The impact of spatiotemporal variability in atmospheric CO 2 concentration on global terrestrial carbon fluxes

Land carbon fluxes, e.g., gross primary production (GPP) and net biome production (NBP), are controlled in part by the responses of terrestrial ecosystemsto atmospheric conditions near the Earth's surface. The Coupled Model Intercomparison ProjectPhase 6 (CMIP6) has recently proposed increased spatial and temporal resolutions for the surface CO 2 concentrations used to calculate GPP, and yet a comprehensive evaluation of the consequences of this increased resolution for carbon cycledynamics is missing. Here, using global offline simulations with a terrestrial biosphere model, the sensitivity of terrestrial carbon cyclefluxes to multiple facets of the spatiotemporal variability of atmospheric CO 2 is quantified. Globally, the spatial variability of CO 2 is found to increase the mean global GPP by 0.2 PgC year −1 , as more vegetated land areas benefit from higher CO 2 concentrations induced by the inter-hemisphere gradient. The temporal variability of CO 2 , however, compensates for this increase, acting to reduce overall global GPP; in particular, consideration of the diurnal variability of atmospheric CO 2 reduces multi-year mean global annual GPP by 0.5 PgC year −1 and net land carbon uptake by 0.1 PgC year −1 . The relative contribution of the different facets of CO 2 variability to GPP are found to vary regionally and seasonally, with the seasonal variation in atmospheric CO 2 , for example, having a notable impact on GPP in boreal regions during fall. Overall, in terms of estimating global GPP, the magnitudes of the sensitivities found here are minor, indicating that the common practice of applying spatially-uniform and annually increasing CO 2 (without higher frequency temporal variability) in offline studies is a reasonable approach – the small errors induced by ignoring CO 2 variability are undoubtedly swamped by other uncertainties in the offline calculations. Still, for certain regional- and seasonal-scale GPP estimations, the proper treatment of spatiotemporal CO 2 variability appears important.