A field study assessed the accuracy and consistency of crop water stress index (CWSI) estimates determined by two empirical models and one theoretical model for evaluation of bermudagrass [Cynodon dactylon (L.) Pers. cv. Midiron] water status. The simplest empirical model, describing canopy temperature minus air temperature (Tc — Ta) of well-watered turf as a linear function of air vapor pressure deficit, was inadequate for estimating CWSI, since seasonal averages of CWSI for well-watered, moderately stressed, and stressed turf had corresponding coefficients of variation of 1170, 160, and 86, respectively. The CWSI values were also highly influenced by net radiation, as indicated by coefficients of determination for regressions of calculated CWSI on net radiation of 0.68, 0.76, and 0.33 for the three levels of irrigation treatments, respectively. The second empirical model included net radiation in the regression analysis of Tc — Ta. The inclusion of net radiation in the equation for the unstressed baseline increased the coefficient of determination from 0.76 to 0.90 as compared with Model 1. The coefficient of determination for the upper baseline (which included net radiation) was 0.872, indicating a strong dependency of CWSI on net radiation. The theoretical CWSI model yielded the most accurate estimates of turfgrass water stress, as indicated by coefficients of variation for seasonal averages of CWSI for well-watered, moderately stressed, and stressed turf of 76, 3, and 86, respectively. Midday estimates of CWSI were related to percent available extractable water, with coefficients of determination for a regression of midday CWSI on percent available extractable water of 0.988, 0.989, and 0.991 for the empirical, modified empirical, and theoretical models, respectively. The theoretical CWSI appears to be the most promising approach for a turfgrass crop water stress index for irrigation scheduling.