ALADIN and WRF forecasting evaluation using satellite imagery

Abstract

Simulated satellite images are a good indicator of the state of the atmosphere described by the fields predicted by numerical weather prediction (NWP) models. Therefore, in order to control NWP operational models used by the Algerian meteorological service, especially over the desert region, Meteosat Second Generation (MSG) simulated images of brightness temperature (BT) were generated from ALADIN (Aire Limitée Adaptation Dynamique Development International) and WRF (Weather and Research Forecasting) outputs using the Radiative Transfer for TIROS Television and Infrared Observation Satellite Operational Vertical Sounder (RTTOV9) model. As reference data, MSG images were used to compute certain deterministic and probabilistic statistical parameters. This version of the RTTOV model assimilates cumuliform clouds, stratiform clouds and those of upper levels, such as cirrus. This comparative study shows that WRF reproduces BTs well where they exist, but raises too many false alarms for very cold BTs with values of bias around 1. The number of false alarms greatly affects the quality of Heidke skill scores (HSS), unlike the ALADIN model, which reveals fewer false alarms but detects events less well. With low values of bias for the lowest temperatures, ALADIN HSS scores are better than those of WRF for the lowest BTs. The double-penalty impact is slightly lessened for local and convective cloud by about 10% with WRF compared with ALADIN.