Assessment of the changes in extreme vulnerability over East Asia due to global warming

Abstract

A number of indices have been employed to describe weather extremes on the basis of climate regimes and public concerns. In this study, we combined these traditional indices into four groups according to whether they relate to warm (T_warm), cold (T_cold), wet (P_wet), or dry (P_dry) extremes. Analysis of the combined indices calculated for the daily temperatures and precipitation at 750 meteorological stations in Korea, China, and Japan for 1960s–2000s shows increasing trends in T_warm and P_dry events and decreasing trends in T_cold events in recent decades, particularly in the northern part of East Asia. A notable regional variation is an increase in the P_wet events in the Korean Peninsula. We applied the same analysis to a 200-year global climate model simulation for 1900–2099 using the National Center for Atmospheric Research-Community Climate System Model 3. During the 20th century, the changes in T_warm and T_cold calculated from the model data are largely consistent with those calculated from the observations, especially in northern East Asia. The model projections for the 21st century indicate statistically significant increasing T_warm and decreasing T_cold trends in extreme events over the region. Results obtained from historical archives and model simulations using our combined weather extreme indices suggest that northern East Asia will be subject to increased warm and dry extremes and the Korea Peninsula will experience more wet extremes.

1 Introduction

Extreme weather events take a heavy toll on human lives, crops, water supply, industries, and ecosystems. Weather extremes, rather than changes in the mean climate, are the most consequential impact of global climate change (Kim 2005; Trenberth et al. 2007), though these extreme events may significantly respond to small changes in the mean climate (Katz and Brown 1992). Assessing the changes in extremes associated with recent climate change is thus one of the most important concerns of climate research.

The impact of climate change on a variety of weather extremes has been examined in a number of studies for North America (e.g., Karl et al. 1996; DeGaetano 1996), Europe (e.g., Heino et al. 1999; Osborn and Jones 2000), and East Asia (e.g., Iwasima and Yamamoto 1993; Zhai and Pan 2003, Park et al. 2011). In particular, East Asia has experienced intense natural disasters accompanied by extreme weather episodes in recent decades. As examples, record-breaking rainfall from typhoon ‘Rusa’ in 2002 resulted in a number of human casualties and economic losses of over 5 billion US dollars in Korea (c.f., Kim et al. 2006; Park and Lee 2007), and in September 2006, a heat wave occurred over an extended period across most of China with daily maximum temperatures well over 35°C (Ding et al. 2009). Such weather extremes cause substantial damages to the heavily concentrated industries and densely populated areas in East Asia. Reducing the damages inflicted by weather-related natural disasters is a priority in the region, and current disaster prevention plans include monitoring weather extremes and investigating the detailed mechanisms that produce such extremes.

Most previous studies have examined the characteristics of a single extreme weather event such as a typhoon, heavy rainfall, or warm/cold incident (Kim et al. 2005; Peterson and Manton 2008; Alexander and et al. 2006; Park et al. 2011). Single-event study is essential, because each event has some unique distribution or other characteristic. However, it is human nature for people to think that these extreme events happen only “elsewhere” and will not occur where they live even though global warming is changing weather worldwide. To provide perspective on the potential for extreme weather events, we here introduce a set of new indices created by integrating the existing various extreme-weather indices. Our new indices may help individuals to see the larger impact of various weather extremes and even the particular vulnerability of their own residential area to weather extremes.

We also investigated historical changes in extreme vulnerability, and modeled projected weather extremes expressed in the new indices, using observations and General Circulation Model (GCM, also known as Global Climate Model) simulations. The gridded values of our new group indices provide geographic visualization of regional and local vulnerabilities to extreme weather, information potentially very useful to both residents and policy makers.

This paper is organized as follows. Section 2 describes the dataset and methodology used for grouping indices to integrate various impacts of extremes. Sections 3 and 4 discuss changes in the extreme vulnerability over East Asia using these newly suggested indices, from observation and climate model projection, respectively. A summary is given in Section 5.

