Monitoring Model Of Saline-Alkali Soil In The Dry And Wet Seasons In Yinbei Area Of Ningxia Based On Multi-Source Remote Sensing

Soil salinization is one of the major soil problems affecting ecological security and stability.The degree of soil salinization is directly related to the sustainable utilization of soil in arid areas.According to statistics,salinization of soil accounts for 7.26%of the world's land area,and the proportion has continued to grow.At present,the problem of soil salinization in the arid and semi-arid regions of Northwest China has severely restricted the sustainable development of local agriculture,In order to further increase the area of reserve arable land in my country,it is urgent to accurately monitor,develop and utilize soil salinization information.Aiming at the problem that the salinized soil in Yinbei area of Ningxia is widely distributed,but the accuracy of remote sensing inversion is not high,this paper takes Pingluo County in this area as the research area,using remote sensing technology,such as soil science and statistics analysis method,combining with the study area field hyperspectral,determination of indoor soil physical and chemical indicators,as well as Landsat 8 OLI image data,analysis of the region's dry season and wet season statistical characteristics of soil salinity index,clear the differences and similarities between the measured hyperspectral and Landsat 8 OLI,parse the dry wet season the sensitivity of different source spectrum data of soil salinity index,and then using the method of pearson correlation coefficient(PCC),stepwise regression(SR),gray relational analysis(GRA)and principal component analysis(PCA)filter sensitive spectrum,Finally,partial least squares regression(PLSR),support vector machine(SVM),ridge regression(RR),Back Propagation neural networks(BPNN),and geographically weighted regression(GWR)method were used to establish soil salinity and pH inversion models in dry and wet seasons in the study area,and the optimization and verification were carried out,which provided a theoretical basis for the identification,inversion,prevention and utilization measures of soil salinization in local and similar areas.The main conclusions of this study are as follows:(1)In Yinbei area,the mean soil salt content and pH value in dry and wet seasons were relatively high on the whole,which were 6.17 g·kg-1 and 9.28,5.28 g·kg-1 and 9.11,respectively,and the soil salinization degree was relatively serious.The mean value of soil salt and pH in the dry season was higher than that in the wet season,and there was a strong variation of soil salt in both dry and wet seasons,and the variation degree of soil salt in the dry season was greater than that in the wet season,while the pH value in both dry and wet seasons was weak variation.The soil spectral characteristic curves under different soil salt content and pH value tended to be consistent in terms of the change trend,and the reflectivity in visible light increased gradually with the increase of the index content.(2)The content of Na+,Cl-and SO42-in the soil in the study area was high,which generally belonged to the salinized soil of sulfate and chloride type.In dry season,except for medium variation of HCO3-value,the others showed strong variability,Na+,Mg2+,Cl-,and CO32-in the wet season show strong variability,and K+,Ca2+,HCO3-,and SO42-show moderate variability.In both wet and dry seasons,Na+accounted for more than 85%and 93%of the total amount of four cations,the sum of SO42-and Cl-accounted for 88.3%and 91.5%of the total amount of anions.(3)In the hyperspectral and image spectra,the bands and salinity indices sensitive to soil salinity and pH value were different in different seasons.Dry and wet season the measured weight sampling frequencies and salt sensitivity index and soil salinity were characterized by significant,dry season Bandl,SI and wet season Bandl,S2 than other bands and salt sensitive index,dry season image only Band7,NDSI,S1 and S2 through inspection,wet season image bands and salt index passed the inspection,including Band7 and NDSI relatively more sensitive.The measured NDSI and image S3 were the most sensitive to soil pH value in dry season,and the image S3 was the most sensitive to soil pH value in wet season.On the whole,both measured and image specific bands and salinity index have good potential to identify soil salinity and pH value in the dry and wet seasons in the study area.(4)Under different soil salt content,the reflectance of the hyperspectral band after resampling was significantly correlated with the reflectance of image band.The SR group model achieved the best inversion effect,whereas,the PCC and GRA groups indicated advantages and disadvantages in different regression algorithms,after compared of the R2,RMSE and RPD of the salt salinity inversion model under the three filter variables of PCC,GRA and SR.In the five inversion models of soil salinity,the GWR model showed a higher accuracy.The SVM and BPNN algorithm performed similarly in the models,based on different variable groups,whereas,the RR performance was the worst,particularly a serious "overfitting" phenomenon in the PLSR model.The evaluation results demonstrated the superiority of the local regression over the global regression model for soil salinity.The measured GRA-GWR model in dry season achieved the best inversion effect,where the values of RP2 and RPD were 0.9355 and 4.49,in the wet season,whereas,the imaged PCC-GWR model obtained the best inversion effect,where the values of RP2 and RPD were 0.9570 and 4.83.(5)In terms of pH value inversion,the number of measured and image characteristic parameters determined based on PCA in dry and wet seasons was relatively stable,while the number of soil characteristic spectra determined based on GRA method in wet seasons was relatively small.Among the pH inversion models based on the three model input variables,the SR group model had the highest accuracy and stability.The inversion accuracy of global regression model was better than that of local regression model,the global regression models(BPNN,SVM,and RR)were superior to the local regression model(GWR),with better inversion results obtained in the wet than the dry season.The BPNN and SVM models performed better than the RR model,and the PCA-SVM model based on measured data in the dry season achieved the best overall performance,where the values of RP2 and RPD were 0.9724 and 5.76,the imaged GRA-SVM model obtained the best inversion effect,where the values of RP2 and RPD were 0.9732 and 5.12.