We apply the Finite Volume Coastal Ocean Model to simulate the Hurricane Ike storm surge using two-dimensional (2-D) and three-dimensional (3-D) formulations. The high resolution, unstructured grid extends over the Gulf of Mexico with open boundaries in the Straits of Florida and the Yucatan Channel. With the same wind and pressure forcing, the bottom drag coefficients for the baseline 2-D and 3-D simulations are determined by spatially varying Manning coefficients and constant bottom roughness, respectively. The baseline 2-D model simulates both the forerunner and the surge, whereas the baseline 3-D model simulates the surge, but underestimates the forerunner. Increasing the minimum Manning coefficient reduces the 2-D forerunner and the surge. Manning coefficient and bottom roughness parameterizations produce different bottom drag coefficients. Using the same bottom drag coefficient, the 2-D simulation yields a smaller surge than in three dimensions. This is investigated for scenarios of either constant or variable bottom roughness where the bottom roughness is determined through Manning coefficient transformation. These sensitivity studies indicate that storm surges, simulated either in two dimensions or three dimensions, depend critically upon the parameterizations and the parameter values used for specifying bottom stress (and similar may be said of surface stress). Given suitable calibration, 2-D and 3-D models may adequately simulate storm surge. However, it is unclear that a calibration for a given storm and location may apply generally. Hence additional experimental guidance is required on the parameterizations and the parameter values used for both the surface and bottom stresses under severe wind conditions.