The influence of the asymmetric structure of hurricane wind field on storm surge is studied with five types of numerical experiments using a three-dimensional storm surge model. The results from the case of Hurricane Floyd (1999) show that Floyd-induced peak surge would have been much higher had the region of maximum wind rotated 30–90° counterclockwise. The idealized cases (the hypothetical hurricanes) with a wind speed asymmetry of 20 m s?1 show that the peak (negative) surge varied from 4.7 to 6.0 m (?5 to ?5.7 m) or equivalent to ?8.8% and 16.3% (2.8% and ?10.4%) differences as compared to the control experiment. The area of flooding varied from 3552 to 3660 km2. The results from two other idealized cases of varying degree of wind speed asymmetry further show that with decreasing (increasing) asymmetry of wind fields, the variations of peak surge and peak negative surge caused by the rotation of wind fields decrease (increase) accordingly. The results suggest that in storm surge simulations forced by winds derived from balanced models, considerable uncertainty in storm surge and inundation can result from wind asymmetries. This is true even if all other storm parameters, including maximum wind speed, the radius of maximum winds (storm size), minimum central pressure, storm translation speed, drag coefficient, and model settings (domain size and resolution) are identical. Thus, when constructing ensemble and probabilistic storm surge forecasts, uncertainty in wind asymmetry should be considered in conjunction with variations in storm track, storm intensity and size.