Integration of Geodetector and PMF improves source factor mapping for soil heavy metal identification

Soil heavy metal pollution is a persistent environmental problem because heavy metals can accumulate in soils, enter food chains, and pose long term ecological and health risks. Accurate source identification is therefore essential for targeted pollution prevention and risk management. Traditional positive matrix factorization (PMF) models rely mainly on receptor data and provide limited source information. This may weaken the interpretability of source factor correspondence in regions with strong spatial heterogeneity. To address this limitation, this study developed an integrated PMF and Geodetector framework for identifying soil heavy metal sources in Jiangxi Province, China, and for linking sources with ecological and human health risks. Six sources were identified: mineral mining (F1, 10.65%), natural background (F2, 12.57%), industrial activities (F3, 9.65%), mining associated concomitant sources (F4, 20.66%), mixed industrial, natural, and transportation sources (F5, 35.40%), and rare earth mining combined with atmospheric deposition (F6, 11.10%). Geodetector provided quantitative spatial evidence for interpreting PMF factors and reduced the uncertainty caused by empirical source matching. The risk assessment showed that mineral mining dominated by cadmium (Cd), industrial sources dominated by arsenic (As), and atmospheric deposition dominated by mercury (Hg) were the main contributors to ecological risk. Overall, 36.42%, 6.29%, and 0.92% of samples exceeded mild, moderate, and severe ecological risk levels, respectively. Health risk assessment indicated that children had higher noncarcinogenic risk than adults, with hazard quotient values of 0.26-9.21 (mean 1.56) for children and 0.04-1.48 (mean 0.26) for adults. Industrial sources dominated the major carcinogenic and noncarcinogenic risks, with As contributing more than 46% of total noncarcinogenic risk and As plus chromium (Cr) contributing more than 99% of carcinogenic risk. The results demonstrate that integrating spatial source information into PMF improves source factor mapping and provides more actionable evidence for priority source control in heterogeneous mining regions.