以陕西省内2019年33个地面气象观测站点的测量数据为真实值,选取相关系数（CC）、均方根误差（RMSE）以及相对误差（BIAS）等多种统计分析指标对GPM（Global precipitation measurement）卫星降水数据进行精度验证,并引入地形因子作为空间参考要素,基于地理加权回归模型（GWR）对GPM降水数据进行降尺度研究。结果表明：（1） 在年际尺度上,GPM降水数据与实测降水数据之间有着明显的相关性,相关性较好（CC=0.89）,相对误差则较低（BIAS=-0.45）;（2） 在季节尺度上,春、夏两季GPM降水数据的降尺度结果与实测降水数据之间的CC值分别为0.92和0.80,而秋季则为0.93;（3） 降尺度降水量结果与高程呈现出明显的负相关性,随着海拔升高,降水相对减少。总体而言,GWR降尺度降水数据在陕西省内有着较好的精度,能够较为准确地反映陕西省内的降水分布。

The downscaling of precipitation based on the geographically weighted regression (GWR) model provides precipitation data with higher spatial resolution, making the spatial scale of precipitation estimation more refined, which is of great importance to regional hydrology and water resources management. Shaanxi Province is located in the central and western region of China, blocked by the Qinling Mountains and Loess Plateau, and exhibits significant regional climate differences. Therefore, the evaluation of the spatial and temporal patterns of precipitation in Shaanxi Province is essential. In this study, the Shaanxi Province was selected as the study area and the precipitation data from ground meteorological observation stations as the real values to verify the accuracy of Global precipitation measurement (GPM) precipitation data from the time scale of the year and spatial scale of the stations. The GWR model was used to downscale the GPM and compare and verify it with the measured precipitation data. Simultaneously, based on the downscaling method of annual precipitation data, the monthly scale was further extended to obtain higher resolution GPM monthly precipitation data and the influence degree of topography on the accuracy of satellite precipitation products was analyzed by combining the digital elevation model (DEM) data. Moreover, the correlation coefficient (CC), root mean square error (RMSE), and relative error (BIAS) were selected to verify the GPM precipitation data accuracy in the Shaanxi Province. Additionally, the GWR model was used to downscale the GPM precipitation data. DEM data were resampled to the same resolution as GPM data. At the same time, the GPM and DEM data under the resolution of 0.1°×0.1° were used as input elements to construct the GWR model and the constant term, DEM coefficient term, and residual of each grid center point were extracted from the regression model. The constant term and coefficient term were rasterized to obtain 1 km data, and the residual was interpolated to the spatial resolution of 1 km×1 km using the ordinary Kriging method. The GPM precipitation data with a spatial resolution of 1 km was calculated and the results show that: (1) The GPM satellite precipitation data demonstrated a good accuracy on the spatial and temporal scales of annual and station, and exhibited certain adaptability in the study region. The downscaling study of GPM precipitation data using the GWR model could greatly improve the spatial resolution of precipitation data, demonstrating a stronger ability to display spatial details. (2) The GWR downscaling results showed a good correlation between the observed precipitation data, CC=0.90 at year scale. The GPM precipitation data of spring and summer between downscaling results and the measured rainfall data of CC values were 0.92 and 0.80, respectively at seasonal scale. In the autumn of 0.93, the month scale correlation was also generally higher and showed better accuracy, meeting the needs of actual precipitation research data. On the whole, however, the GWR downscaling result was higher than the original precipitation data of GPM. (3) The results of downscaling precipitation showed a significant negative correlation with elevation. As the altitude increased, the precipitation decreased. Therefore, the GWR downscaling model was satisfactorily applied to the downscaling study of GPM precipitation data in the Shaanxi Province, which improved the spatial resolution of the data. This study can provide important data for hydrological research and improve the simulation accuracy.