2 Data and method

2.1 Data

We analyzed the observed daily precipitation (PRCP), maximum temperature (TX), and minimum temperature (TN) over East Asia to assess the characteristics of historical extreme events. The daily datasets analyzed were obtained from meteorological stations operated by the governmental meteorological administrations of Korea, China, and Japan. The analysis period covers the recent five decades (1960s–2000s); the beginning and ending years of individual records vary according to stations and variables, but all selected stations included the records for 1970–1999. For data quality control, stations that had missing data for more than seven consecutive days or for a total of 30 days during the five decades were excluded. Although we could not check the individual accuracy of each station used, the data did show a spatially coherent pattern in the variation of each dataset, which we believe confirms the reliability of data used in this study (figure not shown). After filtering out stations that did not meet our quality control criteria, we collected the data from 607 Chinese, 118 Japanese, and 25 Korean stations, for a total of 750 meteorological datasets. The final list of stations included in the analysis is presented in Table 1.

Table 1 Numbers of Chinese, Korean, Japanese, and total meteorological stations used for calculating daily maximum temperature (TX), daily minimum temperature (TN), and daily precipitation (PRCP)

To project the changes in future extreme weather, we performed a future climate simulation using a coupled GCM (CGCM), the National Center for Atmospheric Research (NCAR) Community Climate System Model 3 (CCSM3; Collins et al. 2006), in conjunction with the Intergovernmental Panel on Climate Change (IPCC) Special Report on Emissions Scenarios (SRES) A2, A1B, and B1 (Nakicenovic et al. 2000) for projecting the climate in the 21st century (i.e., 2000–2099). The CCSM3 has been configured with T85 (approximately 1.4° × 1.4°) in the horizontal and 26 layers in the vertical. These scenarios assume that the carbon dioxide (CO₂) emission in the end of the 21st century will range from 5 gigatons of elemental carbon per year (GtC yr⁻¹) for B1, to 6 GtC yr⁻¹ for A1B, and up to 28 GtC yr⁻¹ for A2. The corresponding atmospheric CO₂ concentration is expected to range from 500 to 900 parts per million by volume (ppmv). In addition, a 20th century experiment (20C3M) for the period 1900–1999 was also performed for comparison with observations; forcing agents in this experiment were the historical records of greenhouse gases.

To confirm the reliability of the projected changes in this study, we additionally analyzed several coupled GCMs that participated in IPCC AR4. Among the 25 IPCC AR4 models, six included the required daily maximum and minimum temperature: BCCR-BCM2.0, CCCMA-CGCM3.1 (T42), CNRM-CM3, CSIRO-MK3.0, CSIRO-MK3.5, GFDL-CM2.0. More detailed descriptions of these six models are found in Table 2.

Table 2 Description of the coupled global circulation models (CGCMs) used in this study

2.2 Indices for extreme vulnerability

To evaluate the integrated impact of several weather extremes, we designed four simplified new indices by combining various indices previously used. Table 3 compiles the definitions of the single, previously defined, indices which were adapted from the Expert Team on Climate Change Detection, Monitoring and Indices (ETCCDMI) and have been widely used as tools to estimate and monitor changes in extremes (Peterson and Manton 2008; Alexander et al. 2006). Because combining indices of contrasting characteristics (e.g., warm extreme variables versus cold extreme variables) is likely to result in the loss of inherent characteristics of individual indices, we grouped a total of 24 indices into 4 categories according to their distinctiveness such as warm, cold, wet, and dry, as shown in Table 3. The warm extreme group (T_warm) contains the annual maximum of daily maximum/minimum temperatures (TXx/TNx), summer days (SU25), tropical nights (TR20), the occurrence of warm days/nights (TX90p/TN90p), and the warm spell duration indicators (WSDI). The cold extreme group (T_cold) consists of seven single indices including the minimum of daily maximum/minimum temperatures (TXn/TNn), the occurrence of cold days/nights (TX10p/TN10p), frost days (FD0), ice days (ID0), and the cold spell duration indicator (CSDI). In the wet category (P_wet), 10 single indices were grouped including: the maximum 1-day/5-day precipitation amounts (RX1day/RX5day), the annual precipitation total (PRCPTOT), simple daily intensity index (SDII) that shows precipitation rate, the number of very wet days (R95p), extremely wet days (R99p), heavy precipitation days (>10 mm day⁻¹; R10), very heavy precipitation days (>20 mm day⁻¹; R20), and extreme precipitation days (>50 mm day⁻¹; R50). The category of dry extreme (P_dry) is represented by only one index: the consecutive dry days (CDD).

