Recent Advances in Automated Genus-specific Marine Habitat Mapping Enabled by High-resolution Multibeam Bathymetry

Abstract

There is a great need for accurate, comprehensive maps of seafloor habitat for use in fish stock assessments, marine protected area design, and other resource management pursuits. Recent advances in acoustic remote sensing technology have made it possible to obtain high-resolution (meter to sub-meter) digital elevation models (DEMs) of seafloor bathymetry that can rival or surpass those available for the terrestrial environment. The acquisition and processing of these data are expensive, however, requiring specialized equipment, expertise, and large amounts of both field and laboratory effort per unit area mapped. Further, the interpretation and classification of these data into maps of habitat type is typically (and appropriately) performed only by trained experts that are familiar with both seafloor geomorphology and the nature and limitations of the data sources. Because it is done visually, this interpretation can be very time-consuming and may yield subjective results that are not comparable from site-to-site or between individual interpreters.We applied an algorithmic terrain analysis approach to efficiently and objectively classify seafloor habitats using the quantifiable landscape metric Topographic Position Index (TPI). We used high-resolution multibeam bathymetry, together with precisely geolocated (± 5 m) ROV observations of fish distribution, to produce a preliminary genus-specific habitat suitability model for eight rockfish (Sebastes) species in the Del Monte shale beds of Monterey Bay, California. A high-resolution (2 m) multibeam bathymetry Digital Elevation Model (DEM) was generated and used to produce a derived TPI surface model using repeatable, algorithmic methods. This data layer, together with the positions and counts by species from 229 rockfish observations (2892 total fish) was then used to create preliminary predictive models of habitat suitability and fish distribution, as well as stock estimates for the study area. A second, independent fish observation data set was used to validate the models.