pH-Dependent Responses and Thresholds of Cadmium Fractionation in Two Soil Types with Different Contamination Origins

The toxicity and bioavailability of cadmium (Cd) in soils are governed not only by total Cd concentration but also by soil pH and Cd fractionation. Exchangeable Cd (Ex-Cd) responds nonlinearly to pH, yet its threshold characteristics and their dependence on contamination origins and soil type remain unclear. This partly limits precise risk management of Cd-contaminated soils. In this study, four field soils from southern China, including two red soils and two paddy soils representing geogenic enrichment and anthropogenic contamination, were subjected to 11 target pH levels ranging from 3.5 to 8.5. Cd fractions were determined using the Tessier sequential extraction procedure. LOESS and segmented regression were used to estimate pH breakpoints, with bootstrap resampling used to quantify breakpoint uncertainty, and CatBoost–SHAP models were applied to characterize multivariable associations with Ex-Cd. Ex-Cd decreased strongly with increasing pH in all four soils. Estimated breakpoints were 4.5 and 3.9 for the two red soils and 6.2 and 6.0 for the two paddy soils, respectively, although bootstrap confidence intervals overlapped substantially among soils. Soil pH was consistently the dominant predictor of Ex-Cd in the CatBoost–SHAP models, whereas the contributions of secondary predictors varied among soil scenarios. The lower breakpoint estimates in the two red-soil scenarios and the higher estimates in the two paddy-soil scenarios therefore represent a descriptive pattern rather than statistical evidence of a general soil-type effect. Because soil type and site background were not independently replicated, these comparisons should be interpreted as case-based observations for the four soils examined. The findings provide evidence for pH-dependent changes in Cd lability and identify candidate transition ranges for future multi-site field validation rather than universal management thresholds.