| As a basic component of the earth's water cycle,precipitation has important hydrological significance.However,under the influence of complex topography and severe climatic conditions,precipitation observations in Northwest China are difficult,and it is even more difficult to obtain observation data in a mountainous area.The development of satellite remote sensing precipitation retrieval technology provides the possibility to overcome this limitation.Affected by factors such as sensors,terrain and climate,there are obvious differences in the applicability of different precipitation products in the same area.Based on four sets of precipitation data with good applicability,CHIRPS,GSMAP,TRMM 3B42 V7 and SM2RAIN,this paper evaluated the applicability of each product in the hydrological research of the arid mountainous region of Northwest China from the perspective of precipitation and runoff simulation.In addition,three methods of linear regression model correction,mean deviation correction and proportional coefficient correction were used to calibrate the optimized precipitation products.On this basis,the applicability of the optimized precipitation products was further verified by SWAT hydrological model.The main conclusions are as follows:(1)There are obvious differences in the spatial distribution of the four satellite precipitation data in the Karakoram-Kunlun mountain basin.The average annual precipitation of GSMaP and TRMM in 2008-2018 showed an increase in precipitation from north to south and east to west.The trend is consistent with the results of scientific investigation and measurement of glacier flood.The annual maximum precipitation areas of CHIRPS and SM2RAIN are distributed in the middle of the study area,and the precipitation in the southern high mountainous areas is relatively small,indicating that there is a possibility of underestimating the precipitation in the high-altitude mountainous areas or overestimating the precipitation in the central part.On the seasonal scale,the spatial distribution of precipitation in each season of GSMaP and TRMM is relatively consistent with that on the annual scale,and both show a trend of increasing precipitation from northeast to southwest.However,CHIRPS and SM2RAIN are significantly different from the survey results.(2)GSMAP showed a high correlation with observed precipitation(CC?0.9)on the three time scales of annual,seasonal and monthly.On the annual scale,the root mean square error and bias of GSMaP are 14.29mm and 9.7%,which were relatively small.At the seasonal scale,the RMSE of GSMaP in spring,summer and autumn were all less than 8mm,with BIAS of less than 15%,although it(BIAS?74.12%)was overestimated to a certain extent in winter,this is lower than the CHIRPS(BIAS=151.44%),TRMM(BIAS=357.09%)and SM2RAIN(BIAS=76.40%).At the monthly scale,the root mean square error and bias of GSMaP and measured precipitation are 3.19mm and 9.7%respectively.In addition,the standard deviations of GSMAP at various time scales were close to the actual precipitation.It can be seen that the inversion accuracy of GSMAP is outstanding in all time scales.(3)According to the runoff simulation results of CHIRPS,GSMaP,TRMM and SM2RAIN,NS and R~2 of SM2RAIN exceeded 0.60 and 0.80 in the calibration period and validation period of each basin,showing the best simulation effect.However,the simulated value of SM2RAIN in some winter months was 0m~3/s,which was significantly different from the actual runoff.The NS of GSMaP in the calibration period and validation period of each basin did not exceed 0.55,and the R~2 exceeded0.80,indicating that the hydrological simulation application results were excellent.The simulation effect of TRMM and SM2RAIN precipitation products is slightly worse,and the R~2 of each basin in the calibration period and validation period is basically below 0.80,and the NS is less than 0.60.The relative errors of simulated runoff of the four sets of precipitation products in each basin are all greater than 0,and the simulated runoff value is underestimated,which is related to the lack of glacial melt water calculation process in the model.In general,SM2RAIN and GSMaP precipitation products show great application potential in hydrological simulation of Karakoram-Kunlun mountain watershed.(4)A variety of indexes were used to evaluate the corrected daily precipitation results of three correction methods,namely linear regression model,mean bias correction and proportional coefficient,in the Karakoram-Kunlun Mountains Basin.The results show that the three methods can effectively improve the accuracy of daily precipitation of GSMaP satellite,and the proportional coefficient method has the highest consistency(CC?0.76)and the minimum dispersion level(RMSE<1mm)between the precipitation data and the observation data.In terms of the degree of high and underestimate,the precipitation corrected by the proportional coefficient and linear regression model was significantly overestimated compared with the observed precipitation(Bias>0).The precipitation corrected by the mean bias method was underestimated at Tashkurgan(Bias=-17.07%)and Pishan(Bias=-1.36%)stations,and overestimated at the other stations.(5)The spatial distribution of daily average precipitation after correction by the three methods is basically the same as before correction.The largest precipitation areas are all distributed in the southwest of the study area.However,there are differences in precipitation interval.After linear regression model and proportional coefficient correction,the daily precipitation is between 0.24?0.90mm and 0.18?0.95mm,which is significantly lower than that before correction(0.17?1.17mm).In addition,the mean values of precipitation data after linear regression correction and proportional coefficient correction increased to 0.37mm and 0.24mm in the Yarkand River Basin,respectively,and increased to 0.29mm and 0.27mm in the Hotan River Basin,respectively,indicating that the two methods weakened the heavy precipitation and enhanced the weak precipitation to some extent.After the correction of the mean deviation,the average daily precipitation increased significantly(0.16-1.75 mm),and the average daily precipitation in the Yarkand River Basin and Hotan River Basin increased to 0.51mm and 0.36mm,respectively.Therefore,the corrected precipitation by this method may be overestimated.(6)Use the SWAT model to analyze the effectiveness of the corrected precipitation data in the Karakorum-Kunlun Mountains hydrological application.Based on the corrected precipitation data of the three methods,the simulated runoff values are well fitted with the measured values(NS>0.6).The correlation between the simulated runoff value and the actual runoff value after the correction of the proportionality coefficient increased significantly(0.87?R~2?0.93).The consistency between the simulated value of the linear regression correction(0.73?R~2?0.86)and the mean deviation correction(0.58?R~2?0.63)and the actual value was relatively poor,which reflected that there was a large error between the input precipitation of the model and the actual situation.In addition,the mean bias corrected precipitation simulation runoff is less underestimated in the two sub-basins of the Hotan River(PPBIAS<2%).The lack of calculation of the glacier melt process in the model means that the precipitation may overestimate the actual precipitation.Overall,the precipitation data corrected by the proportional coefficient method has shown great potential in hydrological simulation applications. |
PDF Full Text Request