Answers are grounded in this paper and cite their sources.
Deep soil fertility management is decisive for the productive potential of soybean because it expands the volume of soil effectively explored by roots and reduces physical and chemical constraints to plant growth. This study aimed to quantify the relative contribution of soil physicochemical attributes, sampled from 0 to 200 cm depth, to the prediction of soybean yield in high-performance fields from the Brazilian Soybean Strategic Committee (CESB). A Random Forest classification model was fitted and evaluated using out-of-bag (OOB) error, class purity–based metrics, SHAP values for model interpretation, and partial dependence curves combined with a purity metric across yield classes. The model showed adequate predictive performance (OOB error ≈ 17.5%), and five predictors were consistently important across yield classes and depths: clay content, cation exchange capacity (CEC), phosphorus, pH and copper. We conclude that management decisions targeted at high yields should consider sufficiency levels by depth layer, integrating liming, gypsum application, fertilization and decompaction practices that maximize the volume of soil that can be effectively explored by roots and the extension of the ideally exploitable profile. The results reinforce the need for deep soil diagnosis, with sampling beyond the traditional 0–20 and 20–40 cm layers, and demonstrate the potential of machine-learning approaches to integrate large volumes of soil data and support more accurate, yield-oriented management recommendations.
No takes yet. Share an insight, caveat, or question.
Coutinho et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: