The cause for large variability of hypoxia off the Changjiang Estuary has not been well understood partly due to various nutrient sources and complex physical-biological processes involved. The Regional Ocean Modeling Systems (ROMS) coupled with Carbon, Silicate and Nitrogen Ecosystem (CoSiNE) was used to investigate the 2006 hypoxia in the East China Sea, the largest hypoxia ever recorded. The model performance was evaluated comprehensively by comparing a suite of quantitative metrics, procedures and spatiotemporal patterns between the simulated results and observed data. The simulated results are generally consistent with the observations and are capable of reproducing the development of hypoxia and the observed vertical profiles of dissolved oxygen. Event-scale reduction of hypoxia occurred during the weakening of stratification in mid-July and mid-September, due to strong stirring caused by tropical storms or strong northerly wind. Change in wind direction altered the pathway of Changjiang Diluted Water and consequently caused variation in hypoxic location. Increase in river discharge led to an expansion of hypoxic water during the summer monsoon. Sensitivity analysis suggested that the hypoxia extent was affected by the change in nutrient concentration of the Changjiang as well as that of the Kuroshio. Sensitivity analysis also suggested the importance of sediment oxygen consumption to the size of the hypoxic zone. These results demonstrate that a prognostic 3D model is useful for investigating the highly variable hypoxia, with comprehensive considerations of multiple factors related to both physical and biological processes from the estuary to the shelf break of the East China Sea.