This paper describes storm surge simulations made for Sandy (2012) for the Metropolitan New York (NYC) area using the Advanced Circulation (ADCIRC) model forced by the Weather Research and Forecasting (WRF) model. The atmospheric forecast uncertainty was quantified using 11-members from an atmospheric Ensemble Kalman Filter (EnKF) system. A control WRF member re-initialized every 24 h demonstrated the capability of the WRF-ADCIRC models to realistically simulate the 2.83 m surge and 4.40 m storm tide (surge + astronomical tide) above mean lower low water (MLLW) for NYC. Starting about four days before landfall, an ensemble of model runs based on the 11 “best” meteorological predictions illustrate how modest changes in the track (20–100 km) and winds (3–5 m s−1) of Sandy approaching the New Jersey coast and NYC can lead to relatively large (0.50–1.50 m) storm surge variations. The ensemble also illustrates the extreme importance of the timing of landfall relative to local high tide. The observed coastal flooding was not the worst case for this particular event. Had Sandy made landfall at differing times, locations and stages of the tide, peak water levels could have been up to 0.5 m higher than experienced.