Climate change effects on tropical night days in Seoul, Korea

Abstract

In Seoul (37.57°N, 126.97°E), South Korea (located at mid-latitudes), the frequency of tropical night (TN) days, which have been defined as days with a minimum temperature greater than 25°C, have shown an increase due to the effects of temperature and water vapor. It was found that TN days accounted for almost 10.2% (July) and 22.1% (August) of the total number of days in respective months during the last decade in Seoul, and these figures may be increasing with climatic change. The daytime and nighttime sky on TN days can contain water vapor when the monsoonal southwesterly flow prevails. This strong wind may induce moisture advection from the warm ocean, and consequently, there is much larger specific humidity over the city during TN days in comparison to non-TN days. The effect of climatic change on the specific humidity is related to an increase in the number of TN days, which has shown an upward trend of 13-day/100-year and is significantly modulated by both water vapor and air temperature during July and August. Moreover, the relative role of water vapor in increasing the frequency of TN days has become much more significant after the 1960s in comparison to that of air temperature, which may be attributed to urbanization in Seoul since the 1960s.

1 Introduction

The daily minimum (or night) temperature plays an important role in agricultural yields (Willits and Peet 1998; Peng et al. 2004). Interest in climatic change has increased due to rising surface temperatures, which are caused by global warming and anthropogenic activities. The daily minimum temperature is a good indicator of global warming. Moreover, urbanization can induce nighttime warming in big cities, particularly mega-cities in which more than 10 million people face heat island phenomena and warm nighttime air temperatures, which are related to the capture of heat between tall buildings (Kim and Baik 2004). During the past half century, daily minimum temperatures have shown a rapid increase relative to maximum and mean temperature (Karl et al. 1991; Easterling et al. 1997). This effect has resulted in a remarkable decline in the diurnal temperature range, although the decreased range between the maximum and minimum temperature is insignificant after the late 1970s (Vose et al. 2005).

Since nighttime temperatures during the summer affect sleep comfort, this climate variable can be considered a key factor associated with global warming and urbanization. For example, tropical night (hereafter, TN) has generally been defined as those nights in which the daily minimum temperature exceeds a critical value (Alexander et al. 2006; Trenberth et al. 2007). In Korea, TN occurs when the daily minimum and maximum temperatures exceed 25°C and 30°C, respectively (e.g., Heo and Kwon 2007). The Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC AR4) reported that the increase in TN day frequency is consistent with the increases in greenhouse gas concentrations (Trenberth et al. 2007). The changes in extreme climate events such as TN, heat waves, and heavy precipitation are much more conclusive than those obtained from seasonal or monthly mean states.

In global analysis, many extreme indices have been developed by the World Meteorological Organization Commission for Climatology/CLIVAR Expert Team on Climate Change Detection Monitoring and Indices. Alexander et al. (2006) also used these extreme indices and defined a TN as those days whereby the daily minimum temperature is greater than 20°C. Despite numerous studies on the extreme climate change in global analysis, some problems remain unsolved on a regional-scale, for example, problems to pertaining to TN days. The main question is how global warming and urbanization affect the frequency of TN days. The present study aims to investigate the change in TN days with respect to Seoul, South Korea, which is located at mid-latitudes. Using the daily minimum summer temperatures at Seoul, a megacity with more than 13 million people, temperature thresholds are initially investigated through frequency analysis. The change in TN days and their characteristics are investigated on the basis of the definition of this phenomenon. Moreover, the change in the contributing factors such as moisture and temperature is examined using a variety of statistical methodologies.

2 Data and method

2.1 Data

Seoul, which is located at about 37.5°N and 127.0°E, has a regional network of weather stations (Ha et al. 2009). This network contains one synoptic station and 28 automatic weather stations (Ha et al. 2007) within Seoul city limit. Figure 1 shows the topography in the Seoul region and the distribution of the regional network for these stations. The thick solid line indicates the boundary of Seoul. The Seoul network data include hourly records from 1999 to 2008; data from all the stations are available for this period.

Fig. 1

Topography of the Seoul region and the distribution of the Seoul regional network (28 points) synoptic station (white circles) and automatic weather stations (black circles). The contour indicates the 10-year averaged daily mean temperature during July–August 1999–2008, and the thick solid line indicates the city limits of Seoul

To detect the long-term variability of TN days in Seoul, synoptic station datasets obtained from the Korean Meteorological Administration from 1908 to 2008 were used. For the purpose of evaluating the change in TN days in Seoul, a set of data from the two surrounding stations Incheon (37.47°N, 126.63°E) and Suwon (37.27°N, 126.99°E) were used. Note that, for Suwon, which is a rural city south of Seoul, data are available from 1964 (for 45 years, between 1964 and 2008).

