Abstract
As the ratio of carbon uptake to water use by vegetation, water‐use efficiency (WUE) is a key ecosystem property linking global carbon and water cycles. It can be estimated in several ways, but it is currently unclear how different measures of WUE relate, and how well they each capture variation in WUE with soil moisture availability. We evaluated WUE in an Acacia ‐dominated woodland ecosystem of central Australia at various spatial and temporal scales using stable carbon isotope analysis, leaf gas exchange and eddy covariance (EC) fluxes. Semi‐arid Australia has a highly variable rainfall pattern, making it an ideal system to study how WUE varies with water availability. We normalized our measures of WUE across a range of vapour pressure deficits using g 1 , which is a parameter derived from an optimal stomatal conductance model and which is inversely related to WUE. Continuous measures of whole‐ecosystem g 1 obtained from EC data were elevated in the 3 days following rain, indicating a strong effect of soil evaporation. Once these values were removed, a close relationship of g 1 with soil moisture content was observed. Leaf‐scale values of g 1 derived from gas exchange were in close agreement with ecosystem‐scale values. In contrast, values of g 1 obtained from stable isotopes did not vary with soil moisture availability, potentially indicating remobilization of stored carbon during dry periods. Our comprehensive comparison of alternative measures of WUE shows the importance of stomatal control of fluxes in this highly variable rainfall climate and demonstrates the ability of these different measures to quantify this effect. Our study provides the empirical evidence required to better predict the dynamic carbon–water relations in semi‐arid Australian ecosystems.