Tropical Cyclone Formations in the South China Sea Associated …ntur.lib.ntu.edu.tw/bitstream/246246/174877/1/07.pdf · As in other TC formation studies (e.g., Cheung 2004), formation

Tropical Cyclone Formations in the South China Sea Associated with theMei-Yu Front

CHENG-SHANG LEE AND YUNG-LAN LIN

Department of Atmospheric Sciences, National Taiwan University, Taipei, Taiwan

KEVIN K. W. CHEUNG

National Science and Technology Center for Disaster Reduction, Taipei, Taiwan

(Manuscript received 17 February 2005, in final form 19 January 2006)

ABSTRACT

This study examines the 119 tropical cyclone (TC) formations in the South China Sea (SCS) during1972–2002, and in particular the 20 in May and June. Eleven of these storms are associated with the weakbaroclinic environment of a mei-yu front, while the remaining nine are nonfrontal. Seven of the 11 initialdisturbances originated over land and have a highly similar evolution. Comparison of the frontal andnonfrontal formation shows that a nonfrontal formation usually occurs at a lower latitude, is more baro-tropic, develops faster, and possibly intensifies into a stronger TC. Six nonformation cases in the SCS arealso identified that have similar low-level disturbances near the western end of a mei-yu front but did notdevelop further. In the nonformation cases, both the northeasterlies north of the front and the monsoonalsouthwesterlies are intermittent and weaker in magnitude so that the vorticity in the northern SCS does notspin up to tropical depression intensity. Because of the influence of a strong subtropical high, convection issuppressed in the SCS. The nonformation cases also have an average of 2–3 m s�1 larger vertical wind shearthan the formation cases. A conceptual model is proposed for the typical frontal-type TC formations in theSCS that consists of three essential steps. First, an incipient low-level disturbance that originates over landmoves eastward along the stationary mei-yu front. Second, the low-level circulation center with a relativevorticity maximum moves to the open ocean with the stationary front. Last, with strengthened northeast-erlies, cyclonic shear vorticity continues to increase in the SCS, and after detaching from the stationaryfront, the system becomes a tropical depression.

1. Introduction

The mei-yu season in southeast Asia is from aboutmid-May to mid-June when a stationary front in thenorthern South China Sea (SCS) often occurs. DuringMay and June 2002 two tropical cyclones (TCs) formedin the SCS associated with the mei-yu front, namelyTropical Depression 05W and Typhoon Noguri 07W.Several days before the formation of Typhoon Noguri,a mei-yu front extended from Hainan Island to south-ern Taiwan at 0000 UTC 3 June 2002 (Figs. 1a,b). Alarge area of low pressure was located at the western

end of the front, and a high pressure system was locatedto the north. From 0000 UTC 3 June to 0600 UTC 6June, the low pressure system moved eastward alongthe mei-yu front but was still attached to it when thehigh pressure system to the north moved from land tosea (Figs. 1c,d). The low pressure system deepened andwas declared a tropical depression by the Joint Ty-phoon Warning Center (JTWC) at 0000 UTC 6 June.Curvature in the cloud band of the system as seen in thesatellite imagery also suggests that the vorticity had al-ready developed into a column structure. By 0600 UTC8 June, TC Noguri had further intensified to 998 hPaand detached from the frontal system (Figs. 1e,f). Thepurpose of this study is to examine 1) how often thistype of TC formation occurs, and 2) the associatedmechanisms of formation.

The mei-yu frontal system is part of the East Asia

Corresponding author address: Cheng-Shang Lee, Departmentof Atmospheric Sciences, National Taiwan University, 1, Section4, Roosevelt Rd., 106 Taipei, Taiwan.E-mail: [email protected]

2670 M O N T H L Y W E A T H E R R E V I E W VOLUME 134

© 2006 American Meteorological Society

MWR3221

monsoon system. After the summer monsoon starts ev-ery year, organized precipitation associated with con-vection along the monsoon trough first occurs in theSCS, and is sometimes stationary or has a northward

migration. When the subtropical high in the westernNorth Pacific (WNP) retreats to the east, there is oftena subsequent frontal system approaching southeastAsia (Tao and Chen 1987). The persistent southwest-

FIG. 1. Surface weather maps and GMS-5 visible satellite imageries associated with Typhoon Noguri (in 2002)at (a), (b) 0600 UTC 3 Jun, (c), (d) 0600 UTC 6 Jun, and (e), (f) 0600 UTC 8 Jun 2002.

