We focus on the improvement of accuracy of sea surface wind over complex coastal area during the warm season. Local Analysis Prediction System (LAPS) was used to improve the initial values in Mesoscale Meteorological model (MM5). During the clear summer days with weak wind speed, sea surface wind simulated with LAPS was compared with the case without LAPS. The results of modeling with LAPS has a good agreement mesoscale circulation such as mountain and valley winds on land and in case of modeling without LAPS, wind speed overestimated over the sea in the daytime. And the results of simulation with LAPS indicated similar wind speed values to observational data over the sea under influence of data assimilation using BUOY, QuikSCAT, and AMEDAS. The present study suggests that MM5 modelling with LAPS showed more improved results than that of without LAPS to simulate sea surface wind over the complex coastal area.