Abstract
<title>Abstract</title> <p>Heavy metal contamination in agricultural soils poses potential risks to the ecosystem and human health, however, the linkage between pollution sources and health risks remains poorly understood in complex alluvial environments. This study established an integrated framework combining spatial analysis, Absolute Principal Component Score-Multiple Linear Regression (APCS-MLR) source apportionment, Monte Carlo simulation, and Sobol sensitivity analysis to investigate heavy metal contamination and associated health risks in 741 farmland soil samples from the Dashetai region of the Hetao Plain, Inner Mongolia. The results showed that concentrations of eight heavy metals (As, Cd, Cr, Cu, Hg, Ni, Pb, and Zn) were generally below the national risk screening values but exhibited significant spatial heterogeneity. The APCS-MLR model identified three major sources: mixed natural-agricultural inputs (dominant for Cr, Ni, Zn, and Cu), industrial-traffic activities (contributing 34.01% to As, Cd, and Pb), and localized sources (17.19% for Hg). Monte Carlo simulation indicated that non-carcinogenic risks were acceptable, whereas carcinogenic risk was mainly associated with As exposure. Approximately 64.86% of children exceeded the acceptable carcinogenic risk threshold, with oral ingestion accounting for more than 94% of total exposure. Sobol analysis further revealed that soil As concentration was the dominant factor controlling carcinogenic risk uncertainty. These findings highlight the importance of integrating source contribution and risk contribution analyses to identify priority contaminants and support targeted heavy metal risk management in agricultural regions.</p>