COS 5-4
A synthesis of LTER biodiversity using simulations to guide the inference of community assembly processes from diversity patterns
The metacommunity concept has gained traction in ecology because it provides a general framework to incorporate the many processes that influence biodiversity patterns at local and regional scales. A metacommunity is a group of assemblages of interacting species that are linked over broad spatial scales by dispersal. Metacommunity biodiversity is typically represented by measures of local (alpha) and regional (gamma) diversity, and species turnover (beta-diversity). These biodiversity metrics are hypothesized to be emergent patterns linked to both landscape and species pool characteristics, which set the stage for community assembly dynamics to unfold across a network of connected assemblages. Our objectives in this study were (1) to determine which biodiversity metrics and metacommunity characteristics exhibit the strongest empirical relationships, and (2) synthesize the metacommunity characteristics of different ecosystem (e.g., polar deserts, urban ponds, subtropical wetlands) and organism types (e.g., bacteria, nematodes, fishes) represented in the Long Term Ecological Research (LTER) network. We designed a package for the R statistical environment to simulate different types of metacommunities (MCSim) and used simulation outcomes to evaluate the sensitivity of biodiversity metrics to model parameters. We then used neural networks that were trained on simulated data to characterize metacommunities across the LTER network.
Results/Conclusions
A sensitivity analysis of simulation outcomes showed alpha, beta, and gamma diversity and variation partitioning outcomes (spatial and environmental components of beta diversity) were each sensitive to different metacommunity characteristics. Alpha diversity was most sensitive to assemblage size and gamma diversity was influenced by assemblage size and invasion by novel taxa. Beta diversity and variation partitioning outcomes were most sensitive to regional species pool richness and functional diversity. The metacommunity characteristics that neural networks were best able to predict from biodiversity patterns included historical regional diversity, metacommunity size, assemblage connectivity, and functional diversity. Analysis of observed LTER biodiversity patterns using neural networks showed significant variation in metacommunity connectivity and species sorting dynamics among ecosystems and organism types. For example, macroinvertebrates in Baltimore streams and cyanobacteria in the McMurdo Dry Valleys in Antarctica were more strongly influenced by niche-based species-sorting than other LTER sites. This approach offers much potential to understand how context (e.g., study design and methods, ecosystem type, organism type) influences our interpretation of biodiversity patterns, as well as a predictive framework to forecast shifts in biodiversity associated with climate change.