You are here

Harvard Forest >

Harvard Forest Symposium Abstract 2006

  • Title: The State of the Regional Carbon Cycle: Results from a Constrained Coupled Ecosystem-Atmosphere Model
  • Primary Author: David Medvigy (Harvard University)
  • Additional Authors: Paul Moorcroft (Harvard University); Steven Wofsy (Harvard University)
  • Abstract:

    For over a decade, eddy covariance data have been consistently showing net uptake of carbon dioxide at Harvard Forest. Independently, biometric studies at Harvard Forest have found consistent increases in ecosystem carbon storage. On the regional scale, forest inventories have suggested that other parts of the northeast U.S. also may have been acting as carbon sinks during the 1980s and 1990s. The biological mechanisms responsible for this carbon cycling have been a matter of debate, in part because many of the spatiotemporal scales on which these mechanisms operate are not yet accessible to direct, continuous measurement. To help fill this gap in our understanding, we have developed a new land surface model, ED2, and constrained it with existing observations. ED2 was formulated by coupling the Ecosystem Demography (ED) ecosystem model with the LEAF-2 biophysical model. The resulting coupled model was designed to link process-level biological information obtained for individual plants or ecosystems over short time scales with observations that characterize the large spatial domain and long time scales of regional and global concern.





    We used the model to simulate two years'' worth of carbon fluxes, water fluxes, growth rates and mortality rates at Harvard Forest. These simulation results were compared to eddy covariance data and biometry data, and model parameters were adjusted until the model was in accord with the data. Parameter uncertainty and covariances were also estimated. We evaluated the optimized model by simulating ten years at Harvard Forest. The optimized model gives greatly improved predictions of NEP (Fig. 1a), growth (Fig. 1b) and mortality. In addition, the model successfully predicted GPP and respiration separately, and also predicted an accurate breakdown of growth and mortality into conifer and hardwood components. We then ran the model using the composition and climate forcing of Howland Forest but without any further changes to model parameters. The resulting predictions of the optimized model are in remarkable accord with the data considering the different composition of the two forests (Fig. 1c-d).





    We have used the model to predict regional vegetation dynamics and carbon fluxes. These runs also were done without any further optimization. The simulated region encompassed the northeastern U.S. and southern Quebec (from 41N to 52N and from 80W to 65W). This region included hardwood, coniferous and mixed temperate forests as well as boreal forests. The simulations ran for 10 years at 25~km resolution. The initial forest composition was obtained from FIA and Quebec forest inventory data. The observed decadal-scale growth rates are compared to the modeled growth rates in Fig. 2a and Fig. 2b. Considering that only data from Harvard Forest were used to constrain the model, the model shows very good agreement. The mean bias was only and the RMS error was.





    Currently we are using ED2 to investigate how interannual variability in regional carbon fluxes could be attributed to individual ecological and meteorological factors. In addition, we have recently coupled the ED2 to the RAMS mesoscale atmospheric model and are working on constraining the coupled land-atmosphere model with atmospheric measurements of carbon dioxide concentrations.

  • Research Category: Forest-Atmosphere Exchange
    Large Experiments and Permanent Plot Studies
    Regional Studies