We estimated changes in temperature-related extreme events over South Korea for the mid and late 21st Century using the 122 years (1979-2100) data simulated by RegCM4 with HadGEM2-AO data as boundary conditions. We analyzed the four extreme events (Hot day: HD, Tropical day: TD, Frost day: FD, Icing Day: ID) and five extreme values (Maximum temperature 95/5 percentile: TX95P/TX5P, Minimum temperature 95/5 percentile: TN95P/TN5P, Daily temperature range 95 percentile: DTR95P) based on the absolute and relative thresholds, respectively. Under the global warming conditions, hot extreme indices (HD, TD, TX95P, TN95P) increase, suggesting more frequent and severe extreme events, while cold extreme indices (FD, ID, TX5P, TN5P) decrease their frequency and intensities. In the late 21st Century, changes in extremes are greater in severe global warming scenario, RCP8.5 rather than RCP4.5. HD and TD (FD and ID) are expected to increase (decrease) in the mid 21st Century. The average HD is expected to increase by 14 (17) days in RCP4.5 (8.5). All the percentile indices except for DTR95P are expected to increase in both RCP4.5 and RCP8.5. In the late 21st Century, HD and TD are significantly increased in RCP8.5 compared to RCP4.5, but FD and ID are expected to be significantly reduced. HD is expected to increase mainly in the southwestern region, twice (+41 days) in RCP8.5. TD is expected to increase by 17 days in RCP8.5, which is 5 times greater than that in RCP4.5. TX95P, TN95P and TX5P are expected to increase by about 2°C and 4°C in RCP4.5 and RCP8.5, respectively. TN5P is expected to increase significantly by 4°C and 7°C in RCP4.5 and RCP8.5, respectively.

We investigated on the proper combination of physical parameterization schemes of RegCM4.0 for the simulation of regional climate over CORDEX-East Asia Phase 2 domain. Based on the Lee(2016)’s sensitivity experiments for the four combination using two land surface schemes and two cumulus parameterization schemes during 5 years (1979-1983), we selected the two combinations (CE: CLM+Emanuel and BG: BATS+Grell). The ERA-Interim was used as lateral boundary data of RegCM4.0 for the two experiments during 25 years (1981-2005). Simulation skills of temperature were similar in the two combination of physical processes irrespective of seasons and locations. However, there were a substantial differences in the simulation skills of precipitation according to the combination of physical processes, which were better in CE than BG combination. In general, the CE combination better simulated the precipitation characteristics in July and August over South Korea than BG combination, in terms of frequency and amount of precipitation according to the intensity. The superior skills of CE in simulating precipitation over South Korea can be related to the better simulation of seasonal march of the East Asian summer monsoon including the location and intensity of the North Pacific high pressure system than BG. The results suggested that the CE combination can simulate the climate characteristics in the CORDEX East Asia Phase 2 region better than the BG combination.

We conducted dynamic downscaling using the RegCM4.0 with 25 km of horizontal resolution over CORDEX-East Asia phase 2 domain for the current climate (1981-2005) and evaluated its performance using various reference datasets. The HadGEM2-AO with about 110 km of horizontal resolution provided by the Korea Meteorological Administration (KMA) was used as the forcing data of RegCM4.0. the RegCM4.0 generally well simulated the spatial patterns of temperature and precipitation by representing geographical and topographic conditions, compared with HadGEM2-AO. In particular, the warm biases in Mongolia, northern China, and Russia during summer and cold biases in Tibetan Plateau, Mongolia, northeast China, and northern India during winter were reduced. However, the systematic wet (dry) biases of summer precipitation in the most of model domain (South Korea) were still remained. It was associated with the southward shifting of low-level jet caused by the weakened North Western Pacific High in HadGEM2-AO. In South Korea, the RegCM4.0 showed a better performance than HadGEM2-AO in terms of the magnitude and spatial variability of both temperature and precipitation. In particular, the RegCM4.0 performed better in simulating the ratio of extreme precipitation over 100 mm/day to total precipitation than HadGEM2-AO. The RegCM4.0 reasonably reproduced the frequency and inter-annual variability of heavy rainfall, frost day, and tropical night over South Korea.