Clustering-Based Method for Automatic Shoreline Extraction from Landsat and Sentinel-2 Satellite Imagery in South Maldives

For small and low-lying countries, having a thorough understanding of shoreline’s position and how it changed over the years is essential for effective coastal conservation, management, and national land protection efforts, particularly given the heightened risk of coastal hazards and sea-level rise. This study presented an automated method for extracting shorelines from open-source satellite imagery in South Maldives using unsupervised machine learning. The method involves resampling the Near-Infrared (NIR) image to improve spatial resolution, georeferencing, and followed by applying the K-means clustering algorithm to distinguish between land and water areas. The resulting boundary line is then corrected and georeferenced to match the actual shoreline position and transformed into a smooth line using a new modification algorithm. The accuracy of the automated method was evaluated by comparing its results to those obtained through manual extraction from the highest resolution WorldView-2 satellite images, in Laamu Atoll. The main finding of this study highlights the potential of automated shoreline extraction using freely available satellite images. The automated approach achieved a relatively good level of accuracy from both Sentinel-2 image and Landsat image, falling within acceptable error of 0.95 and 0.83 R-squared, respectively. This approach offers a quick and cost-effective means of extracting shorelines from open-source satellite imagery, enabling coastal engineers and scientists to explore shoreline changes at a regional scale with high accuracy while saving time, money, and labor compared to manual extraction or high-cost satellite imagery purchases.