中文摘要：

中国国家气象信息中心基于2400多个国家级台站观测日降水量和CMORPH卫星反演降水产品,采用概率密度匹配和最优插值相结合的两步数据融合方法,研制了中国区域1998年以来的0.25°×0.25°分辨率的逐日融合降水产品（CMPA_Daily）。通过该数据集与广泛应用于中国天气气候领域的两种国际上降水融合产品TRMM 3B42（Tropical Rainfall Measuring Mission,3B42）和GPCP（Global Precipitation Climatology Project,1 degree daily）的对比评估,考察CMPA_Daily产品的质量,评价其能否合理体现中国降水的天气气候特征。首先利用2008-2010年5-9月独立检验数据定量对比了CMPA_Daily、TRMM 3B42和GPCP三种降水产品的误差,结果表明,在误差的时间变化和空间分布上,CMPA_Daily均具有最高的相关系数和最小的平均偏差及均方根误差,TRMM 3B42其次,GPCP的误差相对较大。CMPA_Daily只低估了大暴雨,TRMM 3B42低估了大雨以上量级的降水,而GPCP低估了除小雨以外的所有降水。CMPA_Daily产品因融入了更多的站点观测信息,不论在中国东部沿海,还是中西部地形复杂区,其精度均优于TRMM 3B42和GPCP产品,即使在站点稀疏的青藏高原地区,CMPA_Daily降水量也更加接近站点观测,呈现明显的高相关。CMPA_Daily与独立检验数据的高相关在地形起伏时效果也较稳定,TRMM和GPCP的相关系数则随着地形变化幅度陡变而非常明显地降低。进一步通过对比分析各降水产品1998—2012年的气候平均降水特征表明,3种资料对中国区域气候平均降水量、降水强度、频率分布以及年际变化的总体描述基本一致,因有效融入了更多的中国站点观测信息,不论降水空间分布还是降水量,CMPA_Daily与地面观测均最为接近,在中国的中东部大部分地区对降水的估计精度明显更高,而在站点分布较稀疏的青藏高原地区,CMPA_Daily的降水分布型与TRMM、GPCP卫星融合资料类似,较地面站点插?

英文摘要：

The two-step merging algorithm called the Probability Density Function and Optimal Interpolation （PDF-OI） tech- nique has been applied to create a China Merged Precipitation Analysis at a daily and 0.25° resolution （CMPA_Daily） since 1998 which is quality-controlled based on more than 2,400 gauge observations over China and the CPC （Climate Prediction Cen- ter, NOAA） Morphing （CMORPH） satellite QPE. The world widely-used datasets of TRMM 3B42 （Tropical Rainfall Meas- uring Mission, 31342） and GPCP （Global Precipitation Climatology Project, 1 degree daily） are introduced to conduct compara tive analysis over the China domain. The errors of CMPA Daily, TRMM 31342 and GPCP are calculated based on the independ- ent validation data from May to September, 2008- 2010. The results show that CMPA_Daily has the highest correlation coeffi- cient and the lowest Bias and Root-Mean-Square Error both in temporal variations and spatial distribution while GPCP has the relatively higher errors among the three products. Underestimation of heavy storm only, heavy rain or above, and moderate rain or above is observed by CMPA_Daily, TRMM 31342 and GPCP, respectively. For the rainfall rate higher than 25 mm/d, the relative Bias is within 10%, 30% and 60% for CMPA_Daily, TRMM 31342 and GPCP, respectively. The precision of CM- PA Daily is observably promoted no matter in the eastern coastal areas of China or in the mid-western complex terrain com- pared with that of both the TRMM 3B42 and GPCP as a result of merging more gauge observations. Even over the sparse ob- servations area, such as the Tibetan Plateau, the precipitation value from CMPA_Daily is closer to the gauge observations. The high correlation between CMPA_Daily and gauge-based precipitations is stable even in areas with great hypsography, while the correlation coefficients for TRMM 31342 and GPCP are falling obviously with topographic changes. At seasonal and annual ti- mescales, the three datasets for 1998 - 2012 can be matched very well to represe