Large increase in heavy rainfall associated with tropical cyclone landfalls in Korea after the late 1970s



[1] This study presents a new demonstration of the abrupt increase in the heavy rainfall events (≥100 mm day−1) during August–September in Korea around the late 1970s. The accumulated heavy rainfall averaged for the two months over 12 stations was 57 mm during 1954–77 (ID1); however, it changed to 103 mm during 1978–2005 (ID2). This change is found to be associated with landfalling tropical cyclones (TCs). The most plausible mechanism that accounts for the TC–heavy rainfall relationship is an enhanced TC–upper-tropospheric trough (UTT) interaction, which results from a southward shift of the upper-tropospheric jet in East Asia during ID2. While the intensity and duration of the landfalling TCs in Korea does not appear to exhibit such an interdecadal change based on the data available, the enhanced TC–UTT interaction increases the upper(lower)-tropospheric divergence (convergence) and coherent ascending motion, which strengthen the frontal zone around Korea.

1. Introduction

[2] During the boreal summer, tropical disturbances continuously move poleward to the northern extratropics, inducing many severe weather events. As they meet the midlatitude westerlies, they change into extratropical cyclones in most cases. An important tropical disturbance is tropical cyclone (TC), which is characterized by intense cyclonic winds, well-organized deep convection, and spiral rain bands. The approach of a TC toward coastal areas results in severe social and economic damage over midlatitude regions.

[3] China, Taiwan, Japan, and Korea suffer severe damages associated with the several TCs that strike East Asia every year. Hence, there have been continuous efforts to investigate the climatic characteristics, variations in TC movement [e.g., Kim et al., 2005a, and references therein], and landfalling activity in East Asia [e.g., Wu et al., 2004; Kim et al., 2005b]. Since the flood damage caused by landfalling TCs is most critical, a number of case studies have been carried out to clarify the distribution of the TC rain band and the heavy rainfall area induced by the TC landfall (e.g., Shimazu [1998] and many others). From the climatic aspect, however, the heavy rainfall induced by the landfalling TCs has not been sufficiently explored so far.

[4] It is well established that the summer rainfall in East Asia has experienced a regime shift with respect to the late 1970s [e.g., Hu, 1997; Chang et al., 2000; Wu and Wang, 2002; Ho et al., 2005]. This interdecadal change is related to the large-scale atmospheric circulation features such as the southward shift of the summertime East Asian subtropical westerly jet (EASW) and the westward expansion of the North Pacific subtropical high (NPSH) [Chang et al., 2000; Gong and Ho, 2002]. Ho et al. [2005] pointed out that the interdecadal change in the rainfall exceeding 30 mm day−1 accounted for a significant part of the total rainfall variability in China. Coinciding with these changes, the summer rainfall has increased in the Yangtze River valley in recent decades. The region with enhanced rainfall extends toward the northeast and further to the Korean peninsula. In Korea, the summer rainfall also showed an abrupt increase in the late 1970s [Ho et al., 2003]. This regime shift is prominent in August, and is mainly explained by the changes in the frequency and intensity of heavy rainfall events.

[5] With regard to such interdecadal variability in the summer rainfall in Korea, the change in the TCs that struck the country should be considered. This is because a landfalling TC is often accompanied by torrential rain and flooding. Climatologically, about 60% of the landfalling TCs in Korea appear in the two months of August and September [Lee et al., 1992]. Therefore, the present analysis only focuses on these two months. The objective of this study is to identify and explain the interdecadal change in the heavy rain in relation to the TC landfall in Korea. Relevant large-scale changes in the heavy rainfall are analyzed to search for possible mechanisms responsible for these changes.

2. Data and Definitions of TC Landfall and Heavy Rainfall

[6] This study has primarily utilized total daily rainfall in Korea and six-hourly TC archives for the western North Pacific (WNP). The daily rainfall data was obtained from 12 weather stations across the Korean peninsula including two islands. The data exhibited a reasonably uniform distribution across the country, and each of the stations recorded relatively long observations (see dots in Figure 1). Although the beginning year of the data obtained from the stations varied from the 1900s to the 1940s, the data was available up to the present, except that for 1950–1953 which corresponded to the Korean War. The TC information for the WNP was obtained from the archives (dating back to 1951) of the Tokyo-Typhoon Center. In the present study, a TC refers to a tropical storm and typhoon when its 10-min sustained wind speed is greater than 17 m s−1 and 33 m s−1, respectively. The temperature, specific humidity, horizontal wind, and vertical velocity data obtained from the European Center for Medium-range Weather Forecast (ECMWF) reanalysis were examined. The analysis period for the rainfall and TC was restricted to 1954–2005 in order to circumvent the period for which the Korean rainfall data was unavailable; on the other hand, the analysis using the ECMWF reanalysis included all the years for which the data was available, i.e., 1958–2002.

