Author:
Siddons Joseph T.,Irwin Andrew J.,Finkel Zoe V.
Abstract
Species-rich communities are structured by environmental filtering and a multitude of associations including trophic, mutualistic, and antagonistic relationships. Graphs (networks) defined from correlations in presence or abundance data have the potential to identify this structure, but species with very high absence rates or abundances frequently near detection limits can result in biased retrieval of association graphs. Here we use graph clustering analysis to identify five sub-communities of plankton from the North Atlantic Ocean. We show how to mitigate the challenges of high absence rates and detection limits. The sub-communities are distinguished partially by their constituent functional groups: one group is dominated by diatoms and another by dinoflagellates, while the other three sub-communities are mixtures of phytoplankton and zooplankton. Diagnosing pairwise taxonomic associations and linking them to specific processes is challenging because of overlapping associations and complex graph topologies. Our approach presents a robust approach for identifying candidate associations among species through sub-community analysis and quantifying the aggregate strength of pairwise associations emerging in natural communities.
Subject
Ocean Engineering,Water Science and Technology,Aquatic Science,Global and Planetary Change,Oceanography
Cited by
1 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献