Research On Soil Moisture Inversion In Rocky Desertification Area Of Guizhou Province Based On The Coordination Of Multi-source Remote Sensing Data

Soil moisture,an important component of the surface-atmosphere water cycle,is not only an important indicator in meteorological research,agricultural irrigation,drought monitoring and forecasting,but also as an important water resource in rocky desertification areas,and it determines the ecological restoration and reconstruction of rocky desertification areas.Therefore,retrieval of soil moisture in rocky desertification areas is of great significance to the comprehensive management and ecological restoration of rocky desertification in Guizhou.Using remote sensing technology can achieve a large area,high spatial-temporal resolution,long-term dynamics of soil moisture monitoring,which can overcome the disadvantages of traditional soil moisture measurement methods,such as the limited temporal-spatial resolution,the time-consuming,the high cost and so on.However,the optical remote sensing method is easily affected by cloud cover and weather,which is not conducive to long-term continuous monitor soil moisture;and the active microwave remote sensing has the characteristics of not affected by the weather,the strong penetrability,the high spatial-temporal resolution,etc.Therefore,scholars have established a large number of global and regional soil moisture inversion model by combining the advantages of the above two kinds of data,but the research on karst rocky desertification areas in Guizhou is still very scarce.In view of this,this paper takes the rocky desertification area in Guizhou as the research area,based on Sentinel-1 SAR and Landsat-8,the backscattering coefficient,vegetation index,and rocky desertification information were extracted as model parameters,and linear regression analysis,general regression neural network and BP neural network optimized by genetic algorithm were used to establish soil moisture retrieval model in the rocky desertification area of Guizhou,respectively,and the model retrieval accuracy was verified and evaluated by ground measurement data.The main results of the paper are as follows:(1)In the same polarization mode,NDWI is more suitable for soil moisture retrieval than,which can better separate the vegetation scattering from the total backscattering;in different polarization modes,the model retrieval accuracy of Sentinel-1 SAR VV polarization is better than VH.Comparing the four combinations,it is found that the linear model developed by NDWI and Sentinel-1 SAR VV data was the best for soil moisture retrieval in rocky desertification areas in Guizhou,the coefficient of determination between the estimated soil moisture by the model and measured value is 0.40,the root mean square error is 6.84%,the unbiased root mean square error is 6.25%,and the bias is-2.80%.(2)Based on general regression neural network and BP neural network optimized by genetic algorithm,the radar backscattering coefficient,rocky desertification indicators,vegetation index and land surface temperature were used as input variable of training models,and established GRNN and GA-BP for retrieval soil moisture,respectively.The test results show that the neural network model that takes into account the rocky desertification indicators(vegetation coverage and rock bare rate)has a higher accuracy.(3)Linear model,GRNN model and GA-BP model under different schemes or combinations of retrieval accuracy evaluation results show that: the GA-BP model has the best accuracy,followed by the GRNN model,and the linear model is the worst.Among them,the GA-BP model that takes into account rocky desertification factors has the highest accuracy and is more suitable for soil moisture retrieval in rocky desertification areas in Guizhou,the coefficient of determination between the estimated soil moisture by the model and measured value is 0.72,the root mean square error is 4.38%,the unbiased root mean square error is 4.29%,and the bias is-0.84%.(4)Analyzed the relationship between soil moisture and surface environmental factors,and the results showed: when the soil moisture level is "extremely severe drought",the occurrence rate of rocky desertification is 56.91%.In new soil and fluvo-aquic soil,the incidence of drought is the highest,the brown soil is the lowest.For topographical factors,soil moisture content decreases as the slope rises,"extremely severe drought" and "heavy drought" gradually become dominant positions;as for elevation and aspect,soil moisture are affected by the combination of terrain,vegetation,climate,and man-made factors,the situation is complicated and there is no obvious rules between soil moisture and the above two.