This study has calculated the change of wind speed according to the features of land surface roughness using the surface wind data provided by the Korean peninsula data of HadGEM3-RA and has analyzed the characteristics of the future upper wind over South Korea driven by several climate change scenarios. The simulation found that the more the time passes, the more the wind speed increases in the previous time period of upper wind and annual average wind speed time series analysis of three kinds of Representative Concentration Pathways (RCP). The wind speed of all three kinds of RCP increased in the summer and winter but decreased in the spring and fall in the analysis of seasonal time series and spatial distribution. The wind speed would be expected to increase in most months except April and November in the analysis of the monthly mean maximum wind speed. The histogram analysis shows the mean wind speed of upper wind over 3m/s. As the time passes, the wind speed increases more than in the past. Certain areas such as the areas under the urbanization development would be anticipated to raise the wind speed throughout all months.

We focused on effects on data assimilation of simulated wind fields by using upper-air observations (wind profiler and sonde data). Local Analysis Prediction System (LAPS), a type of data assimilation system, was used for wind field modeling. Five cases of simulation experiments for sensitivity analysis were performed : which are EXP0) non data assimilation, EXP1) surface data, EXP2) surface data and sonde data, EXP3) surface data and wind profiler data, EXP4) surface data, sonde data and wind profiler data. These were compared with observation data. The result showed that the effects of data assimilation with wind profiler data were found to be greater than sonde data. The delicate wind fields in complex coastal area were simulated well in EXP3. EXP3 and EXP4 using wind profiler data with vertically high resolution represented well sophisticated differences of wind speed compared with EXP1 and EXP2, this is because the effects of wind profiler data assimilation were sensitively adjusted to first guess field than those of sonde observations.