Mapping hotspots and identifying drivers of lead bioaccumulation in Oryza sativa L. in tropical agroecosystems

Tropical rice systems exhibit high annual rates of heavy metal accumulation, requiring accurate identification of accumulation drivers in rice-growing ecosystems to ensure regional food security. Therefore, we collected 229 paired soil and rice samples across Hainan Island, China, revealing distinct spatial patterns of Pb contamination hotspots. By integrating 21 environmental variables, nine machine learning models were systematically compared, with the random forest algorithm emerging superior (R² = 0.842, RMSE = 0.010). Analyses showed that despite a low mean bioaccumulation factor (0.011), 11.4 % of the soils and 22.7 % of the brown rice grains exceeded national Pb limits. Geospatial profiling identified northwestern regions as primary soil Pb hotspots, whereas rice Pb bioaccumulation exhibited tripartite clustering in northwestern, southwestern, and western zones. Mechanistic interpretation via Feature importance and Shapley additive explanations highlighted soil Pb concentrations, annual precipitation, and mean temperature as pivotal bioaccumulation regulators. To mitigate risks, targeted agronomic strategies-including pH optimization and calcium-selenium enrichment-are proposed to immobilize Pb in tropical soils. This predictive framework quantifies Pb concentrations in rice, explains contamination mechanisms, and provides actionable insights for heavy metal management in island agroecosystems.