Abstract

Coastal communities, land covers and inter-tidal habitats in East Anglia are particularly vulnerable to both erosion and flooding. This vulnerability is likely to increase with sea level rise and greater storminess in the near future. The determination of shoreline position, and its landward migration, is therefore imperative for future coastal risk adaptation. This project is developing an automated tool, VEdge_Detector, to extract the coastal vegetation line from high spatial resolution (Planet’s 3-5m) imagery, training a very deep convolutional neural network (VGGNet-16) to predict sequential vegetation line locations. Red, Green and Near-Infrared (RG-NIR) was found to be the optimum image band combination during neural network training and validation. VEdge_Detector outputs were compared with in-situ vegetation line positional measurements to ascertain a mean distance error of 4ma-1. Vegetation line change outputs derived from this tool are produced far quicker than other non-automated methods, and have the potential to inform coastal risk management decisions in East Anglia and other global locations.