Assessment Of The Utility Of Hydrological Simulation Of The Upper Kaidu River With Multi-source Precipitation Data

Precipitation,as a fundamental component of the Earth's water cycle,has important hydrological significance.However,it is difficult to conduct high-density surface precipitation observation in northwest China,where the topography is complex,the land area is large,the economy is weak,and the density of ground-based meteorological observation stations is sparse,and the lack of high-precision spatial and temporal distribution of precipitation constrains the related hydrological utility.Therefore,it is crucial to find high-precision and high-accuracy precipitation data for hydrological utility assessment in data-starved areas.Satellite remote sensing precipitation products are an emerging means of monitoring precipitation,which not only effectively compensate the limitations of radar and ground stations in technical monitoring means,but also fill the data gaps of precipitation data in sparsely populated areas.In order to find a reliable spatial data source for precipitation,this study takes the upper reaches of the Kaidu River basin as the study area.Firstly,the historical observed precipitation data from Bayinbruck meteorological station is used as the benchmark to evaluate the accuracy and error of five precipitation products,TRMM,GPM,PERSIANN,CHIRPS and ERA5,from different spatial and temporal scales in the upper reaches of the Kaidu River.Secondly,a distributed hydrological model is constructed to evaluate the applicability of SWAT model in the upper Kaidu River.Finally,the applicability of the five precipitation products in hydrological simulations is evaluated by driving the SWAT model with multi-source precipitation data.The main findings of the study are as follows.(1)The accuracy and error characteristics of the five precipitation products on multiple time scales were quantitatively evaluated by selecting three continuity indicators and three classification indicators using ground-observed precipitation data as the benchmark.The results show that the multi-source precipitation products perform better on the annual and monthly scales,and the correlations of all precipitation products are greater than 0.5 except for the TRMM product,which has a low correlation on the annual scale.On the daily and monthly scales,the best performers are the GPM and PERSIANN precipitation products.On the annual scale,there are large overestimates and underestimates for the five precipitation products,with the highest correlation for the ERA5 product and the largest error at the same time;the errors for the PERSIANN and CHIRPS precipitation products are smaller.In the seasonal scale,each precipitation product has the best effect on the detection of spring precipitation,and the best performance of seasonal precipitation detection is found in the GPM,CHIRPS and PERSIANN precipitation products.Overall,the better performers in multi-source precipitation multi-timescale assessment are GPM,PERSIANN and CHIRPS precipitation products.(2)Comparing the spatial distribution characteristics of the five precipitation products in the upper Kaidu River basin,it was found that the spatial distribution of precipitation for the five precipitation products did not differ significantly,and generally showed the spatial distribution characteristics of more precipitation in the west and less in the east,with precipitation mainly concentrated in summer and the least in winter.(3)Based on the DEM,land use and soil data and CMADS meteorological data of the upper Kaidu River basin,the SWAT distributed hydrological model was constructed.The results show that the rate periodic coefficient of determination R2=0.79,Nash coefficient NSE=0.78,and relative deviation PBIAS=2.9%,and the validation period R2=0.82,NSE=0.75,and PBIAS=4.7%,all of which achieve satisfactory results,indicating that the SWAT model has good applicability in the upper Kaidu River.(4)The SWAT model was used to verify the runoff simulations of five precipitation data in the upper reaches of the Kaidu River.The results show that the four precipitation data,TRMM,GPM,CHIRPS,and ERA5,can obtain qualified results except for PERSIANN data.Among them,CHIRPS data performed the best in runoff simulation and also had better performance in accuracy assessment,indicating that CHIRPS precipitation products can replace ground observation data for hydrological studies in the upper reaches of the Kaidu River.