Abstract: As contemporary ecosystems experience increasing anthropogenic pressures, there is growing concern over how ecosystems will respond to such novel pressures and environmental conditions. Palaeoecology plays a crucial role in understanding how ecosystems have responded to changing environmental conditions in the past. Insights from palaeoecology can be used to inform management of contemporary ecosystems and to forecast future ecosystem responses to change. However, palaeoecological data are subject to sources of uncertainty that affect inferences drawn from them. These uncertainties should be considered to help improve predictions of future ecosystem states and their application to managing contemporary ecosystems. Much of the palaeoecological literature has relied on qualitative interpretation of proxy records without quantifying uncertainties or applying statistical methods.
We introduce a state-space modelling approach to analysing palaeoecological proxy data. This approach is an advancement on previous state-space methods by fitting multinomial count data, i.e., data typical of quantifying pollen fossils or diatom frustules. A key advantage of state-space modelling is fitting multiple parameters simultaneously; thus, offering exploration of possible underlying dynamics that may drive the observed patterns in the proxy records. Whether change in ecosystems is primarily driven by intrinsic dynamics (e.g., species competition) or extrinsic forcing (e.g., climatic forcing) is a fundamental question to understanding ecosystem trajectories through time. We demonstrate results using sites across North America as case studies. Preliminary results indicate that the newly developed model is appropriate to fitting multivariate timeseries data with irregular sampling intervals using both species count data and external covariates as inputs. We discuss results in the context of contrasting and complimentary methods, advocating for the application of multiple tools for examining ecosystem-level change.
