Effect of cover crops on soil quality, yield, and prediction using machine learning in papaya (Carica papaya L.)

Abstract

The integration of vegetative cover crops and machine learning-based predictive models represents an innovative strategy to enhance the sustainability and productivity of tropical fruit production systems. This study evaluated the effects of four soil cover treatments, spontaneous vegetation, Arachis pintoi, Canavalia ensiformis, and Centrosema macrocarpum, in addition to a no-cover control, on yield performance and soil quality in papaya (Carica papaya L.) cultivation. Agronomic variables, vegetation indices derived from multispectral imagery, and meteorological factors were integrated to develop yield prediction models using Random Forest, K-Nearest Neighbors, and Extreme Gradient Boosting algorithms. Analysis of variance revealed significant differences among treatments (p < 0.05), with Centrosema macrocarpum achieving the highest yield (102.22 t ha-1), representing a 37% increase compared to spontaneous vegetation. Furthermore, cover treatments improved soil pH, suggesting reduced acidity and a positive contribution to the long-term sustainability of the production system. Among the evaluated models, Extreme Gradient Boosting demonstrated the best predictive performance (R² = 0.85; RMSE = 11.56 t ha-1). These findings indicate that the combined use of vegetative cover strategies and precision agriculture tools can optimize decision-making, enhance resource-use efficiency, and strengthen the resilience of papaya production systems.