Table 3 Indices for daily climate extremes based on ETCCDMI

Table 4 Changes (%) in T_warm and T_cold averaged over 80°E–145°E, 30°N–50°N for 20C3M (1900–1999), A2, A1B, and B1 (2000–2099) outputs

All indices above were calculated at each of the 750 meteorological stations. The station values were then converted into gridded values by averaging over the stations within a rectangular area of 5° × 5° centered at the grid point. After gridding all single indices, the individual indices were divided by the spatial and temporal means of the corresponding variables, because taking an average of all indices in their original values can exaggerate the impact of certain indices due to their different units and magnitudes. This standardization by the spatiotemporal mean also allows the resultant gridded indices to represent the relative strength of the extreme in a region at a given day compared to other regions and/or other days. That is, a grid point value exceeding 1 means that the type of extreme event gridded is more frequent (or stronger) at the grid point than the climatological value over the entire analysis domain (i.e., East Asia), implying that the grid point is more vulnerable to the extreme.

After the individual indices were gridded, they were combined in the groups detailed above, with the seven warm indices grouped as index T_warm, seven cold indices grouped as index T_cold, nine wet indices grouped as index P_wet, and the single dry extreme index noted as index P_dry. These group index grids visually show the change in the vulnerability of geographic residential regions to the corresponding extremes due to global warming. This visualization could be of significant use to policy makers, who can now easily compare the vulnerability of their region at present to the past vulnerability in the whole of East Asia, and can plan accordingly.

3 Extreme vulnerability in the historical archive

To examine the historical variations in the vulnerability to extreme weather, we have investigated the climatology for the entire period (i.e., 1960s–2000s) and the anomalies of our four new group indices for each decade. The value of T_warm varies from 0.2 to 1.8 with the maximum value in the southern China coastal region (Fig. 1a). In the zonal band along 30°N and 40°N, T_warm is generally larger in the eastern region with maxima over eastern China, Korea, and Japan, perhaps caused by the influence of the subtropical North Pacific high that covers the area during summer. Anomalies for each decade (Fig. 1b–f) show robust negative values in most of the region until the 1980s, and then they become positive in the 1990s and 2000s with a 95% confidence level (see gray dots). This pattern clearly indicates that warm temperature extremes represented by T_warm have been increasing during recent decades. In particular, as shown in the area marked by a rectangle in Fig. 1f, the increase in T_warm is especially large over northern China and Korea with the values ranging from 0.05 to 0.3 in the 2000s, 25% higher than the climatological averages for the region; compared to the 1960s, the increment in this area is 82% higher than the climatological average for the region.

Fig. 1

Climatology (a) and anomalies of warm temperature extremes (T_warm) for the 1960s (b), 1970s (c), 1980s (d), 1990s (e), and 2000s (f). Regions significant at the 95% level are indicated by gray dots

Figure 2 shows both the mean and decadal anomalies of group index T_cold, which is characterized by larger values in the northern inland region and smaller values in the southern regions, especially along the coastline (Fig. 2a). In the decadal anomalies, positive anomalies are largely distributed in the inland region for the 1960s (Fig. 2b), then become weak and scattered in the 1970s and the 1980s (Fig. 2c and d). In the recent two decades of the 1990s and the 2000s, negative anomalies are dominant (Fig. 2e and f). This pattern clearly indicates that the cold extremes represented by T_cold have been decreasing over East Asia, particularly over the northern regions marked by a rectangle in Fig. 2f. This decreasing trend of T_cold seems to correspond to the increase of T_warm in the boxed region. Although warming due to urbanization over northeastern China would be one of the important factors contributing to the increase in T_warm and decrease in T_cold (Ren et al. 2008), the significant changes are consistently found in the rural region in the western China. We believe this indicates that changes in vulnerabilities to warm/cold extremes during recent decades can be interpreted as signals related to global warming (Zhai and Pan 2003; Gong and Ho 2004; Alexander et al. 2006; Meehl et al. 2007).

