model NOWVIV has been developed at the Deutsches Zentrum für Luft- und Raumfahrt, DLR, to forecast weather parameters in airport environments. The German Meteorological Service, DWD, employs his COSMO-DE model (COnsortium for Small scale Modelling-DEutschland) for operational forecasts in Germany. A systematic comparison of model output from NOWVIV and a derivate of COSMO-DE, named COSMO-FRA, is presented. Both models are centred at Frankfurt Airport with horizontal resolutions of 2.1 km and 2.8 km, respectively. In the DLR Project Wetter & Fliegen the COSMO-FRA model will replace the NOWVIV model and become a key component in the future rapid update cycle for adverse weather predictions at the airports of Frankfurt and Munich. The forecast vertical profiles of runway crosswind, head/tail wind, temperature, and turbulent kinetic energy (TKE) are validated against Wind and Temperature Radar (WTR/RASS) measurements operated by the German Air Navigation Service Provider, DFS, within a 40 day period during fall 2004 and a 60 day period during winter 2007. Model and WTR output is provided every 10 minutes. In general it was found that the predictions of both models yield similar skills based on the root mean square error (RMSE) and mean bias statistics of the crosswind as well as the false alarm-rate (FAR) statistics in forecasting different crosswind thresholds. The RMSE of crosswind between ground and 1600 m altitude ranges between 2.2 and 3.0 ms-1 for NOWVIV in fall and winter. For COSMO-FRA this error ranges between 2.0 and 2.5 ms-1 during winter and 2.5 and 3.5 ms-1 during fall, respectively. The FAR for exceeding a crosswind threshold of 3 ms-1 is about 23 % in fall and 17 % in winter for both models.