Comparative Study On EC Prediction Ability Of Soda Saline-Alkali Soil Based On Different Crack Characteristics

The surface of soda saline-alkali soil in Songnen Plain shows obvious shrinkage and cracking with the evaporation of water,and salt content is an important factor for the cracking of soda saline-alkali soil in Songnen Plain.Fractal dimension can quantitatively describe the self-similarity of complex phenomena in nature,and the texture feature of gray level co-occurrence matrix(GLCM)contains important information about the surface structure of objects and the relationship with the surrounding environment.Both of them can describe the development of surface cracks in soda saline-alkali soil,but there are differences in quantifying the cracking degree of saline-alkali soil with different characteristic parameters.The purpose of this study is to analyze the relationship between the characteristic parameters of different types of cracks and the salt content of soil samples,and to establish the best prediction model of electrical conductivity(EC).In order to achieve this goal,201 soil samples with different salinities were selected by means of ground-based remote sensing and CCD high-precision camera,and the samples were pretreated to quantitatively extract the fractal dimension of cracks in the binary image of the soil surface of the samples and the texture features of gray level co-occurrence matrix in different gray levels,different step sizes and different directions,and the texture features were processed by principal component analysis.Then,this study comparatively analyzed the correlation between fractal dimension and principal component of texture characteristics and electrical conductivity,and established three different models of exponential regression,multiple linear regression and multiple stepwise regression of fractal dimension and principal component of texture characteristics on soil electrical conductivity.Then,the prediction effects of the three models on soil electrical conductivity were analyzed and compared,and the models were verified.The best model was selected to predict soil electrical conductivity,which reduced the time and cost of field sampling and measuring electrical conductivity.At the same time,this method can lay a foundation for remote sensing monitoring of salt content in cracked saline-alkali soil at different scales in the future,and provide guidance for ecological restoration,agricultural production and engineering construction.The main results are as follows:(1)Clay content and clay mineral composition of soda saline-alkali soil in Songnen Plain have a weak effect on soil shrinkage and cracking,but soil salt content has an important influence on the process and degree of soil surface shrinkage and cracking.(2)There is a good correlation between the fractal dimension characteristics of local cracks in the sample and EC value.When calculating the fractal dimension of local cracks in the sample,the image size of local cracks in the sample is large enough,and the cracks in the soil are widely distributed.The box covering the image can basically cover the cracks in the soil well.Therefore,the correlation between the fractal dimension and EC value under different bisection conditions is not significant,and the larger the image size,the smaller the difference.(3)There is obvious correlation between different texture parameters of soil cracks.Principal component analysis can compress the data,while retaining most of the characteristics of different texture features.The first principal component feature value of texture features is higher than the second principal component,and the correlation between first principal component and EC value of local texture features is higher than that of the second principal component.(4)The characteristics of soil cracks have good prediction ability for soil salinity level.Exponential regression model,multiple linear regression model and multiple stepwise regression model have good prediction accuracy for EC,and their correlation coefficients r are all greater than 0.6.Especially,there is a good exponential relationship between fractal dimension characteristics and EC.Exponential regression model can effectively predict soil salinity below 2 ds/m.In order to improve the calculation efficiency and ensure the prediction accuracy,the exponential regression model for predicting EC with fractal dimension is the best.