Figure 1.

Tropical cyclone tracks for (a) 1954–1977 (ID1) and (b) 1978–2005 (ID2) that made landfall on Korea. Dots in the large map indicate the TC genesis positions and those in the embedded circle are the 12 weather stations in Korea.

[7] Normally, a “TC landfall” occurs when the TC center encounters a coastal line on the surface weather chart. The number of TC landfalls in Korea during 1954–2005, as counted by this method, is 46. However, many TCs glancing off the Korean peninsula have provoked torrential rainfall nationwide because their effective radii sufficiently covered the entire territory. Thus, a TC landfall is defined here as an event where the TC center enters a circle with a radius of 5° from central Korea (i.e., 128°E and 36°N) (Figure 1). The circle encompasses all territories including adjacent oceans (parts of the Yellow Sea and the East Sea). The total number of TCs for the two months during 1954–2005 is 90, which is nearly twice as many compared to the number that encountered the coastal line. All the TC tracks that made landfall on Korea are separately depicted for the two periods: 1954–1977 (ID1, Figure 1a) and 1978–2005 (ID2, Figure 1b). The breaking point was set to 1977–78, which is in line with findings by Ho et al. [2003]. However, the selection of the reference year is still a priori, so the authors apply the change-point analysis [Ho et al., 2004] to the time series of total and TC-induced heavy rainfall shown in Figure 2, respectively. The results give the largest t-values, 2.95 for the former and 3.62 for the latter when the year 1977–78 is set to the change-point, which are significant at the 99% confidence level using the two-sided Student's t-test. Namely, the change-point analysis justifies our selection of the reference year. The number of landfalling TCs in ID1 and ID2 are 40 (on average 1.7 yr−1) and 50 (on average 1.8 yr−1), respectively, indicating a negligible increase in recent decades. A heavy rainfall event is defined as an event where at least one station out of 12 stations exceeding 100 mm during one day. The sensitivity of this threshold value was studied by changing it from 70 mm day−1 to 150 mm day−1. It was found that these changes do not affect the conclusion of the present study (figure not shown).

Figure 2.

(a) Time series of the accumulated rainfall averaged over the 12 stations in Korea during heavy rainfall events for August–September. Filled bar denotes the accumulated heavy rainfall influenced only by the landfalling TCs. (b) Precipitation intensity at stations that have a rate equal to or greater than 100 mm day−1 and are influenced by the landfalling TCs. Note that a log scale is used for the y-axis.

3. Results and Discussion

3.1. Interdecadal Change in Heavy Rainfall Associated With TC Landfalls

[8] Figure 2a presents the temporal variations in the total heavy rainfall (open plus filled bar) and the partial contribution by the TC landfalls (filled bar) averaged across the country for the two analyzed months. Due to the definition of the heavy rainfall event, the magnitude of the all-station mean can be below 100 mm day−1. As seen in Figure 2a, there is an obvious interdecadal change in the total heavy rainfall with respect to the late 1970s: 57 mm 2-month−1 in ID1 vs. 103 mm 2-month−1 in ID2. The difference is statistically significant at the 99% confidence level using the two-sided Student's t-test. The number of years having ≥100 mm 2-month−1 station−1 is 2 in ID1 vs. 13 in ID2. A similar feature is also observed in the partial contribution by the TC landfalls: none in ID1 vs. 7 in ID2.

[9] How many heavy rainfall events have increased in association with the TC landfalls in recent decades? In Figure 2b, heavy rainfall events induced by landfalling TCs are plotted if any station among the 12 has a rainfall rate of over 100 mm day−1. There are 127 cases in the analysis period, i.e., an average of 1.4 cases for each TC landfall. The rainfalls recorded at the stations also indicate an apparent regime shift in the late 1970s. The number of events having a rainfall rate of over 200 mm day−1 has greatly increased from 2 in ID1 to 23 in ID2. A similar increase was observed in the heavy rainfall events with rates between 100 mm day−1 and 200 mm day−1: 26 events in ID1 vs. 78 events in ID2. It is interesting to note that the time series of heavy rainfalls independent of the TC landfalls (corresponding to the unfilled portion of the bar in Figure 2a) do not show any distinct change with respect to the late 1970s. At the same time, the number of heavy rainfall events independent of the TC landfalls does not show any significant change either (figure not shown). These results suggest that more frequent and stronger heavy rainfall events in recent decades are explained by the events induced by the landfalling TCs.

