In late autumn and winter, climatologically active cyclogenesis in East Asia is observed in two marginal sea areas (Figure 1): over the Japan Sea and from the East China Sea to the south coast of Japan (e.g., Chen et al., 1991; Takano, 2002; Adachi and Kimura, 2007). Extratropical cyclones that occur in the northern and southern coastal areas of the Japanese Islands have been classified as Japan Sea and South Coast lows, respectively. Many previous studies have discussed cyclogenesis on the lee side of the central mountains of Japan (e.g., Sakakibara, 1983; Takayabu et al., 1996) and have considered the effects of latent heating on cyclogenesis around the Japanese Islands (e.g., Yamagishi et al., 1984; Chen and Dell'Osso, 1987; Kuwano-Yoshida and Asuma, 2008). Takayabu (1991) simulated cyclogenesis with coupling development of upper- and lower-level vortices and compared this with the observed cyclones around Japan. Mesocyclones over the Yellow Sea (e.g., Heo and Ha, 2008; Heo et al., 2009) and over the Japan Sea (e.g., Fu et al., 2004; Yanase et al., 2004; Guo et al., 2007) have also been investigated by data analysis and simulations. Previous studies have examined the impacts of ocean processes on these synoptic- and meso-scale cyclones (Xie et al., 2002; Yamamoto and Hirose, 2007).
Some of the cyclones in the East Asia marginal seas are organized into twin cyclones that travel eastward along sub-parallel paths. Pairs of cyclones that develop around the Japanese Islands occasionally cause heavy rainfall across a large area of Japan. Hitsuma (2006) reported a case that occurred on 6–7 November 2005 and Ogura et al. (2006) reported a case on 6–9 December 1998. Both studies focused on the twin cyclones travelling eastward around the Japanese Islands and reviewed the weather charts and cloud patterns separated by the two different lows. In these cases, the northern and southern lows of the twin cyclones are associated with the polar and subtropical upper jets leading to the two baroclinic zones, respectively. These previous studies emphasized that the presence of such double upper jets is important in the cyclogenesis of wintertime twin cyclones around Japan. In contrast to these previous investigations, the present study discusses the twin cyclone in the situation that a single upper jet is located around Japan.
Ziv and Alpert (1995) investigated the observational and theoretical features of binary interactions for 17 313 cyclone pairs (including extratropical cyclones), while Ziv and Alpert (2003) discussed binary interaction from the perspective of potential vorticity. Renfrew et al. (1997) examined binary interactions between polar lows over the Barent Sea. Although several cases of twin extratropical cyclones have been reported by them, complex binary dynamics of twin extratropical cyclones remain poorly investigated, in contrast to twin tropical cyclones (e.g., Brand, 1970; Chang, 1983).
The cyclogenesis and precipitation processes of twin extratropical cyclones, leading to agriculture damage, electrical outage and traffic disturbance, are socially important in weather forecasting. However, the significance of the dynamics of the twin extratropical cyclones has not yet been fully recognized. In particular, we are interested in the merger and re-cyclogenesis processes in the East Asian marginal sea area, where the coastlines and topography are complex. To discuss these issues, the present study focuses on the complex binary dynamics (including rapid merger and re-cyclogenesis processes) of twin cyclones around the Korean Peninsula and Japanese Islands on 9–10 October 2001, which had asymmetric structures between the northern and southern cyclones. Such a pair of asymmetric lows, with contrasting sizes and intensities, is also identified as a twin cyclone. Although the twin cyclones were not strong in this case, they led to heavy precipitation around the Pacific coast of Japan. This is an interesting case in that the weak twin cyclones caused heavy rain in a complex binary system. In general, weak cyclones over the Yellow and East China Seas are important as the initial stage of the cyclones which fully develop around Korea and Japan. However, the dynamics of the weak lows have not yet been fully examined. In the present study, the weak twin cyclones are investigated using reanalysis data of the Japan Meteorological Agency (JMA) and model outputs of the fifth-generation mesoscale model of the Pennsylvania State University–National Center for Atmospheric Research, ver. 3 (PSU/NCAR MM5V3). The analysis area and place names are shown in Figure 1.
