Mapping the Oceans: Progress in Semi-Automated Seabed Classification Required

Over the last three decades we have witnessed significant progress in our capabilities to map the ocean’s seabed. An increasing number of countries has recognized the importance of seabed mapping for safe navigation, resource assessment, marine spatial planning and nature conservation, which has led to the establishment of national mapping programmes and the Atlantic Ocean Research Alliance.
The ever increasing amount of collected seabed data, the complex relationships between seabed properties and environmental drivers and the need for maps fit for multiple purposes renders the classic and still widely used data analysis method of manual interpretation ‘by eye’ unsuitable for such tasks. However, progress in semi-automated seabed mapping has been fairly slow and only started in earnest a few years ago, as indicated by the limited number of published studies. Arguably, progress in data analysis methods is of equal importance to fully exploit the potential of the collected data in a timely manner. Additionally, such research leads to findings that will have a bearing on how the data should be collected. To unlock the potential of the collected seabed data, it will be necessary to revise current practices of data collection and processing, such as agreed standards for acoustic backscatter data collection, requirements for sample positioning accuracy and quantitative data extraction from seabed imagery.
This contribution will explore recent developments and challenges in semi-automated seabed mapping. Progress made to date will be demonstrated with case studies, illustrating potential avenues towards repeatable, flexible and validated results and highlighting topics still requiring progress. It is argued that to fully exploit the potential of seabed data will require an improved understanding of the interdependencies of the various elements of the seabed mapping workflow from survey design to map production.