Reconciling leaf physiological traits and canopy flux data: Use of the TRY and FLUXNET databases in the Community Land Model version 4

The Community Land Model version 4 overestimates gross primary production (GPP) compared with estimates from FLUXNET eddy covariance towers. The revised model of Bonan et al. (2011) is consistent with FLUXNET, but values for the leaf-level photosynthetic parameter Vcmax that yield realistic GPP at the canopy-scale are lower than observed in the global synthesis of Kattge et al. (2009), except for tropical broadleaf evergreen trees. We investigate this discrepancy between Vcmax and canopy fluxes. A multilayer model with explicit calculation of light absorption and photosynthesis for sunlit and shaded leaves at depths in the canopy gives insight to the scale mismatch between leaf and canopy. We evaluate the model with light-response curves at individual FLUXNET towers and with empirically upscaled annual GPP. Biases in the multilayer canopy with observed Vcmax are similar, or improved, compared with the standard two-leaf canopy and its low Vcmax, though the Amazon is an exception. The difference relates to light absorption by shaded leaves in the two-leaf canopy, and resulting higher photosynthesis when the canopy scaling parameter Kn is low, but observationally constrained. Larger Kn decreases shaded leaf photosynthesis and reduces the difference between the two-leaf and multilayer canopies. The low model Vcmax is diagnosed from nitrogen reduction of GPP in simulations with carbon-nitrogen biogeochemistry. Our results show that the imposed nitrogen reduction compensates for deficiency in the two-leaf canopy that produces high GPP. Leaf trait databases (Vcmax), within-canopy profiles of photosynthetic capacity (Kn), tower fluxes, and empirically upscaled fields provide important complementary information for model evaluation.