“频率匹配法”在集合降水预报中的应用研究

Applications of "Frequency-Matching" Method to Ensemble Precipitation Forecasts

基于"频率匹配法"的思路,采用两种方法进行了集合降水预报的订正研究,一种方法是利用集合成员降水频率订正简单集合平均平滑效应的"概率匹配平均"法,另一种方法是利用实况降水频率订正集合成员降水预报系统偏差的"预报偏差订正"法,通过个例和批量试验,结果表明:(1)概率匹配平均法可以矫正简单集合平均的平滑作用所造成的小量级降水分布范围增大而强降水被削弱的负作用,这种改进对强降水区更显著,并且集合系统离散度越大这种改进也越大;但该方法对预报区域内总降水量的预报没有改进作用,不能改善预报的系统性偏差。(2)虽然预报偏差订正法对降水落区预报的改进有限,但可以订正模式降水预报的系统性误差,改进雨量预报以及集合预报系统的离散度特征和概率预报技巧;直接对集合平均预报进行偏差订正的效果优于单个成员偏差订正后的简单算术平均。(3)在对每个集合成员的降水预报进行偏差订正后,概率匹配平均仍可改善其简单平均的效果,因此在实际业务中,应该综合采用上述两种方法,以获得在消除系统性偏差的同时各量级降水分布又合理的集合平均降水预报。

Following the 'frequency matching' concept,two approaches have been tested to improve ensemble-based precipitation forecasts for the case of Beijing '7.21'(21 July 2012) severe rainstorm and the period of rainy season from 1 May to 31 July 2010 in China.One approach is 'probability-matched ensemble mean(PM)',which uses ensemble member-based precipitation frequency to calibrate the frequency of simple ensemble mean to correct smoothing effect caused by ensemble averaging;another is 'bias correction',which uses observed precipitation frequency to calibrate the precipitation frequency of each ensemble member to remove systematic forecast bias caused by model deficiency.The results show that(1) PM can effectively correct the problem of 'too-wide-spread light precipitation and too-much-reduced heavy precipitation' in a simple ensemble mean,especially so for extremely heavy rains.The larger ensemble spread is,the more improvement will be achieved.However,PM can barely improve total precipitation amount forecasts and has no ability to correct the systematic bias of the model.(2) Although bias correction is not effective in correcting precipitation position errors,it can effectively remove systematic bias in precipitation amount forecasts and,consequently,improve ensemble spread and probabilistic forecasts significantly.Bias correction of ensemble mean forecast needs to be done independently rather than through bias-corrected ensemble members via simple averaging.(3) PM approach also works well even after each ensemble member's bias has been removed in advance.Therefore,both PM and bias correction need to be done jointly to maximize benefit through model forecast post-processing in an operational environment.