The specific humidity (hereafter, SH) and daily mean temperature (hereafter, Tmean) are considered as important factors contributing to TN days. Although the minimum temperature (Tmin) is an important component of Tmean, the nighttime temperature is sensitive to SH changes. This characteristic of Tmin can induce different Tmin and Tmean. For example, during July–August from 1979 to 2008, the maximum temperatures are relatively independent of SH (r ~ 0.31), while Tmin is significantly modulated by SH (r ~ 0.75). This indicates that Tmin is associated with SH in addition to Tmean. Composite analyses for TN days were performed to find the relationship of SH and Tmean. The analysis involved the use of the National Center for Environmental Prediction/National Centers for Atmospheric Research (NCEP/NCAR) reanalysis data recorded between 1979 and 2008 (Kalnay et al. 1996).

2.2 Method

The daily minimum temperature can provide information on heat island events in a mega city. To estimate the occurrence percentile of TN days, the frequency diagram of the daily minimum temperature and peaks over threshold methods were used. To understand the spatial characteristics in TN days, the horizontal regional aspects of the daily minimum temperature were investigated in Seoul, because of the availability of the regional network dataset. In addition, to determine the effects of climatic change on TN days, the changes in SH and Tmean (important factors contributing to TN days) are investigated.

To investigate the change associated with TN days, a variety of statistical time series analyses, including the composite, correlation coefficient, partial correlation coefficient, and two-way analysis of variance (ANOVA) tests were applied. Here, the partial correlation indicates the linear correlation between two variables after the relationship between these two variables and another third variable are excluded. For example, if a partial correlation is approximately zero, it is considered that the linkage between two variables is considerably affected by the third variable. The two-way ANOVA test is a useful tool for measuring the main effects of both SH and Tmean, in addition to the interaction effect between the two factors (e.g., von Storch and Zwiers 1999). The statistical significance of the two-way ANOVA test is shown by three F ratios, one for each factor independently and one for measuring the interaction between the two factors. For this, the two factors (i.e., SH and Tmean) are simply divided by two treatments: a normalized value of greater than zero (i.e., high) and that of less than zero (i.e., low). The four groups for the ANOVA test consisted of high and low temperatures and high and low moistures.

3 Characteristics of daily minimum temperature and TN days

Seoul faces urban warming events associated with heat island phenomena; the phenomena occur in the region within the circle, which corresponds to 25°C (Fig. 1). The circle was determined through 10-year averaged daily mean temperatures during July and August from 1999 to 2008. Figure 2 shows the frequency distribution of daily surface minimum air temperatures over the 28 stations in the Seoul regional network during the same period. In July and August, daily minimum air temperatures were observed to show symmetrical, near-normal distribution against the mean. For temperatures greater than 25°C, the frequency percentiles were 10.2% and 22.1% in July and August, respectively. Mean (standard deviation) temperatures for July and August were 21.79°C (1.71) and 22.38°C (2.00), respectively. Therefore, a threshold of 25°C corresponds to a TN, which is an uncomfortable sleep temperature characterized by sultry and sweltering conditions. For daytime and nighttime continuity, TN days are defined as days in which the daily minimum temperature is greater than 25°C. The threshold of 25°C is consistent with that considered by Heo and Kwon (2007). Here, the daily minimum temperature is measured during the 24-h period from 0900 LST of the previous day to 0900 LST of the day.

Fig. 2

Frequency distribution of daily surface minimum air temperature over the Seoul regional network (28 stations) for the 10 years from 1999 to 2008: a July and b August

In this study, the air temperature and SH are considered as important factors for TN days. Figure 3 displays the diurnal cycle of the air temperature and SH averaged for TN and non-TN days in the Seoul synoptic station during the previously mentioned 10-year period. Here, non-TN days were identified as those days with a Tmean greater than 25°C and Tmin less than 20°C. For nighttimes, larger differences were observed in the air temperature between TN and non-TN days, in comparison to those for daytimes. The SH for TN days was greater than that for non-TN days during daytime and nighttime. This figure suggests that the two parameters Tmean and SH are good indicators of TN days.

Fig. 3

Diurnal cycle averages of the tropical night (TN), defined as days with the daily minimum temperature greater than 25°C, and non-TN days in the Seoul synoptic station during 10 years from 1999 to 2008: a air temperature and b specific humidity

Because the difference in the air temperature between TN and non-TN days was mainly observed in the nighttime, the spatial distribution of the nighttime air temperature is important in determining the characteristics of TN days in Seoul. For comparison, Fig. 4 shows nighttime air temperatures for TN and non-TN days over the Seoul region during the 10-year study period. These plots were averaged from 0500 LST to 0700 LST for maximum occurrences of Tmin. The thick solid line indicates the boundary of Seoul, and the gray circle denotes the Seoul synoptic station. The low topography of the region located southwest of the city’s center leads to the maximum nighttime temperatures being observed in this area. This figure indicates the more significant increase in the nighttime temperature during the TN days compared with non-TN days.

