Spatiotemporal Mangrove Dynamics in Cacheu, Guinea-Bissau (2020-2025): Remote Sensing-Based Change Detection and Driver Analysis

Mangrove forests are vital coastal ecosystems both ecologically and climatically, yet they are declining globally. Guinea-Bissau is home to some of West Africa’s greatest mangrove extents, yet contemporary assessments of how, and why, these forests are changing remain limited. This study mapped mangrove dynamics throughout the Cacheu province in Northern Guinea-Bissau between 2020 and 2025. This study utilized Blue Ventures’ Global Mangrove Mapping Methodology (GEM); a Machine Learning mangrove classification tool hosted in Google Earth Engine (GEE). Sentinel-2 satellite imagery was collected during the dry-season and classified at two timesteps. Three canopy density classes of mangrove were mapped alongside four non-mangrove classes. Classification reference areas were selected via high resolution imagery, and 30% of reference areas were kept for validation, achieving an overall accuracy above 95% at both time steps. Total mangrove extent declined by 13% over the study period from approximately 123,000 ha to 107,000 ha. Within this net change, opposing dynamics unfolded across the region: the eastern sub region was characterized by mangrove losses at forest margins, while the western sub region showed canopy densification representative of mangrove recovery. Rice paddies are historically the major source of mangrove deforestation within Cacheu but their extents remained stable across both time steps. Achieved accuracies may partly reflect the classifier's internal consistency rather than true land cover correspondence. Reference areas were derived exclusively from remote sensing interpretation, with no field-based ground truthing conducted. Future studies integrating in-situ validation would better constrain classification uncertainty and strengthen confidence in the reported patterns of change.