Seminar Series in Quantitative Life Sciences and Medicine

"Predicting the architecture of biodiversity"

Dominique Gravel (University of Sherbrooke)
Tuesday September 25, 12-1pm
McIntyre Building, Room 1027

Abstract: Ecologists have always been fascinated by the spatial distribution of species richness. But biodiversity is more than the distribution of species, it is also the collection of interactions among them driving ecosystem processes and supporting diversification of life. The representation of communities as networks of interacting species, populations and individuals, is a convenient formalism to describe the spatial distribution of biodiversity. It is however fairly limited by our capacity to document such networks over space and time. Another challenge is to predict the emergence and re-assembly of communities following species responses to global changes, such as range shifts, invasions and extinctions. Predicting interactions among species that never co-occurred proves challenging, as traditional empirical methods of food web sampling such as gut content analysis cannot be performed.

Inferring potential interactions among species of an arbitrary defined pool is a major step to predict the structure of emergent communities and their functioning. Development of predictive models of trophic interactions could greatly improve our understanding of large-scale food web structure and our capacity to anticipate major changes in ecosystem functioning. I will first review current approaches used to predict ecological networks of interactions, using information such as functional traits, phylogenies and co-occurrence. I will show there are essentially two broad categories of methods, phenomenological and mechanistic, each of them with pros and cons. Then I will present an analysis I performed to reconstruct trophic interaction networks among marine pelagic fishes, at the global scale. This unprecedent effort reveals that, surprisingly, marine food webs are remarkably well connected because of significant range overlap between species. While this connectivity provide considerable robustness to node extinction, it also suggest that disturbances can rapidly spread across the globe and has the potential to generate significant indirect interactions. I expect that, with the upsurge of interaction data availability, such methods will be more common and could potentially change our view of the conundrum of biodiversity's architecture.