Centennial-scale ecological interactions between U.S. forests and climate: Using models and data to constrain the historic biosphere-atmosphere carbon cycle
Interactions between ecological systems and the atmosphere are the result of dynamic processes with system memories that persist from seconds to centuries. Adequately capturing long-term biosphere-atmosphere exchange within earth system models (ESMs) requires an accurate representation of changes in plant functional types (PFTs), vegetation biomass, and leaf area index (LAI) through time and space, particularly at timescales associated with ecological succession. However, most model parameterization and development has occurred using datasets than span less than a decade. We tested the ability of ESMs to capture the ecological dynamics observed in paleoecological and historical data spanning the last millennium. Focusing on an area from the Upper Midwest to New England, we examined differences in the magnitude and spatial pattern of PFT distributions and ecotones, vegetation biomass, and LAI between historic datasets and the CMIP5 and MsTMIP inter-comparison project’s large-scale ESMs. We then conducted a 1000-year model inter-comparison using six state-of-the-art biosphere models at sites that bridged regional temperature and precipitation gradients.
The distribution of ecosystem characteristics in modeled climate space reveals widely disparate relationships between modeled climate and vegetation that led to large differences in long-term biosphere-atmosphere fluxes for this region. We hypothesized that much of the difference between data and models was due to the modeled response of vegetation dynamics to infrequent, but extreme climate responses such as drought stress. To test this hypothesis, we examined model interactions through our own model intercomparison. Model simulations revealed that both the interaction between climate and vegetation and the representation of ecosystem dynamics within models were important controls on biosphere-atmosphere exchange.