A Novel Framework to Predict Relative Habitat Selection in Aquatic Systems: Applying Machine Learning and Resource Selection Functions to Acoustic Telemetry Data From Multiple Shark Species

Resource selection functions (RSFs) have been widely applied to animal tracking data to examine relative habitat selection and to help guide management and conservation strategies. While readily used in terrestrial ecology, RSFs have yet to be extensively used within marine systems. As acoustic telemetry continues to be a pervasive approach within marine environments, incorporation of RSFs can provide new insights to help prioritize habitat protection and restoration to meet conservation goals. To overcome statistical hurdles and achieve high prediction accuracy, machine learning algorithms could be paired with RSFs to predict relative habitat selection for a species within and even outside the monitoring range of acoustic receiver arrays, making this a valuable tool for marine ecologists and resource managers. Here, we apply RSFs using machine learning to an acoustic telemetry dataset of four shark species to explore and predict species-specific habitat selection within a marine protected area. In addition, we also apply this RSF-machine learning approach to investigate predator-prey relationships by comparing and averaging tiger shark relative selection values with the relative selection values derived for eight potential prey-species. We provide methodological considerations along with a framework and flexible approach to apply RSFs with machine learning algorithms to acoustic telemetry data and suggest marine ecologists and resource managers consider adopting such tools to help guide both conservation and management strategies.

Coyote diet and spatial co-occurrence with caribou

The interplay of predator encounters and anti-predator responses is an integral part of understanding predator-prey interactions and spatial co-occurrence and avoidance can elucidate these interactions. We conducted hard-part dietary analysis of coyotes (Canis latrans Say, 1823) and space use of coyotes and caribou (Rangifer tarandus Gmelin, 1788) to test two competing hypotheses about coyote and caribou predator-prey spatial dynamics using resource selection functions. The high encounter hypothesis predicts that coyotes would maximize encounters with caribou via high spatial co-occurrence, whereas the predator stealth hypothesis predicts that through low spatial co-occurrence with caribou, coyotes act as stealth predators by avoiding habitats that caribou typically select. Our dietary analysis revealed that ~46% of sampled coyote diet is composed of caribou. We found that coyote share space with caribou in lichen-barren habitat in both summer and winter and that coyotes co-occur with caribou in forested habitat during summer, but not winter. Our findings support predictions associated with the high encounter predator hypothesis whereby coyotes and caribou have high spatial co-occurrence promoting caribou in coyote diet.

A reformulation of the selection ratio shed light on resource selection functions and leads to a unified framework for habitat selection studies

The selection ratio (SR), i.e. the ratio of proportional use of a habitat over proportional availability of this habitat, has for long been the standard metric of habitat selection analyses. It is easy to compute and directly estimates disproportionate use. Its apparent restriction to habitat selection analyses using categorical predictors led to the development of the resource selection functions (RSF) approach, which has now become the norm.The RSF approach has however led to debates and confusion. For instance, what functional form can be used remains debated, and the concept of relative probability of selection is often misunderstood.I propose a reformulation of the SR demonstrating that it can be estimated in a regression context, and thus even with continuous predictors. This reformulation suggests that RSF can be seen as an intermediate step in the calculation of SR. This reformulation also clarifies some long-standing debates about RSF and data-selection/fitting practices.I further suggest that SR estimates the strength of habitat selection, but that the contribution of selection in determining use, which should be more directly linked to fitness than selection per se, should be estimated by another metric, the selection effect on use (SE). SE could be estimated simply as the difference between proportional use and proportional availability, and can be computed from SR and a density estimation of availability.I conduct a habitat selection analysis of plains zebras to demonstrate the added-value of going beyond RSF scores and using SR estimated in a regression context, and of combining SR and SE.Overall, I highlight the inter-relation between various metrics used to study habitat selection (i.e., SR, other selection indices, RSF scores, marginality). I conclude by proposing that SR and SE can be the unifying metrics of habitat selection, as together they offer a comprehensive view on the strength of habitat selection and its effect on habitat use.