Multi-Source Remote Sensing Modeling Inversion Of Soil Moisture,Fertilizer And Salt In Typical Agricultural Demonstration Areas Located In He Tao Irrigation District

Soil salinization is an important cause of crop production reduction,and the intensified global scale of soil salinization brings great challenges to human food production.Hetao irrigation area is one of the three major irrigated areas in China.Soil salinization and secondary salinization problems are serious,and the sustainable development of agriculture is greatly threatened.Therefore,it is of great practical significance to monitor important indicators such as soil moisture,fertilizer and salt in Hetao irrigation area.Based on the remote sensing data of Landsat8 OLI and Sentinel-2 in April and October from 2018 to 2020,various related variables were used to conduct correlation analysis on the sampling data of soil moisture,fertilizer and salt in the study area.The regression model of soil moisture,fertilizer and salt was established by multiple linear regression analysis and BP neural network regression analysis.The regression model was used to conduct batch inversion and the reasons for the change of soil moisture,fertilizer and salt in the study area were discussed.The following main results were achieved:(1)In Landsat8 OLI and Sentinel-2 satellite images,B1,B2,B5 bands and infrared bands have high correlation with fertilizer,B1,visible light band and thermal infrared band have high correlation with moisture,and blue light,green light and red light bands have high correlation with salt.(2)Binary cooperative modeling of Landsat8 OLI and Sentinel-2 has higher prediction accuracy than Landsat8 OLI and Sentinel-2 alone.In April 2019,the moisture inversion model R~2was 0.600 with a root mean square error of 4%,in October 2018,the fertilizer inversion model R~2was 0.529 with a root mean square error of 2.678g/kg,and in October 2020,the salt inversion model R~2was 0.620 with a root mean square error of 3.537g/kg.(3)At the six research time points from 2018 to 2020,the R~2of salt inversion model remains in the range of 0.711 to 0.850,the root mean square error is from0.490g/kg to 1.273g/kg,and the R~2of fertilizer inversion model is above 0.74,in October 2018,The R~2of the fertilizer inversion model is 0.862 with the root mean square error of 0.636g/kg.At the six research time points from 2018 to 2020,the mean R~2of moisture inversion model is 0.753 with the mean root mean square error of 2%.(4)Inversion images are generated based on inversion models of moisture,fertilizer and salt,to predict and analyze the research area of annual variability in April and October of 2018,2019 and 2020.The moisture is mostly between 15%and20%,showing a trend of first increase and then decrease,and the overall trend of slow growth.The influence of atmospheric precipitation is great in April,and the fertilizer is mostly between 10g/kg and 20g/kg,showing a small fluctuation change,while the salt is mostly between 0g/kg and 4g/kg,showing a trend of annual decrease.