As human populations increase around the world, so does pressure on the marine environment from anthropogenic activities. There is now a high demand for high-resolution maps of benthic substrates and habitats to help in development of effective management measures. This study presents a methodology integrating multibeam echo-sounder (MBES) data, object-based image analysis (OBIA) and video data to produce an accurate habitat map for Refuge Cove, Wilsons Promontory National Park, Australia. A number of derivatives traditionally used in the habitat mapping literature were first computed from MBES bathymetry and backscatter data. A subset of these derivatives was then used in an OBIA resulting in a segmentation of the site. Finally, five different models implementing a Quick Unbiased and Efficient Statistical Tree (QUEST) on various combinations of MBES derivatives and segmentation layers were run to identify the layers’ suitability to predict marine habitats. The accuracy of the habitat map resulting from each model was assessed by computing the overall accuracy and kappa measures on its associated error matrix.
The differences between the maps were analyzed by calculating pair-wise Z-statistics and using a map comparison tool to identify the local agreement between each pair of maps. The best overall accuracy was achieved by the model using bathymetry, bathymetry derivatives and the OBIA segmentation layers (78.34% overall accuracy). The maps showed large differences in the spatial distribution of habitats across the sites, but the pair-wise Z-statistic computations revealed that the models did not differ significantly in performance. A main result is that the OBIA segmentation layers performed well, presumably because they smooth-out the inherent noise in the original MBES layers. A secondary advantage is that they provided smooth homogenous boundaries between habitat classes, making the map suitable for habitat management application. The seagrass Amphibolis Antarctica was observed in the high-use area of Refuge Cove, making it susceptible to anchorage scars. This study provides a map of the areas where anchorage should be discouraged to prevent future impacts on these fragile but yet important habitats.