Fig. 2

Same as Fig. 1 except for cold temperature extremes (T_cold)

Climatological vulnerability to wet extremes is generally large in the coastal regions in China, Korea, and Japan, which are strongly affected by the East Asian monsoon system and tropical storms (Fig. 3a). Unlike the two temperature extreme indices (T_warm and T_cold), P_wet anomalies are not spatially and/or temporally organized (Fig. 3b–f), which implies that wet extremes vary in a more complex manner spatially and temporally than do temperature extremes. Nevertheless, the increase in P_wet from the 1960s to 2000s in the vicinity of the Korean peninsula (the box area in Fig. 3f) is clearly defined in temporal variations of the index, increasing by about 50% during the period. This increase in precipitation extremes over Korea adds to a consensus of previous studies (e.g., Ho et al. 2003; Wang et al. 2006; Kim et al. 2006) which conclude that Korea has become more vulnerable to wet extremes in recent decades.

Fig. 3

Same as Fig. 1 except for wet precipitation extremes (P_wet)

The climatology and anomalies of P_dry are plotted in Fig. 4. Note that, unlike the other three indices, P_dry is composed of a single index, CDD. Climatologically, the largest value appears over the desert in northern China, which is nearly three times as large as values over the eastern region including eastern China, Korea, and Japan that are strongly affected by monsoon rainfall (Fig. 4a). Dry extremes are influenced mostly by large-scale circulations rather than regional convections, so the anomalies of P_dry show more coherent spatial patterns than P_wet (Fig. 4b–f). In northern inland China, which is affected by large variations in temperature extremes (the box area in Fig. 4f), the vulnerability to dry extremes intensified from the 1960s to 2000s, consistent with the documented recent desertification of northern China (Yang et al. 2007; Wang et al. 2008).

Fig. 4

Same as Fig. 1 except for dry precipitation extremes (P_dry)

Having analyzed the spatial distributions and decadal variations of the four indices from the climatology and the anomalies of each index per decade, we were able to recognize the locations of the core variability region of each index. However, the anomalies cannot show how much the core variability regions are vulnerable to these extremes compared to other regions, because the anomalies are simply the departures from the climatological averages. To approach this problem, we examined the annual time series of the four group indices averaged over the core variability region for each index. We also investigated the time series of the original individual indices that were grouped into the new indices, to examine whether the signal of change is exaggerated by combining several similar indices.

The black solid line in Fig. 5a shows the T_warm time series averaged over the core variability region, a rectangular region of 80°E–150°E and 35°N–45°N. T_warm values were about 0.8 in the 1960s, exceeded 1 after the 1990s, and since the 1990s have continuously increased with a statistical significance of 95%, implying that the impact of warm extremes in the region has become stronger as compared to the climatological mean for the entire East Asia region. As seen in the lower panel of the figure, this change is consistently found in the time series of the individual indices used in the construction of the integrated index. All indices show the increase in the corresponding warm extremes in the region, especially after the 1990s. In particular, the frequency indices (i.e., WSDI, TX90p, TN90p) have increased more than the intensity indices (i.e., TXx, TNx). Figure 5b presents the time series of T_cold and the individual components averaged over the same area with T_warm. Even though T_cold has gradually decreased until the 2000s with a statistical significance of 95%, the values remain greater than unity. This implies that T_cold responds more gradually to the global warming than T_warm does. For each index, an obvious decreasing trend in found in CSDI, TX10p, and TN10p.

Fig. 5

Time series of domain averaged indices and individual components for T_warm (a), T_cold (b), P_wet (c), and P_dry (d). The lower part of the panel (d) shows what the individual indices used in the construction of the integrated index indicate and a color bar for the time series of the individual indices in (a)–(c)

Figures 5c and d represent the averaged time series of P_wet and P_dry, respectively. Since P_wet is significant only in one grid that includes Korea, the index is averaged only for the region 125°E–130°E and 35°N–45°N. From 1961 to 2008, P_wet has steadily increased by 50%, with more rapid increases after the 1990s. All the individual indices are greater than 1, indicating that Korea is strongly affected by precipitation extremes in general. In particular, the heavy extremes such as R99p and R50 show much greater values than the moderate extremes such as R95p and R10 over the region. The time series in Fig. 5d shows that P_dry values averaged over the same area as the temperature indices, including selected regions of inland China that are climatologically dry, have increased in recent decades. In particular, the time series of P_dry shows a large shift after the late 1970s with a statistical significance of 95%, which means that the number of consecutive zero precipitation days has increased after the late 1970s (Ho et al. 2005).

