Statistical Downscaling Study Of The Surface Temperature Forecast In The Northern Hemisphere Using The TIGGE Data

Based on the ensemble mean outcomes of the1-7-day ensemble forecasts of the surface temperature in the Northern Hemisphere during the summer of2007which are provided by ECMWF, JMA, NCEP and UKMO in the TIGGE data archive, the statistical downscaling forecast was investigated by using the interpolation, linear regression in conjunction with multimodel superensemble (SUP), bias-removed ensemble mean (BREM) in order to provide technical references of fine forecasting for operational meteorological centers. The trial is designed to provide some method to realize the fine weather forecast.After using downscaling of each single center, the RMSE and AE of each time of1-7day forecast limitation has a significant improvement. And more significantly improved the skill of the surface temperature forecast. After the introduction of the ensemble forecasting of SUP and BREM, the effect of statistical downscaling prediction has been further improved, the forecast skill of D-SUP is better than the skill of D-BREM.After using respective downscaling for each centers, the average RPS score of seven times has a more significant decrease, the techniques of probability forecasting can be improved, the average improvement percentage of RPS of each forecast time up to43%, and the improvement percentage of score was also slow decreasing with the extending of each forecast time compared with the model prediction of original center. Each order BS score is revised after downscaling, and the improvement scale is different of each order. After introduction of a set of D-S-SUP and building a "POOL-Man" probabilistic forecasting system (D-S-POOL) to make a downscaling probability forecast of surface temperature of East Asia, the probability forecast skill has been further improved.For the four single-centers, after using the method of LAF, the BS score has a certain level improved of1-6day forecast time. As far as downscaling program of D-S-SUP, after using the LAF method, the forecasting technique has been further improved, especially for1-3day forecast time.Using the intensive station in Hunan Province as a live field to do the fine weather forecast, the average ACC score by using the D-SUP method is up to0.75on the1-7day forecast limitation, the average RMSE score by using the D-SUP method is1.46℃on the1day forecast limitation, and can be stable at2.46℃on the7day forecast limitation.