Flood backwater generally occurs where main and branch rivers meet,resulting in a prolonged high water level and posing significant challenges to flood simulation.This paper focuses on Dalai station,located the downstream of Nenjiang River,which is obviously affected by flood backwater.Firstly,the backwater types of historical flood events re-corded at Dalai Station was analyzed.And then these flood events were categorized based on the identified backwater types.On this basis,a flood simulation model was developed using long short-term memory(LSTM)networks,and its performance was further evaluated.The results show that the change rate of observed flow and water level can effectively identify the types of backwater,and the Nenjiang River experiences the backwater effect in most years.Meanwhile,for the LSTM model input,considering only upstream inflow has a relatively minor impact on the accuracy of simulated flow,while it significantly affects the accuracy of simulated water level.The LSTM model obtains high accuracy in both simulated flow and water level when considering downstream backwater effects.The developed LSTM model accurately captures the flood dynamics at the selected Dalai station affected by downstream backwater effect,which offers reference for flood simulation in similar basins or stations.
维普
万方数据