2. Overview of twin cyclones on 9–10 October 2001
The present study used the JMA (Japan Meteorological Agency) weather charts and Regional-ANALysis (RANAL) dataset with a horizontal resolution of 20 km and 21 levels at 0000, 0600, 1200 and 1800 UTC. The JMA-RANAL data are produced by the JMA operational model, into which observational data (fixed land stations, ships, moored and drifting buoys, rawinsondes, pilot balloons, aircrafts, wind profilers, and satellites with polar and geostationary orbits) are assimilated (Fu et al., 2004). In addition, the JMA radar-AMeDAS (Automated Meteorological Data Acquisition System) precipitation data were used in this study. The dataset is the radar precipitation estimation assimilated with precipitation gauges (Makihara et al., 1996).
Figure 2 shows JMA weather charts of the twin cyclones on 9–10 October 2001. Two lows are found over the Yellow and East China Seas (YECS) at 0000 UTC 9 October, having first appeared at 1800 UTC 8 October (data not shown). The sea-level pressures of the northern and southern lows were the same (1008 hPa at 1800 UTC 8 October), and the locations of the two lows were similar those at 0000 UTC 9 October (Figure 2(a)). Hereafter, this pair of lows is termed the ‘YECS twin cyclones’. The northern low of the YECS twin cyclones stayed in the Yellow Sea area until 1200 UTC 9 October, whereas the southern low travelled northeastward through the Tsushima Strait (around 34°N, 130°E) during 0000–1200 UTC 9 October, before moving northward along the east coast of the Korean Peninsula after 1200 UTC 9 October. Finally, the developing southern low merged with the northern low at 1800 UTC 9 October (Figure 2(c)). Subsequently, a new weak low formed around the occlusion point in the southern coastal area of the Japanese Islands at 0000 UTC 10 October (Figure 2(d)). Hereafter, the pair of Japan Sea and South Coast lows is referred to as the ‘JSSC twin cyclones’.
According to JMA regional objective analysis data (JMA-RANAL; Figure 3), the troughs at the 300 hPa level were located west of the twin cyclones analyzed on weather charts, and only a single strong jet (velocity > 50 m s−1) was observed, in contrast to the cases reported by Hitsuma (2006) and Ogura et al. (2006), in which twin lows were associated with both the polar and subtropical upper-level jets. Since the southern lows in the present case were shallow, the minima in geopotential height were relatively weak at the 850 hPa level at around 32°N, 127°E (Figure 3(c)) and at around 35°N, 137°E (Figure 3(d)). Near the centres of the weak southern lows, specific humidity was greater than 10 g kg−1. The fully developed deep lows were located north of the upper-level jet, while the weak shallow lows were located south of the jet.
The vertical components of the relative vorticities at the centres of the YECS twin cyclones (Figure 4(a) and (b)) were considerably different between the northern and southern lows. High relative vorticity was observed along the trough for the northern low, but was confined to below the 850 hPa level for the southern low. The vorticity of the southern low increased and extended to the middle troposphere at 38.5°N at 0000 UTC 10 October (Figure 4(c)). Coupling of the southern lower-level and northern upper-level relative vorticities of the YECS twin cyclones led to the formation of a deep structure in the northern low of the JSSC twin cyclones. This process is similar to a previously reported merging of northern and southern lows over Canada (Juang and Ogura, 1990; Ogura and Juang, 1990) and around Japan (Takayabu, 1991), which led to the development of the merged cyclone after the coupling. Takayabu (1991) reported that seven cases of the coupling developments were observed around Korea and Japan in 1986. A 36 year climatological analysis (Adachi and Kimura, 2007) showed two different cyclone tracks over the YECS merged over northern Korea in winter. The previous analyses suggested that the merger process of the YECS twin cyclones sometimes occurs around Japan.
