Precision management of Cd-contaminated paddy fields with high geochemical backgrounds in karst regions: integrating Bayesian decision tree and spatial zoning

Cadmium (Cd) contamination in paddy fields with high geochemical backgrounds in karst regions poses significant challenges. This study aims to explore practical technologies applicable to different risk causal zones throughout karstic Cd-polluted paddy fields, by integrating a Bayesian decision tree-based risk warning model with spatial zoning. In the study area, 92.1 % of soil and 30.1 % of rice samples exceeded Cd limits, yet the synergy between elevated Cd levels in soil and rice was weak. The model achieved an accuracy of 93.4 % in predicting Cd risk in rice and generated 13 risk classification rules for evaluating Cd risk levels in paddy fields. Notably, the identified risk zones were primarily concentrated in areas with relatively low total Cd concentrations in the soil. Regional management framework, including passivation techniques, phytoremediation, agronomic regulation, was proposed according to 7 causal rules across the fields. Each zone was associated with a customized method for rice safe production, subsequently developing a spatially precise, differentiated solution. This study offers new insights into managing Cd-contaminated paddy fields in karst regions.