Comparisons were made among Douglas-fir forest, aspen (broad leaf deciduous) forest and wheatgrass (C3) grassland for ecosystem-level water-use efficiency (WUE). WUE was defined as the ratio of photosynthetic CO2 assimilation rate and evapotranspiration (ET) rate. The ET data measured by eddy covariance were screened so that they overwhelmingly represented transpiration. The three sites used in this comparison spanned a range of vegetation (plant functional) types and environmental conditions within western Canada. When compared in the relative order Douglas-fir (located on Vancouver Island, BC), aspen (northern Saskatchewan), grassland (southern Alberta), the sites demonstrated a progressive decline in precipitation and a general increase in maximum air temperature and atmospheric saturation deficit (Dmax) during the mid-summer. The average (±SD) WUE at the grassland site was 2.6±0.7 mmol mol−1, which was much lower than the average values observed for the two other sites (aspen: 5.4±2.3, Douglas-fir: 8.1±2.4). The differences in WUE among sites were primarily because of variation in ET. The highest maximum ET rates were approximately 5, 3.2 and 2.7 mm day−1 for the grassland, aspen and Douglas-fir sites, respectively. There was a strong negative correlation between WUE and Dmax for all sites. We also made seasonal measurements of the carbon isotope ratio of ecosystem respired CO2 (δR) in order to test for the expected correlation between shifts in environmental conditions and changes to the ecosystem-integrated ratio of leaf intercellular to ambient CO2 concentration (ci/ca). There was a consistent increase in δR values in the grassland, aspen forest and Douglas-fir forest associated with a seasonal reduction in soil moisture. Comparisons were made between WUE measured using eddy covariance with that calculated based on D and δR measurements. There was excellent agreement between WUE values calculated using the two techniques. Our δR measurements indicated that ci/cavalues were quite similar among the Douglas-fir, aspen and grassland sites, despite large variation in environmental conditions among sites. This implied that the shorter-lived grass species had relatively high ci/ca values for the D of their habitat. By contrast, the longer-lived Douglas-fir trees were more conservative in water-use with lower ci/ca values relative to their habitat D. This illustrates the interaction between biological and environmental characteristics influencing ecosystem-level WUE. The strong correlation we observed between the two independent measurements of WUE, indicates that the stable isotope composition of respired CO2 is a useful ecosystem-scale tool to help study constraints to photosynthesis and acclimation of ecosystems to environmental stress.