On the other hand, the relative vorticity of the newly generated southern low of the JSSC could not be separated from the high relative vorticity of the fully developed front (Figure 4(d)), as the southern low was weak and shallow. The formation of the southern shallow low in the vicinity of the front will be discussed in Section 3.3.
The northern low stayed over the Yellow Sea on 9 October, while the trough was near-vertical in Figure 4(a) (i.e., the low did not develop). In general, sea-level pressure is low over the sea around Japan in winter, compared with that over the land. Thus, it may be difficult for weak cyclones to travel eastward over the Korean Peninsula (with elevations of more than 1000 m) without any development mechanism. Instead, in this case, the Yellow Sea cyclone (the northern low of the YECS twin cyclones) developed and advanced eastward with the help of the merger with the East-China Sea cyclone (the southern low).
Figure 5 shows the 300 hPa potential vorticity (PV) and the 900 hPa potential temperature. The deep northern low of the YECS twin cyclones corresponds to high PV, while the shallow southern low of the YECS twin cyclones corresponds to the tip of the warm tongue (W in Figure 5). The warm southern low advanced northward around the cold northern low with the PV maximum, which slowly migrated south-southeastward around the warm southern low. This situation is the same as the mid-latitude contact binary system described by Ziv and Alpert (2003). Both the vorticity coupling (e.g., Ogura and Juang, 1990; Takayabu, 1991) and the contact binary rotation (Ziv and Alpert, 2003) of the upper- and lower-level cyclones occurred at the same time in the present case. Thus, the rapid merger of the YECS cyclones is caused by dynamical coupling of the southern lower-level vorticity with the northern upper-level vorticity (Figure 4) in the contact binary rotation (Figure 5).
Precipitation was observed around the central areas of the lows and along the fronts, as shown in weather charts (Figure 2). During the rapid change from YECS to JSSC twin cyclones, 12 h precipitation around the Japanese Islands exceeded 40 mm (Figure 6). The heaviest precipitation was recorded over the Tsushima Strait at 1200 UTC 9 October, and around the Shikoku Island, Kii Peninsula (for locations, Figure 1(c)), and the Japan Sea side of the Japanese Islands at 0000 UTC 10 October. Particularly heavy rainfall of 102 mm h−1 was locally observed at Owase, Japan (34.070°N, 136.193°E) at 2300 UTC 9 October. The dynamics of the precipitation will be discussed in Section 3.2.
3. Numerical experiments
3.1. Model setup
Numerical simulations were performed to investigate the influences of latent heating and topography on the successive dynamical processes involved in forming the YECS and JSSC twin cyclones. Both sets of twin cyclones are considered together, as the successive processes were continuous and cannot be separated. All experiments started from 0000 UTC 8 October 2001, using the nonhydrostatic mesoscale model PSU/NCAR MM5V3 (Dudhia, 1993; Dudhia et al., 2005) with 23 sigma layers from the surface to 50 hPa. The initial and boundary conditions are produced from JMA-RANAL data with a 20 km horizontal resolution and 6 h interval.
The model domain (shaded area in Figure 1(a)) has a horizontal resolution of 9 km (401 × 434 grid points) and a central latitude and longitude of 38°N and 133°E, respectively. The lower boundary employed terrain height data from the United States Geological Survey (USGS) and weekly optimum interpolation sea surface temperature (SST) data from NOAA (National Oceanic and Atmospheric Administration) (Reynolds and Smith, 1994) (Figure 1(a)). Land-surface temperature was calculated using a five-layer soil model based on a vertical diffusion equation, and the cloud-radiation option was chosen in the model (Grell et al., 1994). An explicit moisture scheme including simple ice (Dudhia, 1989), the Medium-Range Forecast Planetary Boundary Layer scheme (Hong and Pan, 1996), and the Grell cumulus parameterization (Grell, 1993) were applied to the simulation.
