Typical performances of mesoscale meteorology models

Abstract

Reliable meteorological model results are one pre-condition for a good air quality simulation. The achievable quality of the meteorological information determines how accurate a concentration simulation can be. Many meteorological services as well as research institutions perform model evaluations on a routine basis, but the outcomes are not always published in refereed journals and little is known on typical model performances. This paper summarizes results of quantitative model evaluations that were published in refereed journals by statistically analyzing the published values for bias, root mean square error, rmse, as well as correlation coefficient, r. The 50 percentile of the quality measures rmse and r is used as threshold to derive typical performances. For r the 50 percentile is 0.47, 0.62, 0.89 and 0.87 for wind direction, wind speed, temperature and specific humidity, respectively. While bias values are small compared to their average values, rmse values are large.