Toward Cross-Species Crop Se Content Prediction Using Random Forest Modeling

Abstract

Selenium is an indispensable trace element in the human body that plays an important role in maintaining life activities. The consumption of Se-rich crops provides a practical and effective way for the body to supplement Se. However, the Se content in crops is affected by the soil Se content and the interactions between other elements in the soil. In this study, the Tibetan Plateau of China was chosen as the study area. The random forest algorithm was applied to select four key indicators—selenium (Se), bioavailable phosphorus (P), cadmium (Cd), and bioavailable copper (Cu)—from 29 soil variables to predict the Se content in rapeseed, wheat, potato, pasture, and chrysanthemum crops. The results showed that, despite the rich soil Se resources in the Tibetan Plateau, only 20% of the crop samples met the national Se enrichment standard (>0.07 mg kg−1). Compared with the traditional multiple linear regression method, the random forest model is more accurate, efficient, and reliable in predicting the Se content of crops. In cross-species crop prediction, which refers to the simultaneous cultivation and analysis of multiple distinct crop species within the same agricultural setting, the random forest model demonstrated superior performance, marking a significant breakthrough in cross-species crop research. This approach effectively eliminates the tedious process of conducting repetitive individual evaluations for different crop types in the same region, highlighting its innovative significance. Meanwhile, the Tibetan Plateau, known as the “Roof of the World”, is also of great research value. These results provide valuable references for the planning and management of Se-enriched farmlands, which will help improve the yield and quality of Se-enriched crops and promote the growth of farmers’ interests.