Since the cyclones were weak (∼1000 hPa) and their pressure depressions were small in this case, the model conditions were largely altered in the following sensitivity experiments (Table I). In a fake-dry experiment (Exp. FD), condensational heating was set to zero (Dudhia et al., 2005) to indicate the significance of condensational heating. The topography of the Korean Peninsula (34–39°N, 124–130°E) was flat (i.e., the elevation was set to zero) in Exp. FK, and the topography of the Japanese Islands (30–40°N, 130–145°E) was flat in Exp. FJ, in order to examine the effect of topography.
Table I. Experiment conditions
Latent heat release
Removed from 34 to 39°N, 124 to 130°E
Removed from 30 to 40°N, 130 to 145°E
3.2. Influence of condensational latent heating on cyclogenesis
The YECS and JSSC twin cyclones are well simulated in Case CL. Figure 7 shows the horizontal distributions of the simulated outgoing longwave radiation and satellite IR images. At 0000 UTC 10 October, the cloud areas are located north of the Korean Peninsula and over the Japanese Islands, and the dry-air intrusion by the northern deep cyclone is seen in the eastern coast of Korea (black area around 38°N, 131°E in Figure 7(b)). These cloud patterns agree well with the observed ones. As will be shown later, the sea-level pressure and precipitation patterns in Case CL (Figures 8 and 9) are also similar to the observations (Figures 2 and 6). Thus, the output data are used for the analysis of the twin cyclones in the present study.
In this subsection, the focus is on the dynamical effect of condensational latent heating on the successive cyclogenesis of the YECS and JSSC twin cyclones. Figure 8 shows the distributions of sea-level pressure and precipitable water on 9–10 October. In order to compare the total vapour amount between the experiments, the vertically integrated water vapour amount (i.e., precipitable water) is used rather than mixing ratios at lower levels, because the mixing ratio at a given level is strongly influenced by local topography and vertical flow. The YECS and JSSC twin cyclones form at the surface in the full-physics simulation (Exp. CL): however, they are not simulated in the experiment without condensational latent heating (Exp. FD). The weak lows over the Bohai and Yellow Seas in Exp. FD are significantly different from the YECS twin cyclones at 0000 UTC 9 October. The central pressure of 1012 hPa in Exp. FD is weaker than those of the YECS twin cyclones (1006 and 1008 hPa) in Exp. CL. The low with a central pressure of 1012 hPa is not developed in Exp. FD. In contrast to the JSSC twin cyclones in Exp. CL, a single cyclone is located over the Tsushima Strait at 0000 UTC 10 October in Exp. FD, with a weaker central pressure than those of the YECS twin cyclones (1000 and 1010 hPa) in Exp. CL.
Juang and Ogura (1990) reported that the merger of the two cyclones was insensitive to the latent heating over Canada on 25 April 1979. The effect of latent heating on the merger was not simulated by Takayabu (1991). In contrast, the present study finds the importance of the latent heating in the development and merger processes of the YECS twin cyclones, based on a sensitivity simulation (Case FD). However, it is difficult to separate dynamical effects of the latent heating and coupling development of the upper- and lower-level vortices quantitatively, because the pressure depressions are weak and their two processes interact.
In Exp. CL, the tongue-shaped humid area (hereafter referred to as the ‘moist tongue’) moved over the Japan Sea area with northward migration of the developing southern YECS low (Figure 8(c)), whereas in Exp. FD it followed a linear trend to the south of the Japanese Islands (Figure 8(f)). The linear humid region is located almost parallel to the upper-level jet in Exp. FD, but the moist tongue is located below the jet in Exp. CL. In Exp. FD, the humid air does not cross over the mountains of the Japanese Islands because the lows are not yet fully developed over the East China and Japan Seas. On the other hand, in Exp. CL, the merger of the YECS twin cyclones, of which the southern low brings the humid air to the Japan Sea, results in formation of the moist tongue across the Japanese Islands. In the tongue-shaped humid area, lower-level moisture is transported by the cyclonic flow.
