From Pixels to People: Satellite-Based Mapping and Quantification of Riverbank Erosion and Lost Villages in Bangladesh

TL;DR

This work tackles the crucial problem of riverbank erosion in Bangladesh by building a first-of-its-kind, manually annotated dataset of disappeared settlements from 2003–2025 and by adapting a general-purpose vision model, SAM, for fine-grained, time-resolved segmentation of land–water boundaries. A decoder-only fine-tuning approach frozen the ViT-H encoder, enabling efficient adaptation with strong erosion-change detection capabilities and pixel-based area quantification. The method achieves high segmentation accuracy (e.g., $mIoU=0.863$, $Dice=0.926$) and robust erosion mapping, with validation on unseen regions showing substantial improvement over zero-shot baselines; however, accretion is underrepresented, highlighting areas for methodological refinement. The study offers a practical, scalable tool for disaster management and policy planning, providing visual evidence and quantitative land-loss metrics to monitor erosion trajectories and support proactive protection of vulnerable communities, with potential to evolve into an all-weather, multi-modal Digital Twin of the delta.

Abstract

The great rivers of Bangladesh, arteries of commerce and sustenance, are also agents of relentless destruction. Each year, they swallow whole villages and vast tracts of farmland, erasing communities from the map and displacing thousands of families. To track this slow-motion catastrophe has, until now, been a Herculean task for human analysts. Here we show how a powerful general-purpose vision model, the Segment Anything Model (SAM), can be adapted to this task with remarkable precision. To do this, we assembled a new dataset - a digital chronicle of loss compiled from historical Google Earth imagery of Bangladesh's most vulnerable regions, including Mokterer Char Union, Kedarpur Union, Balchipara village, and Chowhali Upazila, from 2003 to 2025. Crucially, this dataset is the first to include manually annotated data on the settlements that have vanished beneath the water. Our method first uses a simple color-channel analysis to provide a rough segmentation of land and water, and then fine-tunes SAM's mask decoder to recognize the subtle signatures of riverbank erosion. The resulting model demonstrates a keen eye for this destructive process, achieving a mean Intersection over Union of 86.30% and a Dice score of 92.60% - a performance that significantly surpasses traditional methods and off-the-shelf deep learning models. This work delivers three key contributions: the first annotated dataset of disappeared settlements in Bangladesh due to river erosion; a specialized AI model fine-tuned for this critical task; and a method for quantifying land loss with compelling visual evidence. Together, these tools provide a powerful new lens through which policymakers and disaster management agencies can monitor erosion, anticipate its trajectory, and ultimately protect the vulnerable communities in its path.

Figures (5)

• Figure 1: Satellite imagery of Naria Upazila showing riverbank changes between 2003 (top) and 2020 (bottom). The marker indicates the site of Notun Masjid, beside which a drastic loss of land area has occurred due to river erosion. The comparison reveals significant channel migration and reduction of landmass over time.
• Figure 2: An evaluation of the process
• Figure 3: Base Architectural Components
• Figure 4: Comparison of Naria Upazila erosion patterns and SAM-generated masks. Top row: Satellite imagery (2010 vs 2020). Middle row: Base SAM zero-shot segmentation results. Bottom row: Fine-tuned SAM predictions with improved delineation of erosion zones.
• Figure 5: Comparison of riverbank change analyses between (a) ground truth and (b) SAM-predicted segmentation.