Change of extreme events of temperature and precipitation over Korea using regional projection of future climate change



[1] This study investigates possible changes of extreme events in global warming over Korea with MM5 downscaling simulation during the period 1971–2100. Frequency distribution of daily temperatures over Korea shows an increase in the mean by about 5.5°C from 1971–2000 to 2071–2100 while change in the variance is negligible. Increasing temperature results in changes in the frequency and intensity of temperature extremes. Under the current climate change scenario, hot events are expected to be more frequent and intense, while cold events will be rare and weaker. The increasing trend of temperature is associated with an increasing trend of precipitation. The increasing trend produces an increase in the number of the days of heavy precipitation as well as the corresponding amount. Better resolved topography in MM5 produces bigger changes in local precipitation than in the temperature field. Consequently increasing tendency is obvious in the northern part of Korea.

1. Introduction

[2] During the last century, global mean surface temperature has risen by about 0.6°C [Intergovernmental Panel on Climate Change (IPCC), 2001]. Coupled global climate models based on the Special Report on Emission Scenario (SRES) also showed the temperature has continued to increase. The global mean surface temperature is projected to rise by 1.4–5.8°C over the period 1990 to 2100 due to the projected increases of greenhouse gas concentrations in the atmosphere. Global mean temperature increase is likely to change characteristics of extreme climate event [Meehl et al., 2000; Griffiths et al., 2005]. So the IPCC Third Assessment Report was interested and comprehensively dealt with possible future changes of extreme climate events in global warming.

[3] Climate change gives different effects on the geographically diverse region. Therefore, changes in extreme events show large regional variations. The large variations are found in the studies of observations as well as regional climate projections. For example, Easterling et al. [2000] mentioned that frequency of heavy precipitation events exhibits regionally different tendency between the northern and southern parts of Japan as well as China. The regional dependency in the change of extreme events is demonstrated in the future projections. In Bell et al.'s [2004] study, trends of frequency of 1-day extreme temperature events show regional differences as large as three-fold magnitude in the increasing rate of the frequency. Hennessy et al. [1997] reported substantial increases in the frequency of extreme daily rainfall and maximum temperature over Australia. Increases in the number days of heavy rainfall were also documented over East Asia [Gao et al., 2001]. Similar study was carried out for Britain [Jones and Reid, 2001].

[4] Based on the previous studies, changes of extreme events can be found in complex regions such as the Korean Peninsula in global warming. The changes are likely to appear at regional scale. Therefore the present study addresses possible changes in extreme events over the Korean Peninsula based on the global warming scenario. We focused mostly on changes in the frequency and intensity of the daily mean temperatures and daily amounts of precipitation.

2. Data and Method

[5] The data sets are the daily temperature at 2 m height and daily precipitation simulated by MM5 at 27 km horizontal resolution for the period 1971 to 2100. The model domain is centered at 38°N and 125°E and covers an area of East Asia and the whole Korean Peninsula. The initial and boundary conditions for MM5 are provided from a long-term simulation based on IPCC SRES A2 scenario using the ECMWF Hamburg Atmosphere Model Version 4 coupled with the Hamburg Ocean Primitive Equation-Global (ECHAM4/HOPE-G) model of Max Planck Institute for Meteorology. The ECHAM4/HOPE-G model is described by Zorita et al. [2003] and Min et al. [2003] and details of downscaling methodology are documented by Oh et al. [2004] and Kwon et al. [2004].

[6] For comparison with the simulated data, we used observed temperature data in Korea Meteorological Administration (KMA) for the period of 1971 to 2000. Temperature data of North Korea is collected from Global Telecommunication System (GTS) for the period of 1981 to 2000.

[7] In the analysis of daily temperature, we used data in four seasons. For daily precipitation, we used data in rainy season; June, July, and August. Frequency of daily precipitation is represented by probability density function of a gamma distribution. The function is calculated with the algorithm based on Hosking and Wallis [1997].

3. Results

3.1. Temperature

[8] Griffiths et al. [2005] showed that correlations between mean temperature and the frequency of extreme temperatures were strongest in the Asia-Pacific region. Meehl et al. [2000] and Mearns et al. [1984] also mentioned that the small change in the mean temperature can result in a large change in the frequency of extremes. Therefore, to investigate possible changes in the frequency of temperature extremes, frequency distribution of daily mean temperature in Korea is presented in a statistical sense for the entire data record length (Figure 1). Frequency distribution of daily mean temperature simulated by MM5 tends to be asymmetric with respect to the mean value. The mean value by the MM5 simulation for the period 1971 to 2000 has a cold bias compared with that of the KMA observation.

Figure 1.

Probability density function of the distribution for daily mean temperature in Korea. Solid and dashed lines indicate the MM5 simulation during 1971–2000 and 2071–2100, respectively. Dash-dotted line means KMA observations for southern part during 1971–2000.

[9] In Figure 1, simulated daily temperature for the period of 1971 to 2000 is projected to show a shift in the mean of a distribution by about 5.5°C for the period of 2071 to 2100. Compared to the change in the mean value, change of the standard deviation appears to be negligible. Considering temperature extreme with specific temperature thresholds, hot and cold events can be defined with the days above and below the threshold. Rising in the daily mean temperature results in increase in the number of hot days exceeding a predefined threshold value. In a similar way, the number of cold days below specific threshold is expected to decrease. So, a change in the distribution in Figure 1 can give influence on probabilities of extreme hot and cold events. Changes in frequency of hot and cold events are related to change in length of the events [Bell et al., 2004]. Their results indicate that the increase in daily minimum temperature is associated with decrease in days below freezing and prolonged cold events occur less often and are shorter and warmer on average.

[10] To estimate change in intensity of temperature extremes, we define intensity of hot and cold events with the 95th percentile of daily maximum temperature in summer and the 5th percentile of daily minimum temperature in winter, respectively. Horizontal distribution of the 95th percentile of observed daily maximum temperature shows that the percentile is high over the southern part and low over the eastern and northeastern part (Figure 2). The spatial features also appear in the distribution of the 5th percentile (Figure 2). Since the eastern and northeastern part of the Korean Peninsula is characterized by complex mountains, the lower percentile in the eastern and northeastern part seems to be related to the topographical effect. The observed distributions in Figure 2 are similar to those of MM5 simulated distribution of Figure 3. The similarities between the observation and MM5 simulation were also shown by Boo et al. [2004a, 2004b].

Figure 2.

(a) The 95th percentile of daily maximum temperature in summer and (b) 5th percentile of daily minimum temperature in winter. Southern part is based on the period of 1971–2000 in KMA observation and northern part is on the period of 1981–2000. Contour interval is 2°C.

Figure 3.

(a) The 95th percentile of daily maximum temperature in summer and (b) 5th percentile of daily minimum temperature in winter for the period of 1971–2000 in MM5 observation. (c) and (d) Changes in the percentiles from 1971–2000 to 2071–2100. Contour interval is 2°C for (a), (b) and 0.5°C for (c), (d).

[11] Regional projection of this study shows increases in the 95th and 5th percentiles. The 95th percentile of daily maximum temperature increases by about 4.–5.°C from 1971–2000 to 2071–2100 over the entire domain of Korea. As the 95th percentile are rising, so are the 5th percentile of daily minimum temperatures. The 5th percentile increases by about 7.–9.°C toward the north over Korea. The increases in 5th percentile of daily minimum temperature are about twice as large as the increase in the 95th percentile of daily maximum temperature (Figures 3c and 3d).

3.2. Precipitation

[12] Regional projection in this study shows increases in temperature as well as in the amounts of precipitation over the Korean Peninsula in the end of 21st century. Simulated precipitation is highly variable on spatial scales. Since high-resolution regional model simulation embedded in the global models is affected by better-resolved topography [Giorgi et al., 1994; Kato et al., 2001] and smaller-scale atmospheric dynamics to be simulated [IPCC, 2001], regional climate change presents detailed patterns of precipitation different with the driving general circulation model simulation [IPCC, 2001]. To consider local change, we divided the analysis area over Korea into southern and northern parts. Southern part is defined as the area of 126.3°E–129.5°E and 34.7°–38.1°N. Northern part covers the area of 125.1–129.3°E and 38.8–41.5°N.

[13] Figure 4 presents changes in frequency distribution of daily precipitation from 1971–2000 to 2071–2100. Changes in frequency of daily precipitation are slightly different between the southern and northern parts. For the northern part, the number of days with daily precipitation above 20 mm/day tends to increase and that of daily precipitation below 20 mm/day decreases. For the southern part, the number of days with daily precipitation below 5 mm/day slightly increases and that of daily precipitation between 5 mm/day and 40 mm/day slightly decreases.

