Refining source-oriented ecological-health risk assessments of soil potentially toxic elements using region-specific geochemical baseline values in a high geological background region.
Naturally mineralized regions exhibit elevated geochemical backgrounds of soil potentially toxic elements (PTEs). However, the use of uniform geochemical baseline values (GBVs) frequently leads to misclassification between natural enrichment and anthropogenic contamination, hindering accurate source identification and risk control. To address this issue, this study established a region-specific GBV framework and integrated multi-receptor models with the Geo-detector and Partial Least Squares Structural Equation Modeling (PLS-SEM) to achieve source-oriented comprehensive risk assessments and elucidate underlying driving mechanisms in a high geological background region of northwestern China. Results showed that soil PTE enrichment was governed by four distinct pathways, including soil parent material characterized by Cr, Ni, and Co (A-F1: 26.61%), grazing-related inputs characterized by Cu and Zn (A-F2: 21.09%), mineralization-related background anomalies dominated by Hg (A-F3: 16.09%), and domestic emissions characterized by Pb and As (A-F4: 7.62%). PLS-SEM further demonstrated that soil PTE enrichment resulted from interactions among geological background (0.894, p < 0.001), climatic regulation, and anthropogenic modification (0.448, p < 0.001), revealing hierarchical mechanisms linking environmental drivers, source formation, and PTE accumulation. Meanwhile, low and moderate potential ecological risks occurred in 99.6% and 0.4% of the area, respectively. The risk was mainly driven by mineralization-related anomalies, which accounted for 67.41% of the ecological risk. The total non-carcinogenic health risks were generally within acceptable levels (Total hazard index < 1.00). Deterministic assessment indicated unacceptable carcinogenic risk (Total carcinogenic risk > 1.00 × 10⁻4) in 6.0% (children) and 3.9% (adults) of the study area, whereas probabilistic Monte Carlo simulation revealed higher exceedance probabilities at the 95th percentile (25.97% for children). Natural geogenic sources (A-F1 and A-F3) dominated the non-carcinogenic risk, contributing 65.48% and 68.36% of the THI for adults and children, respectively. Carcinogenic risk was driven by both soil parent material (A-F1: 44.34% for adults, 36.32% for children) and anthropogenic inputs (A-F2 and A-F4: 46.88% for adults, 56.48% for children) of the TCR. Among the characteristic PTEs, Cr and Ni primarily represented concentration-driven risks, whereas Hg posed toxicity-driven risks despite its relatively low concentrations. These findings demonstrate that ecological-health risks in high geological background regions are commonly controlled by source characteristics, elemental toxicity, and exposure pathways rather than contamination levels alone. They provide a scientific basis for source-oriented risk management under region-specific geochemical backgrounds.