Long-term Landsat time series reveals mangrove vegetation fluctuations and drivers in the Sundarbans

• Best-available-pixel annual time series vegetation indices from Landsat were used. • High to very high vegetation fluctuations were observed in 80 % of the forest over 1988–2020. • Mann-Kendall analysis of EVI data shows degraded mangrove health in 77 % of the Sundarbans. • GBM model identified precipitation and longitude as key drivers of vegetation change. Mangrove forests represent an essential yet delicate ecosystem on Earth. The Sundarbans, the largest mangrove forest in the world, is under considerable threat from natural and human-induced stressors. This study uniquely combined annual Landsat satellite Best Available Pixels (BAP) composites and Generalized Boosted Model (GBM) to examine the spatiotemporal relationships between vegetation dynamics and various driving factors across the Sundarbans from 1988 to 2020. We applied the innovative BAP algorithm on Google Earth Engine to produce annual time series of three vegetation indices: Normalized Difference Vegetation Index (NDVI), Green Normalized Difference Vegetation Index (GNDVI), and Enhanced Vegetation Index (EVI). While the total forest area remained stable, substantial fluctuations in vegetation health were observed. Approximately 80 % of the Sundarbans’ area experienced high to very high vegetation fluctuations from 1988 to 2020, with the central region undergoing the most pronounced changes. Results from the Mann-Kendall test indicated a pronounced declining trend, particularly with the EVI time series; 77 % of the area exhibited signs of browning, which has significant implications for different ecosystem services, including biodiversity conservation, carbon sequestration, storm buffering, and people’s livelihood. The variations in longitude, latitude, precipitation, and land use land cover changes accounted for approximately 30 %, 13 %, 33 %, and 11 %, respectively, of the changes detected by EVI. By integrating the advanced BAP algorithm, multi-decadal Landsat time series images, and machine learning techniques, this study offers a novel framework for detecting spatiotemporal vegetation changes and their drivers across the Sundarbans. The findings highlight critical zones of degradation and resilience, providing actionable insights for region-specific mangrove management and ecosystem restoration.