Advanced Tree Detection and Classification for Afforestation Monitoring Using High-Resolution Satellite Imagery: A Case Study from Medina, Saudi Arabia

This study presents an innovative monitoring methodology developed for Saudi Arabia’s national afforestation campaign, with a focus on a key implementation site in Medina, where 50,000 trees are being planted. The initiative forms part of a broader national goal to plant 5 million native desert plants by 2030, supporting Saudi Vision 2030’s environmental and sustainable development objectives. The methodology combines advanced remote sensing, machine learning, and distributed computing to provide a scalable solution for vegetation monitoring and land management. Using high-resolution Pléiades Neo 3 satellite imagery (0.27-meter spatial resolution) covering 1.20 km2, the analysis employed Principal Component Analysis (PCA), multiple vegetation indices (NDVI, SAVI, MSAVI2, GNDVI), and Random Forest classification. Post-processing included the Circular Hough Transform for accurate tree detection and counting. The results identified 18,442 individual trees, classified into three size categories: 1,074 large trees (average crown diameter of 3.8 meters), 3,181 medium trees (2.3 meters), and 14,187 small trees (1.1 meters). Validation procedures confirmed 92% overall detection accuracy for mature vegetation, demonstrating the methodology’s robustness and its potential for large-scale application across Saudi Arabia’s afforestation efforts.

This research was led by Geosystems Middle East (GSME) as prime contractor, in collaboration with Planetek Italia, Geophysical Applications Processing (G.A.P.), and the National Center for Vegetation Cover (NCVC). It reflects effective international and local cooperation in support of Saudi Arabia’s environmental goals. NCVC provided strategic oversight and facilitated fieldwork, while GSME coordinated technical partners and stakeholder engagement. The study was supported by the Saudi Green Initiative and Vision 2030. Ground truth data collection was key to validating the monitoring methodology.