The Three Gorges Reservoir (TGR) is one of the world's largest hydropower projects and plays an important role in water resources management in the Yangtze River. For the sake of disaster prevention and catchment management, it is crucial to understand the regulation capacity of the TGR on extreme hydrological events and its impact on flow regime in a changing climate. This study obtains historical inflows of the TGR from 1961 to 2019 and uses a distributed hydrological model to simulate the future inflows from 2021 to 2070. These data are adopted to drive a machine learning-based TGR operation model to obtain the simulated outflow with TGR operation, which are then compared with the natural flow without TGR operation to assess the impact of TGR. The results indicate that the average flood peaks and total flooding days in the historical period could have been reduced by 29.2% and 53.4% with the operation of TGR. The relative declines in drought indicators including duration and intensity were generally less than 10%. Faced with more severe extreme hydrological events in the future, the TGR is still expected to alleviate floods and droughts, but cannot bring them down to historical levels. The impact of TGR operation on flow regime will also evolve in a changing climate, potentially altering the habitats of river ecosystems. This study proposes feasible methods for simulating the operation of large reservoirs and quantifying the impact on flow regime, and provides insights for integrated watershed management in the upper Yangtze River basin. The Three Gorges Reservoir (TGR) is located in the upper Yangtze River basin and is one of the world's largest hydropower projects. This study combines hydrological modeling and machine learning methods to quantify the effects of TGR operation on flow regime in the historical and future periods. The results indicate that the operation of TGR could play an important role in mitigating historical droughts and floods. Specifically, the flood peak in 1998 could have been reduced by 24.7% if the TGR was in operation. In the future, the extreme hydrological events will become more severe. While the TGR will continue to play a significant role, it cannot fully control floods and droughts to the same extent as it did in the historical period. In a changing climate, the Yangtze River basin will still face higher flood and drought risks with the operation of TGR. Under future climate scenarios, the operation of TGR will also alter the flow regime metrics and have new implications on the river ecosystem environment. An ensemble learning model based on two neural networks is constructed to simulate Three Gorges Reservoir (TGR) operations TGR can mitigate floods and droughts, but this capability might be weakened by more severe future extreme events TGR can offset most flow regime changes caused by climate change, but adaptive operation strategies are still needed
