Drought is the main factor affecting crop grain yield. Increasing grain yield under drought and crop water use efficiency (WUE) is essential for enhancing world crop production and food availability. The objective of this study, carried out in India on 20 durum wheat cultivars, under three water regimes (full irrigation, limited irrigation and residual soil moisture) and during two seasons, was to investigate the potential use of plant traits (particularly carbon isotope discrimination, ?, and ash content, ma) to predict grain yield and WUE in wheat. WUE components were estimated using a soil water balance model (Budget) allowing comparison of environments in data scarce situations. A highly significant correlation was noted between grain yield and grain ? across water regimes. However, the associations between grain yield, ? and ma were found to depend highly on the water regime and environmental conditions. The association between grain yield and grain ? was significant under full irrigation in season 1 and under residual soil moisture in season 2. Significant positive correlations were noted in both seasons between grain yield and leaf ? under residual soil moisture and between grain yield and leaf ash content at anthesis under limited irrigation. A significant correlation was found across environments between grain and leaf ? and T, the quantity of water transpired during the growth cycle, as estimated by the soil water balance model. T also significantly correlated to grain and leaf ma. Variation in WUE across environments was driven more by runoff, drainage and soil evaporation than by harvest index and transpiration. The associations between WUE and transpiration, runoff and ? were negative but not significant. WUE was significantly correlated with leaf and grain ma at maturity. The study indicates that ? and ma can be used as indirect selection criteria for grain yield and suggests that ma is a good predictor of transpiration, grain yield and WUE across environments. The use of mechanistic models that allows differentiating between cultivars should permit in a next future to analyze the relationships between WUE, ? and ma across cultivars and evaluate the possibility to use these traits as predictors of WUE in wheat breeding programs