Advanced matter-element extension model and machine learning for source-specific probabilistic health risk assessment of heavy metals/metaloids in soil-rose systems in Kushui, Northwest China

Abstract

Heavy metals/metaloids (HMs) contamination in soil-agricultural systems is a global environmental concern, with significant implications for food safety and public health. However, research focusing on accurate contamination assessment and precise source identification with source-specific probabilistic health risk in soil-rose systems, particularly in semi-arid regions, is scarce. This study introduces a comprehensive framework utilizing the contamination factor (CF), geo-accumulation index (Igeo), pollution loading index (PLI), and improved matter-element extension model (IMEM) for enhanced contamination evaluation, combines correlation analysis (CA), self-organizing map (SOM), and positive matrix factorization (PMF), for precise source analysis, and applies source-specific probabilistic health risk assessment in the soil-rose system. The results showed that approximately 13.40 %, 63.92 %, 84.54 %, 89.69 %, 89.69 %, 92.78 %, 94.85 %, and 65.88 % of Hg, Cr, Cd, Ni, Pb, As, Zn, and Cu in soils exceeded their background values. Despite low bioconcentration in roses, Hg levels in 5.62 % of samples surpassed safety standards. Soils displayed varying levels of HM contamination, especially from As, Cd, Cu, and Zn, largely attributed to household coal burning, industrial emissions, fertilization and pesticide application, and road traffic based on multiple approaches. Source-specific probabilistic risk assessment indicated soil HMs posed carcinogenic threats to adults and children, while roses were considered safe for consumption. Road emissions were identified as the principal contributors to health risks, with Cr and Ni identified as priority control elements. This framework refines contamination assessment, enhances source apportionment, and provides a scalable model for global agricultural and environmental risk management.

Keywords

• Metal pollution
• Monte Carlo simulation
• Positive matrix factorization
• Risk assessment
• Soil-rose system
• Source apportionment