Nutrient dynamics in lakes are determined by the combined effects of external and internal nutrient fluxes. However, the feedback loop of nutrient fluxes and water quality changes is still an open question. An integrated three-dimensional flux tracking approach based on the Environmental Fluid Dynamics Code model was established to quantify the long-term dynamic changes in external and internal processes in Lake Dianchi (one of the three most hyper-eutrophic lakes in China), and to explore the potential causes of water quality improvement during 2012–2018. The long-term trends and relative influences of nutrient fluxes on water quality were identified. The results showed that the inflow flux was the largest input source and declined by 50% between 2003 and 2012, which was followed by a stable trend from 2012 to 2018. The second largest input source was benthic release, which exhibited a significant interannual reduction. Algae sedimentation was the largest removal process, and declined by 45% between 2012 and 2018. An integrated analysis demonstrated that, following an external loading reduction prior to 2012, the positive feedback of internal fluxes promoted water quality improvement during 2012–2018. Considering the long timescale of the nutrient-flux feedback mechanism, reducing external loading is still the top priority for a long-term virtuous cycle of water quality in the process of eutrophic lake restoration.