基于TIGGE多模式降水量预报的统计降尺度研究

Statistical downscaling research of precipitation forecast based on TIGGE multimodel ensemble

利用TIGGE资料中欧洲中期天气预报中心、美国国家环境预报中心、英国气象局以及日本气象厅4个中心,1~7 d预报时效的降水量预报资料,以TRMM/3B42RT降水量作为“观测值”,对东亚地区降水量进行统计降尺度处理。首先利用逻辑回归方法将天气分为有雨和无雨,再对有雨的情况,利用线性回归方法对插值后的预报结果进行降尺度订正,最后将4个中心的预报值进行消除偏差集合平均,得到多模式集成的降水量预报场。结果表明：逻辑回归能够有效地改善预报中小雨的空报情况,统计降尺度订正后的预报结果比直接插值更加准确,多模式集成的预报效果优于单模式结果,其改进效果随预报时效的延长逐渐减小。

By taking the TRMM/3B42RT rainfall amount as ＂observed value＂, the 1-7 day＇s pre-cipitation forecasting data from the European Centre for Medium-Range Weather Forecasts, the Japan Me-teorological Agency, the National Centers for Environmental Prediction and the UK Met Office in the TIGGE datasets were used to statistically downseale the precipitation over East Asia（90-140 °E, 15-45°N）. Firstly, the weather phenomenon was divided into rain and no-rain by logistic regression. Then, the linear regression was subsequently used to statistically downscale the interpolated precipitation forecast for rainy weather. Finally, the bias-removed ensemble mean was applied to gain the multimodel ensemble precipitation forecast. Results show that the logistic regression can effeetively eliminate the false alarms for light and moderate rain; the forecasts by statistical downscaling are more accurate than that by interpola- ting. The muhimodel ensemble forecast is superior to that by individual model in terms of the root-mean- square errors and the anomaly correlation coefficients of the precipitation forecasts, whose improvement effects decrease as forecast leading time increases.