| The water distribution of soil is an important research direction of regional soil erosion.The traditional monitoring of soil moisture is mainly based on spot sampling.It is difficult to fit the needs of large-scale regional research.While,Radar remote sensing has specific advantages in the study of large-scale regional distribution of land features in Guizhou Province.This paper takes the Chabai River basin in Pingba District,Anshun City as the research area.It utilized Satell-1A Radar Data and Landsat 8 Optical Data in March 2018,based on Soil moisture inversion which improved water cloud model and BP neural network model with vegetation coverage in the study area.It uses Landsat 8 optical remote sensing data to establish TVDI model with different vegetation indices for inversion of surface(0-5cm)soil moisture,and it compared and analyzed the accuracy of soil moisture inversion by different models.The main research results are as follows:(1)BP neural network algorithm based on BAS optimization The soil water content in the study area is synergistically inversion using Sentinel-1A radar and Landsat 8 optical data,which is better than other models in this study.The back scattering coefficients of VV,VH and VV/VH from three bands of red,green and near infrared in Landsat 8 and entinel-1A data are used as input data,which is researched that the method is feasible.The fitting degree(R~2)of model inversion and measured data is 0.57 and RMSE is 3.48.(2)Based on different sizes of filtering windows and methods to Process Sentinel-1A radar data,It shows that the filtering effect of window size 5*5 is the best.Comprehensive evaluation index(statistical image mean,standard deviation,mean preservation index,smoothing index,equivalent visual number and edge preservation index)considers that Lee filter has the best effect in suppressing speckle noise of radar data.(3)Based on the Water Cloud Model which draw into vegetation coverage extracted the backscattering coefficient for reducing the scattering effect of vegetation layer.It used the measured data to establish mathematical model and invert soil moisture.Use different vegetation water content to measure water cloud models with accuracy verification and evaluation analysis.The results show that the water cloud model of vegetation water content based on NDWI inversion is more suitable.The inversion results verify that the fitness R~2 reaches 0.49(VH)and 0.51(VV)respectively.It is found that VV polarization is more suitable for water retrieval in vegetation-covered areas.(4)By extracting different vegetation indices(NDVI,MSAVI,EVA,RVI)and surface temperature from Landsat 8 data to construct the TVDI model to retrieve soil water values,by comparing different TS-VI eigenvalues,it was found that the determinant coefficients of dry-wet edge equation constructed by EVI and MSAVI were as high as 0.88,0.91 and 0.82.The precision of soil moisture inversion of TVDI model based on EVI and MASVI is 0.50 and 0.54.(5)According to the results of soil moisture inversion,and the spatial characteristics of soil moisture gradually increase from north to south,which is consistent with the spatial changes of vegetation cover and surface temperature.Slope is one of the three topographic factors,which mainly affects the distribution of soil moisture,followed by altitude.With the increase of altitude,the main distribution areas of wetting and wetting grades also increase.The gradient of slope between5°-15°is mainly drought and drought grade distribution and the gradient of slope over15°is mainly wet and wet grade distribution. |
PDF Full Text Request