Verification of forecasts against analysis - some problems Ulrich Damrath
Overview: Cross verification against analysis Verification of GME and ECMWF forecasts against own analysis and against analysis of the other centre Verification of operational forecasts of GME and experimental forecasts of GME Verification of COSMO-EU forecasts against own analysis and against analysis of ECMWF for shortest ranges Verification for different aggregation types Use of area means Use of different climatologies UD
Verification of GME and ECMWF forecasts against the own and the other analysis (one of the best cases) UD
Verification of GME and ECMWF forecasts against the own and the other analysis (one of the worst cases) UD
Verification of COSMO forecasts against the own and the analysis of ECMWF UD
Conclusions: Verification against analysis may lead to useful results if: Forecast times are greater than about 24 hours. Analysis schemes of the models to be compared have nearly the same properties. Otherwise: Numbers are calculated - but no results! Results concerning the use of averaged fields could be expected. When using a new climatology changes in long term time series concerning the anomaly correlation coefficient should be mentioned an interpreted. UD