Multi-tiered soil property thresholds for mitigating cadmium accumulation in subtropical rice: a big-data analysis using improved CART.

Cadmium (Cd) accumulation in rice threatens food safety in subtropical paddy systems, where soil properties strongly influence Cd mobility and uptake. However, current assessment often relies on single-factor criteria and fails to capture the hierarchical and region-specific controls of soil properties on Cd enrichment in rice. This study aimed to identify the key soil properties governing Cd bioaccumulation in rice grains and to establish multi-tiered thresholds for safer rice production. Using 85,000 paired soil-rice samples from Hunan Province, China, we combined big-data analysis, logistic regression, an optimized classification and regression tree (CART), variance decomposition, and structural equation modeling to quantify the contributions and pathways of major soil properties affecting the Cd bioaccumulation factor (BCF). The selected soil properties showed negative logistic relationships with BCF, and their importance ranked as pH > SOM > Zn > CEC, with total contributions ranging from 60.3% to 90.2%. Structural equation modeling showed that pH had the strongest total negative effect on rice Cd (-0.694), both directly and indirectly through SOM, CEC, and Zn. CART analysis revealed that when pH < 5.5, BCF remained high and safer production required raising pH while maintaining SOM≥33.5 g kg-1 and Zn ≥ 97.7 mg kg-1; when pH was 5.5-6.0, CEC≥12.0 cmol kg-1 and SOM≥34.1 g kg-1 were needed. These findings provide a novel regional-scale framework for soil management and Cd-risk mitigation in safer rice production.