Figure 9 shows the horizontal distributions of 12 h precipitation and surface latent heat flux. Precipitation of more than 10 mm (12 h)−1 is seen near the centres of the YECS twin cyclones at 0000 UTC 9 October 2001. As the developing cyclones migrate eastward (1200 UTC 9 October), precipitation exceeding 40 mm is located over the Korean Peninsula in Exp. CL, but is not simulated in Exp. FD. At 0000 UTC 10 October, crossing of the moist tongue (warm colours in Figure 8) over the Japanese Islands leads to high precipitation in the western part of the Japanese Islands and in the eastern coastal area of the Korean Peninsula, as is shown in Exp. CL (Figure 9). The areas of high precipitation at 40° and 34°N correspond to the locations of the JSSC twin cyclones at 0000 UTC 10 October. In contrast, precipitation in Exp. FD is weak and confined to around the Korean Peninsula. Evaporation (i.e., latent heat flux) in Exp. CL exceeds that in Exp. FD in the YECS areas at 1200 UTC 9 October and at 0000 UTC 10 October. Intensification of the cyclones by latent heating results in enhanced evaporation in YECS (latent heat supply from the seas in Figure 9(b) and (c)), thereby increasing the amount of water vapour (precipitable water in Figure 8(b) and (c)).
Figure 10 shows vertical profiles of 12 h mean vertical wind averaged over the areas of high precipitation around the centres of the cyclones in Exp. CL. The average vertical wind velocity observed in precipitation areas at 0000 UTC 9 October is 0.05–0.1 m s−1 (Figure 10(a) and (b)), while that observed at 0000 UTC 10 October exceeds 0.1 m s−1 (Figure 10(c) and (d)). The upward flows of the JSSC twin cyclones are twice as large as those of the YECS twin cyclones. The enhancement of vertical wind on 10 October is associated with high precipitation located in the central and eastern areas of the cyclones. In Exp. FD, in which the vertical wind cannot be enhanced by latent heat release, the upward flow is weak (0–0.05 m s−1) and thus the cyclone is not fully developed. Enhanced precipitation and upward flow due to condensational heating are important factors in the formation of the twin cyclones in these areas.
Figure 11 shows the horizontal distributions of 300-hPa jets and 500-hPa upward flows. As seen in the JMA-RANAL data, the upper-level jets with speeds more than 50 m s−1 located southeast of the upper cyclone are similar in Exp. CL and Exp. FD, though the locations of the jets differ between the two experiments. The upward wind is located east of the centre of the sea-level low pressure (Figure 8), as seen in the case of a typical baroclinic wave (e.g., Holton, 2004). However, the upward wind speeds in Exp. CL are markedly different from those in Exp. FD. The 500-hPa upward flows are enhanced by condensational heating in coastal areas around 40°N and in the area of intersection between the upper-level jet stream (Figure 11) and moist tongue (Figure 8). North of the baroclinic zone below the upper-level jet, condensational heating (precipitation) contributes to the strong pressure depression of the northern low (e.g., Yamagishi et al., 1984). South of the intersection area between the upper-level jet stream (Figure 11) and tongue-like humid air (Figure 8), upward flow is enhanced by latent heat release, and the southern lows of the JSSC twin cyclones are formed in this region. However, the pressure depression of the southern low is much weaker than that of the northern low.
At 0000 UTC 10 October, the warm and humid air (potential temperature of > 294 K in Figure 5(c); precipitable water of > 50 mm in Figure 8(c)) is transported by the strong lower-level jet (Figure 12(a)), which intersects the upper-level jet near 40°N, 134°E. Accordingly, the maximum equivalent potential temperature of ∼325 K is located around the lower-jet core of more than 20 m s−1 at the 800–900 hPa level (Figure 12(b)). Upward flows in the region between the upper and lower jet cores are observed near conditional unstable or neutral areas where the contours of equivalent potential temperature are near-vertical (Figure 12(b)). The rapid merger of the YECS cyclones leads to the intersection of the upper-level jet and moist tongue (lower-level jet), which induces a coupling of the upper- and lower-level jets via upward flow (e.g., Shapiro, 1982; Keyser, 1999) over the Japan Sea. Such an enhancement of the upward flow is not simulated in Exp. FD (Figure 12(c) and (d)), though a strong upper-level jet of more than 60 m s−1 is formed. In addition to condensation heating, the presence of the lower-level jet is also important in the enhancement of upward flow.