[10] More heavy rainfall events associated with the TC landfall can result from intensified TCs and/or increased number of landfalling TCs. Some recent literatures have demonstrated that TCs tend to sustain their intensity over the midlatitude oceanic basin despite the continuous sea surface temperature warming [Emanuel, 2005]. To find out whether there are any changes in the statistics of TCs when landfalls occur in Korea, the minimum central pressure of all the landfalling TCs every six hours and the time duration within the defined circle are plotted in Figure 3. Overall, the lowest pressure value and the number of six-hourly observations do not show any significant interdecadal change with respect to the late 1970s (Figure 3a). These patterns are consistent for heavy rainfall cases (≥100 mm day−1, filled circles) as well as for lesser heavy rainfall (<100 mm day−1, open circles). The time duration does not show such a change either (Figure 3b). In fact, it exhibits decadal fluctuation along with a strong interannual variation. All these results show that the statistical characteristics represented in the TC best track data do not show any remarkable regime shift with respect to the late 1970s.

Figure 3.

(a) The six-hourly minimum central pressure of the landfalling TCs and (b) their time duration with an influence on Korea. Cases producing heavy rainfall (≥100 mm day−1) are denoted by filled circles.

[11] Intuitively, the prominent regime shift in the heavy rainfall may be considered to be a consequence of the enhanced TC activity around the Korean peninsula, which cannot be confirmed by the present results. However, it would be too premature to conclude that the TC activity did not change significantly around the late 1970s. This is because the central pressure measurements of the TC best track data before the 1980s may have considerable observational errors owing to the lack of reliable observations.

3.2. Related Large-Scale Atmospheric Circulation

[12] How can the increase in the heavy rainfall around the late 1970s be explained? Was it mainly caused by the increased convective instability that resulted from the changes in large-scale climate? This does not seem to be the case because the heavy rainfall events that were not associated with the TC landfall do not show any significant change. The most distinctive interdecadal change in the summer-mean large-scale atmospheric circulations in East Asia is reflected in the southward shift of the summertime EASW and the westward expansion of the NPSH [Gong and Ho, 2002; Ho et al., 2005]. Given these changes in the mean fields, it is necessary to show the large-scale feature of the interdecadal change around the date of TC landfall. To demonstrate this, the composite differences between ID1 and ID2 in terms of the 200-hPa horizontal wind, 850-hPa moisture flux, their divergence fields, and the 500-hPa vertical velocity on day −2, −1, 0, and +1 with respect to the date of TC landfall are displayed (Figure 4).

Figure 4.

The difference (ID2 minus ID1) of (a–d) horizontal wind (arrows, unit: m s−1) and its divergence (shaded, unit: 10−6 s−1) at 200 hPa and (e–h) vertical velocity (contour, unit: 10−2 Pa s−1) at 500 hPa, moisture flux (arrows, unit: m s−1) and its divergence (shaded, unit: 10−8 s−1) at 850 hPa. The statistically significant differences (at the 95% confidence level using the two-sided Student's t-test) in horizontal wind and moisture flux are plotted with thick black arrows. TC symbol indicates the mean location of TC center.

[13] It is identified that the upper-tropospheric outflow jet and the divergence (Figures 4a–4d) over the downstream side of the westerly trough become stronger north of the mean TC center in ID2 when the TCs approach Korea. The difference between the horizontal wind velocities of ID1 and ID2 is more than 10 m s−1, which appears to be very large since the average velocity of the summer upper-tropospheric jet is about 25 m s−1 in East Asia. The 500-hPa vertical velocity (Figures 4e–4h; dotted contour indicates ascending motion) shows an anomalous ascending motion below the anomalous divergence region at 200 hPa. In particular, the Korean peninsula is always located in the ascending area. An apparent convergence of the lower-tropospheric moisture flux around the date of TC landfall, up to 7 × 10−8 s−1 on day 0 (Figures 4e–4h), is also shown. The convergence is caused by the southerly moisture flux from the mean TC center, which is considered to be a compensatory lower-tropospheric signal, to the upper-tropospheric outflow jet. All these consistent changes in the dynamic fields support the abrupt increase in the heavy rainfall in Korea.