Fig. 4

Nighttime air temperature over the Seoul region during the 10 years from 1999 to 2008: a TN days; b non-TN days. These plots are averaged from 5 LST to 7 LST. The thick solid line and gray circle indicate the city boundaries of Seoul and the Seoul synoptic station, respectively. The solid line indicates an interval of 1°C, while the dotted line indicates an interval of 0.5°C

Because the difference in SH between TN days and non-TN days is evident during daytime and nighttime, the possible water vapor and wind conditions during both periods are presented in Fig. 5. Nighttime is considered between 1900 LST and 2300 LST and the daytime between the 1200 LST and 1600 LST. The SH during the daytime for TN days is dominated by windy conditions, which indicates that the synoptic condition for wind is important in determining TN days. On TN days, during daytime and nighttime, the monsoonal southwesterly flow is persistent, and consequently, it may induce very large SH over the city through the TN days. On the other hand, on non-TN days, mountain and valley winds prevail over the city with relatively calm winds. The maximum SH is specific to the region east of Seoul near the ocean, rather than the region to the west. This result indicates that the moisture advection from the ocean is critical for providing water vapor during TN days.

Fig. 5

a, c The daytime (1200–1600 LST) specific humidity (grams per kilogram, contour) and wind (meters per second, vector) distributions for a TN days and c non-TN days. b, d Same as (a) and (c), except for nighttime (1900–2300 LST)

For the validation of synoptic station datasets in relation to the temperature and SH, a composite analysis of the factors contributing to TN days versus non-TN days was performed for the 30-year period from 1979 to 2008, by using the NCEP/NCAR reanalysis dataset. In conclusion, the SH and Tmean in the lower-level (i.e., 1,000 hPa) are significantly responsible for TN days (Fig. 6). On TN days, a considerable amount of moisture persisted over the Korean Peninsula, unlike non-TN days. A higher Tmean compared with that of the surroundings occurred around Korea. On non-TN days, cold air makes incursions into the Korean Peninsula. Consequently, marked moisture and temperature signals were observed over the East Asian region.

Fig. 6

Composite map of a–b daily special humidity and d–e air temperature at 1,000 hPa for a, d TN days and b, e non-TN days. c, f Composite difference between TN days and non-TN days. Shading indicates the value significant at the 95% confidence level

TN days are characterized by the high SH and high Tmean. To understand the monthly difference in relation to TN days in greater detail, composite analyses of the SH and Tmean were also performed for the strong and weak TN years (June to September in each year). For this, the strong (weak) TN years are defined as those having anomalies such that the standard deviation of normalized values is greater than 0.7 (less than −0.7). The seven chosen strong TN years were 1994, 1995, 1996, 1997, 2004, 2005, and 2007; the 11 weak TN years were 1979, 1980, 1982, 1986, 1987, 1989, 1991, 1992, 1993, 1998, and 2003. Because the most significant signals appeared during July and August, the composite differences between the strong and weak TN years were examined for these months (Fig. 7). The strong TN years revealed remarkable increases in the moisture and temperature in both July and August, which is in reasonable agreement with the results shown Fig. 6. This is particularly evident for August. Note that the TN phenomena in Korea mostly occurred after the summer rainy season during July which is often referred to as the Changma. Thus, TN days could be closely linked with the SH and Tmean during August.

Fig. 7

Composite difference of the a–b monthly special humidity and c–d air temperature at 1,000 hPa between strong and weak TN years during (a, c) July and (b, d) August. Shading indicates the value significant at the 95% confidence level