3.3. Influence of topography on cyclogenesis
Figure 13 shows the horizontal distributions of sea-level pressure and precipitable water for the twin cyclones in Exp. FJ (without the topography of the Japanese Islands) and in Exp. FK (without the topography of the Korean Peninsula), respectively. The results of Exp. FJ and Exp. FK are similar to those of the control run (Exp. CL), in contrast to Exp. FD. However, the topography influences the merging of the YECS twin cyclones and the formation of the JSSC twin cyclones. The southern low along the east coast of the Korean Peninsula is not seen at 1200 UTC 9 October in Exp. FK, because the YECS twin cyclones have already merged in Exp. FK over the Korean Peninsula (over the Japan Sea at 1800 UTC in Exp. CL). The location and intensity of the Japan Sea lows are somewhat different from those in the control run, as the centres of the cyclones in Exp. FJ and Exp. FK are shifted westward relative to those in the control run at 0000 UTC 10 October.
Exp. FJ and Exp. FK differ from Exp. CL in terms of 12 h precipitation at 0000 UTC 10 October (Figure 14). The decrease in precipitation due to removing the Japanese topography (Exp. FJ—Exp. CL) is located along the south coast of the Japanese Islands. Around 40°N, 129°E, precipitation in Exp. FJ is 60 mm more than that in Exp. CL. This difference around northern Korea is more pronounced than that around the Japanese Islands, even though the topography of the Japanese Islands is removed. The topographic influence of the Japanese Islands extends over a wider area than the area of influence of the Korean Peninsula. Thus, the topography of the Japanese Islands has a strong influence on precipitation around Japan and northern Korea. In Exp. FK, the area with more than 60 mm difference in precipitation is limited to a northern area of the Korean Peninsula, associated with vertical motions related to mountain waves along the coast.
Figure 15 shows vertical profiles of the 12 h mean vertical flow and horizontal moisture flux in the two central areas of the JSSC twin cyclones. Here, the moisture flux is defined as the horizontal flux of the water vapour mixing ratio. In these areas, although the vertical flows in Exp. FK are almost the same as those in Exp. CL, the differences with Exp. FJ are significant. The vertical flow of the Japan Sea cyclone in Exp. FJ (Figure 15(a)) is enhanced by removing the topography of the Japanese Islands, whereas that of the South Coast cyclone (Figure 15(d)) is weakened. The absence of the topography results in a slight strengthening of the moisture flux around the Japan Sea low (Exp. FJ > Exp. CL in Figure 15(b)) and a weakening around the South Coast low (Exp. CL > Exp. FJ in Figure 15(e)).
In the area of 40–42°N, 128–132°E around the Japan Sea low, large moisture convergences (negative) are observed below the 800 hPa level (Figure 15(c)). Such convergence near the surface has the potential to induce condensation. The convergence in Exp. FJ reaches the 600 hPa level, while that in Exp. CL is limited to the near-surface region. Enhanced upward flow above the 900 hPa level in Exp. FJ is possibly caused by the convergence of large moisture flux at 600–900 hPa. The absence of the topography of the Japanese Islands results in intensification of the upward flow around the Japan Sea cyclone, because the moisture convergence that results in condensation extends to the lower part of the enhanced upward flow. Because the topography of the Japanese Islands results in reduced moisture transport across the mountainous parts of the islands, precipitation and upward flow in Exp. CL are weaker than those in Exp. FJ at around 41°N, 128°E.
