The effects of climate change on forest ecosystems have been investigated through both empirical paleoecological data and ecosystem models, but the results of these two approaches are often difficult to reconcile. Fossil pollen has been used to describe past shifts in forest composition, but the centennial-scale resolution of these data make it difficult to identify the ecological processes causing that change. In contrast, process-based ecosystem models provide a formalized structure for investigating the mechanisms through which climate change impacts forests. However, simplifying assumptions inherent in the modeling process confound model data-comparisons, particularly at the large spatial and temporal scales at which past climate-driven ecosystem change has occurred. We have reconciled some of the fundamental discrepancies between models and data by: (i) developing novel statistical methods for estimating composition and biomass from pollen and (ii) using generalized additive mixed modeling to generate time series from ecosystem models that reflect the properties of the composition and biomass pollen products. We have combined these advances in an analysis of empirical data and ecosystem model simulations to identify how and why shifts in climate from 850-1850 AD cause changes in forest composition and biomass in the Midwestern and Northeastern United States.
Results/Conclusions
The effects of climate change on ecosystem dynamics in models and data must be compared in environmental rather than geographic space due to reconcile simplifying assumptions inherent in each approach. When temporal stability of drought is used as a measure of climate change through time, pollen products and all models show a weak, but positive correlation between climate and forest composition stability. However, pollen products do not show any correlation between climate stability and biomass. Only ecosystem models with detailed representation of plant ecophysiology (ED2, LPJ-GUESS, LPJ-WSL) are consistent with the lack of correlation between climate and biomass stability. The two models with simpler vegetation dynamics (JULES-TRIFFID and LINKAGES) show tight coupling among climate and stability responses of ecosystem states such as gross primary productivity, leaf area index, biomass, and composition. The models with patterns consistent with pollen products suggest that shifts in forest composition facilitate biomass stability in the face of climate change. A regional, long-term perspective is necessary to ensure that climate and ecosystem variability across space and time is sufficient for feedbacks among ecosystem processes in models to produce the emergent patterns seen in empirical data.