Demersal fishes constitute an essential component of the continental shelf ecosystem, and a significant element of fisheries catch around the world. However, collecting distribution and abundance data of demersal fish, necessary for their conservation and management, is usually expensive and logistically complex. The increasing availability of seafloor mapping technologies has led to the opportunity to exploit the strong relationship demersal fish exhibit with seafloor morphology to model their distribution. Multibeam echo-sounder (MBES) systems are a standard method to map seafloor morphology. The amount of acoustic energy reflected by the seafloor (backscatter) is used to estimate specific characteristics of the seafloor, including acoustic hardness and roughness. MBES data including bathymetry and depth derivatives were used to model the distribution of Abalistes stellatus, Gymnocranius grandoculis, Lagocephalus sceleratus, Lethrinus miniatus, Loxodon macrorhinus, Lutjanus sebae, and Scomberomorus queenslandicus. The possible improvement of model accuracy by adding the seafloor backscatter was tested in three different areas of the Ningaloo Marine Park off the west coast of Australia. For the majority of species, depth was a primary variable explaining their distribution in the three study sites. Backscatter was identified to be an important variable in the models, but did not necessarily lead to a significant improvement in the demersal fish distribution models’ accuracy. Possible reasons for this include: the depth and derivatives were capturing the significant changes in the habitat, or the acoustic data collected with a high-frequency MBES were not capturing accurately relevant seafloor characteristics associated with the species distribution. The improvement in the accuracy of the models for certain species using data already available is an encouraging result, which can have a direct impact in our ability to monitor these species.
Temporal transferability of stream fish distribution models: can uncalibrated SDMs predict distribution shifts over time?
