Study On Water Body Extraction And Discharge Inversion In The Source Region Of The Yellow River Based On Multi-source Remote Sensing Data

River discharge is one of the most important hydrological factors on land.Precisely obtaining discharge information plays an important role in the evaluation and management of regional water resources,as well as in flood control and other aspects.It provides significant support for the development,utilization,saving and protection of water resources.With the rapid growth and development of economy,the shortage of water resources,drought and other problems in China are becoming increasingly prominent.Sudden floods have caused great losses to the safety of people's lives and property.Therefore,the monitoring of river discharge has very important practical significance.Remote sensing technology has the advantages of less ground limitation,and fast data acquisition,which provides an important support for river discharge inversion.Based on the Landsat8 images,Sentinel-1/2 images provided by the Google Earth Engine(GEE)cloud platform,this thesis carried out water extraction,including the use of the modified normalized water index(MNDWI)and Otsu method to extract water from Landsat8 and Sentinel-2 images,and the use of the adaptive threshold method combined with the canny edge detection and Otsu method to extract water from Sentinel-1 images.On this basis,the hydraulic parameters such as river length,river width,roughness,slope,river depth and flow velocity were estimated combined with digital elevation model(DEM),and then the discharge of the reach near Tangnaihai station in the source area of the Yellow River was inversed by using the relationship fitting method and the improved Manning formula method.At the same time,this thesis discusses the influence of the length difference of the river reach on the accuracy of discharge inversion.By establishing the average river width relationship between the station reach and the upstream and downstream reach,it realized the expansion and supplement of the discharge monitoring time series of the station reach.The main conclusions are as follows:(1)Through the analysis of water extraction and classification results of Landsat8,Sentinel-2 and Sentinel-1,it is found that the effect of water extraction is better,and the kappa coefficients are 0.825,0.890,0.759,respectively.(2)Through the analysis of discharge inversion results of the relation fitting method and the improved Manning formula method,it is found that the two models can effectively simulate and estimate the discharge,and the values of the Nash efficiency coefficient(NSE)are above 0.80;the values of the root mean square error(RMSE)are244.849 m³s^-1 and 253.925 m³s^-1 respectively,and the values of the relative root mean square error(RRMSE)are 17%and 22%respectively,which indicates that the precision of discharge inversion by using the relation fitting method is better than that of the improved Manning formula method.(3)Through the comparative analysis of different length river sections and the corresponding discharge inversion results,it is found that when the relationship fitting method is used for the discharge inversion,the larger the average river width in the fluctuation area is,the higher the accuracy of the discharge inversion is;when the improved Manning formula method is used for the discharge inversion,choosing the non braided reach section can reduce the uncertainty of discharge simulation and improve the accuracy of discharge inversion.(4)There is a strong correlation between the average width of the station reach and that of the upstream and downstream reach,and the correlation coefficients are all above0.96.Therefore,the image of the upstream and downstream reach can provide important data supplement for the discharge inversion of the station reach,and realize the intensive monitoring of the discharge of the station reach.