Spatial Downscaling Of Satellite Derived Precipitation Data And Its Spatio-temporal Variation Over The Tibetan Plateau

Indirect observation through space-based remote sensing techniques became an effective way to collect precipitation data over those areas.Still,the spatial resolution was not able to meet the requirements of many applications in regional scales.So,it was necessary to develop downscaling algorithms for satellite datasets to improve their spatial resolutions.In this study,precipitation data with high spatial resolution were obtained by establishing downscaling models and used to analyze the spatio-temporal variation of precipitation over the Tibetan Plateau.The main conclusions are as follows:(1)On the monthly scale,TRMM data and GPM data demonstrated good agreement with the rain gauge data,but on the whole,they obviously overestimated the precipitation.The satellite precipitation at most rain gauge stations was accurate,but the accuracy at some stations was lower.The consistency between two satellite products and rain gauge stations was also comparable in arid and semi-arid areas over the Tibetan Plateau,suggesting GPM improved rainfall estimates significantly relative to TRMM.(2)The R~2 between the simulated precipitation of XGBoost model and Random Forest model and the original TRMM Precipitation data was 0.91,and the Bias values were low.The accuracy of 1km precipitation obtained by the two models was basically the same.The Random Forest model could reasonably describe the spatial distribution of precipitation.Therefore,the Random Forest model was choosen to downscale the precipitation data.The downscaling methods with calibration improved the accuracy with increased R~2,reduced RMSE,MAE,and Bias values.The results of monthly precipitation estimates were basically consistent with the original precipitation data in spatial distribution,and the mean Bias corrected by rain gauge stations was significantly reduced.(3)From 2003 to 2017,the precipitation over the Tibetan Plateau had a weak increase trend,with obvious interannual change and greater fluctuation in the regions with abundant precipitation.The annual precipitation showed a spatial pattern of increasing from northwest to Southeast,with obvious variation in the spatial distribution of precipitation.Obvious difference was found in the beginning and ending time of rainy season in different ecological geographical regions.The precipitation mainly occurred in summer while the seasonal distribution of precipitation in different regions was different...