In order to quickly realize the traceability of sudden water pollution in river basin, a cloud source identification system in river basin based on the cloud-edge-terminal collaboration architecture was constructed. By comparing the two-dimensional hydrodynamic water quality prediction results of the whole basin in the cloud with the water quality monitoring data of the terminal in real-time, sudden water pollution accidents could be identified. After the accident warning, rapid water quality simulation of multi pollution scenario in accident area was conducted at the edge end to form a scenario database based on source-quality response. By comparing and analyzing the process fitting degree of monitoring information and pollution scenario simulation sequence, the scope of pollution source screening was determined. The method was applied to the source identification of sudden water pollution in Chongqing section of the Yangtze River. The results show that under a certain pollution scenario, the Nash-Sutcliffe efficiency coefficient of pollutant discharge enterprises A and B for class 1 pollution process of downstream monitoring section are 0. 91 and 0. 87 respectively, and for class 2 pollution process are 0. 84 and 0. 95 respectively. It indicates that enterprises A and B are possible accident sources, and the source of class 1 pollution process is more likely to be enterprise A, and the source of class 2 pollution process is more likely to be enterprise B.