Vegetation biogeography is a main source of uncertainty in modelling the land carbon cycle.

The terrestrial biosphere exchanges a large amount of CO2 with the atmosphere through photosynthesis and respiration, determining the magnitude of land carbon sink and consequently influencing the rate of global warming. The magnitudes of global photosynthesis and respiration, however, vary widely across models (100-200 PgC/year), constituting a key and persistent source of uncertainty in carbon cycle and climate modelling. Here, we argue that the uncertainty in the land carbon cycle modelling is largely attributable to the uncertainty in biogeography - the distribution of plant functional types (PFTs). Using an ensemble of dynamic global vegetation models (DGVMs), we find a strong dependence of total photosynthesis on total area for each PFT. The dependence allows us to reduce the spread of land carbon cycle estimates by ~75% using remote sensing-based PFT maps. We further find that 56 ± 21% of climate-driven changes in global photosynthesis modelled by DGVMs are caused by changes in PFT distribution in the last two decades. Our study identifies vegetation biogeography as a main controlling factor of uncertainty in land carbon cycle modelling and highlights the importance of biogeography-climate interactions in carbon cycle and climate studies.