Study On The Methods Of Soil Salt Extraction In The Yellow River Delta Based On Multi-Source Data

Soil salinization is an important challenge to the sustainable use of land resources,as it has a great impact on crop growth and agricultural production.Monitoring of soil salinity and its spatial distribution over time is conducive to securing efficient agricultural development.With its high resolution and long time series,remote sensing has played an invaluable role in the dynamic monitoring of soil salinization over large areas.In this study,Kenli County,a typical region of the Yellow River Delta,was selected as the research area for a study of soil salinity extraction methods.The main conclusions are as follows:(1)Based on the 2014 Landsat 8 OLI image data and field sampling data,the characteristic parameters involved in soil salt inversion(Albedo,MSAVI,SI,and NDVI)were extracted,and the two-dimensional feature spaces of Albedo-MSAVI,SI-Albedo,and SI-NDVI were constructed and quantified.The relationships between soil salinity and surface biophysical parameters are discussed.An optimal inversion model of soil salinity in the Yellow River Delta was established.The accuracy of the structured monitoring model was verified using the measured salinity data,and the optimal model was obtained.The analysis showed that the SI-Albedo model is the most suitable model for inversion of salinity in coastal areas.The Albedo-MSAVI and SI-NDVI models have a certain utility for extracting salinization information for arid and semi-arid inland areas.(2)Using the 2019 Sentinel-1A radar data and field sampling data,the backscatter coefficients of the radar data were extracted,and the backscatter coefficients VV,VH,VV+ VH,and VV/VH,which had high correlations with soil salinity,were screened.Combined with the measured data of soil organic matter,p H value,and soil salinity,an inversion model was established using multiple linear regression and the BP neural network model method.Comparative analysis shows that the BP neural network model method matches the characteristic relationships between soil organic matter data,p H value,radar image backscattering coefficient,and combination value with soil salinity during the modeling process.A comprehensive comparison of the multiple linear regression model method,the BP neural network model method,and the feature space model method reveals that the feature space model method is more suitable for the inversion of soil salinization in the Yellow River Delta.(3)A more appropriate model was used to invert the soil salinity of Kenli County in 2019,and the spatial changes of soil salinity there were analyzed.Through overlay analysis and other methods,we obtained maps of spatial changes in soil salinity levels in Kenli County from 2014 to 2019.Across the whole region,the salinity generally shows an increasing trend from the southwest to the eastern coastal areas.The degree of salinization increases with proximity to the Bohai Sea,which is consistent with the mechanism of saltaccumulation in this area.