Figure 4.

Probability distribution functions for daily precipitation (mm/day) in summer during 1971–2000 (solid) and 2071–2100 (dash-dotted) in Korea based on MM5 simulation. (left) Southern part (34.7°–38.1°N, 126.3°E–129.5°E) in Korea and (right) northern part (38.8°–41.5°N, 125.1°–129.3°E).

[14] The local change in frequency of precipitation events was discussed by Easterling et al. [2000]. Heavy precipitation events in Japan show different long-term trends of the frequency between northern and southern part. The results are similar with those in China. Accordingly, we expect changes in the frequency of extreme events are so variable on regional scale for complex regions such as the Korean Peninsula.

[15] The above results for heavy precipitation events are obviously confirmed in Figure 5. Figure 5 represents trends of precipitation amount deviated from the mean of 1971–2000. Annual precipitation amount during 2071–2100 increases by about 11% over the southern part of Korea with respect to the period of 1971–2000. For the northern part of Korea, precipitation amount is estimated to increase by about 28%. Considering change in the amount contributed by the heavy precipitation above 30 mm/day, the amount increases are 34% and 98% for the southern and northern parts, respectively. For the events above 50 mm/day, the amount increase is 57% and 168% for the southern and northern parts, respectively. Accordingly, the northern part of Korea will experience remarkable increases in the amounts of heavy precipitation events compared to the southern part. Increasing tendencies of the amount of heavy precipitation in the excess of 30 mm/day and 50 mm/day in Figure 5 also accompany increasing tendency of the ratio of the heavy precipitation amount contributing to the annual precipitation (figure not shown).

Figure 5.

Trends of the annual precipitation amount over southern (bar) and northern region (line) in Korea deviated from the mean during 1971–2000. Trends of the amount due to daily precipitation above 30 mm/day, above 50 mm/day, and below 10 mm/day are also presented.

[16] On the other hand, the moderate precipitation events below 10 mm/day show decrease by 3% for the southern part and 1% for the northern part. The frequency of light and moderate precipitation changes locally different in Figure 4, and the total amount decrease for both regions even with slightly different magnitude.

[17] In Figure 4, the possibility of increasing frequency of days of heavy precipitation events appears for the northern part. Considering the precipitation amounts of individual heavy precipitation events, the increasing trend is more distinct (Figure 5).

4. Conclusions

[18] To estimate possible future climate changes in extreme events over Korea, we have analyzed the regional projection data produced by MM5 downscaling simulation techniques based on SRES A2 scenario.

[19] Simulated frequency distributions of daily temperature show a shift in the mean by about 5.5° while holding the variance similar from 1971–2000 to 2071–2100 (Figure 1). Rising in the mean implies frequency change of extremes based on specific threshold. Daily mean temperature increase is associated with rising in the 95th and 5th percentiles of daily maximum and minimum temperatures. It results in changes in the intensity of temperature extremes (Figure 3). Hot events become hotter by about 4.–5.°C. Therefore, hot events are expected to be more severe under the current climate change scenario. Cold events are warmer by about 7.–9.°C in a future climate.

[20] Large-scale forcing arising from global warming may locally change the precipitation distribution over the Korean Peninsula typified with complex terrain topography. Trenberth et al. [2003] have mentioned that the changes in total precipitation arising from global warming are likely to change the intensity and frequency of precipitation events. Simulated precipitation distribution demonstrates local change in the moderate and heavy precipitation occurrence (Figure 4). The increasing trend in annual precipitation is associated with increasing trend of frequency and intensity of the heavy precipitation. The increasing tendency is distinct in the northern part compared to the southern part (Figure 5). The local dependency of changes in heavy precipitation events in this study is consistent with the previous observational results for the regions of Japan and China [Easterling et al., 2000].


[21] This study is supported by the projects of METRI “Research on the Development of Regional Climate Change Scenarios to Prepare the National Climate Change Report”. The authors are grateful to Dr. Hyun-Kyung Kim and anonymous reviewers for helpful comment on our manuscript and Dr. B. P. Kirtman for pdf fortran code.