Fractal Approach to Measuring Electrical Conductivity Values of Soda Saline-Alkali Soils with Desiccation Cracks in the Songnen Plain, China

The objective of this study was to quantitatively analyze the ability of fractal dimension in characterizing the extent of desiccation cracking, and to develop a new method for salinity prediction of soda saline-alkali soil in Songnen Plain, China. Cracked soil samples with varying salinity levels were selected to quantitatively analyze the relationship between soil salinity and box-counting dimensions extracted from binary crack patterns under different denoising thresholds and four equal-partitions. A series of correlation analyses were also carried out between the electrical conductivity (EC) values and box-counting dimensions derived from different conditions. The relationship between EC values and box-counting dimensions reveals a clear correlation trend of an initial increase followed by a decrease in response to an increase in the denoising threshold (optimal threshold of 400 pixels); however, equal-partitions had no discernible effect on the correlation analysis results. Considering the factors including calculation time and accuracy, threshold of 400 pixels was selected to compute the box-counting dimensions and to develop exponential model for the prediction of soil EC with R-2 of modeling test above 0.61 and R-2 of validation test above 0.63. In addition, the results also indicate that the prediction accuracy of EC values decreased seriously given the box-counting above 1.8 since the soil salinity was obviously underestimated given high salinity levels. All the findings in this study contribute to the understanding of fractal dimensions can be selected as an effective indicator to characterize desiccation cracking for soil salinity estimation.