Abstract:

 The meso-scale numerical weather prediction system in the middle region of China has been updated recently with WRF model. In order to evaluate the prediction performance of the updated forecasting system, 24 and 48 hours precipitation prediction during the flood season of 2013 were analyzed in terms of TS score, forecast accuracy rate, missing alarm rate, false alarm rate, bias and ETS score. The results show that for 24 hours forecast, the distribution of daily mean precipitation rate, precipitation intensity and center position are closer to those of observation, while those from the 48 hours forecasts are significantly overestimated. Precipitation verification during the flood season suggests that most precipitation forecasts are overestimated, and TS and ETS scores decrease gradually and ETS score is gradually close to TS score with the increase of precipitation grade. Monthly precipitation verification suggests that accuracy rate of rainy and shine weather forecast is in a positive correlation with rainy day rate. Verification of forecasts from two models (WRF and GRAPES_Meso) shows that both have good prediction scores. It is worth noting that forecast results by the WRF are getting better with the increase of precipitation grade. In general, precipitation prediction products of this model are of a certain references to the precipitation forecast.

Key words: Verification, Numerical weather prediction model, Precipitation prediction