Abstract: This study conducts parametric uncertainty and sensitivity analysis of large shallow lake hydrodynamic models. The Environmental Fluid Dynamics Code (EFDC) model is applied to simulate the flow velocity and water level in Lake Taihu that is characterized by its shallowness and large surface area. The Latin hypercube sampling (LHS) method is used to sample values for five uncertain parameters in the EFDC model, which including the wind drag coefficient, the roughness height, the eddy viscosity coefficient, the turbulent diffusion coefficient, and the wind shelter. The results show that uncertainties of simulated hydrodynamics process exist due to the contributions of model parameter uncertainties. The extents and ranges of uncertainties in large lake hydrodynamic models (e.g., EFDC) are highly associated with the spatiotemporal distribution of winds, shorelines, lakes bottom topography, and geography around the lake. Among those parameters, the wind drag coefficient and the wind shelter play the most important role in the spatial distribution of modeled velocity and water level, especially in those semi-closure bays and the lake regions with complex topography in the lake. Vertically, the velocity in the surface layer is also largely influenced by the two wind parameters, followed by the velocity of bottom layer, and the middle velocity has a minimal impact. The roughness height also makes a contribution to the uncertainty of simulated hydrological process. However, the uncertainties of viscosity coefficient and turbulent diffusion coefficient have no clear effect on model simulations. Therefore, the wind drag coefficient, the wind shelter, and the roughness height should be paid much attention when calibrating a hydrodynamic model of larg shallow lakes. Additionally, LHS is a cost-effective sampling method to reduce the number of parameters needs to be calibrated and to improve the accuracy of numerical simulations.