Title: Inference of ecological networks from a sparse ocean dataset
Authors: Joseph Siddons - Dalhousie University (Canada) [presenting]
Andrew Irwin - Dalhousie University (Canada)
Zoe Finkel - Dalhousie University (Canada)
Abstract: The continuous plankton recorder (CPR) survey is a large, multi-decadal, plankton monitoring programme spanning the North Atlantic, and is a powerful tool for investigating the effects of climate change. Using a network inference approach, including sparse neighbourhood selection and correlation analysis, we build ecological interaction networks to investigate the planktonic community structure. We are particularly interested in observing trophic level interactions (such as grazing), and the evolution of communities over time (e.g. between seasons or decades). We attempt to account for the environmental predictors in order to separate the relationships that are driven by niche effects from those that arise as a consequence of biological forcing. A difficulty of the CPR dataset is that it is heavily zero-inflated. Consequently, we need to make use of appropriate correlation estimators for sparse data. We have identified a number of species that are important for community structure, acting as either hub species, or as a bridge between clusters of taxonomic groups. Results indicate that species within plankton taxonomic groups are more positively correlated than species between groups (in particular zooplankton-phytoplankton relationships).