Multimodel Downscaling And Bias Correction Of The Surface Precipitation

Based on the TIGGE data and the TRMM/3B42RT rainfall product during the period from June1through August27,2007, several statistical downscaling schemes are applied to the ensemble outcomes of the1-7d ensemble forecasts for the global surface temperature and1d accumulated total precipitation taken from Medium-Range Weather Forecasts (ECMWF), the Japan Meteorological Agency (JMA), the National Centers for Environmental Prediction (NCEP) and the UK Met Office (UKMO) models. And according to the characteristics of the ensemble forecasting, using the1d accumulated precipitation information of all ensemble members from ECMWF, JMA, NCEP, and UKMO Decaying Average correction and super-threshold events correction were adopted to further reduce the bias of downscaling.The downscaling research results show that the statistical downscaling may improve the forecast skills of the each single model. After the downscaling, the root-mean-square errors (RMSE) of the forecasts are significantly reduced, and the Anomaly Correlation Coefficients (ACC) between the forecasts and the TRMM data become larger. The downscaling of the multimodel ensemble forecasts is superior to that of the single model. Using the observational data of97station Hunan Province to further test the effect of the statistical downscaling. RMSE of forecast reduce to6mm/day from10mm/day. ACC become larger compared to direct interpolation results. The multi-model downscaling of the revised position is more accurate, the center is a clearer, closer to the real rainfall.The study of Decaying Average correction found that the correction effectively reduce the system error and to adjust the dispersion of ensemble prediction system, greatly reduces the probability of an empty forecasts, and with the prediction time growth to improve the effect is more pronounced. The corrected downscaling ETS score of each order of magnitude of precipitation has increased, which the drizzle magnitude ETS scores improved significantly, the corrected B significantly reduced. The figure of regions where there have precipitations in term of the calibration is similar to the figure of that before the calibration.The revised method of super-threshold event is applied to the storm event of more than50mm precipitation. After the correction TS score has improved significantly. Quantitative forecasts of rainstorm is closer to the TRMM data which is considered to be the true precipitation.