Coastal Features Segmentation and Assessing their dynamics Using Machine Learning: Random Forest

Estuaries represent complex morphodynamic systems where interactions between tides, waves, and sediment processes control coastal stability and its ecological resilience. One such estuary, located along the bank of the Purna River in Navsari District, Gujarat, India, is currently experiencing severe erosion, with nearly two-thirds of the estuarine coastline affected.  Understanding spatio-temporal evolution of key coastal features is essential, including tidal flats, salt marshes, mangrove cover, and anthropogenic infrastructures within the study region. In this study, the coastal features segmentation is performed using the Random Forest on derived Landsat satellite imagery spectral indices spanning 2005–2024. The results indicate that over the past two decades, mangrove cover has increased by more than twofold, particularly near the estuary mouth. In contrast, tidal flat areas exhibited significant spatial variability, while salt marshes showed a considerable decline.

Shoreline change analysis shows extensive coastal erosion with the Net Shoreline Movement (NSM) exceeding 150 m in certain stretches, while the End Point Rate (EPR) ranged from 1.5 to 17 m/year (mean: 9.5 m/year). The analysis further indicates significant accretion in the estuaryward region and pronounced erosion along the seaward coast near its mouth. Further the coupled tide-wave numerical modelling was carried to attribute the observed changes. Overall, the findings highlight the complex interplay between natural coastal processes and anthropogenic pressures in this dynamic estuarine coastal system and provide valuable baseline information for coastal zone management and conservation planning.

Keywords: Estuary Dynamics, Random Forest, Shoreline changes, Tide Modelling, Wave Modelling, Remote Sensing.