广义线性统计降尺度方法模拟日降水量的应用研究  

An application of the generalized linear statitical downscaling method to simulating daily precipitation

作　　者：曹经福[1,2] 江志红[1] 任福民[2] 徐振亚[2,3] 

机构地区：[1]南京信息工程大学,南京210044 [2]中国气象科学研究院灾害天气国家重点实验室,北京100081 [3]南京大学灾害性天气研究所,南京210093

出　　处：《气象学报》2013年第1期167-175,共9页Acta Meteorologica Sinica

基　　金：全球变化重大科学研究计划(2010CB950501);国家自然科学基金项目(40875058;41175075)

摘　　要：利用1960—2010年青藏高原23个台站和长江下游25个台站的日降水量观测资料及NCEP再分析资料,采用广义线性模型的统计降尺度方法模拟台站日降水量,并评估了广义线性模型对日降水量的模拟能力。在建模期(1960—2005年)广义线性模型对日降水量表现出良好的模拟能力,两区域模拟结果与观测值1月平均相关系数0.75左右,7月也均超过0.5。模拟结果大部分台站日降水偏大,但偏大的量值较小;模拟的无降水准确率较高,最高值在高原区域,1月平均达85.2%。检验期(2006—2010年)广义线性模型模拟的日降水与建模期具有较好的一致性。此外,对两区域代表站的分析显示,广义线性模型模拟降水极值和降水0值的效果较好,且较好地还原了主要降水过程。总之,广义线性模型对日降水量的降尺度效果良好,适合应用于气候领域的相关研究。An applicational study of the statistical downscaling method of Generalized Linear Model（GLM） was carried out on downscaling daily precipitations. Applying the observational daily precipitation data and the NCEP reanalysis data from 1960 to 2010, the study focuses on two regions - the Tibetan Plateau and the lower valley of the Yangtze River. The GLM method shows good ability in simulating daily precipitation during the simulation period（1960-2005）： the correlation coefficients for the two regions between the simulations and the observations are around 0.75 in January and above 0.5 in July, and the simulations are generally greater than the observations with small biases, while the accuracy of simulating no precipitation is much higher with the biggest value being 85.2% for the Tibetan Plateau in January. Meanwhile, the simulative daily precipitation during the test period is good consistent with that during the simulative period. In addition, a further analysis shows that the GLM has good ability in simulating heavy precipitation as well as no precipitation. And, the GLM retrieves the main rainfall processes successfully. In short, the good performance of the GLM method in downscaling daily precipitation makes it suitable for applying to relevant climatological researches：

关 键 词：统计降尺度 日降水量 广义线性模型 

分 类 号：P467[天文地球—大气科学及气象学]