Large scale inter-basin water diversion projects would have a set of ecological and environmental impacts on aquatic ecosystems. However, knowledge regarding water transfer induced water quality changes in determining fish communities in the impounding ecosystems remain largely unknown. In the current study, we filled this research gap by using a set of machine learning algorithms to ensemble modelling fish community indices with water quality indicators in the impounded lakes along the eastern route of China's South-to-North Water Diversion Project (SNWDP). Overall, our results realed that water quality changes can be good predictors for the variation of fish community structures in these lakes. Although different model techniques highlighted different variable importance of water quality in determining fish communities, there is generally a consensus that the hydrological related water quality indicators like: water clarity (e.g., total suspended solids) and nutrient loading (e.g., phosphate) contributed the most to the changes of fish communities. Our results also indicated that water diversions could bring knock-on effects on fish communities. Thus, more attention should be paid to long-term ecological effects from future water diversions.