Machine learning methods to predict cadmium (Cd) concentration in rice grain and support soil management at a regional scale

Abstract

Rice is a major dietary source of the toxic metal cadmium (Cd). Concentration of Cd in rice grain varies widely at the regional scale, and it is challenging to predict grain Cd concentration using soil properties. The lack of reliable predictive models hampers management of contaminated soils. Here, we conducted a three-year survey of 601 pairs of soil and rice samples at a regional scale. Approximately 78.3% of the soil samples exceeded the soil screening values for Cd in China, and 53.9% of rice grain samples exceeded the Chinese maximum permissible limit for Cd. Predictive models were developed using multiple linear regression and machine learning methods. The correlations between rice grain Cd and soil total Cd concentrations were poor (R ² < 0.17). Both linear regression and machine learning methods identified four key factors that significantly affect grain Cd concentrations, including Fe-Mn oxide bound Cd, soil pH, field soil moisture content, and the concentration of soil reducible Mn. The machine learning-based support vector machine model showed the best performance (R ² = 0.87) in predicting grain Cd concentrations at a regional scale, followed by machine learning-based random forest model (R ² = 0.67), and back propagation neural network model (R ² = 0.64). Scenario simulations revealed that liming soil to a target pH of 6.5 could be one of the most cost-effective approaches to reduce the exceedance of Cd in rice grain. Taken together, these results show that machine learning methods can be used to predict Cd concentration in rice grain reliably at a regional scale and to support soil management and safe rice production.

Keywords: Cadmium; Food safety; Heave metals; Machine learning; Predictive model; Soil contamination.

Graphical abstract

Fig. 1

The study area and sampling sites of paired soil and rice grain samples in 2016 (n =200 pairs), 2019 (n =201 pairs), and 2020 (n =200 pairs). Map of China is edited on Chinese standard map GS(2019)1697.

Fig. 2

Soil Cd concentrations (a), rice grain Cd concentrations (b), the correlations between soil Cd and rice Cd concentrations (c), and soil pH values (d) in the 2016, 2019 and 2020 surveys. The numbers in (a) and (b) are exceedance rates of Cd in soil and rice samples. Different letters represent significant differences in the median values between different years at a level of 0.05.

Fig. 3

Prediction of rice grain Cd concentrations in 2019 and 2020 surveys based on the machine learning-based support vector machine (SVM) model. (a) Scatter plots of measured grain Cd concentrations versus predicted values (n = 401), consisting of 280 training datasets and 121 test datasets. (b) The relative importance of variables in the SVM model, including soil amorphous Fe-Mn oxides-bound Cd (Fe/Mn-Cd), air-dried soil pH, soil moisture content, and reducible Mn concentration. Spatial distribution maps of measured and predicted grain Cd concentrations in 2019 (c, d) and 2020 (e, f).

Fig. 4

Scenario simulations based on the SVM model. (a) The acidification trend of soils and the exceedance rate of rice grain Cd concentration in the upcoming 30 years. (b) Predicted grain Cd concentrations under scenarios of further soil acidification (∆pH: -0.5 units), liming to increase soil pH by 0.5 units (∆pH: +0.5 units), or liming to a target pH of 6.5. (c) Predicted grain Cd concentrations under scenarios of dry or wet climatic conditions, with the dry and wet climate being assumed at 10% and 90% percentiles of the soil moisture frequency distribution, respectively. (d) Predicted grain Cd concentrations under the scenario of increasing concentration of soil reducible Mn by 20%. The dashed lines in (b-d) represent the Chinese maximum permissible limit of Cd for rice (0.2 mg/kg). Different letters represent significant differences in the median values between different scenarios at a level of 0.05.

Fig. 5

Scenario simulation of liming with a target soil pH of 6.5. (a) The current soil pH, and (b) the current rice grain Cd concentrations and exceeding rate. (c) The lime amounts required to adjust soil pH to a target pH of 6.5. (d) Predicted rice grain Cd concentrations under the scenario of liming to increase soil pH to 6.5 based on the SVM model. The dashed lines in (b) and (d) are the Chinese maximum permissible limit of Cd for rice (0.2 mg/kg).