Zelda poster 🌊

Predicting Ecological Networks


  1. David Beauchenes
  2. Kevin Cazelles
  3. Guillaume Blanchet
  4. Philippe Archambault
  5. Dominique Gravel


Exhaustively describing the complex nature of ecological interaction networks is a challenging task even under ideal conditions. When confronted with data-poor environments and large scales of analyses, the task becomes even more daunting. Network-level descriptors are thus largely ignored for practical applications, even though we recognize the importance of considering the reticulated nature of complex networks. Significant insights can however be gleaned through the combined study of biotic (i.e. biotic interactions) and abiotic (i.e. environmental factors) constraints effecting the distribution and structure of communities. To address this issue we combine a recently published machine learning approach to predict biotic interactions in data-poor environments with joint species distribution models (JSDM) in an effort to characterize the spatial distribution of ecological networks. We use data from the Department of Fisheries and Oceans Canada’s annual trawl survey in the estuary and gulf of St. Lawrence in eastern Canada gathered during summer between 2011 and 2015 to apply our approach. We evaluate network-level descriptors (i.e. link density, connectance, robustness and richness) to spatially characterize networks throughout the St. Lawrence. Results from this approach could then be correlated to levels of environmental stressors in order to evaluate the cumulative impacts on the communities of the St. Lawrence. We believe that such approaches including the complexities of ecological network are essential assets if we are to properly evaluate the impacts of global changes on the structure and functioning of ecosystems



comments powered by Disqus