Species occurrence is driven by complex networks of interspecific interactions (e.g. predation, competition, disease) and environmental conditions. Current approaches for examining interspecific interactions through multispecies occupancy modelling are limited by an approach that is focused on co-occurrence. A focus on patterns of co-occurrence restricts inferences on the role of interspecific interactions in driving species occurrence to exceedingly strong interspecific actions, best suited to relatively rare species, where the presence of one species results in the exclusion of another species, which is by definition rare. A more common scenario exists, where species interactions occur between relatively abundant species, which are countered by mechanisms for co-existence e.g. spatial or temporal avoidance, with strength and direction of interactions being dependant on abundance of the species and local habitat conditions. Thus, we propose an integrated occupancy-abundance model that models the occurrence of one species in response to the abundance of another. We adapt the single-species occupancy model to include one or more nested n-mixture models for interacting species. We allow the abundance of the dominant species to interact with all parameters of the subordinate, and for these interaction terms to vary as a function of habitat covariates.
Results/Conclusions
As an example, we model the top-down interactions between an expanding dominant mesopredator; the coyote (Canis latrans) and a smaller carnivore; the red fox (Vulpes vulpes) from a landscape-scale camera trap survey conducted between 2013 – 2021 in NY, USA. We compare the outputs of an existing multi-species co-occurrence model with our integrated model. We show that whilst a co-occurrence approach reports positive co-occurrence between the coyote and the red fox, the integrated occupancy-abundance model demonstrates a dynamic interaction that changes as a function of both habitat and time. At mean forest cover or higher, coyote populations reduce the occurrence of the red fox, with this interaction becoming stronger both as forest cover and through time. The increasing interaction strength through time, alongside observed increases in coyote abundance, suggest the relationship to be density dependant, with red foxes being suppressed after coyote abundance reaches a certain threshold. Our model has wide applications for investigating the role of interspecific interactions in structuring ecological communities and thereby informing evidence-based management of animal populations where contemporary methods cannot. This approach represents a key improvement in our ability to make community level inferences on the occurrence of species that are subject to imperfect detection.