OCTOBER 2006 L E E E T A L . 2671

erlies associated with the summer monsoon and thefrontal easterlies create a general cyclonic flow envi-ronment in the northern SCS. Moreover, the onset ofthe summer monsoon is closely related to the eastwardextension of the westerlies from the Bay of Bengal (Liuet al. 2002). Occasionally, disturbances can originatefrom these westerlies to become surface lows in thenorthern SCS. Chen and Chang (1980) showed that themei-yu front and midlatitude fronts have similar struc-tures in that they possess a large horizontal tempera-ture gradient and vertical baroclinic structures. The ma-jor difference is that the mei-yu front is only weaklybaroclinic and there can be strong low-troposphericvertical wind shear, which is especially true for the fron-tal systems west of Japan (Ninomiya and Akiyama1992).

Lander (1996) examined TC formations during 1978–94 associated with the so-called reverse-oriented mon-soon trough, which is similar to the frontal orientationin the East Asia mei-yu season. However, Lander(1996) did not explicitly identify the cases that originat-ed from the mei-yu front. Such mei-yu formations maybe related to the subtropical cyclone (STC), which He-bert and Poteat (1975) defined as a TC-like circulationthat forms in the baroclinic subtropical environmentassociated with a front. The cloud coverage in a STCusually measures 15° latitude across, which is largerthan a typical TC. Similar to the naming convention fortropical storms, a STC is called a subtropical stormwhen its maximum wind first exceeds 17 m s�1.

Different statistics have been given on tropical cyclo-genesis associated with baroclinic or frontal systems.The results of Frank and Clark (1977) showed that lessthan 5% of all TC formations in the Atlantic occurredin close vicinity to a front or the east side of a westerlytrough. However, Gray (1998) indicated that about25% of the global TC formations were not in a puretropical environment. Bosart and Bartlo (1991) studiedHurricane Diana (in 1984) in the Atlantic, and showedthat the pre-Diana disturbance was also a STC thatoriginated from a frontal system east of Florida. Bosartand Bartlo (1991) demonstrated an interaction betweenupper- and lower-level potential vorticity (PV) anoma-lies. The PV anomaly associated with the upper-leveltrough was advected toward the low-level disturbancesuch that the entire system was able to evolve to a morebarotropic warm-core system. Davis and Bosart (2001)simulated the formation of Hurricane Diana using thefifth-generation Pennsylvania State University–National Center for Atmospheric Research (PSU–NCAR) Mesoscale Model (MM5) model and foundthat the essential stage of transformation from a weak

baroclinic disturbance into Diana was strongly modi-fied by the effects of latent heating. In the first stage,the low-level circulation is strengthened through theaxisymmetrization of remote PV anomalies that aregenerated by condensational heating and then advectedtoward the incipient storm. After a warm-core vortexand spiral bands of convection begin to form, the mois-ture in the core of the storm increases. Davis andBosart (2002) explored the model sensitivities in simu-lating Diana. When the upper-level trough was re-moved from the initial condition, cyclogenesis did notoccur, which indicates the importance of appropriatecoupling between the upper and lower levels in trans-forming the baroclinic system into a tropical depres-sion.

In a further study, Davis and Bosart (2003) collected10 cases of transformation from a STC to a TC duringSeptember–November in 2000 and 2001. A rapid de-crease in the 200–900-hPa vertical wind shear occurredbefore the STC became a TC. On the other hand, thosesubtropical cyclones that did not transform to a TCexperienced vertical wind shear of about 15–20 m s�1.Davis and Bosart (2003) also simulated the formationof Hurricane Michael (in 2000) and found that the ver-tical wind shear had to decrease rapidly from 30 m s�1

to realize the warm-core structure. Therefore, the re-duction of the vertical wind shear of the initial baro-clinic system is an essential requirement for its trans-formation to a TC.

