Aiming to improve its ability to simulate crop growth and yield under different scenarios of soil water stress, the current stress factors of water stress index 1 (WSI1) was modified and two more different indices of WSI2 and WSI3 were created for the root zone water quality model (RZWQM2). The index of WSI2 was based on a modification of the soil water factor (SWFAC) for photosynthesisrelated processes in RZWQM2 using daily potential root water uptake calculated by the Nimah and Hanks approach. The index of WSI3 was based on WSI2 but with new terms accounting for the stress due to additional heating of the canopy by unused energy for potential soil evaporation in both the supply and demand terms of WSI2. The RZWQM2 model and these three different water stress indices were evaluated by using the data of soil water, total transpiration, biomass and grain yield obtained from a winter wheat experiment conducted under different water stress scenarios at different growing stages in two consecutive growth seasons of 2012—2013 and 2013—2014. The results showed that these three water stress indices were all able to correctly simulate the dynamic changes of winter wheat growth and soil water content when water stress occurred during the heading and grainfilling stages. However, when water stress occurred earlier and severer, the simulation accuracy became lower. Generally, the water stress index of WSI2 was better than WSI1 under different scenarios, since the relative root mean square error (RRMSE) for biomass was decreased by 2.84 percentage point, and the absolute relative error (ARE) for yield was decreased by 1.43 percentage point, especially when water stress occurred during the wintering and greening stages. In general, there were still some limitations for the RZWQM2 model to simulate winter wheat growth under arid conditions. Thus, the RZWQM2 model can be further improved in the following aspects, such as the response of phenology to soil water stress, the improvement of ET simulation, and the compensation effect of rehydration on winter wheat after long water stress. It was suggested that model users should choose WSI2, if the RZWQM2 model was expected to simulate the response of winter wheat to soil water stress.