Comparison of spatial distribution models to predict subtidal burying habitat of the forage fish Ammodytes personatus in the Strait of Georgia, British Columbia, Canada

The Pacific sand lance (Ammodytes personatus) is a key forage species for many commercially important fish (e.g. salmon and groundfish), marine birds, and whales found in nearshore coastal waters of British Columbia, Canada. Sand lance lack a swim bladder and have a requirement for low‐silt, medium‐coarse sandy sea‐bed habitat for burying. Little information is available describing the distribution of burying habitat, partly because there are no commercial fisheries for A. personatus in British Columbia. This information is required by habitat and wildlife managers to identify and protect uncommon patches of burying habitats from detrimental activities, including dredging, infilling, and oil spills. In this study, habitat distribution results from five suitability modelling algorithms were evaluated: maximum entropy, generalized linear model, generalized additive model, random forest, and an ensemble model of the latter three. The maximum entropy model had the highest performance score (area under the receiver operator characteristic curve was 0.78) and was selected as the model that most accurately identified the presence of suitable A. personatus burying habitat. Model results indicate that suitable burying habitat is primarily influenced by derived sea‐bed substrate, distance to estuary, distance to sand‐gravel beaches, and bottom sea temperature. Overall, the spatial modelling identified only 105 km2 of highly suitable sand lance burying habitat, or 2.6% of the study area (0–150 m), primarily in Haro Strait, along the east coast of Vancouver Island, and in northern regions of the strait near Cortes, Savary, and Harwood islands. Identification of this uncommon and patchy burying habitat will contribute to the ongoing conservation of an important coastal prey species.