Integrating seasonal greenness dynamics for improved forest type discrimination in Lualaba Province (DRC): implications for sustainable forest management

Forests in Lualaba Province, Democratic Republic of the Congo (DRC), are increasingly threatened by logging activities, agricultural expansion, and recurrent bushfires. Yet, little is known about the spatial extent and the structural and functional diversity of the main forest types—dry dense forests, gallery forests, and miombo woodlands. Understanding these differences is crucial, as each forest type faces distinct ecological pressures and provides unique ecosystem services, including climate regulation, biodiversity conservation, and livelihood support. This study evaluates the potential of Sentinel-2 time-series imagery to discriminate forest types by explicitly incorporating seasonal variability in canopy greenness in a seasonally dry tropical context. Monthly NDVI composites for 2024 were used to derive annual and seasonal averages, as well as intra-annual coefficients of variation, capturing vegetation-specific responses to seasonal dynamics. A spatially stratified validation framework implemented in Google Earth Engine was used to assess model robustness. The best-performing classification achieved an overall accuracy of 80.66% and a macro-averaged F1-score of 81% across all eight land-cover classes (Model M2). Results indicate that the dry season (June–September) provides the strongest spectral separation among forest types, underscoring the importance of temporal information for characterising vegetation heterogeneity. By integrating seasonal canopy greenness dynamics into forest-type mapping, this study demonstrates a reproducible, scalable approach for mapping seasonally dry tropical forests. The resulting maps provide an operational framework for forest-type discrimination in seasonally dry tropical landscapes and support sustainable forest management, conservation planning, and climate-resilient land-use strategies in southern DRC and comparable tropical regions.