Inversion Of Sediment Concentration In The Yellow River Based On Multispectral Remote Sensing Images

As the mother river of the Chinese nation,the Yellow River has been plagued by sediment problems for thousands of years.Measuring and monitoring the suspended sediment concentration along the Yellow River is helpful to grasp the sediment variation of the Yellow River in real time and provide important information for high-quality development and ecological protection of the Yellow River basin.However,the traditional river sediment monitoring methods rely on manual sampling and field measurement,which is time-consuming and labor-consuming,and the coverage is limited,so it is difficult to meet the needs of the whole river section monitoring.With the development of remote sensing technology,it has the advantages of large coverage,fast updating cycle and easy access to data,which can provide more scientific,accurate and reliable data support for river sediment management and protection.In this study,the Loess Plateau section,which is the main source of sediment in the Yellow River,and the Xiaobei mainstream section in the lower reaches of the Yellow River(collectively known as the North mainstream of the Yellow River)are selected as the study area,combined with multi-spectral and highresolution Sentinel-2 products and the daily average sediment concentration data of the monitoring stations of the Yellow River,a remote sensing inversion model of suspended sediment concentration is constructed for the north mainstream of the Yellow River,revealing the sensitive band of suspended sediment concentration in this area,combined with the launch time earlier.The Landsat series satellites with stronger detection ability of long time series have realized the remote sensing monitoring of sediment concentration in the north main stream of the Yellow River with longer time series through the correction of data and model.The main research contents and achievements of this paper include the following points:(1)Extraction of sensitive bands for sediment inversion of the Yellow River.Based on the advantages of high spatial resolution,wide range and short revisit period of Sentinel-2 multispectral data,it is found that green band B3 and near infrared band B8 and B8 A are important predictors of suspended sediment concentration through the analysis of sediment concentration data of the Yellow River.And the estimation of suspended sediment in the Yellow River with multi-band and band combination can improve the accuracy of sediment inversion.(2)Comparison and selection of sediment inversion models of the Yellow River.According to the mechanism and applicable conditions of different sediment inversion models,this paper compares and analyzes the sediment inversion efficiency of different inversion models in different seasons,and finds that the random forest model is flexible,intuitive and convenient.It shows many advantages in parameter optimization,variable importance ranking,variable analysis and interpretation,which is generally better than parameter inversion model and other non-parametric models.The random forest regression model constructed by dividing the suspended sediment concentration data into low water period and high water season,the average fitting accuracy is 0.735,the average relative root mean square error is 6.08%,and the estimation effect is the best.it can meet the needs of high-precision inversion of suspended sediment concentration.(3)The generalization of sediment inversion model.The band reflectance of Sentinel-2 and Landsat-8 are in good agreement,so the sediment concentration inversion model can be extended to Landsat-8 data.Finally,it is verified that the fitting accuracy of inversion based on Landsat data is 71.49%.This can be used to realize the long time series monitoring of sediment concentration in the middle and upper reaches of the Yellow River.(4)The application analysis of sediment inversion results.The remote sensing monitoring results of sediment concentration in Qiankun Bay and Yumen Kou reach of the north main stream of the Yellow River from 2013 to 2022 show that there is a positive correlation between rainfall and sediment concentration,and one of the main sources of sediment in the Yellow River is Rain Water erosion.The sediment concentration near the bank is higher,while that in the center of the river is lower.The sediment concentration of the Yellow River fluctuates greatly both within the year and from year to year.In the past ten years,the suspended sediment concentration first decreased,then increased and then decreased.In the last four years,the sediment concentration of the Yellow River has begun to decline gradually,and the sediment control of the Yellow River has made some achievements.Based on the random forest model and multi-spectral image data,this study realized the long-time sequence analysis of sediment variation in the north main stream of the Yellow River.The purpose of this paper is to provide important information reference for the sediment variation of the north main stream of the Yellow River and other inland rivers,and to fill the gap of sediment monitoring in the reach of the Yellow River without monitoring stations.At the same time,this study can also be combined with traditional inversion methods to provide a new idea for remote sensing monitoring of river sediment concentration.