Abstract: Abstract: Crop growth simulation models are important in agricultural planning and management. The simulation of crop phenology is the basis of correct simulation of growth and development processes in crop models. Calculation of accumulative thermal time is a common way of simulating crop phonological development in crop models, while the effects of photoperiod and vernalization are also considered. However, the response of crop phenology to water stress is rarely quantified and often neglected. The main objective of this study was to explore and quantify the mechanism of phenology response of winter wheat (Triticumaestivum L.) to water stress. Experiments were conducted in plastic columns under a rainout shelter for winter wheat growing under water stresses at different growth stages in two growing seasons (from October 2013 to June 2014 and from October 2014 to June 2015). Another independent field experiment was conducted under a rainout shelter for winter wheat under water stresses at different growth stages in three continuous growing seasons (2012-2013, 2013-2014, and 2014-2015). In this study, relative water availability (Aw) was chosen as water stress index. When Aw was below a certain value of A (defined as critical point of accelerating development), crop began to hasten development, while there was no effect on crop phenological development above A. When Aw was below a certain value of S (defined as critical point of ceasing development), crop development stopped. Thus, it was reasonable to propose that there existed a certain value of D (defined as critical point of decelerating development) between points A and S. Thus, Aw would hasten crop development between A and D and delay development between D and S. When Aw did not affect crop phenological development, the value of water modification factor (WMF) was set as 1; when accelerating crop development, WMF was greater than 1; and when decelerating development, WMF was smaller than 1. Then, modified physiological day (MPD) was computed through multiplying WMF with physiological day (PD). The values of MPD were used to quantify the phenology response of winter wheat to soil water stress. The soil column experimental data of 2014-2015 growing season were used to calibrate the phenology water stress response function. The estimated values of relative water availability of points A, D, and S were 0.30, 0.10 and 0, respectively. The root mean square error (RMSE) between simulated and observed jointing and flowering dates were 0.8 and 1.7 d. The values of absolute relative error (ARE) were below 0.68% and 2.09%, respectively. When verifying the phenology water stress response function with the data of 2013-2014 soil column experiment, the RMSE between simulated and observed jointing and flowering dates were 0.9 and 1.1 d and ARE were less than 1.37% and 1.68%, respectively. When verifying the modified phenology algorithm with the data of independent field experiment of three growing seasons, the RMSE between simulated and observed flowering and maturity dates were 2.4 and 2.0 d and ARE were less than 4.21% and 2.67%, respectively. Compared with the simulation results of CERES-Wheat model in the DSSAT, it showed that the modified algorithm was able to reflect the influences of water stress on winter phenology while CERES-Wheat model showed no difference among different treatments in the same year. The RMSE between CERES-Wheat simulated and observed flowering and maturity dates were 4.0 and 5.5 d and the maximum error were 8 and 6 d, respectively. The results of calibration and verification showed that the phenology water stress response function developed in this study could be used to accurately simulate the variations in phenological dates of different winter wheat varieties caused by different scenarios of soil water stress. This response function needs to be evaluated further in more field experiments and then be embedded in current popular crop models, such as CERES-Wheat in the DSSAT model, to improve their simulation accuracy of phenology under water stress conditions. Consequently, modified crop models are supposed to have a better accuracy and applicability in arid and semi-arid areas.