Which Precipitation Product Is Better, Multi-Sources Blended Dataset, Global/Regional Reanalysis Dataset, and Satellite Retrieved Precipitation, in the Qinghai-Tibet Plateau?

Precipitation is a crucial factor in water allocation decision-making, hydrological processes, atmospheric research, and hazards prevention. Compared to surface rain gauge observation, satellite retrieved precipitation (SRP) and numerical weather prediction model (NWP) precipitation can cover ungauged regions that usually has vulnerable environment. However, these precipitation datasets should be comprehensively evaluated before be applied. Previous studies about the evaluation of SRP shows the bias correction of SRP and evaluation with rain gauge observation has autocorrelation phenomenon. Meanwhile, the evaluation is applied in climatological scale. This study evaluates precipitation of high horizontal resolution global/regional NWP, and multi sources blended precipitation in complex terrain of the Qinghai-Tibet Plateau. Results show on annual scale GSMAP_gauge and MSWEP is better than others with higher correlation coefficient, and lower mean error. On seasonal scale, PERSIANN_CCS and the reanalysis of western China has overestimate the precipitation. And mean error in TRMM_3B42、GsMaP_gauge、CMORPH_ADJ、CMORPH_sun、PERSIANN_CDR, and CHIRPS is less 5 mm/day. In evaluation of precipitation intensity, TRMM_3B42、GsMaP_gauge、CMORPH_sun、CHIRPS, and MSWEP outperform than others. On montly scale, the correlation coefficient in gsmap_gauge、cmorph_sun,and mswep is higher than others, which is more than 0.6. The RSME in most precipitation datasets has consist changes. Furthermore, in warm season (May-September), RSME is more than 50 mm/day. In all basins of the Qinghai-Tibet Plateau, the PDF of most precipitation datasets shows overestimate light precipitation, and the underestimate the heavy precipitation. CHIRPS, WCR, CMORPH_RAW, and PERSIANN_CCS have worse performance than other. MSWEP is better than the precipitation of NWP, CMORPH family, PERSIANN family and TRMM 3B42 in general. Multi sources blended precipitation is promising methodology for high resolution precipitation datasets.