Title : Coastal aquaculture expansion and mangrove change in southwestern Bangladesh: A Google Earth Engine assessment of environmental trade-offs (2015-2025)
Abstract:
Coastal aquaculture is an important source of income and food in southwestern Bangladesh, yet its spatial growth occurs alongside mangrove ecosystems, agricultural land and other coastal vegetation that protect shorelines, store carbon and support fisheries. Earlier Landsat-based studies in the Ganges–Brahmaputra delta have documented agriculture–aquaculture transitions and land-cover change around the Sundarbans. However, based on a targeted literature review, few studies appear to combine recent aquaculture mapping, mangrove and surrounding land-cover transitions, district-level comparison across Satkhira, Khulna and Bagerhat, independent accuracy assessment an spatially explicit identification of environmental trade-off hotspots within a single reproducible framework.
This proposed study will (i) quantify changes in aquaculture extent between 2015 and 2025; (ii) map changes in mangroves, other vegetation, agricultural land, built-up areas and open water or exposed land; (iii) identify land-cover transitions associated with aquaculture expansion; (iv) assess the spatial overlap between aquaculture expansion and mangrove or other ecosystem change; and (v) identify environmental trade-off hotspots. Overlap will be interpreted as spatial association, and the study does not assume that aquaculture expansion caused mangrove loss.
All analyses will be run in Google Earth Engine using open-access data. Cloud-masked dry-and wet-season composites from Landsat 8/9 will provide a consistent record for 2015, 2020 and 2025, while Sentinel-2 imagery will support training-sample delineation and visual cross-checking. A Random Forest classifier will use spectral bands, water and vegetation indices, seasonal water-persistence and NDVI-amplitude metrics, and terrain and distance-to-river variables. To separate aquaculture ponds from natural water bodies and seasonal wetlands, the workflow will combine water persistence, seasonal contrast with adjacent land, and patch-level shape and size metrics. Identical class definitions, seasonal windows, features and model parameters will be applied to every epoch to maintain consistency. Accuracy will be assessed with independent validation samples, confusion matrices, overall, producer’s and user’s accuracy, class-specific F1 scores and area estimates with uncertainty. Post-classification change detection, transition matrices, district-level comparisons and grid-based hotspot analysis will follow.
No quantitative results are reported at this stage. The expected outcome is a validated, reproducible evidence base to support sustainable aquaculture zoning, mangrove conservation and coastal spatial planning in southwestern Bangladesh, and a foundation for a subsequent peer-reviewed analysis.
Keywords: Coastal aquaculture; Mangrove change; Google Earth Engine; Land-cover transitions; Southwestern Bangladesh
