Abstract: Agricultural water accounts for more than 60% of the total water consumption in China. Among them, crops irrigation accounts for about 90% of the total agricultural water consumption. Therefore, it is highly demanding to transform the water management in modern irrigation systems. Alternatively, water-saving transformation can be utilized to save irrigation water, where the Water-saving Potential (WSP) is a key parameter to evaluate the performance of system. It is also necessary to consider the scale effect of parameters. Taking the Yang-shu-dang (YSD) watershed in Zhanghe Irrigation System in Hubei Province of China as the study area, a distributed hydrological model was established in this study using the modified Soil and Water Assessment Tool (SWAT). Some datasets were collected to validate the model, including the daily discharge at the outlet of YSD watershed in 2005-2009 and 2016-2017, the daily evapotranspiration in the typical experimental fields in 2007-2008 and 2016-2017, the irrigation water consumptions from the local water source (drainage channels and ponds) and the Zhanghe Reservoir in 2005-2008, and the irrigation amount in the typical experimental field in 2016-2017. The simulation performance of model was also assessed using the Relative Errors (RE), the coefficients of determination (R2), and the Nash-Sutcliffe Efficiency coefficients (NSE). The results showed that the modified SWAT model presented an excellent performance to simulate the hydrological processes and irrigation water consumptions in a multi-source system. For example, the NSE of daily discharge reached 0.81 in the calibration period, whereas, that of daily evapotranspiration reached 0.90, and the RE of irrigation water consumption was only 2.75%. Moreover, the sub-basins nesting was used to divide the study area into six size scales. Three water-saving scenarios were set for the study area, namely the increased drainage area of ponds, the increased water utilization coefficient of canal system, and the rice water-saving irrigation. The traditional and new WSPs were calculated under different scenarios with the modified SWAT model, particularly considering the reuse of return flow in different scales, further to analyze the change rules. In the increased drainage area of ponds, the traditional and new WSPs first increased, then decreased, and finally stabilized with the increase of scale area, while the change rule depended mainly on the variation in water supply proportion of ponds over different scales. In the increased water utilization coefficient of canal system, the traditional and new WSPs gradually decreased, and then stabilized with the increase of the scale area. The reason was that the great variation in the water supply proportion of Zhanghe Reservoir over different scales. In rice water-saving irrigation, the traditional and new WSPs first decreased, and then increased as the scale increased, while finally remained stable. The change rule depended on the changes in the ratio of paddy fields at different scales. Finally, the relationship between two WSPs depended mainly on the changes in irrigation water consumption, and the reused amount of return flow at different scales before and after the implementation of water-saving measures. This study can provide decision-making basis and guidance for water management and water-saving reconstruction project layout in irrigation systems.