Analysis of forest vegetation changes in Vietnam’s Central Highlands and southeast regions using remote sensing and machine learning

Forest vegetation is pivotal for maintaining ecological equilibrium and providing essential ecosystem services, yet faces persistent decline, particularly in ecologically sensitive regions like Vietnam’s Central Highlands and Southeast. This study aimed to assess the status and dynamics of forest vegetation across these regions from 2016 to 2024, providing a scientific foundation for sustainable forest management. Using Sentinel-2 imagery (10–20 m resolution), GIS, and a Random Forest classifier, we mapped 21 vegetation classes and assessed forest change in Vietnam’s Central Highlands and Southeast regions for 2016–2024. The classifier achieved strong performance (mean precision = 0.81, recall = 0.78, F1 = 0.79). Results indicate a net decline in natural forest area (e.g., Rich evergreen broadleaf forest decreased by 2,166.5 ha; Medium coniferous forest decreased by 1,776.8 ha), while «Other lands» increased by 10,232.5 ha and grasslands/shrub/regenerated trees declined by 10,037 ha, reflecting substantial land-use conversion pressures. The study’s novelty lies in a refined training-sample collection protocol and systematic hyperparameter optimization tailored to highly heterogeneous tropical vegetation and complex terrain, improving classification reliability for large, fragmented landscapes. These quantitative findings provide actionable evidence to support targeted forest management and land-use change mitigation measures. Urgent implementation of sustainable forest management practices and robust biodiversity conservation measures is imperative to protect these valuable forest ecosystems for future generations.