The object of this study is to evaluate an ensemble forecast of extreme storm surge by using a case of Typhoon Haiyan (2013) and its associated storm surge. A simple numerical model composed of ARW-WRF, FVCOM and SWAN is employed as a forecast system for storm surge. This ensemble system can successfully forecast storm surge 3-4 days before it happened. However, the typhoons in almost all ensemble members were underpredicted probably because of its difficulty in forecasting a track and central pressure of highly intense typhoon. This leads to the underestimation of a prediction of storm surges around Leyte Gulf. Compensating the underestimation of forecasted extreme storm surge, it can be important to not only examine the ensemble mean among members but also consider the phase-shifted manipulation and the worst ensemble member in the case where the extreme storm surge is forecasted. In addition, the ensemble forecast system can have a potential to determine the time at which the peak of extreme surge appears with a high precision.