The objective of this study was to project the detailed changes of extreme climate over South Korea under the two Representative Concentration Pathways (RCP) scenarios. Long term simulations for current (1979-2010) and future (2019-2050) climate over North-East Asian region were performed using RegCM4 with 12.5 km resolution driven by HadGEM2-AO under two RCPs 4.5/8.5 scenarios provided by the Korea Meteorological Administration (KMA). The simulation skills of RegCM4 for the extreme climate simulations were evaluated with the daily observation data (maximum/minimum temperature, precipitation) of 50 stations from KMA. Two types of definition for the extreme climate were used based on the intensity and frequency of weather event. The one is the percentile method (highest/lowest 5%: four extremes: daily maximum, minimum, diurnal temperature range, and extreme daily precipitation) and the other is absolute threshold method (seven extremes: hot, tropical night, frost, freezing, wet, dry, and heavy rain days). Also, three types (mean, median, and linear regression) of bias correction method were developed to minimize the systematic biases of the simulation results by RegCM4.
RegCM4 driven by HadGEM2-AO simulated comparatively well the spatial distribution of extreme climate in South Korea in comparison with observation. However, the RegCM4 generally underestimated intensity and frequency of most of extreme climate indices. Biases for the TX5%, TN5%, and ETR5% are -1℃, -2℃ and -1℃, respectively. Although the improvements were not consistent and systematic, the applications of bias correction have reduced the RegCM4''s bias and RMSE (Root Mean Square Error), and increased the spatial correlation. The simulation results of RegCM4 with two RCP (4.5/8.5) scenarios showed that the maximum and minimum temperature will be increased in South Korea. Also regardless of RCP scenarios, the freezing (tropical night) days are expected to significantly decrease (increase) by the mid of the 21st century. The correction levels of the bias correction methods is clearly dependent on the types of extreme climate. The results showed that more robust bias correction method is needed to improve the projection skills for the extreme climate indices over South Korea.