Improved soil salinity estimation in arid regions: Leveraging bare soil periods and environmental factors

Authors: MA, WEITONG - Northwest A&f University; CUI, XIN - Northwest A&f University; HAN, WENTING - Northwest A&f University; Zhang, Huihui; ZHANG, LIYUAN - Jiangsu University

Submitted to: iScience
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/25/2025
Publication Date: 6/27/2025
Citation: Ma, W., Cui, X., Han, W., Zhang, H., Zhang, L. 2025. Improved soil salinity estimation in arid regions: Leveraging bare soil periods and environmental factors. iScience. Article e113020. https://doi.org/10.1016/j.isci.2025.113020.
DOI: https://doi.org/10.1016/j.isci.2025.113020

Interpretive Summary: When farmland gets saltier year after year, it is harder to grow crops. A more accurate way to estimate soil salinization is needed. Existing methods can be skewed by long-term environmental patterns. This study explores a novel approach to tackle this challenge. The study was conducted in arid regions in China. We used satellite imagery and ground measurements to analyze how different factors, like soil water content and soil composition, relate to soil salinity. Here's the key findings: looking at the bare soil during two specific times (before planting and after harvest) gave a clearer picture of salinity compared to a single analysis. Additionally, the amount of clay in the soil proved to be a strong indicator of saltiness. By combining these insights and using the best-performing algorithm, we can create highly accurate maps revealing soil salinity across large areas. This could be a powerful tool for farmers to manage their land more effectively and ultimately grow more food.

Technical Abstract: In the context of global land degradation and increasing salinization of cultivated land, obtaining the degree and distribution of soil salinization accurately and sustainably is critical to achieving land management and improving agricultural productivity. This study explores a novel approach to improve soil salt content (SSC) monitoring by minimizing the influence of long-term environmental variations and incorporating relevant environmental factors. Specifically, focusing on exposed cultivated land in the Hetao Irrigation District (HID) of Inner Mongolia, we analyzed the response sensitivity of multispectral indices, soil water content (SWC), and soil mechanical composition (SMC) to SSC during pre-seeding and post-harvest bare soil periods using Sentinel-2 imagery and ground data collected concurrently. We also evaluated the performance of three machine learning algorithms (random forest regression (RFR), support vector regression (SVR), and artificial neural network (ANN)) in SSC estimation models. Our findings revealed the response of spectral indices to SSC in different bare soil periods. While most vegetation indices (VIs) showed minimal sensitivity, salinity indices (SIs) exhibited a strong correlation with SSC, which further increased upon dividing the bare soil periods. Additionally, we observed a significant correlation between soil clay content and SSC (r = 0.50), with this relationship demonstrating greater stability compared to SIs during bare soil periods. The R2 of models was enhanced by 10.2% to 55.7% by dividing bare soil periods. Among the algorithms, the SVR model outperformed RFR and ANN, with an R2 of 0.77, RMES of 0.11%, LCCC of 0.81, and RPD of 1.92. The study proposed a novel and highly accurate method for large-scale soil salinization mapping. By leveraging the optimal SVR model and dividing bare soil periods, this approach offers a powerful tool for precision agriculture and sustainable land management.