Prediction of Selenium-Rich Maize Planting in Selenium-Poor Land Based on Random Forest Model
Selenium (Se) is an essential trace element for human health, and dietary intake through Se-rich crops is the primary route. However, total soil Se content does not directly reflect the bioavailability to plants, which depends largely on soil available Se. This study, conducted in Shipai Town, Longshan County, Hunan Province, used 1:50,000 land quality geochemical survey data to investigate factors influencing the Se bioaccumulation coefficient in maize kernels. Soil pH, CaO, and MgO were identified as significantly positively correlated with the bioaccumulation coefficient and were selected as proxies for soil available Se. A random forest (RF) model was developed to predict maize grain Se content and assess the feasibility of cultivating Se-rich maize in low-Se farmland. Results showed that although soil Se was deficient, 53.64% of maize grain samples met the Se-rich product standard (0.02–0.30 mg·kg−1). Compared with multiple linear regression, the RF model exhibited higher accuracy and reliability. The RF model predicted that 40.91% of farmland in the study area is suitable for natural Se-rich maize cultivation, representing a 25.86% increase over the area identified by soil total Se alone. This study provides a novel methodological framework for planting natural Se-rich maize in Se-deficient regions, validating the potential for such cultivation.