Crop–water interactions define productivity in water-limited dryland agricultural production systems in cold regions. Despite the agronomic and economic importance of this relationship there are challenges in quantifying crop water use efficiency (WUE). To understand dynamics driving crop water use and agricultural productivity in these environments, observations of evapotranspiration, carbon assimilation, meteorology, and crop growth were collected over 17 site-years at 5 agricultural sites in the sub-humid continental Canadian Prairies. Eddy-covariance (EC) derived water and carbon fluxes provided a means to comprehensively assess the WUE of current agricultural practices by both physiological (WUEP: g C kg−1 H2O) and agronomic (WUEY): kg yield mm H2O−1 hectare−1) approaches. Mean field scale WUEY for grain yields were 10.4 (Barley), 10.2 (Wheat), 6.0 (Canola), 19.3 (Peas), 12.2 (Lentils) and for silage/forage crops were 23.0 (Barley), 11.9 (Forage), and 20.7 (Corn) (kg yield mm H2O−1 hectare−1). An assessment of environmental factors and their covariance with WUE, utilising a conditional inference tree approach, demonstrated that WUE decreased when crops were under greater evapotranspiration demands. EC-based areal WUE approaches, measuring fluxes over footprints of hundreds of square metres, were compared with more commonly reported point-scale water balance residual approaches (WUEWB) and demonstrated consistently smaller magnitudes. WUEWB was greater than EC-estimated WUEY by an average of 52% and 65% for grain and forage/silage crops respectively. WUEWB also had greater variability than EC estimates, with standard deviations 188% and 128% greater than Barley and Wheat crops, respectively. This comparison highlights the scale dependency of WUE estimation methods, demonstrates considerable uncertainty in point scale water balance approaches due to spatial variability in crop–water interactions, and shows how this variability can be accounted for by EC observations. This improves the understanding of WUE and quantifies its variability in cold continental water-limited climates and provides a means to diagnose improved agricultural water management.