Estimating community change from sporadic data: A novel statistical technique sheds light on continental-scale ecology of the Pleistocene-Holocene transition

Background/Question/Methods
Massive amounts of ecological data are becoming available through large digitization collaborative projects. These data come from disparate sources with varied methodological biases and yet present the opportunity to address ecological questions at unprecedented spatial and temporal scales. Large scale anthropogenic perturbations of the global system are also in urgent need to being diagnosed. The time is ripe to apply novel statistical techniques to leverage sporadic data to understand how global change drives ecological change. To that end we develop a new hierarchical statistical method to model changes in community composition through time while accounting for variable sampling effort. We validate this method with simulation and apply it to fossil data from the past 20,000 years to evaluate how this period of global change influenced shifts in mammal communities in North America.

Results/Conclusions
Using our method we can accurately detect communities shifting in elevation and composition at the continental scale during the Pleistocene-Holocene transition. Because we take a hierarchical approach we can model multiple species and sites simultaneously, allowing us to test multiple hypotheses about which characteristics of organisms and geographic regions lead to differential ecological responses to the Pleistocene-Holocene transition. We find that geographic factors appear more important than organismal traits. We also explore applications of our method in other situations with large amounts of sporadic data, such as museum specimen databases and citizen science projects.