Print PageModels of beta diversity are important tools for predicting the spatial patterns of biodiversity and community turnover along environmental gradients. Generalized dissimilarity modeling (GDM) is a form of beta-diversity modeling that has been applied to predict the distributions of species and communities, identify survey gaps, set conservation priorities, and assess climate-change impacts. Critically, GDM does not assume that species persist in fixed community types, and can be used to predict the occurrence of communities with no modern analog. Late-Quaternary paleo-records, with their strong signals of community turnover, environmental change, and no-analog community formation, offer a natural opportunity to test GDM and other beta-diversity models. Here, we test the role of climate in controlling past community turnover by asking whether the emergent rates of community dissimilarity along spatial environmental gradients were constant over the past 21,000 years. We use GDM to quantitatively model the spatial relationships among species turnover and potential driving environmental variables. We rely on the dense network of fossil-pollen data in eastern North America, coupled with downscaled CCSM3 transient paleoclimate simulations since the Last Glacial Maximum from the SynTrace project. For every 1000 years from the Last Glacial Maximum to the present, we assembled fossil-pollen data from lakes and mires and the paleoclimate simulations for each location. We then modeled the relationship between community dissimilarity and environmental dissimilarity using GDM and determined which climate variables most strongly predicted community dissimilarity. Finally, we tested how well the model for one time period (e.g. 14 ka) predicted spatial dissimilarity between communities at all other time periods (e.g., 13 ka, 12 ka, etc).
Results/Conclusions
Preliminary analyses show that GDM provided a good fit to late-Quaternary fossil-pollen and paleoclimate data, generally explaining at least 60% of community turnover. The strongest predictor variables differed across time. Models built for one time period explained turnover during nearby time periods relatively well, but decayed across time, such that, e.g., the model for 13 ka only explained 44% of the community dissimilarity at 8 ka. Results were similar regardless of whether GDM was predicted forward or backward through time. Overall, these initial results indicate that GDM holds promise for modeling the climatic drivers of community turnover across space and time. Future work will apply GDM to predict the occurrence of no-analog communities.