4 Future extreme vulnerability projected in the model simulations

In order to assess the impact of global climate change on the weather extremes in East Asia, the new grouped indices have been calculated from the outputs of CCSM3 simulations (Section 2.1). Model-projections of changes in extremes suffer from large uncertainties, in particular, on the regional scale; however, at present, a climate model approach appears to be the best way to estimate the tendency of the changes in extremes.

We analyzed the output of a 20C3M scenario run covering 1900 to 1999 to evaluate the CGCM under the present-day climate condition. Figures 6 and 7 show the climatology and anomalies of T_warm and T_cold, respectively; the spatiotemporal variations of model output are shown in 20-year increments from 1900 to 1999. Compared to the observed climatology of T_warm, the value from the 20C3M run is shown to be greater in general, especially in southern China and Japan (Fig. 6a). The core of the variability occurs over these regions in the model simulation, though the observations locate the variability core in the inland area of China; this discrepancy shows the limitations of the CCSM3. As a whole, the pattern of enhancement of T_warm over East Asia in recent years is nearly the same as that of the observations (Fig. 6b–f). In the case of T_cold (Fig. 7), except for the overestimation in the southern region, the climatological values are similar to observed values in distribution and range, and the core of variability is also shown in the northern part of China, although it tends to be shifted slightly to the east. The pattern of variation is also consistent with that of the observations, showing that using the model results for projection of T_warm and T_cold is reasonable despite some uncertainty in projection of detailed features.

Fig. 6

Climatology (a) and anomalies of T_warm from 20C3M output for 1900–1919 (b), 1920–1939 (c), 1940–1959 (d), 1960–1979 (e), and 1980–1999 (f)

Fig. 7

Same as Fig. 6 except for T_cold

Figures 8 and 9 present the results of P_wet and P_dry, respectively. It is well known that the simulation of precipitation is one of the most difficult problems in GCM study, especially on the local and quantitative approaches, so it is not surprising that though the model climatology of P_wet shows similar features as the observations over the southern coastal region it overestimates over the inland China region (Fig. 8a). In the anomalies (Fig. 8b–f), the spatial distribution is scattered, similar to the observations. Considering only the changes from 1960 to 1999, P_wet seems to increase in the vicinity of Korean Peninsula, however, the significant increase over Korea found in the observations does not appear in the model data. Compared to the observations, P_dry is underestimated in the 20C3M simulation, and the high observed values over the northwestern part of the region cannot be found in the model shown in Fig. 9a. Particularly, the model does not show the signal of desertification over China in the anomalies of P_dry (Fig. 9b–f). The spatial distribution of anomalies in the model shows a dispersive pattern unlike the homogeneous pattern in the analysis of the observation. Given these results, we now know that obtaining reliable precipitation estimates from GCM simulation is difficult not only for mean climate modeling but also for modeling extremes, and we have excluded the modeling results of precipitation extremes from further consideration..

Fig. 8

Same as Fig. 6 except for P_wet

Fig. 9

Same as Fig. 6 except for P_dry

The primary goal of our efforts to project the changes in extremes is to estimate the future climate state. For this work, we used three SRES emissions scenarios including A2, A1B, and B1, for simulating the 21st-century climate. To easily compare the changes in the 21st century to those in the 20th century, we used the same value as the 20C3M (i.e., the averaged value over whole region during 1900–1999 in 20C3M) for the normalization of individual indices. Although the degree of changes varies between the three scenarios, the spatial patterns are similar among them (figure not shown). In further discussion, we will only explain the spatial variation of extreme vulnerability in the climate with the A2 scenario, in which the highest rate of carbon emission is assumed.

