Inversion Of Surface Soil Moisture In Alaer Reclamation Area Based On Multi-source Remote Sensing

Alaer Reclamation Area of the First Division of the Southern Xinjiang Corps in Xinjiang Uygur Autonomous Region is an important cotton production area.Besides cotton,a large number of agricultural products such as fragrant pear,jujube tree and prune are planted in the area,which is also an area suffering from water shortage,desertification and salinization.Soil moisture content directly affects crop production and agricultural output,so soil moisture monitoring is of great significance to agricultural production.Remote sensing technology can realize accurate,real-time and efficient monitoring of soil moisture in large areas,effectively improve the level of soil drought monitoring,and provide effective methods and technologies for realizing agricultural intelligence and implementing precision irrigation in Alar Reclamation area.In this paper,multi-source remote sensing technology was used to extract environmental variables from low vegetation areas and high canopy areas to establish three kinds of water inversion models,so as to achieve high-precision inversion of surface soil water.Moreover,environmental variables and model inversion accuracy in different regions were compared and analyzed,and the conclusions were as follows:(1)The study area was divided into low vegetation area and high canopy area.Environmental variables were extracted from multi-source remote sensing data,and the selected environmental variables were introduced into three models(PLSR,SVM and RF)to establish surface soil moisture models for the two regions respectively,realizing the high-precision inversion of surface soil moisture in Alar Reclamation area in southern Xinjiang.(2)Through the correlation analysis of 11 environmental variables extracted from multi-source remote sensing images and the measured surface soil water data in two regions,it can be seen that different environmental variables have different correlations with surface soil water content,and there are certain differences in the correlation between the same environmental variable and surface soil water content in different regions.Five environmental variables in the low-rise vegetation area were significantly correlated with surface soil moisture,namely,backscattering coefficient(VH,VV),LST,band B11 and band B12.Four environmental variables were significantly correlated with surface soil moisture,namely,NDVI,DEM,band B5 and band B8.In the high canopy area,four environmental variables were significantly correlated with surface soil moisture,namely,backscattering coefficient(VH,VV),LST and NDVI.Four environmental variables were significantly correlated with surface soil moisture: NDWI,DEM,band B11 and B12.Among the environmental variables,the correlation between LST and surface soil moisture content was the highest,and the coefficient of determination in low vegetation area and high canopy area reached 0.54 and 0.68,respectively.At the same time,it can be observed that there is a significant correlation between DEM and soil moisture content,which can be attributed to the influence of topography and landform on surface runoff,solar illumination,vegetation growth environment and other factors,which in turn affect soil moisture content.Therefore,elevation data can be used as one of the important variables to predict soil moisture content,and it is expected to be widely used in many fields,such as agricultural production and water resources management.(3)In the correlation comparison between environmental variables and surface soil moisture content in low-rise vegetation area and high-rise canopy area,the determining coefficients of 7 environmental variables in low-rise vegetation area were higher than those in high-rise canopy area,which were: The back scattering coefficient(VH,VV),band B4,band B5,band B8,band B11 and band B12,DEM was basically the same in the two regions,and the determination coefficient of NDVI,NDWI and LST was higher in the high canopy area than in the low vegetation area.The correlation between NDVI and soil moisture was more obvious in the low vegetation area and the high canopy area,with the determination coefficients of0.22 and 0.54,respectively.The main reason is that the vegetation coverage in the high canopy area is higher,and NDVI is more sensitive in the high vegetation coverage area,which can reflect the richer information of the region.The determination coefficients of the backscattering coefficients VV and VH were higher in the low vegetation area than in the high canopy area,because the backscattering coefficients VV and VH reflected the soil moisture content more strongly in the low vegetation coverage area.(4)The selected environmental variables of low vegetation area and high canopy area were substituted into partial least squares(PLSR),support vector machine(SVM)and random forest(RF)models respectively.Through comparative analysis of the evaluation indexes of the three models,it was found that the R2 of RF model in low vegetation area was 0.82,which was significantly higher than that of the other two models.RMSE was 9.83%,lower than the other two models.The R2 and RMSE of RF model in high canopy area were 0.79 and 10.35%,respectively,and the prediction effect of RF model was higher than that of the other two models.The results show that random forest(RF)is the optimal model for surface soil water retrieval.