Whereas the mei-yu front is a weak baroclinic systemand occurs, on average, at a lower latitude than theextratropical cyclones in the Atlantic, this study is ded-icated to identifying the mechanisms of transformationfrom baroclinic systems to TCs in the SCS. The orga-nization of this study is as follows. The data sourcesused in the study are discussed in section 2, and theoverall TC activity in the SCS and their large-scale forc-ing are examined in section 3. In section 4, the frontal-type TC formation in the SCS is defined, and it is com-pared with the nonfrontal-type formation. It is also im-portant to understand the nonformation cases in whichan incipient storm does not intensify; this is discussed insection 5. Finally, the mechanisms of frontal-type for-mation are summarized in section 6.

2. Data

Various data are utilized in this study. First, the cli-matology of TC formations in the SCS during 1972–2002 is established based on the best-track data fromthe JTWC. Whereas the best-track data from 1950 areavailable, many of the records prior to 1972 begin with


tropical storm intensity (35 kt or 18 m s�1), and are notsuitable for TC formation studies. Thus, only therecords after 1972 are used. Detailed descriptions of theTCs are also extracted from the corresponding JTWCAnnual Tropical Cyclone Reports.

Surface features are defined from the daily weathercharts of the Japan Meteorological Agency (JMA).When JMA charts are missing, the Taiwan CentralWeather Bureau charts are used. In addition, the infra-red and visible satellite imageries from the Geostation-ary Meteorological Satellites (GMS) in the same 31-yrperiod are examined to compare with the surface fea-tures.

The extended reconstructed sea surface temperature(ERSST) analyses from the National Oceanic andAtmospheric Administration (NOAA) with 2° lati-tude–longitude resolution are used (Smith and Reyn-olds 2003). The ERSST data have been statisticallycorrected (bias correction and error estimation) basedon the Comprehensive Ocean–Atmosphere Data Set.The monthly mean interpolated outgoing longwaveradiation (OLR) with 2.5° latitude–longitude resolu-tion is also from NOAA, and the 6-hourly NationalCenters for Environmental Prediction (NCEP) reanaly-ses with the same resolution are used for the upper-level data.

The SCS is defined as the region 0°–22°N, 105°–120°E in this study (the area east of 120°E is called theWNP). Any TCs entering the SCS from the WNP wereomitted, so only TC formations associated with a fron-tal system in the SCS are considered. As in other TCformation studies (e.g., Cheung 2004), formation time isdefined as the first time a TC reaches an intensity of 25kt (�13 m s�1) in the JTWC best-track data.

3. Climatology of tropical cyclone formations inthe South China Sea

a. Spatial and temporal distribution

A total of 119 (793) TC formations in the SCS(WNP) occurred during 1972–2002 with an annual av-erage of 3.8 (25.6). Almost no formations occur in theSCS from January to March, but the number increasesrapidly to 16.8% (of the total number of formations inthe SCS) during May and June. The corresponding per-centage for May and June in the WNP is only 9%,which indicates that the two months of the mei-yu sea-son are particularly favorable for TC formations in theSCS. Both the SCS and WNP have the highest monthlypercentages of formations in August (21% for the SCSand 20% for the WNP), and then the formation ratesdecrease gradually until the end of the year. A contrast

between the two basins is in December in which thepercentage of formations in the SCS is relatively larger(9% versus 5%).

The TC formations in the SCS have a latitudinalvariation during the season (Fig. 2). Formations duringthe first half of the season (May–September) are usu-ally north of 15°N, while those in late season (October–December) are usually south of 15°N. This latitudinalvariation is closely related to the meridional shift of themonsoon shear line (where smaller vertical wind shearis found) in a year (Cheung 2004), and also the en-hanced convection in the tropical convergence zoneduring cold surges related to the winter monsoon.

