Machine learning-based evaluation of seed priming and biostimulant applications in rainfed wheat

Rainfed wheat suffers from water scarcity and micronutrient deficits, calling for innovative practices. This study tests zinc sulphate (ZnSO₄) seed priming combined with foliar iron (Fe), zinc (ZN), and manganese (Mn) (various formulations) and uses multiple machine-learning models to predict agronomic outcomes.

A field experiment was conducted in northwestern Iran using a factorial randomized complete block design with four replications. Treatments included three ZnSO₄ priming concentrations (0.1%, 0.2%, 0.3%) and five foliar sprays (conventional, magnetized, and nano-formulations of Fe, Zn, and Mn), plus water controls. Agronomic traits (e.g., spike emergence, plant height, yield, protein content) were measured. Data were analyzed with ANOVA and modelled using eight regression algorithms (Linear, Ridge, Lasso, Elastic Net, support vector regression (SVR), Random Forest, eXtreme Gradient Boosting (XGBoost), CatBoost).

The results demonstrated that seed priming with ZnSO₄ considerably expedited spike emergence, enhanced plant height, and increased biological yield, with elevated ZnSO₄ concentrations intensifying these effects. Foliar application of biostimulants enhanced yield components, grain yield, and protein-related characteristics. The maximum grain production (1,648 kg ha^-1) was attained with 0.3% ZnSO₄ seed priming in conjunction with nano foliar application of Fe, Zn, and Mn, indicating synergistic nutrient uptake/use or, alternatively, the potential of prediction approaches such as regularized regressions (Ridge, Lasso, Elastic Net) were most accurate for biological yield, protein content, and harvest index, whereas XGBoost/CatBoost captured nonlinearities but were less consistent for seed-related features. Overall, ZnSO₄ priming combined with nano biostimulants markedly enhances rainfed wheat performance.