4 Long-term change

To detect the long-term climatic trend for TN days over Seoul, the long-term records from the Seoul synoptic station were analyzed. Figure 8 displays the interannual variability of the frequency of TN days. In Seoul, Incheon, and Suwon, TN days have evidently increased over recent decades. To compare the change in TN days at the three stations, a linear trend coefficient and the statistical significance from the Mann–Kendall test (Mann 1945; Kendall 1975) were calculated for the 45-year period from 1964 to 2008. While the TN days in Incheon and Suwon showed an increasing trend of 4-day/100-year and 9-day/100-year, respectively, those in Seoul exhibited an upward trend of 13-day/100-year. In particular, this upward trend in Seoul is found to be significant at the 95% confidence level (i.e., P < 0.05). The increase in Seoul appeared more distinctly than that in Incheon and Suwon, pointing to the urban effect in Seoul. In some epochs (e.g., for the periods 1964 to 1993 and 1994 to 2008), the linear trend is negative due to decadal fluctuations in the number of TN days. Note that the frequency of TN days shows a maximum value in 1994, which was the year of the hottest recorded summer in Korea (Heo et al. 1997). Kwon et al. (2007) reported a notable decadal change in the East Asian summer monsoon around 1993/1994. The number of TN days also shows decadal change after this period. For convenience, the period was divided into two parts: before and after 1993/1994. The mean and standard deviation for TN days are presented in Table 1 for the different time periods. TN days have increased by nearly three times during 1994–2008 compared with those during 1979–1993. The statistics also indicate that the frequency of TN days in Seoul have increased, with significant deviations.

Fig. 8

The year-to-year frequency (bar) of TN days in a Seoul, b Incheon, and c Suwon synoptic stations. The solid line indicates a 5-year moving averaged value. The dashed line denotes the linear trend for the frequency during the 45 years from 1964 to 2008

Table 1 The statistical mean and standard deviation of the TN days in Seoul

The significant increase in TN days is closely connected to the long-term change in Tmin, Tmean, and SH. Figure 9 shows the 5-year moving averaged anomalies and the linear trend coefficient for Tmin, SH, and Tmean for July and August. While the Tmin showed a significant upward trend, SH exhibited only decadal fluctuation without showing an increase or decrease. Tmean showed a slightly increasing trend during July and August. A greater increase in Tmin compared with Tmean is consistent with the global tendency of Tmin and Tmean, which is likely to contain contributions from the local effect of urbanization in Seoul and global warming (Easterling et al. 1997). Moreover, the long-term change in the temperature is remarkably different from that in the SH. It has been noted in several earlier studies that the relative roles of temperature and water vapor in the long-term climate vary (Um et al. 2007). In the present study, the relative roles of these parameters on TN days were investigated to understand the climate change in TN days as well.

Fig. 9

The 5-year moving averaged value (solid line) and the linear trend (dashed line) in monthly Tmin, SH, and Tmean during the 101 years from 1908 to 2008

5 The change in the relationship

If the relative humidity remains constant, the SH has a linear relationship with temperature. However, the relative humidity varies with time, and it changes the relationship between the two parameters. Although daily SH is closely correlated with daily Tmean, the relationship between temperature and water vapor on a longer time scale is somewhat vague (Gaffen et al. 1992). In the present study, the relationship between TN days, SH, and Tmean was assumed to have changed over recent decades. To confirm this theory, the sliding correlation coefficient between the daily Tmin, Tmean, and SH on TN days with a 21-year window was initially calculated (Fig. 10a). During TN days, the correlation between the daily Tmin and Tmean is roughly indicated by a 5% significant value during most of the interdecadal periods. In contrast, the relationship between the daily Tmin and SH revealed a rapid interdecadal change from the 1940s to the 1970s. This link showed a significant correlation coefficient only for the recent decadal periods (i.e., after the 1990s). This result implies that the increasing trend in TN days is sensitive to changes in the SH as well as Tmean.

Fig. 10

a Sliding correlation between the daily Tmin, Tmean (gray line), and SH (black line) on TN days, obtained with a 21-year window. The open circle indicates the correlation coefficient significant at the 95% confidence level with varying degree of freedom. b Sliding correlation between the number of TN days and monthly SH, with a 21-year window, during July (gray line) and August (black line). The dashed line denotes the value significant at the 95% confidence level

To manifest the notable effect of SH on the TN days, the sliding correlation coefficient between the number of TN days and monthly SH was also calculated with a 21-year window (Fig. 10b). Overall, the SH during July was more closely related to the TN days than those during August. In July, the TN days were closely linked to the SH, with relatively significant correlation, for all decadal periods. In August, however, the TN days were closely related to the SH since the 1960s, with a rapid decadal change. These findings correspond well with the work of Um et al. (2007), who reported that the influence of water vapor in the urban effect was significant in the 1960s and has dramatically increased after the mid-1970s. On the other hand, the relationship between the TN days and Tmean is found to show a positive correlation, with significance at the 95% confidence level, for both July and August (Fig. 11a, b). This indicates that the physical linkage of the TN days with SH, particularly during August, has noticeably changed since the 1960s, while that with Tmean has not varied over this epoch.

