Model transferability in ecology and conservation: The what, the how and the where to now?
Speaker: Philippe Bouchet (formerly University of Western Australia)
Improved knowledge of the distribution of species and habitats is needed to support conservation efforts worldwide. Yet despite decades of extensive surveying, our understanding of global patterns in terrestrial and marine biodiversity is still woefully incomplete. In practice, available ecological data are often insufficient to support robust scientific inference, and additional sampling in the field can prove prohibitively expensive or logistically impossible. Projecting (‘transferring’) predictive models in time and/or space offers a potential solution to the challenges posed by data deficiency. However, the determinants of ecological predictability are still little understood, and a consensus method for assessing the predictive performance of transferred models (i.e. their “transferability”) is lacking. In this talk, I synthesise recent advances in the field of ecological transferability to give an overview of how model transfers work, why they are critical to wildlife management and decision-making, what factors underpin their success (or failure), and to propose ways in which they can be broadly improved. I also highlight outstanding knowledge and methodological gaps that should receive priority attention in the future.