Building on a robust and scalable approach, and in collaboration with Swedish regional and national authorities, the DHI team developed an SAV monitoring tool to monitor the extent and dynamics of SAV at scale for Sweden in a cost-efficient and timely manner. The project included the entire shallow coastal zone of Sweden, an area covering almost 50,000 km2. The cloud-based web platform allows non-EO specialists to apply advanced machine learning algorithms and the latest earth observation data to conduct scalable and detailed SAV classification on demand.
The team first combined Copernicus Sentinel-2 satellite imagery, novel machine learning techniques and advanced data processing to create the first spatial overview of the distribution of SAV at a national scale in Sweden. A training dataset constructed with more than 30,000 manually labelled polygons was used to build the classification model, with which more than 3,800 km2 of underwater habitats of the entire shallow coastal zone of Sweden was mapped.
With this novel tool, authorities can now execute the entire mapping process in a few clicks – from the selection of suitable imagery to the final classification – just by interactively adding relevant training data. The user-friendly platform enables SAV mapping to be conducted on demand to ensure that data quality and update frequency meets the needs of Swedish authorities.