Evidence for non-primary macronutrient limitations on cherry yields in young robusta agroforestry systems in the Ecuadorian Amazon

Robusta (Coffea canephora Pierre ex Froehner) is a vital cash crop for smallholder farmers in the Ecuadorian Amazon. However, fertility recommendations for robusta production are highly variable across contexts, necessitating regionally tailored recommendations to better diagnose yield-limiting nutrients. Across a gradient of input intensities and agroforestry reflective of local practices, we employed the least absolute shrinkage and selection operator (LASSO) regression to identify which soil fertility measures and leaf nutrients best explained robusta yields across replicated management system treatments in the Ecuadorian Amazon. Leaf nutrients, particularly calcium and magnesium, were stronger and more parsimonious predictors of yields than soil inorganic nitrogen and Mehlich-3 extractable phosphorus and potassium. Although the LASSO model provided reasonable yield estimates (R² = 0.74; root mean square error = 0.23 kg tree⁻¹), model underestimation of yields >1.0 kg tree⁻¹ suggests that other factor(s) not captured by soil and foliar nutrient measures may limit cherry production in higher-yielding systems.