Monitoring Urban Expansion and Green-Space Deficits in Jakarta and Bandung (2017–2025) Using Sentinel‑2, Random Forest, and GEE

Abstract

Rapid urban growth in Indonesia is increasing pressure on the legal requirement that cities maintain at least 30% Green Open Space (Ruang Terbuka Hijau). This study examines land use/land cover change and vegetation dynamics in Jakarta and Bandung from 2017 to 2025 using Sentinel‑2 Level‑2A imagery, multiple spectral indices, and Random Forest classification implemented in Google Earth Engine. Four classes were mapped annually: Vegetation, Urban, Water, and Open Land. The classification performed well, with mean overall accuracy above 94% and Kappa around 0.9. Despite similar technical performance, the two cities show contrasting ecological trajectories. Jakarta remains structurally under‑greened, with vegetation increasing only from 8.1% in 2017 to 9.3% in 2025, far below the 30% benchmark. Bandung instead undergoes rapid green space loss, with vegetation declining from 22.17 km² to 8.53 km², a net loss of over 61% in less than a decade. These findings indicate that a single national target can mask divergent local realities: Jakarta faces a restoration challenge in a saturated built environment, whereas Bandung must prioritize protection of remaining vegetated land at its urban outskirts. The workflow presented in this study provides a GeoAI-based monitoring approach that can be replicated for other Indonesian cities.