Vertical gradients in the canopy represent a major challenge for scaling from foliar photosynthesis to ecosystem-level CO₂ fluxes. We tested whether accounting for independent gradients of carboxylation capacity (V_cmax) and photosynthetic electron transport (J_max) improves estimates of forest net ecosystem CO₂ exchange (NEE). We modified the process-based Soil-Plant-Atmosphere (SPA) model to represent different gradients of foliar N allocation to V_cmax and J_max in the canopy, and inversely calibrated the model via Bayesian inference using eddy covariance measurements of NEE and evapotranspiration from a mixed-deciduous forest. Inversely calibrated N allocation resulted in highest V_cmax at the top and highest J_max at the bottom of the canopy, which is similar to N allocation gradients from measured foliar traits. These vertical gradients resulted in the best fit of simulated CO₂ fluxes to measured NEE compared to alternative N allocation schemes, due to a higher, more realistic foliar CO₂ uptake in the lower canopy. Canopy gradients of V_cmax and J_max are thus important drivers of NEE and need to be considered to improve simulations of ecosystem-level CO₂ fluxes of forests.