Quantitative Inversion Of Soil Moisture And Salinity According To Different Salinization Grades In Coastal Saline Soil Of Yellow River Delta

At present,the protection of soil resources and the management of land degradation have become the focus of global attention.Salinized soil not only has low yield and low fertilizer,but also has poor sustainable utilization.As the main land degradation problem,the management and utilization of soil salinization is of great significance to regional food production and sustainable development.It is the premise of its management and utilization to grasp the distribution of soil moisture and soil salinity quickly and accurately.Remote sensing has become an important mean of monitoring soil salinization information,and improving the quantitative optical remote sensing analysis accuracy of salinized soil information is also a research hotpot.In this paper,according to the coastal salinized soil in the YRD,Kenli District of Shandong Province was taken as the study area.Firstly,according to the different salinization grades,the test areas of the mild or medium-severe salinized soil were selected and the field soil samples were collected,soil moisture content?SMC??soil salinity content?SSC?of three groups samples were measured:mild?M?,medium-severe?S?and whole samples?all samples,I?,the unmanned aerial vehicle?UAV?was equipped with Sequoia multispectral camera to obtain the multispectral image of the two test areas on April 26 to 28,2018,and the Sentinel-2A multispectral image of the study area in the same period was obtained;Secondly,based on the UAV multispectral image,spectral parameters of soil moisture and soil salinity in the three groups were constructed and screened,SMC?SSC models based on the three groups were built to realize the inversion of SMC and SSC in the test areas;Finally,the UAV-based models were applied to the satellite images of different salinization grades areas to realize the regional SMC?SSC inversion.The specific research contents and results are as follows:?1?Spectral characteristics and parameters of SMC and SSC according to different salinization grades were determined.Based on the correlation analysis and grey correlation analysis,the sensitive bands of SMC in the three groups all were Red?R?and Near IR?Nir?band,and the spectral index was also the combination of R and Nir bands.As for SSC,the sensitive bands selected by the correlation analysis were R and green?G?bands,and the spectral index was the combination of R and G bands,while the R and Nir bands selected by the grey correlation analysis,and the spectral index was also the combination of two bands.Two analysis methods showed that the spectral response degree of SMC and SSC according to different salinization grades was different.As for SMC,the spectral response degree of mild salinized soil was the strongest and that of the whole group was the faintest;while in the spectral analysis of SSC,the spectral response degree of medium-severe salinized soil was the strongest and that of the whole group was the faintest.In conclude,whatever SMC and SSC,the response degree according to different salinization gardes were different,and the spectral response degree of grouped sample?M?S?is better than that of the whole sample.?2?Inversion and accuracy analysis of SMC and SSC in test areas were determined.Based on the UAV image,the spectral parameters selected by correlation analysis and grey correlation analysis were used as independent variables,and three groups'SMC?SSC inversion models were established by multiple linear regression?MLR?,partial least square regression?PLSR?,support vector machine?SVM?and back propagation neural network?BP?.The results showed that the best SMC and SSC models of the three groups all were SVM models based on grey correlation analysis.As for SMC models,the models of the group M had the highest precision and that of the group I had the lowest.On the contray,the SSC models of the group S had the highest precision and that of the group I had the lowest.It can be seen that constructing the inversion models according to different salinization grades can improve the inversion accuracy in a certain extent.Based on the best models,the SMC in area M is 7.823-62.622%,the average value is 33.211%,the SSC is 0.323-21.210 g.kg^-1,the average value is 6.871 g.kg^-1,which area with mild saline soil and abundant water content;the SMC in area S is 5.551-62.930%,the average value is 23.424%,and the SSC is 0.136-5.93g.kg^-1,the average value is 2.158 g.kg^-1,and the area S with medium-severe saline soil and insufficient water content,which the result is consistent with the field survey results.?3?Inversion and accuracy analysis of SMC and SSC in research area were determined.Based on Sentinel-2A image,according to the relationship between normalized difference vegetation index?NDVI?and SSC,the study area according to different salinization grades?mild?medium-severe?were divided.Based on the corresponding waveband of UAV image and sentinel-2A image,the best UAV-based models of grouped sample were substituted into the mild and medium-severe salinization grade area respectively;The best model of the whole group was substituted into the study area,and the accuracy of the inversion map is analyzed.The results showed that the inversion accuracy R² and RMSE of SMC in the study area based on different salinization grades models were 0.671 and 14.268,with the absolute residual value was 0.050-13.199,as for SSC,the R² and RMSE were 0.692and 8.562,with the absolute residual value was 0.059-4.802.Comparatively,the inversion accuracy R² and RMSE of SMC based on the whole sample model were 0.624?14.581,with the absolute residual value was 0.086-22.094,as for SSC,the R² and RMSE were 0.622?9.013,with the absolute residual value was 0.255-6.994.The residual value of SMC and SSC for the grouped model was smaller than that of the whole model,and the residual value was more closer to 0,the statistical results were also more close to the statistical analysis results of field soil samples.Therefore,grouped models according to different salinization grades can improve the regional inversion accuracy of SMC and SSC.The spatial SMC distribution in the study area is from southwest to northeast,decreasing first and then increasing,while the SSC distribution is gradually increasing from southwest to northeast.Most of the agricultural production areas in the southwest were non or mild saline soil with moderate soil moisture content;most of the cultivated land in the middle were mild or medium saline soil,with insufficient soil moisture content;and most coastal areas in the east were severe saline soil,with abundant soil moisture content,which was consistent with the statistical analysis results.It can be seen that the UAV-based SMC and SSC models can realize the inversion of satellite image,with the inversion effect of the SMC and SSC model according to different salinization grades is better and more reliable.