As shown in Figure 4, before the TGR was put into operation, the flow of the reconstructed Yichang, Luoshan, and Hankou Stations was basically consistent with the measured flow, and the peak flow was well captured. The MAPE during the training period was all lower than 10%. It shows that the model performs well (Moriasi et al. 2007). It can be considered that this model can accurately reflect the natural flow process in the MRYR without considering the reservoir impoundment. By comparing the measured and reconstructed values during the post-TGR period, the impact of the TGR on rivers during the non-flood season is more pronounced. If there was no TGR, the peak flow rate would be larger, and the magnitude of flow during the dry season would be smaller. The Yichang Station is located downstream of the TGR, and its flow process is most severely affected by the TGR. In terms of the fit between simulated and measured values, the MAPE and RMSE at Yichang Station increased from 13.33% and 1,522 m³/s during the pre-TGR period to 18.18% and 3,083 m³/s during the post-TGR period, and R² decreased from 0.980 to 0.870 (Table 5).