Mapping Ancient Forests: Bayesian Inference for Spatio-temporal Trends in Forest Composition Using the Fossil Pollen...

Area of Science:

• Ecology
• Paleoecology
• Statistical Modeling

Background:

• Ecologists reconstruct past forest composition using fossil pollen relative abundance in sediments.
• Predicting historical forest composition and quantifying information requires robust modeling approaches.

Purpose of the Study:

• To build a Bayesian hierarchical model for predicting historical forest composition in central New England using pollen data.
• To estimate critical relationships between pollen and actual vegetation abundances for modern and historical periods.
• To assess the reliability of spatial patterns and infer changes in tree species distribution over the last 2500 years.

Main Methods:

• Developed a Bayesian hierarchical model integrating fossil pollen data from a pond network.
• Estimated modern and colonial period vegetation-pollen relationships using a latent multivariate spatial process.
• Utilized estimated parameters to predict past forest composition conditional on pollen data.
• Implemented a graphical assessment of feature significance to evaluate reliable spatial pattern estimation.

Main Results:

• The model successfully estimated the spatial distribution and relative abundances of tree species over the last 2500 years, including uncertainty.
• Cross-validation indicated reliable estimation of large-scale spatial features for many taxa, though features smaller than 50 km were difficult to distinguish.
• The approach allowed quantitative investigation of ecological hypotheses, such as covariate effects and pollen dispersal characteristics.

Conclusions:

• The Bayesian hierarchical model provides an explicit spatio-temporal representation of forest composition, quantifying tree abundance and uncertainty.
• This approach offers critical advantages over current ecological analyses by providing tree-scale abundance estimates and uncertainty characterization.
• The feature significance assessment aids in inferring reliably estimated spatial patterns in paleoecological reconstructions.