The interannual variation of TC formations in theSCS is also large compared with those in the WNP (Fig.3). No formations occurred in the SCS in 1991, while 10formations occurred in 1998. The latter is highly relatedto the La Niña during that year that persisted until early1999. For instance, the formation in January 1999 wasthe first in that month since 1972. The enhanced con-

FIG. 2. Latitudinal distribution (average per 2° latitude) of TCformations in the SCS in May and June (solid), July–September(dotted), and October–December (dashed). The abscissa indi-cates percentage.


vective activity in the western (eastern) portions of theentire WNP basin tends to shift the mean positions ofTC formations westward (eastward) substantially dur-ing the La Niña (El Niño) (Wang and Chan 2002). TheLa Niña (or post–El Niño) effect can also be seen in1973 and 1983. Therefore, the TC activity in the SCSand WNP are out of phase during the phases of the ElNiño–Southern Oscillation. Perhaps a trend exists foran increasing number of formations in the SCS sincethe average number of cases during 1972–94 is only2.81, while the average of 6.33 after 1994 is far higherthan the climatological average of 3.8. Certainly, thisincreasing trend might be related to interdecadal vari-ability in TC activity (e.g., Ho et al. 2004) and should besubjected to further statistical verification.

b. Correlations with SST and OLR

The interannual variability of TC activity in the SCScan be partially explained by the corresponding vari-ability in SST and OLR in the area. For instance, thelong-term mean of SST (during 1972–2002) in the re-gion 8°–21°N, 110°–120°E is 27.8°C. The correlationcoefficient between the SST anomalies (differencesfrom the long-term mean) in this region and the num-ber of TCs in the SCS is 0.55 during 1972–2001. How-ever, this correlation coefficient changes substantiallyin different decades: 0.4 in 1972–79; �0.6 in 1980–88;and 0.9 in 1989–2001. The high correlation in the latestperiod is associated with the much above-average TCactivity in the SCS. For OLR data averaged in the re-gion 0°–22°N, 105°–120°E, the correlation betweenOLR anomaly and SCS TC activity is �0.5 during

1979–2001, which is similar in magnitude to that of SST(negative because low OLR anomaly represents strongconvection).

The monthly mean SST and OLR anomalies (in thesame regions defined above) are also correlated withthe TC activity in the SCS (Fig. 4a). During the firstfour months of a year, the SST anomaly is still low,and OLR anomaly still high (minimum deep convec-tion) in the SCS. The TC activity in the basin beginsin May when both the SST and OLR anomalies changedramatically to be favorable for TC formation. Thesefavorable conditions in May persist until Septemberbefore the number of formations decreases. The meanSST is above 29°C in the SCS during May and June.The cold OLR anomaly well below 200 W m�2 is be-tween the equator and the Borneo Island (latitudinalrange of the island is about 3°S to about 5°N) fromJanuary to April (Fig. 4b), the cold OLR anomalyshifts sharply after May to about 13°N, and stays atthis latitude until September. This OLR migration isconsistent with the latitudes of TC formations in theSCS during the summer (Fig. 2). Although the con-vective clouds in the TCs certainly also contributeto the low OLR values, this should not be the dominat-ing factor in the climatological values because the TCslast only a few days and occur at different times everyyear.

c. Synoptic analyses

The monthly mean wind fields from the NCEP re-analyses are consistent with the relatively active TCformations in the SCS during May and June. Duringwinter, the dominant low-level (850 hPa) circulationover the SCS is the high pressure center associated withthe winter monsoon (not shown). Very often this anti-cyclone extends eastward from mainland China toabout 120°–125°E, which brings persistent northeaster-lies with negative relative vorticity over the SCS. In thewinter, positive 850-hPa relative vorticity is only foundnear the equator (e.g., near the west coast of theBorneo Island). In May, southerlies and southwester-lies associated with the summer monsoon start to be thedominant circulation in the SCS (Fig. 5a). Positive 850-hPa relative vorticity is also found at higher latitudes(5°–10°N) such that TC formation is possible. In June,southwesterly flow covers almost the entire SCS, andpositive cyclonic vorticity associated with the south-westerlies is common in the region (Fig. 5b).

The midlevel (500 hPa) circulation depends mainlyon the location of the subtropical high. During the win-ter, the subtropical anticyclone is over the SCS suchthat the entire basin is associated with negative relative

FIG. 3. Interannual variation of the number of TC formationsfrom 1972 to 2002 in the SCS (solid) and the WNP (dashed).


vorticity (not shown). In May, the subtropical highcenter begins to retreat eastward to about 140°E,which gives way to the southwesterlies from the west(Fig. 5c). In June, the subtropical high shifts slightlynorthward to about 20°N (Fig. 5d), which leads to aconfluent-type flow between the southwesterlies andthe subtropical high over the SCS, which is favorablefor TC formation.

