Dataset for Physiological-Driven Irrigation Scheduling in Ananas comosus via Hybrid Machine Learning (PIML-GB)

Dataset Description This dataset provides a comprehensive multimodal matrix containing N=150 independent observations. It is specifically designed to validate a Physics-Informed Machine Learning (PIML) framework for precision irrigation in 5 hectares of commercial pineapple crops (Ananas comosus var. MD2) located in Tauramena, Colombia. The data covers a six-month vegetative cycle. The repository includes high-resolution multispectral imagery (0.82 cm/px) obtained from UAV flights, microclimatic records from a modular IoT agrometeorological station recorded at 10-minute intervals, and mechanistic soil water balance simulations derived from the FAO-56 Penman-Monteith standard. These resources enable the reproduction of the Gradient Boosting (XGBoost) architecture, which achieved a predictive performance of R^2=0.851 and an overall accuracy of 91.1% in hydric status classification.