Dynamic global vegetation models (DGVMS) vary widely in their simulation of the terrestrial carbon balance, and this poor performance is the major stumbling block in our ability to predict the…

Dynamic global vegetation models (DGVMS) vary widely in their simulation of the terrestrial carbon balance, and this poor performance is the major stumbling block in our ability to predict the global carbon balance under climate change. Modelintercomparison studies have pointed out the common weaknesses or process differences among models, but there has been little systematic effort to improve model parameterizations or account for their uncertainty for global scale simulations. There is high natural variability in as well as different approaches to deriving the parameters that are used to define a very simplified set of plant functional types (PFTS). Because of computational constraints, global vegetation models cannot simulate individuals or all species diversity when coupled to climate models, and must capture at best the statistics of geographic variability in plant diversity and behavior. Thus far, the mean, let alone the variability, in biophysical, biochemical, and behavioral characteristics for a so-called plant functional type has not been well-defined. We have been developing a dynamic global vegetation model, the ENT terrestrial biosphere model (ENT TBM) (kiang ET AL. 2006), which is the first to represent vegetation heterogeneity with the statistical-mechanical approach of the ecosystem demography (ED) model (moorcroft ET AL., 2001) specifically for coupling with general circulation models (GCMS). ENT augments ED by quantifying foliage clumping for better accuracy in canopy radiative transfer and the surface energy balance. ENT's biophysics and seasonal growth dynamics have been well-tested for several PFTS at fluxnet sites. ENT is coupled to the NASA goddard institute for space studies GCM (giss GCM) MODELE2, capable of both off-line experiments with only land surface hydrology, or fully coupled with the GCM and carbon tracers. We would like now rigorously to determine parameter sets that are appropriate for global scale simulations. Because the ED approach statistically describes vegetation structure, IT is appropriate that parameter sets should also define limited plant functional types according to their statistics across a global scale. This proposal is to do a systematic mining of several large, recently compiled, globally ranging databases of plant traits and phenological timings to generate parameter fits and their variances for relations used in ENT. These include relations between leaf traits that affect photosynthesis and conductance, spectral albedoes of leaf and stem, geometric allometry, biomass allometry, plant tissue quality, specific respiration, turnover, and ecophysiology. We will use the technique of linear manifold clustering also to identifyalternative sets of PFTS. We expect to develop parameter sets that distinguish the ENT PFTS, as well as identify global and climate relations that may help reduce parameter sets or the number of PFTS. Propagation of variances in these parameter sets will be approached through latin hypercube sampling methods in simulations using observed climate drivers (e.g. 50-year dataset of sheffield ET AL. 2006). GCM coupled simulations will be conducted to identify how GCM biases alter the carbon balance, how uncertainties propagate through future climate change scenarios, and how natural vegetation cover change may have different trajectories given parameter variance. A community version of ENT will be released for use in global scale experiments, with technical reports and publications providing quantification of model performance and estimation of uncertainty in the terrestrial carbon balance given observed climate. This proposal is responsive to the program call for investigations utilizing the NASA giss GCM and observations aimed at identifying, characterizing, and reducing uncertainties in dgvm predictions of climate change on seasonal time scales. Form nress