Soil moisture supply and atmospheric demand for water independently limit-and profoundly a ect-vegetation productivity and water use during periods of hydrologic stress [1][2][3][4] . Disentangling the impact of these two drivers on ecosystem carbon and water cycling is di cult because they are often correlated, and experimental tools for manipulating atmospheric demand in the field are lacking. Consequently, the role of atmospheric demand is often not adequately factored into experiments or represented in models 5-7 . Here we show that atmospheric demand limits surface conductance and evapotranspiration to a greater extent than soil moisture in many biomes, including mesic forests that are of particular importance to the terrestrial carbon sink 8,9 . Further, using projections from ten general circulation models, we show that climate change will increase the importance of atmospheric constraints to carbon and water fluxes in all ecosystems. Consequently, atmospheric demand will become increasingly important for vegetation function, accounting for >70% of growing season limitation to surface conductance in mesic temperate forests. Our results suggest that failure to consider the limiting role of atmospheric demand in experimental designs, simulation models and land management strategies will lead to incorrect projections of ecosystem responses to future climate conditions. Ecosystem moisture stress is often characterized by changes in soil water availability 10,11 . Declining soil moisture impedes the movement of water to evaporating sites at the soil or leaf surface 12 , reducing the surface conductance to water vapour (G S )-a key determinant of carbon and water cycling-and thereby evapotranspiration (ET). However, atmospheric demand for water, which is directly related to the atmospheric vapour pressure deficit (VPD), also affects G S and ET. Plants close their stomata to prevent excessive water loss when VPD is high [13][14][15][16] and thus, increases in VPD during periods of hydrologic stress represent an independent constraint on plant carbon uptake and water use in ecosystems.While the plant physiological community has long recognized the critical role of VPD in determining plant functioning, VPD is often overlooked in many fields of hydrologic and climate science. For example, precipitation manipulation experiments are frequently used to draw conclusions about ecosystem response to drought stress, even though VPD is unaffected by precipitation manipulation 10 . Some terrestrial ecosystem and ecohydrological models do not permit stomatal conductance to vary with atmospheric demand 5,11 . Many models designed to capture these impacts rely on empirical parameterizations for soil moisture and VPD stress that promote compensating effects and model equifinality 5 , and/or use relative humidity instead of VPD as the primary driver, with significant consequences for projections of