In the area of 33–35°N, 134–138°E around the South Coast low, the terrain height differs between Exp. CL and Exp. FJ. The topography has a direct influence on convergence of the horizontal wind, as also observed for moisture convergence in the lower troposphere (Figure 15(f)). The convergences of wind and moisture in Exp. CL around the South Coast low are stronger than those in Exp. FJ below the 700 hPa level. Thus, the topography of the Japanese Islands results in enhanced convergence of wind and moisture, generating strong upward wind in Exp. CL (Figure 15(d)). The topographically forced convergence of horizontal flow induces upward flow, and convergence of moisture flux may contribute to the enhanced upward motion via latent heat release. Although the effects of wind and moisture convergence cannot be separated, the presence of topography leads to intensification of the upward flow via topographic convergence around the South Coast cyclone.
Figure 16 shows the distributions of sea-level pressure in Exp. CL and Exp. FJ. At 1800 UTC 9 October, a minimum of sea-level pressure is located in the southern coastal sea areas of the Shikoku Island (33.5°N) in Exp. CL, whereas it is located in the northern coastal sea area (34.5°N) in Exp. FJ. After 6 h, a weak pressure depression is found around the Kii Peninsula in Exp. CL. The low pressure is maintained along the southern coast of the Japanese Islands at 0600 UTC 10 October. In contrast, the minimum is not seen in Exp. FJ at 0000 UTC 10 October. These differences with the control run are caused by the topography of the Japanese Islands.
In Exp. CL at 0000 UTC 10 October (Figure 16(b)), a weak pressure depression occurs east of the Kii Peninsula (around 34.0°N, 136.5°E) and a convergent front trends south-southwestward from 35.0°N, 137.0°E. The westerly surface wind blowing into the front is blocked by the Kii Mountains (altitude, ∼1000 m; 34.0°N, 135.5°E) and meanders along the coast of the Kii Peninsula. As a result, the surface wind becomes cyclonic around 33.5°N, 136.0°E. In the case that the characteristic height H is assumed to be 1000 m, the Froude number (≡V/NH) obtained from horizontal velocity V (<10 m s−1) and buoyancy frequency N (0.01–0.015 s−1) in the upstream area around 34°N, 135°E is less than unity. This enhances the blocking of horizontal flow by the Kii Mountains, leading to cyclonic flows near the surface.
4. Discussion and summary
This study investigated a case of the Yellow and East China Seas (YECS) and Japan Sea and South Coast (JSSC) twin cyclones on 9–10 October 2001. The northern surface lows of the YECS and JSSC twin cyclones reached the upper troposphere and were located north of the upper-level jet. The southern shallow lows were located south of the upper-level jet, which intersected with the lower-level jet accompanying the moist and warm air. Although the twin cyclones were not strong, heavy precipitation was observed in Japan. The shallow southern low of the YECS merged with the northern low via coupling of the southern near-surface relative vorticity with the northern upper-level vorticity in the asymmetric contact binary rotation. Because the distance between the deep northern and shallow southern lows is short in the contact binary rotation compared with the distance reported by Ziv and Alpert (2003), the twin cyclones are likely to merge in the small YECS area. Furthermore, the coupling between the shallow southern and the upper northern lows is enhanced by the latent heating via the amplified vertical flow.
During the rapid merger process of the YECS twin cyclones, the developing southern YECS cyclone travelled northward along the east coast of the Korean Peninsula, thereby extending the moist tongue from the Pacific to the Japan Sea. The crossing of the moist tongue over the Japanese Islands resulted in heavy precipitation in Japan. After merger of the YECS twin cyclones, the JSSC twin cyclones were newly organized. The deep northern low was amplified by the merger of the YECS twin cyclones, while a new, shallow southern low formed around the occlusion point.
The twin cyclones formed around a baroclinic zone that involved only a single upper-level jet, in contrast to the twin cyclones associated with two baroclinic zones formed by double upper-level jets described in previous works (Hitsuma, 2006; Ogura et al., 2006). Although a strong upper-level jet (similar to that in the control run) is formed in the fake dry simulation, the twin cyclones are not formed solely as a result of the baroclinicity of the strong upper-level jet. Thus, condensational heating is essential for the formation of the twin cyclones, and enhances the evaporation over the Yellow Sea and the precipitation around Japan. The low on the northern side of the baroclinic zone is more strongly amplified by condensational heating than is the low on the southern side. This causes the asymmetric structure between the northern and southern lows of the twin cyclones.
