Related Experiment Video
Updated: Sep 19, 2025

Two-Dimensional Visualization and Quantification of Labile, Inorganic Plant Nutrients and Contaminants in Soil
Published on: September 1, 2020
Diffusive gradient in thin films combined with machine learning to discern the accumulation characteristics and
Zhen Wang1, Chuangchuang Zhang2, Danlei Lv2
1Institute of Agricultural Product Quality, Safety and Nutrition, Tianjin Academy of Agricultural Sciences, Tianjin 300381, China.
Abstract:
The dietary exposure risk of cadmium (Cd) in rice is significantly higher than that of copper (Cu), while the co-migration of Cd and Cu in the soil-crop system may enhance the bioavailability of pollution, thus making rapid and accurate prediction of their accumulation in crops essential for assessing the risks associated with cultivated land. However, the previous prediction methods based on the total content of heavy metals in soil or in a single form have certain limitations. We collected 82 pairs of soil and rice samples from Hunan Province in China and used diffusive gradient in thin films (DGT) technology combined with chemical extraction (ultrapure water, acidic ammonium oxalate, calcium chloride) to explore the bioavailability characteristic of Cd and Cu, while machine learning (ML) models for Cd and Cu accumulation in rice were developed by integrating soil physicochemical parameters and heavy metal speciation. Results revealed that when the pH exceeded 7, the efficiency of Cd extraction by CaCl2 (CaCl2-Cd/Total Cd) decreased sharply from 43.5 % to 7.8 %, whereas Cu extraction was not affected. The difference between DGT and chemical extraction revealed that the migration risk of Cd in soil is much greater than that of Cu. The release of Cd is coupled with the reduction of Mn oxide (p < 0.01), The Cd/Mn ratios of different forms are highly consistent, and the Cd/Mn ratios in the ultrapure water and DGT are significantly positively correlated with those in the CaCl2 (r = 0.77, 0.89, respectively). The Random Forest model showed excellent performance for Cd in the rice roots (R2 = 0.72), whereas the Extreme Random Tree model exhibited the best performance for Cu (R2 = 0.38). Feature importance analysis revealed that the exchangeable Cd and Mn had the greatest impact on the Cd risk in root, whereas dissolved Cu and pH had the greatest impact on Cu. This study provided a theoretical basis for understanding the accumulation characteristics and driving factors of Cd and Cu in the soil-rice systems of rice-growing areas.

