Evaluating interannual variability in modeled global carbon fluxes: the impacts of uncertainties in model parameters and driver data

Confidence in predictions of future greenhouse gas levels depends on the ability to accurately model recent and current atmospheric CO₂ variability. Observed global atmospheric CO₂ growth rate for the past few decades has exhibited large interannual variability (IAV), most of which can likely be attributed to the differential response of terrestrial net primary productivity (NPP), heterotrophic respiration (Rh), and fire emissions to climate anomalies such as those associated with ENSO events (Nemani et al., 2003; van der Werf et al., 2006; Zeng and Qian, 2005). Terrestrial carbon models generally do not capture CO₂ growth rate IAV or do so for suspect reasons. In this study, we characterize the Carnegie-Ames-Stanford Approach - Global Fire Emissions Database (CASA-GFED) modeled carbon flux anomalies in terms of timing and magnitude in relation to atmospheric observations of CO₂ growth rate. Through sensitivity analysis we identify model parameters that have a large impact on modeled flux IAV and we use Monte Carlo methods to propagate uncertainty in parameters and driver data to the modeled fluxes. Finally, we compare modeled flux anomalies with recent large regional climate anomalies known to influence carbon fluxes (e.g. Ciais et al., 2005). The results identify strengths and weaknesses in the ability of the model to capture IAV in net carbon flux and atmospheric CO₂ growth rate, and highlight ways to improve model performance. Satellite-constrained diagnostic models such as CASA-GFED provide a benchmark for prognostic models used to predict future terrestrial carbon fluxes. References: Ciais et al. (2005) Nature 437, 529-533, doi: 10.1038/nature03972. Nemani et al. (2003) Science 300, 1560-1563, doi: 10.1126/science.1082750. van der Werf et al. (2006) Atmos. Chem. Phys. 6, 3423-3441. Zeng and Qian (2005) Geophys. Res. Lett. 32, L22709, doi: 10.1029/2005GL024607.