Parameterization of a diagnostic carbon cycle model for continental scale application

Diagnostic carbon cycle models depend on parameterization to establish model sensitivity to climate variables and site factors. Here we acquired meteorological and carbon flux data from a diverse set (N = 18) of eddy covariance (EC) flux towers, along with MODIS data on FPAR (the fraction of incident photosynthetically active radiation that is absorbed by the plant canopy) at the sites, and used the data to develop a parameter set for the application of a diagnostic carbon cycle model over North America. The parameter optimization approach relied on goodness of fit between model simulations and tower estimates of gross primary production and net ecosystem production (NEP). Parameters such as light use efficiency (LUE) and base rate of heterotrophic respiration varied widely between sites representing different plant functional types (PFTs), thus supporting the value of stratification by PFT when parameterizing the model. Where multiple EC sites were available within a PFT, overall prediction error and bias in mean NEP was reduced by cross-site optimization as opposed to reliance on a single site. Optimization with the MODIS Enhanced Vegetation Index (EVI) instead of MODIS FPAR resulted in a similar goodness of fit, however, LUE values were pushed to levels that were not physiologically realistic when using EVI. The increasing availability of gap-filled EC tower data is rapidly improving the opportunities for direct coupling of satellite and ground observational data for parameterizing of diagnostic carbon flux models.