After the last glacial maximum (c. 21k years ago), rapid warming and retreating ice sheets opened new areas for colonization, allowing species to expand their ranges through migration. Paleoecological evidence, such as fossil pollen data, indicates that many tree taxa responded to this rapid warming during deglaciation by shifting their ranges to match their climate tolerances. Increasingly large and standardized online repositories of fossil pollen data allow us to compile data from many sites to look at tree occurrences over large geographic areas using pollen as a proxy for vegetation presence and abundance. Traditionally spatio-temporal pollen abundance is estimated using generalized linear or additive modeling approaches. But these approaches typically do not permit formal quantification of uncertainty. In this study, we use pollen samples for 20 North American tree genera over the past 21,000 years to estimate the relative abundance of pollen with uncertainty. We do this using a Bayesian framework, which allows for the characterization of uncertainty in: sampling, ecological processes (unaccounted for variation), and the age-depth relationship. This approach also allows for the formalization of spatio-temporal dependence, making it possible to infer relative abundance of pollen by borrowing information across the full network of pollen records.
Results/Conclusions
We estimated the spatio-temporal relative abundances of tree distributions from the last glacial maximum through to the modern era with associated uncertainty. Using these estimates, we calculated biotic velocity for each genus over 1000-year intervals from 21,000 years ago to present. We found that pollen distributions shifted contemporaneously with climate during warming events, particularly between 15k and 10k years ago. The Bayesian hierarchical framework allowed us to partition the sources of uncertainty, permitting inference about the impact of the age-depth model uncertainty (low impact at 1000-year temporal resolution). As expected, we also found uncertainty increased with increasing distance from a pollen sample in space and time. Quantifying uncertainty is important for improving predictions of the capacity of trees to respond to rapid climate change, allowing for informed decisions about strategic sampling of pollen cores. The ability to quantify the posterior distributions of biotic velocity allows for inference about ecologically significant changes in biotic velocity, refining understanding about the migration patterns of plant taxa. Ultimately, more accurate predictions of tree migration response to modern climate change can result in improved conservation strategies such as assisted migration, ex situ conservation, and ecosystem restoration.