为了获得对中国区域降水刻画能力较好的再分析资料,以长江中下游地区为例,将PREC/L、CMAP、GPCP及NCEP2四种再分析降水资料与台站降水观测资料进行了多年和逐年平均的年、季、月降水量距平的时空相关分析和方差分析,比较了再分析资料对该区降水刻画能力的差异。首先对该区域的观测资料、CMAP、GPCP及NCEP2资料进行插值,使所有资料具有相同的空间分辨率,在此基础上比较再分析资料与观测资料年、季、月平均的降水变化特征。结果表明,四种再分析资料与观测资料的距平相关系数(均方根误差)由大到小(由小到大)的变化次序为:PREC/L→cMAP→GPCP→NCEP2,四种再分析资料均能较好地刻画出长江中下游地区降水的时空分布特征,尤以PREC/L资料对该区域降水变化的时空演变规律描述得最好。最后,利用PREC/L资料与观测资料对该区域降水的时空特征进行了相关分析、EOF分析和功率谱分析,结果显示夏季平均降水量具有2.3和3.1年的显著周期,EOF分析的前两个模态的时间变化分别具有2.3和2.8年的显著周期,这些显著周期都通过了95%的红噪声检验。
In order to find a proper reanalysis dataset which can best capture the precipitation characteristics of China,the multi-year and yearly annual,seasonal and monthly mean anomaly correlation coefficients and the root mean square errors of four reanalysis precipitation datasets (PREC/L,CMAP,GPCP and NCEP2) and the rain-gauge records were compared from January 1979 to December 2003 based on the area of the middle-lower reaches of the Yangtze River (26°N-34°N, 105°E-122°E). All the four reanalysis datasets and observation are monthly data. The data of observation,CMAP,GPCP and NCEP2 were interpolated to the same resolution as PREC/L with resolution 0.5×0.5 (longitude×latitude) and the comparison of the precipitation characteristics of annual, seasonal and monthly variations was based on the interpolated data. Results indicated that all of the four reanalysis datasets can capture the spatial-temporal characteristics of precipitation over this region quite well,in which the PREC/L data were the best datasets to consist with the observational data. The anomaly correlation coefficients (the root mean square errors) between four re- analysis precipitation datasets and the rain-gauge records were descending (increasing) gradually from PREC/L,CMAP,GPCP to NCEP2. Finally,the spatial and temporal characteristics of precipitation in the middle-lower reaches of the Yangtze River were analyzed by the methods of correlation,Empirical Orthogonal Function (EOF) and spectrum analysis using PREC/L and observational data. The correlation coefficients of precipitation, the modes and principal components of first two EOF eigenvectors between PREC/L and the observational were above 0.96. Spatial pattern of the first two EOF modes exposed that the sum- mer precipitation in the southeastern China was out of phase with that in the northwestern China,and that in the southern China was out of phase with that in the northern China. Spectrum analysis indicated two significant periods with 2.3 and 3.1 years existed