Figures 10a–f show the mean of T_warm during the last 20 years in the 20C3M simulation and the deviation of the projected T_warm into the 21st century from the 20-year mean, respectively. As expected, the amplitudes of changes in the projected index are very large; for instance, the mean of T_warm during the last 20 years of 20th century has a range between 0 and 2, but its increment during the 100 years of the 21st century is more than 1.8 in the region of maximum anomaly. Figure 11 shows the same calculations for T_cold; although the amplitude of changes in T_cold is not as much as that in T_warm, the decreasing trend is apparent over the whole region.

Fig. 10

Mean of T_warm during the last 20 years in the 20C3M simulation (a) and the deviation of the projected T_warm into the A2 simulation from the 20-year mean during 2000–2019 (b), 2020–2039 (c), 2040–2059 (d), 2060–2079 (e), and 2080–2099 (f)

Fig. 11

Same as Fig. 10 except for T_cold

It is noteworthy that the negative trend in T_cold exists even in the southernmost region where the value in the last 20 years of 20C3M is nearly zero. This means that T_cold over the region during the 21st century is the negative value itself, which could be induced by the fact that T_cold includes the index using the absolute value of daily minimum temperature, TNn. In other words, when the value used for normalization (i.e., averaged value over the temporal and spatial domain) is negative and the original TNn is positive in a certain grid, the normalized TNn can have a negative value in the grid. Then, if this normalized TNn contributes more to T_cold than the other six individual indices, T_cold may have a negative value. We could not find the negative value of T_cold in the observations or 20C3M simulation, suggesting that the influence of TNn may not be very large, but, in the A2 scenario for the 21st century, TNn over the southern region rapidly increases to a large positive value, while the value used for normalization shows a small negative value. In that interval, the influence of TNn is apparently sufficient to make T_cold negative.

Following the analysis of the observed data, we examined the temporal variation of domain-average indices using model outputs. As noted in Figs. 6 and 7, the selected domain is confined to 80°E–145°E, 30°N–45°N, which is an extensive region including a large variability compared to that of the observation analysis. To compare the changes in the different climate conditions, we also analyzed B1 and A1B scenarios in which the lowest and intermediate rates of carbon emission are assumed, respectively. In Fig. 12, red, green, and blue solid lines display the changes in extreme vulnerability for A2, A1B, and B1 scenario simulations, respectively. To show historical change, the result of the 20C3M scenarios is also plotted together as a black line. Consistent with the anomalies for each two-decade interval, T_warm shows increasing trends, and T_cold shows decreasing trends in all scenarios. The amplitude of changes is largest in A2 and lowest in B1 scenario, meaning that the further global warming progresses, the more T_warm and T_cold will change.

Fig. 12

Time series of domain averaged values for T_warm indices (a) and T_cold indices (b) from various model scenario run. Black, red, green, and blue solid lines indicate the results from 20C3M, A2, A1B, and B1 scenarios, respectively

Additional detailed assessment of changes in extreme vulnerability with various climate scenarios is summarized in Table 4. For T_warm, values have risen 37% from 1900 to 1999 in the 20C3M simulation and risen 95% from 2000 to 2099 in the result of A2 scenario. Although the increment in the B1 scenario is small compared to the others, it is still 29%. T_cold has dropped by 14% during 100 years of 20th century observations, and by 70% during 100 years of 21st century in the A2 scenario. In the future climate simulated by B1 scenario, the total decrement in T_cold is about 31%, however, this change mainly occurs in the first 50 years of the 21st century and the tendency of decline has disappeared after the 2040s.

To evaluate the projected changes in CCSM3 simulation, we additionally analyzed the output of six IPCC AR4 models for T_warm and T_cold. Figure 13 shows the mean of T_warm and T_cold averaged over the region of 80°E–145°E, 30°N–45°N during 1980–1999 and 2081–2100 from seven IPCC AR4 models including CCSM3. Although T_warm shows the largest changes and T_cold in 20C3M only shows values less than 1 in the CCSM3 result, the differences across models are not very large. All models including CCSM3 consistently show the increase in T_warm and the decrease in T_cold in the 21st century compared to the present day. Especially, the maximum and minimum changes are consistently modeled in all A2 and B1 scenarios. Therefore, we can conclude that the projected changes in T_warm and T_cold in CCSM3 in this study are reliable.