Fig. 11

Sliding correlation between the number of TN days and monthly Tmean with a 21-year window during a July and b August (black line). The gray line shows the partial correlation excluding the SH effect. The dashed line denotes the value significant at the 95% confidence level

The increasing role of the SH in determining the frequency of TN days is supported by the partial correlation between TN days and Tmean, excluding the SH effect (Fig. 11). In July, when the effect of SH was removed from the TN day-Tmean relationship, the correlation coefficient dropped below 0.3 during most interdecadal periods, except for the period of 1930s–1950s. Before the 1960s, the partial correlation did not reveal any significant difference for August, while after the 1970s, the correlation excluding the SH effect exhibited a significant difference from the plain correlation. This partial correlation result indicates that the relative role of the SH during the recent period has become greater than during that of the Tmean, although the change in the total amount of SH is insignificant.

To illuminate the effects of the two factors (i.e., Tmean and SH) on TN days, a two-way ANOVA test was performed. This test has been a useful statistical tool in many previous studies for potential predictability and climate variability (e.g., Shukla 1981). In this study, the method was applied to analyze the change in the main effects of the SH and Tmean and the effects of their interaction. The ANOVA test was individually performed for an analysis period of 40 years, with a time lag of 10 years (Fig. 12). The test of the daily Tmin for the daily Tmean and SH presents the increasing role of the main effect of Tmean and SH and the effect of their interaction. It should be noted that the effect of SH has been significantly increasing over the past 40 years, which is in reasonable agreement with the results shown in Fig. 12a. In July, the F ratios for the monthly SH and Tmean on TN days exceeded the 5% critical value for the effect of the interaction between the two factors as well as in both main effects. This relationship was relatively consistent for all decadal periods. In August, while the F ratios for the main effect of Tmean exceeded the 5% significant value for all periods, the F ratios for the main effect of SH and the effect of their interaction showed 10% significant values only after the period close to the 1970s. Although the F ratios were relatively small, the ANOVA result indicated that after the 1960s moisture played a primary role in modulating the frequency of TN days in the main effect and the interaction effect between the SH and Tmean.

Fig. 12

a F ratios of the ANOVA test for the daily Tmin with two factors of daily Tmean and SH on TN days, respectively, for a 40-year period with a 10-year moving window. The F ratios of the ANOVA test for the number of TN days with two factors of monthly Tmean and SH during b July and c August, respectively, for a 40-year period, with a 10-year moving window. The dashed horizontal lines indicate the values significant at the 90% and 95% confidence levels

6 Discussion and conclusion

There is increasing concern that regional climate temperature extremes may be changing in frequency and intensity due to global warming and urbanization. This study aimed to provide a regional perspective of climatic change in terms of Tmin during summers in Seoul. In particular, our main objective was to show the effects of water vapor and temperature on TN days by considering long-term records. For this, TN days were defined as days when the daily minimum temperature is greater than 25°C.

From the Seoul regional network data, TN days have a relatively greater SH during the daytime and nighttime, which may be induced by moisture advection from the warm ocean due to monsoonal low-level westerly flow. During non-TN days, mountain and valley winds prevail over the city with relatively calm winds. During the last decade in Seoul, TN days amounted to 10.2% and 22.1% in July and August, respectively. Long-term records showed a gradually increasing trend of 13-days/100-years, which results from the changes in the SH and Tmean. It should be noted that the influence of the SH is significant after the 1960s, particularly during August. In Korea, industrialization began in the 1960s, and a significant decrease in relative humidity occurred after the 1960s (Um et al. 2007). This increasing role of moisture content might be due to urbanization in Seoul. One of the most striking characteristics of urbanization is nighttime warming. As shown in Fig. 13, the difference between Tmean and Tmin abruptly decreased after the 1960s, which is in good agreement with the city’s urbanization. This change reveals a larger nighttime warming relative to Tmean after the 1960s. Note that the changing time after the 1960s is highly consistent with the period in which the SH significantly correlated to TN days. Although the SH does not increase after the 1960s, modifications on the land surface due to urbanization may induce larger warming at night than during the day. It seems that the surface water vapor acts to effectively trap more heat from the surface. This should be investigated in detail in future work.

Fig. 13

The 5-year moving averaged value of Tmean minus Tmin during July–August from 1908 to 2008

In Seoul, the increases in the daily Tmin and TN days can be contributed by global warming and the urbanization effect. Although the problem of to what degree the increase is caused by these parameters remains to be resolved, it can be concluded that higher SH and Tmean have important roles in increasing TN days. The unsolved problem for robust conclusion still remains. In particular, it is important to verify the relative contributions of the effects of the SH and Tmean on the increase in TN days to confirm this conclusion. This study on the extreme climate change in TN days can contribute to improving scientific understanding of observed changes in extreme climate conditions.