Spatial Downscaling Study Of Three Remotely Sensed Precipitation Products In Qinghai Lake Basin And Analysis Of Temporal And Spatial Changes

Precipitation is an important parameter in the climate system,and accurate precipitation data are important for understanding climate change and hydrological processes.With the development of remote sensing detection technology,remote sensing precipitation products have been widely used in meteorology,hydrology and other fields.However,the spatial resolution of remote sensing precipitation products is generally relatively low,which is difficult to meet the research needs of small watersheds.In view of the above background,the Qinghai Lake Basin is used as the research area,and the principal component-stepwise regression model(PCSR),geographically weighted regression model(GWR)and random forest model(RF)are selected to downscale TRMM,GPM and CMORPH precipitation products to obtain more accurate high-resolution precipitation data in the Qinghai Lake Basin,and further give the temporal and spatial variation pattern of precipitation in the basin.The main conclusions are as follows:(1)Through the correction of the measured data of the stations,it is found that the variation trend of the measured precipitation before and after the correction is consistent,but the change is not obvious.After the correction,the precipitation data of each month is higher than the measured data,and the precipitation difference increases with the increase of the monthly precipitation.After correction,the average annual precipitation increased by 7.304 mm.The measured annual precipitation in 2005 increased the most after correction,while that in 2006 increased the least.(2)The variation trend of TRMM,GPM and CMORPH data was basically consistent with the measured precipitation data,and there was an overestimation phenomenon on the whole.The correlation coefficients between TRMM and GPM data and observed data were high,which were 0.831 and 0.732,respectively.The root mean square error of TRMM data is the lowest,0.169.Generally speaking,TRMM data has the best applicability in Qinghai Lake Basin.(3)After downscaling,the spatial distribution of TRMM,GPM and CMORPH data on the annual,seasonal and monthly scales was basically consistent.Through accuracy verification,the PCSR model in TRMM data is slightly better than GWR model in annual and seasonal scale.The accuracy of the two models is basically the same in monthly scale,and the RF model has the worst performance.The accuracy of GWR model is the highest in the annual,seasonal and monthly scales of GPM and CMORPH data.The comparison of TRMM-PCSR,GPM-GWR and CMORPH-GWR data on monthly scale and accuracy of each station shows that GPM-GWR data is the best downscaling data in the study area.In summary,among the 9 kinds of downscaling data of TRMM,GPM and CMORPH,GPM-GWR data had the best performance in the study area.(4)Based on the GPM-GWR high-resolution precipitation data,the temporal variation of precipitation in the Qinghai Lake Basin from 2001 to 2019 was analyzed.It was found that the annual and seasonal precipitation in the Qinghai Lake Basin showed an increasing trend,and the interannual and summer precipitation changes were significant(α<0.05).At the monthly scale,except for March and May,the other months showed an increasing trend,and July,August and October showed a significant increase(α<0.05).(5)In terms of spatial variation,the annual precipitation showed no significant change trend in the northwestern part of Qinghai Lake Basin,the northern and northeastern edges of Qinghai Lake Basin,and a small part of the southeastern part of Qinghai Lake Basin,while the other regions showed a significant increase trend(α<0.05).In the four seasons,precipitation in spring,autumn and winter showed no significant change in the Qinghai Lake basin,while precipitation in summer showed a significant increase trend(α<0.05).The increasing area of precipitation was distributed in the south of Qinghai Lake,the west area near Qinghai Lake,and the north and northeast area of Qinghai Lake Basin.There was no significant change in precipitation in January,February,March and December.There were a few areas showing a significant increase trend in April,May and September,and a wide range of significant increase trends in the rest of the months.There are differences in the growth distribution of precipitation in each county area,among which the precipitation in Tianjun and Gangcha areas shows a significant growth trend.