Hyperspectral classification of tree species for precision water management in Mediterranean agriculture

Effective agricultural water management relies on innovative tools like remote sensing to assess irrigation needs at basin scales. This study, part of the WAter DIgital Twin (WADIT) project, aims to support water balance modeling by precisely identifying tree species using airborne hyperspectral imagery. Conducted at an experimental farm in Valenzano (Southern Italy), the research employs the CASI sensor to acquire high-resolution data of irrigated tree crops (e.g., olive, fig, grapevine). It evaluates supervised classification techniques—Support Vector Machines (SVM), Linear Discriminant Analysis (LDA), and Random Forest (RF)—to map species from hyperspectral images after pre-processing and feature extraction. Results demonstrate exceptional accuracy (up to 99%), enabling detailed hydrological modeling for tailored irrigation strategies. The study advances precision agriculture and sustainable water management in Mediterranean regions, offering a methodological framework for integrating remote sensing into digital twin systems like WADIT to address water scarcity.