GSA Annual Meeting in Phoenix, Arizona, USA - 2019

Paper No. 67-1
Presentation Time: 1:30 PM


SWAIN, Anshuman1, BUTTERFIELD, Nicholas J.2, DEVEREUX, Matt3 and FAGAN, William F.1, (1)Department of Biology, University of Maryland, College Park, MD 20742, (2)Department of Earth Sciences, University of Cambridge, Downing Street, Cambridgeshire, Cambridge, CB2 3EQ, United Kingdom, (3)Department of Earth Sciences, Western University, 1151 Richmond St, London, ON N6A 3K7, Canada

Network science methods have improved our fundamental understanding of community structure and function. Despite the power and utility of network approaches, only a handful of studies have employed these tools to explore paleocommunities, and these studies have their own shortcomings due to sample size constraints, coarse timescales, and biases in preservation, collection, and identification. Moreover, most prior applications of network approaches to paleocommunity data have focused on the network metrics themselves with relatively less attention to the mechanistic biological processes involved. To help fill this gap, we used network tools to explore fine-scale paleocommunity structure and dynamics using a new and extensive fossil abundance dataset from the Middle Cambrian Burgess Shale of SE British Columbia - one of the most complete views of early animal community structure preserved in the fossil record. We sought to (a) find an analytical way to study bias in fossil assemblages, (b) identify possible interactions among the taxa using abundance data, (c) categorize possible interactions into ecological roles, and (d) understand changes in community dynamics over time. We used exponential random graph models (ERGMs) and stochastic block models (SBMs) to show no preservation bias due to habitat or phyla membership in our dataset overall, and later validate our findings using agent-based models (ABMs). We employed networks based on partial correlation statistics and on graphical least absolute shrinkage and selection operators (LASSO) to identify possible interactions. In addition to predicting links between taxa, the analysis of correlation networks also revealed niches, key species and alternative community configurations. We assigned species to ecological roles (e.g., specialist predator, generalist predator, mutualist and competitor) by applying functional decomposition and maximum likelihood approaches to the occurrence / abundance data. We also explored aspects of prey-predator dynamics over evolutionary time. Overall, our work gives a novel framework that can be employed to study fossil communities, especially lagerstatten assemblages, to elucidate general ecological characteristics of the time-periods involved.