What question did this study set out to answer?

The aim is to identify environmental factors influencing soil contamination by potentially toxic metals in southeastern Iran.

April 23, 2026Scientific Reports0 citationsOpen Access

Environmental drivers of soil potentially toxic metals identified through multi-scenario machine learning in an arid region of Southeastern Iran

ZEZohre Ebrahimi-Khusfi MSMojtaba Soleimani-Sardo GJGhobad Jalali

Key Points

The aim is to identify environmental factors influencing soil contamination by potentially toxic metals in southeastern Iran.
Measured soil properties and potentially toxic metals in 104 surface soil samples.
Utilized Random Forest and Extreme Gradient Boosting models under five different scenarios.
Applied Shapley Additive Explanations to determine the influence of various factors on predictions.
Optimal predictions for arsenic and chromium yielded R² values of up to 0.55 using XGB with RSAD.
Lead, cadmium, and zinc showed higher predictive strength with R² up to 0.65 in a combined scenario.
Identified critical contamination zones, especially in areas with human settlements.

Abstract

Soil contamination by potentially toxic metals (PTMs) in dust-prone regions poses significant risks to public health and the environment. Accurate prediction of their spatial distribution and identification of controlling factors are essential for effective risk mitigation. This study determined the most accurate model–scenario combinations for predicting arsenic (As), cadmium (Cd), chromium (Cr), nickel (Ni), lead (Pb), and zinc (Zn) concentrations in the northwestern Jazmourian Basin, southeastern Iran. PTMs and soil properties were measured in 104 surface soil samples (0–10 cm) in the laboratory. Modeling was performed using Random Forest (RF) and Extreme Gradient Boosting (XGB) under five scenarios based on soil properties (SP), anthropogenic factors (AF), geo-based factors (GBF), meteorological elements (ME), and remote sensing auxiliary data (RSAD). Critical regions were identified by combining binary maps of the six PTMs, classified according to their crustal mean concentrations, where higher cumulative values indicated multi-metal contamination zones. The contribution of controlling factors was quantified using Shapley Additive Explanations (SHAP). Optimal predictions for As and Cr were obtained using XGB–Scenario (I) (RSAD) and XGB–Scenario (III) (RSAD + ME + GBF) with R² values of 0.40 and 0.55, respectively. Higher R² values of 0.52, 0.34, and 0.65 were achieved for Cd, Pb, and Zn using XGB–Scenario (V) (RSAD + ME + GBF + AF + SP). RF–Scenario (V) provided the best spatial prediction for Ni (R² = 0.31). Human settlements in the central region were identified as critical zones. SHAP analysis showed that RSAD contributed most to predicting As and Zn, whereas GBF, SP, and ME were dominant for Cr, Ni, Cd, and Pb. Among predictive variables, inverted difference vegetation index (IPVI), band ratio (3/4), geological formations, slope, silt, and wind speed were key controlling factors. These findings provide valuable guidance for environmental planning and soil contamination risk reduction in dust-prone areas.

AIに質問

Bookmark

View Full Paper