Forecasting Shoreline Change and Identifying Erosion Hotspots to Propose Site Specific Setback Regulations: A Machine Learning Approach For Coastal Planning In East Coast of Sri Lanka

Coastal erosion poses a persistent and intensifying threat to Sri Lanka’s eastern shoreline, affecting both ecological systems and socio-economic infrastructure. This study combines remote sensing techniques, historical shoreline data, and machine learning models to analyze shoreline dynamics from 1990 to 2024 across six key coastal segments. Shoreline changes were assessed using the Digital Shoreline Analysis System (DSAS), supported by NDWI and vegetation index layers to map land cover transitions. A range of machine learning models was tested to predict future shoreline positions at erosion hotspots, with the Random Forest model yielding the highest accuracy (R² = 0.93). Projections show significant shoreline retreat across several vulnerable locations by 2035. Additionally, the study evaluates coastal setback policies outlined in Sri Lanka’s Coastal Zone Management Plans (2018–2024), highlighting a critical misalignment between increasing erosion risk and reduced setback distances. A site-specific setback model is proposed for the Reservation Area, incorporating historical erosion rates and safety margins to support more adaptive and risk-responsive coastal planning. These findings emphasize the importance of data-driven strategies for enhancing long-term coastal resilience.