Source-oriented probabilistic risk assessment of sediment heavy metal(loid)s in a typical fireworks-production basin with mixed industrial and agricultural influences.

Fireworks production introduces persistent heavy metals and metalloids into river sediments, yet source-specific ecological and health risks under mixed land uses remain unclear. Here, we developed an integrated framework coupling Positive Matrix Factorization (PMF) with Fractional Moment-Maximum Entropy (FM-MEM)-based probabilistic health risk assessment for Liuyang River Basin, the world's largest fireworks base, to trace source-to-risk pathways of eight metalloids: Cr, Mn, Co, Cu, Zn, As, Cd, and Pb. Seasonal variation, spatial patterns, chemical speciation, source contributions, and health risks were systematically investigated. Results showed that concentrations were generally higher in winter, with Cd, Mn, and Pb identified as priority ecological risk metals due to high bioavailable fractions. PMF resolved five sources: natural, traffic, industrial, fireworks, and agricultural. Probabilistic health risk assessment indicated acceptable non-carcinogenic risks but non-negligible carcinogenic risks; children were the most vulnerable, with Cr and As as the main carcinogenic drivers. FM-MEM efficiently captured risk distributions with substantially lower computational requirements than Monte Carlo simulation. Source-oriented analysis revealed decoupling between metal loading and health risks: traffic dominated non-carcinogenic risk, while agriculture dominated carcinogenic risk. These findings provide guidance for source-specific remediation and risk management in fireworks clusters and mixed land use watersheds.