Pollution risk assessment and source apportionment of metals in surface sediments of Dongping Lake, China: Quantitative identification using PMF assisted by SOM.

Lake sediments act as important sinks for watershed-derived materials and contaminants, preserving integrated signals of both natural processes and anthropogenic activities. To improve source identification within complex environmental matrices, the robust nonlinear topological clustering capability of self-organizing maps (SOM) was utilized to spatially constrain and aid the identification of factors derived from positive matrix factorization (PMF). The framework was implemented in Dongping Lake, a key water-transfer lake along the Lower Yellow River, China. The pollution characteristics, ecological risks, and source contributions of 12 metals were systematically investigated. All metals exhibited varying degrees of enrichment relative to regional background values, with Hg and Cd showing the strongest accumulation, reaching 2.9- and 1.7-fold the provincial background levels, respectively. Hg displayed extremely high spatial variability (CV = 285.71%), indicating strong control by localized sources. Pollution assessment suggested an overall slight contamination status; however, the mean potential ecological risk index (PERI) reached 254.01, and 88.57% of sampling sites exceeded the low-risk threshold, spanning moderate to very high ecological risk categories. Hg was identified as the dominant ecological risk driver. The topological structure of the SOM revealed a nonlinear spatial pattern characterized by "geological process-diffuse deposition-localized enrichment" alongside an isolated Hg anomaly. This spatial pattern aligned strongly with the three main sources resolved by the PMF model-namely natural rock weathering (51.80%), agricultural input (32.79%), and mixed industrial sources (15.42%)-thereby achieving topological constraint and accurate identification of the major pollution sources. The industrial source, dominated by coal combustion and residential emissions, was identified as the primary driver of extreme Hg enrichment. These findings demonstrate that SOM-derived topological constraints can improve the environmental interpretability of PMF solutions and provide a transferable framework for source identification and risk-oriented management of metals in shallow lake ecosystems.