[14] It is also important to analyze front development (i.e., frontogenesis) since it is closely related to the rain intensity. In this study, an increased horizontal potential temperature (θ) gradient is regarded as an indication of front intensification. If there is an increased horizontal θ gradient around the Korean peninsula, it is thought to contribute to the increased heavy rainfall. The difference in the magnitude of the horizontal θ gradient at 1000 hPa between ID1 and ID2 on day 0 shows positive values from the East China Sea to the East Sea through the southern region of the Korean peninsula (figure not shown). The southwest-northeast-oriented increased region of the horizontal θ gradient coincides with the entrance region of the outflow jet where the ascending motion is located. This indicates that the development of frontal features becomes more active in ID2 as compared with that in ID1.

3.3. TC-Upper Tropospheric Trough Interaction

[15] The interaction between the TC and the approaching upper-tropospheric trough (UTT) has been a matter of serious concern for understanding the TC intensity change and induced rain band structure [Holland and Merrill, 1984; Rodgers et al., 1991; Shi et al., 1997; Zhu et al., 2000]. Holland and Merrill [1984] pointed out that an outflow jet formed by the interaction between the TC and the approaching UTT directly affects the intensity change. Rodgers et al. [1991] and Shi et al. [1997] studied the intensification of Hurricane Florence. They also suggested that this intensification was caused by the approach of the UTT from the west. The westerly jet might have provided a channel for the formation of the outflow jet of Florence and initiated the development of the second convective cell north of Florence's center, thereby resulting in enhanced precipitation there. Further, Zhu et al. [2000] showed that the rainfall in front of the UTT would increase dramatically if a TC approaches south of it. A low-level southeast jet flow in the right-hand sector of the TC would transport an abundance of moisture to the rainy region in front of the UTT.

[16] In light of the previous studies, the abrupt increase in the heavy rainfall in Korea could have been caused by the enhanced TC–UTT interaction. Given the southward displacement of the summertime EASW in recent decades, the recurving TCs approaching East Asia would have more opportunities to interact with the EASW as compared to earlier decades. The considerably stronger outflow jet north of the TC center in ID2 may reflect the stronger TC–UTT interaction (Figures 4a–4d). In other words, the development of the outflow jet helps to increase the divergence in the outflow layer and initiates the ascending motion and second convection induced by the circum-jet secondary circulation at the entrance region of the outflow jet [Rodgers et al., 1991; Shi et al., 1997]. The corresponding ascending motion is clearly evident in the composite difference of the 500-hPa vertical velocity at the entrance region of the outflow jet (Figures 4e–4h).

4. Concluding Remarks

[17] It was suggested that the TC–UTT interaction is the primary cause of the interdecadal change in heavy rainfall in Korea in the late 1970s. The dominant features of the interaction are the strong poleward outflow into the divergent (convergent) region ahead of an approaching westerly trough in the upper(lower)-troposphere. The ascending motion in ID2 becomes stronger as compared to that in ID1 in the southern region of the Korean peninsula, where the entrance region of the outflow jet is located. Thus, the increased heavy rainfall in Korea is mainly caused by the enhanced TC–UTT interaction in ID2. Although the increased TC–UTT interaction in ID2 is strongly supported by the current observational study, it is necessary to prove this mechanism through a high-resolution numerical model that resolves the mesoscale phenomena realistically. It would be highly interesting to see whether a model simulation with the background state of ID2, in comparison to that with a background state of ID1, will lead to a stronger TC–UTT interaction.

[18] The TC–UTT interaction has been commonly considered as a factor of TC intensification in many previous literatures [Rodgers et al., 1991; Shi et al., 1997]. However, it was found in this study that there is no evidence of interdecadal change in the TC intensification when the TC approaches the Korean peninsula. This result is supported by a study of Merrill [1988], who argued that the TC intensification caused by the TC–UTT interaction is not apparent because of the increased westerly shear. This issue cannot be resolved in this study because of the issue of data reliability. Therefore, whether the TC characteristics have changed or not remains an open question.


[19] This work was funded by the Korea Meteorological Administration Research and Development Program under grant CATER 2006–4204. J.-H. Kim and M.-H. Lee are also supported by the BK21 project of the Korean government. The authors thank J. E. Saiers and two anonymous reviewers for valuable comments.