Fig. 13

T_warm indices (a) and T_cold indices (b) averaged over the region of 80°E–145°E, 30°N–45°N during 1970–1999 and 2081–2100 from various IPCC AR4 model runs. Black, light gray, dark gray, and white bars indicate the results from 20C3M, A2, A1B, and B1 scenarios

5 Summary and discussion

Given climate changes as represented by global (and/or regional) warming, the attendant changes in extremes have generated major concern in the scientific and public society. It is especially necessary for answering questions of public and decision makers related to the extreme vulnerability to know overall influence of various extremes as well as changes in individual extreme events. To satisfy these curiosities, we have developed new indices that integrate the influences of various existing extreme indices based on ETCCDMI. Our four indices are characterized by warm (T_warm), cold (T_cold), wet (P_wet), and dry (P_dry) extremes. The changes in these four indices have been examined using meteorological station observations for the period from the 1960s to the 2000s and model simulations using CCSM3 for 1900–2099.

In the observed data, three of the four extreme indices (all except P_wet) show significant changes in northern regions of East Asia during the recent five decades. The temporal variations characterized by increase in T_warm and P_dry and decrease in T_cold may be considered as consistent warming signals due to an anthropogenic climate changes. In particular, the results, which show the change in climatological average values of T_warm is large in the southern region but the change in anomaly is large in the northern region, are line with previous studies concluding that the response to global warming is more obvious in the extratropics rather than the tropics (Dai et al. 2001). Correspondingly, T_cold decreases over the northern region and P_dry increases, because over the northern inland region the available moisture is not sufficient despite the increase of temperature. The spatial variation of P_wet in the East Asia region presents a very sparse pattern and the temporal variation shows an insignificant pattern. However, in the area of Korea, observed P_wet has steadily increased during the past 40 years, consistent with previous studies that have examined the large increase of heavy rainfall over the same region in recent years.

In the model simulation using CCSM3 for 1900–2099, changes consistent with those in the observations are detected in the two temperature extremes (T_warm and T_cold) of the 20th century simulation, and consistent changes are continuously shown in the projection for the 21st century. In the additional analysis of six IPCC AR4 models, the changes in T_warm and T_cold are projected in similar ranges to the CCSM3. From the results of both the historical and the model analyses, we can conclude that planning for increases in the vulnerability of the northern East Asia to the warm and dry extremes and in that of the Korea Peninsula to the wet extreme is strongly urged.

Indeed, many weather extreme events could be mitigated reasonably well by planning responses to the increased vulnerability of the East Asia region. In the absence of planning and proper policy, even a moderate weather event can become a disaster. It is thus important to assess the impact of extremes and establish the proper policy for minimizing the damage from disaster. By using our combined indices, people can easily understand the integrated impact of various extremes on their residential area, not taking care of the detail characteristics. In addition, these indices directly show how vulnerable the region is to extremes compared to other regions, since they are represented by the normalized value using the spatiotemporal mean of extremes over East Asia. Areas in the region shown to be vulnerable to greater dry extremes can for instance establish planning based on proven policies in currently dry areas, and so on. These indices can be helpful in preparing response to extreme weather by showing the vulnerability to the extremes with quantitative values.

Prior to applying these indices to public policy, we should note several limitations of the analysis. First, some indices obtained from ETCCDMI, especially the indices using a fixed threshold such as TR20 and SU25, may be insufficient to fully reflect the characteristics of extremes over East Asia. In addition, CDD is only one index for dry extreme in ETCCDMI. It is possible that we could confirm better-optimized indices for East Asia by applying this method on a sub-regional scale for East Asia. Second, to show the vulnerability more realistically, socio-economic factors such as population density and national economy should be considered as well. Finally, it must be noted that the conclusions of this study are limited to extreme weather events based on the indices represented by absolute variations in temperatures and precipitation. It is, however, dependent upon the perception of people affected by an event whether the event is considered as an extreme or not. For example, although the minimum temperature has increased in recent years, cold events suddenly occurring after consecutive warm days can be still considered as an extreme event by some people. Therefore, in further study, it is necessary to supplement our indices by considering even such a relative concept of the extremes as personal perception, to assess more realistically the extreme vulnerability.