Automated detection of river mouth opening and closure using cloud-based processing of Sentinel-2 imagery

Intermittently open/closed river mouths are abundant on micro-tidal, wave-dominated coasts subject to strong interannual climatic variability. Monitoring and understanding their dynamics is essential for informed coastal management. We present a novel, semi-automated and calibration-free framework that uses Sentinel-2 imagery and cloud-based processing in Google Earth Engine (GEE) to classify river-ocean connectivity status through index-derived water masks and geometric intersection analysis. The method is applied to the Western Mediterranean basin (∼ 1,500 km of coast) over the entire Sentinel-2 imagery timeseries. We evaluated the performance of different index–threshold combinations—NDWI, AWEI, and MNDWI using fixed thresholds and the Otsu method—against expert-labelled imagery (∼600 samples). Fixed thresholds exhibited overall better agreement than the Otsu method, with NDWI demonstrating optimal performance for this application (∼80% agreement). Most classification uncertainty is associated with river mouth widths near the Sentinel-2 spatial resolution limit. We extracted geomorphic features and hydro-maritime time series for each river mouth and analyzed their correlation with open/closed dynamics. Results confirm the opposing roles of river discharge (promoting breaching) and wave action (promoting bar formation). The workflow also captures event-scale dynamics, exemplified by the sustained opening of the Tordera River (NE Spain) following Storm Gloria, highlighting the role of extreme events within seasonal and interannual cycles. This study provides the first regional assessment of intermittently open/closed river mouths in the Western Mediterranean and is readily extendable to other regions. It demonstrates the potential of satellite-based, cloud-computing tools for scalable monitoring of coastal morphodynamics, with clear implications for coastal management.