21,000 years of shifting species, communities, and biotic associations in response to past climate change
Potential negative impacts of the high rate and magnitude of future climate change on biodiversity are of increasing concern to the conservation and biogeography communities. In response, many studies aim to predict the distributions of species based on scenarios of future climate change, and from these predictions, estimate other biodiversity properties such as species richness or extinction risk. Many such models use only climatic or habitat variables as predictors, but other factors such as dispersal lags and interactions between species can greatly influence the distributions of species and communities through time and across space. However, quantifying the relative influence of climate or habitat, dispersal limitation, and biotic interactions on species and communities is not straightforward. Here, I use species lists from late Quaternary fossil localities in North America paired with downscaled paleoclimate simulations to assess potential causes of species and community changes across space and time. Using complementary analyses such as generalized dissimilarity modeling, analyses of species pairs, and species distribution modeling, I focus in particular on disentangling the relative contributions of climate versus other mechanisms of change.
Climate strongly structures both species distributions and community attributes across space and time at broad spatial and temporal scales. Generalized dissimilarity models show that climate influences dissimilarity within fossil pollen assemblages across eastern North America and that communities are structured similarly across spatial and temporal climate gradients. Additionally, most of the non-randomly associated species pairs can be explained by climatic or spatial attributes of sites (mean = 83% of the aggregated pairs and 93% of the segregated pairs across all time slices), and biotic interaction is not the most parsimonious explanation of the non-random species associations. However, the influence of climate is variable across space and time. In the latest Pleistocene, climate explains less variation in community dissimilarity than in the Holocene, indicating that dispersal limitation or species interactions may be more important. Additionally, most non-random species associations that are potentially attributed to a biotic interaction occur during the latest Pleistocene. This implies that, at least at broad scales, climate-based models are relatively good for predicting changes in species and communities. However, care needs to be taken when predicting changes far into the future or across large magnitude climate changes, when there is greater potential for no-analog conditions.