Estimating and mapping the dynamics of soil salinity under different crop types using Sentinel-2 satellite imagery

Abstract

Soil salinization is one of the main factors contributing to land degradation, affecting ecological equilibrium, environmental health, and the sustainable development of agriculture. Due to the spatial and temporal heterogeneity of soil properties and environmental conditions in a large-scale region, the monitoring accuracy of soil salinization can be challenging. This study investigated whether the classification of diverse crop types on a time series can improve the prediction accuracy of regional soil salinity levels. Specifically, we evaluated the changes in soil salt content (SSC) under diverse vegetation cover over time in the Hetao Irrigation District (HID) using multi-phase Sentinel-2 imagery and ground-truth data collected from June to September 2021 and 2022. Focused on sunflower and maize fields, this study analyzed the impact of classifying these two crop types and examining four distinct time series on the accuracy of SSC estimation. Five indices were selected as characteristic parameters from a pool of 17 vegetation indices (VIs) and 13 soil salinity indices (SIs) derived from satellite images. Moreover, three machine learning algorithms were used to establish SSC estimation models. The findings underscored the efficacy of classifying crop types and considering different time series in enhancing the response sensitivity of spectral indices to SSC and improving modeling accuracy. Among the spectral indices, VIs made more contributions to the SSC estimation model than SIs, achieving the highest coefficient of determination (R2) of 0.71. The artificial neural networks algorithm outperformed the other two machine learning algorithms in terms of accuracy and stability, yielding an optimal R2 of 0.72 and a Root Mean Square Error (RMSE) of 0.15%. This study proposed a modeling and mapping approach that considers crop types and various time series, offering valuable insights for accurately assessing soil salinization, guiding strategies for its prevention and remediation.