The benefit of advanced mapping techniques to address the spatial distribution of soil sulfur—total, gypsum-based, and total-reduced—in the dried Sawa Lakebed

This study employs digital soil mapping (DSM) and Sentinel-2A satellite imagery to investigate the impact of Sawa Lake’s desiccation in Iraq on the spatial variability of soil sulfur. The findings assist in risk assessments. For this, two distinct strata were identified: St1 (never inundated) and St2 (which dried between 2016 and 2023). Ninety-nine surface soil samples (0–20 cm) were taken in July 2022. Total sulfur (Sₓ₎ₓ), was directly measured using an X-ray fluorescence (XRF) spectrometer. The gypsum-based sulfur (S₆ₘ₏) was also directly determined by measuring sulfate in a dilute soil extract following dissolution. The total-reduced sulfur (Sₑ₄₃^r) was then roughly estimated by Sₑ₄₃^r= (Sₓ₎ₓ-S₆ₘ₏) -. Hence, further evaluation is needed to understand how sulfur accounts for the ɛ (non-gypsum sulfate). A total of 17 predictors were calculated from the spectral bands data. Given the small size and homogeneous topography of the study area, climatic and terrain-related attributes were not included as predictors. Regarding the concordance correlation coefficient (CCC) and RMSE values in the calibration-fitted model, the Cubist outperformed the Random Forest. Accordingly, the highest performance was for S₆ₘ₏ (CCC = 0. 79; RMSE = 0. 70), followed by Sₑ₄₃^r (CCC = 0. 63; RMSE = 0. 56), and Sₓ₎ₓ (CCC = 0. 54; RMSE = 0. 47). Multicollinearity analysis and variance inflation factor for feature selection did not improve the model accuracy. Overall, the digital maps showed that the studied sulfur-bearing soil materials were directly related to the distance from the shoreline, demonstrating the dynamic behavior of the study area. Between the two strata, Sₑ₄₃^r increased by 4. 07% (w/w) in contrast, S₆ₘ₏ decreased by 1. 07% (w/w), illustrating that the sulfur speciation and chemistry analysis following the lake desiccation would be of interest. Ultimately, this research demonstrates that DSM can address the knowledge gap regarding the spatial distribution of hard-to-measure properties (e. g. , sulfur-bearing soil materials) using remote sensing data alone, even in dynamic environments such as the dried Sawa Lakebed.