Combining acoustic and visual detections in habitat models of Dall’s porpoise

Abstract

Habitat-based distribution modelling is an established method for predicting species distributions and is necessary for many conservation and management applications. Cetacean habitat models have primarily been developed using data from visual surveys. However, numerous techniques exist for detecting animal presence and each capture a portion of the true population. Combining detection data gathered from multiple survey methods, such as visual and acoustic surveys, may lead to a more robust picture of a species distribution and ecology. We compare habitat models for Dall’s porpoise built with visual versus acoustic survey data from a line-transect survey in the California Current and develop a combined model, utilizing both acoustic detections and visual sightings. Combining acoustic and visual detections increases sample size and allows for detections under a greater range of oceanographic conditions. Consequently, the combined model shows a modest expansion of predicted distribution of Dall’s porpoise compared to either single-source model. However, this study reveals that acoustic and visual methods appear to be more complementary, rather than directly additive. Models built with acoustic data display differences from those built with visual data. Different predictor variables were selected across models and the acoustic model predicts a distribution shifted slightly south of the visual distribution. Results from the current study show promise for incorporating acoustics into habitat models but also identify discrepancies in population sampling between these two methods that should inform future population assessments and modelling efforts.