中文摘要：

在新疆天山山区,利用重建时间序列后的NDVI和DEM数据,基于人工神经网络对TRMM3B43月降水数据进行了校正。并采用研究区25个站点实测降水数据对仅考虑地理因子校正后的TRMM值与同时考虑地理因子和NDVI进行校正后的TRMM值分别进行精度检验,结果表明：仅考虑地理因子校正后的TRMM数据的校正效果明显,同时考虑地理因子与NDVI进行校正后的TRMM数据效果更好,R2显著提高,δ和RMSE明显降低。对于单个站点而言,仅考虑地理因子校正后的TRMM数据和综合考虑地理因子和NDVI校正后的TRMM数据精度大部分有所提高,个别站点与实测值之间有一定差异,降水量相对较小的站点差异较明显。

英文摘要：

Based on the NDVI and DEM data after reconstructing the time series,in this study the artificial neural network approach was used to mainlY revise the TRMM31343 monthly precipitation data in the Tianshan Mountains. The precipitation data from 25 meteorological stations in the study area were used to test the accuracy of adjusted TRMM values under only considering the geographical factors and taking into account both the geographical factors and NDVI. The results showed that the correction effect of TRMM data was significant under only considering the correction of geographical factors, and the results of TRMM data were more significant when both the geographical factors and NDVI were taken into account, the R2 was increased significantly, and the 8 and RMSE were decreased obviously. For single stations, the accuracy of the corrected results of most stations was increased under both the two methods except some differences between the corrected TRMM data and the observed data at several stations, and such differences were relatively low at the stations with low precipitation.