Machine learning-driven multi-technique source tracing and source-specific probabilistic ecological risk assessment of heavy metal(loid)s in urban river sediments

• Metal(loid)s pollution with sources was investigated in an urban river in China. • An integrated framework was proposed to analyze metal(loid) sources and risks. • Pollution hotspots are prevalent in highly urbanized areas. • Pollution sources were identified through correlation, SOM, and PMF analyses. • Industrial and traffic activities present the greatest ecological risks. Heavy metal(loid)s (HMs) pollution in urban rivers is an urgent environmental concern, but addressing source-specific risks within complex urban environments remains a critical challenge. This study introduces a machine learning-driven framework designed to characterize pollution risks, delineate specific sources, and assess the source-oriented probabilistic ecological risks of HMs accumulating in surface sediments from the Kongtong section of the Jinghe River, Northwest China. Results indicated that Cd and Hg levels (average concentrations of 0.19 mg kg −1 and 0.03 mg kg −1 , respectively) in the Kongtong section were significantly above multiple background values, with hotspots associated with intense human activity. The integration of the geological accumulation index ( I geo ), enrichment factor (EF), pollution load index (PLI), and modified Nemerow integrated ecological risk index ( mNIRI ) confirmed Cd and Hg as dominant ecological risk drivers. Based on mNIRI values, 47.25 % of sites were at moderate risk, while 16.49 % posed higher risk levels. By integrating correlation analysis, super-clustering of self-organizing maps (SOM), and positive matrix factorization (PMF), five major pollution sources were identified: industrial and traffic activities (33.33 %), agriculture (27.21 %), metal manufacturing (15.49 %), natural sources (12.95 %), and smelting/electroplating (11.02 %). The source-oriented probabilistic risk assessment using Monte Carlo simulation identified industrial and traffic activities as the primary contributors to ecological hazards. This study provides a robust framework for accurately tracing multiple pollution sources, offering scientific guidance for targeted risk management and pollution control in urban river systems.