CoastSentinel AI: an end-to-end system for coastal health monitoring and community emergency response

Coastal areas are facing two major environmental problems: marine plastic pollution and shoreline erosion. At the same time, information from satellite monitoring, risk models, and local response teams is often handled separately. This paper presents CoastSentinel AI, an end-to-end system that brings these parts together in one platform. The system collects Sentinel-2 satellite data through openEO and uses a UNet++ model with a ResNet-50 backbone to identify marine plastic. It also uses Random Forest models to estimate the Coastal Erosion Risk Index (CERI) and Plastic Pollution Density Score (PPDS), while a CNN-LSTM model is used to study changes in the shoreline over time. The platform also connects citizens, NGOs, and response teams through geotagged reports, photo uploads, automatic alerts, and a simple leaderboard. Tests were carried out across eight coastal sectors in India. The results show an F1-score of 81.8 percent and an IoU of 69.3 percent for plastic detection, with R-squared values of 0.935 for CERI and 0.926 for PPDS. The system is designed as a practical framework for combining AI-based monitoring with local community response.