River water temperature has been regarded as an important ecological and environmental assessment factor. Intervention of human activities (such as dam construction) often causes abnormal changes in river water temperature and significant effects on river ecosystem. The effective water temperature prediction can provide a good indication for the long-term evolution trend of the river thermal dynamics. In this paper, the Air2stream model, which is a newly data-driven water temperature prediction model, has been used in the Yangtze River to explore its temperature variation before and after the operation of Three Gorges Dam. The performance of the model is assessed by the root mean square error (RMSE) and the Nash efficiency coefficient (NSE), with a comparison with other empirical regression models. The results show that the Air2stream model has the advantages of higher accuracy and better stability in the medium and long-term water temperature prediction for the river water temperature that would have occurred in absence of the large-scale engineering works or other intensive human interventions. In addition, the difference between observed and simulated values of river water temperature can be used as a useful indicator for assessing the impacts of human interventions. The case study applied to the Yangtze River shows that Air2stream can be used to determine the extent of variation of water temperature and main controlled factors, as well as quantitatively analyzes the contributions of both human interventions and climate change to water temperature mutation. Therefore, it can provide an important technical support for monitoring and management of river ecosystem.