Mangrove Forest loss and Future Risk Analysis for Southeast Asia using Satellite-derived data and Machine learning

Mangrove forests are a critical part of our ecosystem, which works continuously to fight the carbon footprint, coastal erosion and provides support for biodiversity. However, these forests are encountering notable loss across tropical areas. To save these diverse ecosystem it is important to identify the mangrove loss factors and future risk zones. In this study, I have worked on a multi-country data-driven framework to understand the mangrove loss in Southeast Asia and to forewarn future loss risks. Global Mangrove Watch(GMW) data set was used to extract the mangrove data for 1996, 2007,2010,2015,2020 timeline. Lateron, Google Earth Engine was used to extract Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), elevation, and slope data for Bangladesh, Thailand, Myanmar, and Indonesia. After sampling and balancing the data, four ensemble machine learning models were applied to model mangrove loss. Random Forest, Gradient Boosting, XGBoost, and LightGBM models were tested, and among them, Random Forest outperforms others with a ROC-AUC of 0.96 and an F1-score of 0.90. Elevation was identified as the primary cause of mangrove decline, followed by NDVI and NDWI, suggesting increased vulnerability in low-lying, water-stressed mangrove areas. While country-level comparisons revealed similar loss intensities throughout the research region, temporal analysis revealed comparatively steady loss rates between 1996 and 2020. About 36% of mangrove habitats were classified as high or very high risk in future risk projections for 2025–2030, with low-elevation coastal zones being specifically vulnerable. The suggested methodology facilitates the evidence-based prioritizing of conservation and management measures and offers a scalable and adaptable method for regional mangrove monitoring.