Upper level (200 hPa) divergent flow is known to befavorable for TC formation. During the winter (notshown), the Pacific anticyclone is centered at around150°E such that southwesterlies are over the SCS.These strong southwesterlies over low-level northeast-

erlies induce large vertical wind shear in the region,which is not favorable for development of any tropicaldisturbances that might exist. However, an anticyclonebuilds over Indochina during May (Fig. 5e) and is lo-cated farther to the northwest in June (Fig. 5f). Thedivergent flow over the SCS associated with this anti-cyclone during these months provides a favorable up-per-level environment for TC formation.

4. Frontal-type formations

Among the 119 TC formations in the SCS during1972–2002, nine occurred in May and 11 in June. One of

FIG. 4. (a) Monthly variation of averaged OLR anomaly (left vertical axis; W m�2), SSTanomaly (right vertical axis; °C), and anomaly number of TC formations in the SCS (TCnumber; also using the left vertical axis) in the SCS. (b) Spatial distribution of OLR within thedomain 0°–26°N, 105°–123°E from January to December.


the most interesting features about these formations inMay and June is that some of them are closely relatedto the quasi-stationary front at the northern part of theSCS, which occurs in every mei-yu season (15 May–15

June). To identify the mechanism for this type of for-mation, a TC formation is defined as a frontal-typeformation when the following conditions are satisfiedduring the 4 days before the formation:

FIG. 5. Monthly mean flow in (left) May and (right) June from 1972 to 2002 at (a), (b) 850; (c), (d) 500; and (e), (f) 200 hPa. Heavycontours in (a), (b) indicate wind speed greater than 6 m s�1 (interval: 2 m s�1). Shaded contours are (a)–(d) positive relative vorticity(interval: 0.5 � 10�5 s�1) and (e), (f) divergence (interval: 0.2 � 10�5 s�1).


1) A closed surface isobar is located at the southwestend of the stationary front;

2) The 925-hPa relative vorticity associated with thefrontal zone is greater than 10�5 s�1 and is orientednortheast–southwest; and

3) A substantial north–south-oriented gradient ofequivalent potential temperature exists at 1000 hPa,above the frontal zone.

If all of these conditions are not met, the formationcase is classified as a nonfrontal type formation. In to-tal, 11 frontal-type formations occurred in the 31-yrperiod and the remaining nine cases were of the non-frontal type (Table 1), which indicates that the frontaltype is not uncommon during the mei-yu season. Themean formation location of the 11 frontal-type cases isat 18.5°N, 115.6°E (Fig. 6a). The mean motion of theseTCs is toward the east-northeast, which is similar tothose associated with a reverse-oriented trough forma-tion analyzed by Lander (1996). Among these systems,only two (Ike in 1981 and Noguri in 2002) reached ty-phoon intensity and five never reached tropical stormintensity in their lifetimes. In addition, the average timeinterval for these TCs to develop from TC formationalert (TCFA) issued by the JTWC to tropical storm is47 h, which indicates that they were not in an environ-ment that was favorable for rapid intensification. Nev-ertheless, the formation location of these frontal-typeTCs is close to cities in southeast Asia and could affectthese cities in a short time after formation. Therefore,understanding their formation mechanisms is importantfor forecasters.

The nine nonfrontal type cases in May and June hada mean formation location of 14.0°N, 114.2°E, which ismore than 4° latitude south of that of the frontal type.Moreover, the mean motion of these nine TCs wasnorthwestward because they were mainly steered bythe easterlies of the subtropical high. On average, thesenonfrontal TCs intensified more than the frontal ones[average maximum intensity reached was 50 kt (�26m s�1)], and they also developed faster as the averagetime interval from TCFA to tropical storm was 32 hcompared with 47 h for the frontal type.

a. Composites

To compare in detail the differences between frontal-type and nonfrontal-type formations, composites of the11 frontal cases and the nine nonfrontal cases are com-puted using the NCEP reanalyses. Because all of the 20cases formed near the same location (Fig. 6), the com-posites with a fixed geographical domain have similarcharacteristics to those composites with respect to thesystem center (sea level pressure minimum) of each

formation case. Therefore, only the composites with fixeddomain are presented here. The zero time of each caseis the first time that Vmax exceeded 25 kt (�13 m s�1).

