中文摘要：

摘要：在地形比较复杂的石羊河流域上游，基于2001—2010年TRMM3843降水数据和1km空间分辨率DEM数据，采用回归方程＋残差法的插值方法，对石羊河流域上游10年平均TRMM3843和2001、2007年两个典型干、湿年份的TRMM3843数据分别进行降尺度操作，将空间分辨率由原来的0．25°×0．25°提高为1km×1km。利用研究区34个雨量站的观测数据作为“真实值”，对2001—2010年的TRMM3843和降尺度结果进行精度检验，结果表明：TRMM3843降水数据整体上具有一定的可信度，但比地面站点观测的降水量偏小；TRMM3843经过降尺度之后，数据在空间分辨率和精度方面都有一定程度的提高，其中10年平均TRMM3843降尺度结果的精度要高于2001年（干旱1，而2001年的降尺度结果精度又高于2007年（湿润）；就单个雨量站而言，位于西部高海拔区域的站点的降尺度结果精度较高，而位于东部低海拔区域的站点的降尺度结果精度较低。

英文摘要：

Abstract： Rainfall data are often obtained by ground-based observatories. However, traditional measurements based on raingauge stations can＇t reflect the spatial variation of precipitation effectively, especially in the Shi- yang River Basin, which is a typical area with complicated terrain and climatic characteristics. There is great po- tential in making hydrological predictions by using satellite-based rainfall estimation. As a precipitation radar satellite, TRMM has been collecting plenty of fine temporal-spatial precipitation data. However, when applied to local basins and regions, the spatial resolution of TRMM products is too coarse, so it is necessary to develop a method to improve the spatial resolution of TRMM before using it. In this paper, a statistical downscaling algo- rithm based on the relationship between Tropical Rainfall Measuring Mission （TRMM） 3B43 dataset and the DEM （GTOPO30） from USGS is presented. A multiple linear regression model was established under the scale of 1 kin. By applying a downscaling methodology based on 1 km resolution, the TRMM3B43 0.25~x0.25~ pre- cipitations were downscaled to 1 kmx 1 km pixel precipitation for each year from 2001 to 2010. The downscaled precipitation estimates were subsequently validated by using the observations obtained from 34 raingauge sta- tions for the duration of 10 years in the Shiyang River Basin. These results showed that： the downscaling proce- dure resulted in significant improvement in spatial resolution and data quality for annual precipitation during 2001-2010, as well as for a typical dry year （2001） and wet year （2007）, and captured the trends of precipitation in spatial distribution and inter-annual variability of annual precipitation with the coefficient of determination R2 ranging from 0.45 to 0.93 at 34 different raingauge stations.