Modelling benthic habitats and biotopes off the coast of Norway to support spatial management.

-Habitat conservation, and hence conservation of biodiversity hinges on knowledge of the spatial distribution of habitats, not least those that are particularly valuable or vulnerable. In offshore Norway, benthic habitats are systematically surveyed and described by the national programme MAREANO (Marine AREAl database for NOrwegian waters). Benthic habitats and biotopes are defined in terms of the species composition of their epibenthic megafauna. Some habitats are of special conservation interest on account of their intrinsic value and/or vulnerability (e.g., long-lived species, rareness, to comply with international regulations such as OSPAR). In Norway, off Nordland and Troms, the following habitats of special interest can be found: Umbellula encrinus Stands, Radicipes sp. Meadows, Deep Sea Sponge Aggregations, Seapen and Burrowing Megafauna Communities, Hard Bottom Coral Gardens. In this paper, we used underwater video data collected within the MAREANO programme to define and describe benthic habitats and biotopes of special interest, and to map the geographic distribution thereof by means of habitat modelling. We first evaluated the community structure of each habitat in the list using a SIMPROF test. We determined that the class Deep Sea Sponge Aggregations, as defined by OSPAR, had to be split into at least three classes. We then re-defined seven new types of ecological features, including habitats and biotopes that were sufficiently homogeneous. Then we modelled the spatial distributions of these habitats and biotopes using Conditional Inference Forests. Since the purpose of the distribution maps is to support spatial planning we classified the heat maps using density thresholds. The accuracy of models ranged from fair to excellent. Hard Bottom Coral Gardens were the most rare habitat in terms of total area predicted (224 km2, 0.3% of the area modelled), closely followed by Radicipes Meadows (391 km2, 0.6%). Soft Bottom Demosponges (Geodid sponges and other taxa) represent the largest habitat, with a predicted area of 9288 km2 (14%). Distribution maps of classes defined by habitat-forming species (Hard Bottom Coral Gardens) were more reliable than those defined by a host of species, or where no single species was a clear habitat provider (e.g. Seapen and Burrowing Megafauna Communities). We also put forward that a scale of patchiness larger than the scale of observation, and homogeneity of the community both play a role in model performance, and hence in map usefulness. These along with density threshold values based on observed data should all be taken into account in marine classifications and habitat definitions