Precision mapping and driving factors of heavy metal(loid)s in agricultural soils of the Yellow River: An integrated machine learning and Geodetector approach.

Heavy metal (loid) (HM) contamination of agricultural soils threatens global food security and sustainable development. Despite the Yellow River Basin's (YRB) status as a major grain-producing region under complex industrial and hydrological pressures, the basin-scale drivers and spatial patterns of HM pollution within the YRB remain poorly resolved. To address this, we integrated 27,003 sampling sites data from 209 publications with multi-source environmental covariates and combined pollution indices with an optimal parameter-based Geo-detector to identify HM (As, Cd, Cr, Cu, Hg, Ni, Pb, and Zn) pollution levels and their spatial drivers. Optimized genetic algorithm-based machine learning models enabled high-precision predictions of HM distribution. Our assessment identified Cd and Hg as the primary pollutants, with high-risk areas concentrated in the upper and middle reaches of the YRB. Crucially, the predictive mapping revealed three distinct, source-specific accumulation zones: 1) a central industrial corridor dominated by Cd and Cr; 2) lower reaches impacted by mixed agro-industrial sources of Cu and Zn; and 3) an upper-reach zone with geogenic Pb and long-range atmospheric Hg deposition. These findings provide the scientific foundation for spatially-explicit pollution control and management strategies to safeguard food security and public health in this vital agricultural region.