In addition to the deep Yellow Sea cyclone with the strong upper-level jet, the shallow low appeared around the Kuroshio front (where SST gradient is large) and baroclinic zone by the jet over the East China Sea. Both the ocean front (e.g., Xie et al., 2002) and upper-level jet might contribute to the cyclogenesis. The YECS twin cyclone formed in the trough which opens to the Pacific in Figure 2(a). The trough is likely to transport humid air from the Pacific to the YECS areas, because it opens to the warm and humid area over the Pacific (Ogura et al., 2005). In such a humid and baroclinic situation over the YECS, the twin cyclones were generated in the presence of a single jet and moist tongue. This is a new cyclogenesis system of twin cyclones with a single upper-level jet.
The topography of the Japanese Islands affects the formation of the JCCS twin cyclones. Around the southern low of the twin cyclones, enhanced vertical flow in the precipitation area is associated with topographic upwelling (due to wind convergence) and latent heating (due to moisture convergence) in the case that considers the topography of the Japanese Islands. The presence of the Kii Mountains leads to topographic convergence and blocking of the horizontal wind, thereby enhancing the upward and cyclonic flows. Compared with Exp. FJ, upward flow is amplified in Exp. CL, but topography has only a minor effect on the pressure depression (sea-level pressure is slightly depressed near the centre of the southern cyclone).
As mentioned above, the present study proposes a complex heavy precipitation process caused by a humid air transport via the merger of twin cyclones (moisture transport process in Figure 17) and by a dynamical coupling of the lower- and upper-level jets (dynamical process in Figure 17) in a binary rotation system of twin cyclones. In addition, a formation mechanism of twin cyclone in a baroclinic zone caused by a single upper-level jet was found, different to the double upper-level jets in previous studies (Hitsuma, 2006; Ogura et al., 2006). In the present case, the latent heating is important in the rapid merger process and the formation of the humid lower-level jet (i.e., moist tongue) across the Japanese Islands, and the topography of the Kii Peninsula is important in the formation of the southern low of the twin cyclones under the condition of the single upper-level jet. A merger process in the contact binary rotation (Ziv and Alpert, 2003), which is caused by coupling of the upper- and lower-level vorticities (Takayabu, 1991) and enhanced by latent heating, is found over the small marginal sea. After the merger, because the humid lower-level jet was located below the upper-level jet (Shapiro, 1982), both the upward wind and precipitation are enhanced by the coupling of the upper- and lower-level jets over Japan.
In the present study, the significance of the merger and binary interaction of twin extratropical cyclones in weather forecasting and analysis of heavy precipitation is recognized. In particular, the rapid YECS merger changes the travelling path from the Kuroshio to the Japan-Sea active cyclogenesis areas (Figure 18) and enhances the Japan Sea cyclone. As a result of this merger, humid air (grey area in Figure 18) is transported over the Japanese Islands where the two active cyclogenesis areas separate, and leads to heavy precipitation over the wide area. In future studies it would be necessary to classify the cases of twin cyclones associated with a single upper-level jet (the present work) and double upper-level jets (Hitsuma, 2006; Ogura et al., 2006), and to examine the differences and similarities in the cyclogenesis and precipitation processes of the two-type twin cyclones.
The author would like to thank S. Ikeda (Wakayama University) for fruitful discussions on her preliminary data analysis and hazard report (which formed the starting point for the present study). This study is part of the project ‘Understanding influences of global warming and rapid economic development on the East Asia marine and atmospheric environment’, Kyushu University, and a Grant-in-Aid for Scientific Research on Innovative Areas (KAKENHI No. 22106003). This investigation analyzed RANAL and radar AMeDAS data provided by JMA. Weekly optimum interpolation SST data were sourced from the data archives of the NOAA Environmental Modelling Center (http://www.emc.ncep.noaa.gov/research/cmb/sst_analysis /).