Explainable machine learning reveals key determinants of biochar soil-application effects on plant growth

Biochar is a key product for waste-biomass valorization and a soil amendment that can enhance fertility and crop growth, yet its agronomic performance varies with feedstock, production conditions, and soil context. Here, we used explainable machine learning to predict biochar-induced plant growth responses by integrating biochar properties (e.g., category, addition amount, pyrolysis temperature, pH, electrical conductivity, and nutrient contents) with soil properties (e.g., pH, electrical conductivity, and nutrient status). Among the evaluated models (random forest, RF; gradient boosting regression, GBR; support vector machine, SVM; and extreme gradient boosting, XGB), RF achieved the balanced train (R 2 = 0.787) and test performance (R 2 = 0.765) and was therefore selected for interpretation. Feature-importance and partial-dependence analyses identified biochar category (35.66%), biochar addition amount (15.25%), and soil pH (7.79%) as the mostly important predictors, with biochar-related variables explaining 72.3% of total importance. Within categories, manure-derived biochar (MB) and sludge-derived biochar (SLB) exhibited the best performance for improving plant growth. Based on partial-dependence analysis, the model suggests a feasible strategy: apply high-temperature biochars (≥700 ℃) at ~25% (w/w), with EC 3,000 μS·cm -1 –4,000 μS·cm -1 , N 3%–4%, and P ≤ 6%, in neutral to mildly alkaline soils, where predicted growth responses are maximized. Leveraging category-specific traits, we translate these patterns into a conditional-optimization paradigm targeting acidity correction, nutrient supplementation, and salinity alleviation, supported by a preliminary cost–benefit and environmental appraisal. This study suggests that explainable machine learning, integrating biochar and soil properties, can help optimize biochar production conditions and application strategies for soil types. • Model-driven optimization recommendations tailored biochar formulations • Biochar category and addition amount dominate, pH is the key soil variable • For soil application, MB and SLB are most suitable to enhance plant growth • In near-neutral soils, 25% biochar (≥700 ℃) is optimal for plant growth