Evaluation of cadmium (Cd) transfer from paddy soil to rice (Oryza sativa L.) using DGT in comparison with conventional chemical methods: derivation of models to predict Cd accumulation in rice grains

The consumption of rice contaminated with soil cadmium (Cd) threatens human health. It is essential to ensure the production of rice that meets food quality standards. Therefore, a large-scale field survey was conducted in Zhejiang province, southeastern China, to investigate the relationship between Cd accumulation in rice grains and Cd bioavailability in soil, and thus to establish a model to predict Cd contents in rice grains based on soil properties. For this purpose, a total of 156 paired rice and soil samples were collected. Pearson's correlation analysis revealed that Cd measurements obtained by diffusive gradient in thin films (DGT) had a higher correlation (r = 0.818, p < 0.001) with the Cd in rice grains as compared to the Cd measured by the DTPA, CaCl2, EDTA, and HCl extraction methods, which indicated that the DGT technique was a reliable method for the assessment of Cd bioavailability in soils. In addition, among the four extraction methods, the DTPA-extractable Cd showed the highest correlation with the Cd contents in rice grains. Therefore, we developed two predictive models (model(DGT) and model(DTPA)) to predict Cd levels in rice grains via Cubist multivariate mixed linear regression, using soil DGT-measured Cd, pH, and oxide contents of Ca, Si, and Fe or soil DTPA-extractable Cd, pH, OM, and oxide contents of Ca and Fe as explanatory variables, respectively. The overall model(DGT) and model(DTPA) had R-2 values of 0.95 and 0.93, respectively, and relative error values of 0.30 and 0.33, respectively. Simple correlation analysis showed direct and close relationships between the measured Cd in rice grains and the Cd concentrations predicted by the Cubist model(DGT) and model(DTPA), with R-2 values of 0.979 and 0.922, respectively. Therefore, Cd levels in rice grains could be predicted very well based on the two prediction models, and thus, the two models derived in this study are effective in identifying soils in which the Cd in rice grains will exceed food safety standards, thereby helping to ensure safe rice production.