Three days before a frontal-type formation (Fig. 7a),the 925-hPa flow in the northern SCS has an east–west-oriented band of positive relative vorticity associatedwith the monsoon trough and the surface front. South-westerlies of up to 10 m s�1 are flowing into the SCSand meeting with northeasterlies associated with an an-ticyclone over mainland China. From 48 to 36 h beforeformation (Fig. 7b), the synoptic situation remains simi-lar. However, the vorticity maximum over the northernSCS intensifies, the southwesterlies into the SCS arestill strong, and the anticyclone to the north is still overland. The surface front (not shown) is oriented north-east–southwest, and is just south of Taiwan.

One day before formation (Fig. 7c), the southwest-erlies increase in magnitude up to 12 m s�1, which fur-

TABLE 1. Maximum intensity (1 kt � 0.5144 m s�1) and timetaken to develop (when data are available) from a TCFA to atropical storm (TS) for the (a) frontal-type and (b) nonfrontal-type formations during May and June 1972–2002. Cases A to Gare the seven typical frontal-type formations used to compute thecomposites in section 5 for comparing with the nonformationcases.

(a)

Yr/month TC # Name

Maxintensity

(kt)Development timeTCFA → TS (h)

(A) 1974/6 TD 05w — 30 —1979/5 TD 05w — 30 —1981/6 TY 04w Ike 65 42

(B) 1985/6 TD 04w — 30 —(C) 1986/5 TS 04w Mac 45 68(D) 1987/6 TS 03w Ruth 35 36(E) 1994/6 TS 05w Russ 55 17(F) 1997/5 TS 05w Levi 45 54

2000/5 TD 03w — 30 —2002/5 TD 06w — 25 —

(G) 2002/6 TY 07w Noguri 85 66Avg �43 �47

(b)

Yr/month TC # Name

Maxintensity

(kt)Development timeTCFA → TS (h)

1972/6 TS 05w Mamie 50 241980/5 TS 06w Georgia 55 52.31982/6 TS 06w Tess 35 60.51984/6 TS 01w Vernon 40 25.81989/5 TY 04w Cecil 75 9.51990/6 TD 04w — 30 —1996/5 TS 05w Cam 60 37.52000/5 TD 04w — 30 —2001/6 TY 05w Durian 75 15Avg 50 �32


ther increases the vorticity maximum in the SCS. Thecold high to the north starts to move off the east coastof China (and so does the surface front), which bringssome strong easterlies or northeasterlies to the north-ern SCS. At the time of frontal-type formation (Fig.7d), the high has moved into the East China Sea, andthe 925-hPa vorticity maximum has intensified as thetropical depression stage is achieved.

At the midlevel (500 hPa), the dominant flows arewesterly and southwesterly over the SCS 3 days beforeformation (Fig. 8a). The subtropical high is east of140°E so that its influence over the SCS is weak. By36 h before formation (not shown), an area of strongerrelative vorticity is located at the northern SCS becauseof a developing short-wave trough. At the time of for-mation (Fig. 8c), this short-wave trough further deep-ens in conjunction with the low-level system. Compari-son with the 925-hPa maps reveals the importance ofthe low-level features in this frontal-type formation, andthe baroclinicity of the system. At the 200-hPa level, ananticyclone is located over northern Indochina threedays before formation (Fig. 8b). By 36 h before forma-tion (not shown), this anticyclone extends eastwardsinto the northern SCS. At the formation time (Fig. 8d),the ridge further develops and leads to a diffluent flowregion over the SCS. Superposing divergence over thelow-level system is favorable for its intensification.

The composites associated with the nonfrontal-typeformation have distinct differences from the frontaltype. Two days before formation (Fig. 9a), the maxi-mum positive relative vorticity at 925 hPa is at a lowerlatitude (about 15°N) associated with the monsoon

trough, and the easterlies over the northern SCS areassociated with the subtropical high in the WNP. At24 h before formation (Fig. 9b), both the southwester-lies associated with the monsoon trough and the east-erlies to the north increase in magnitude, which givesrise to a strong vorticity center. The situation at 500 hPa(Figs. 9c,d) is similar to that at 925 hPa, which indicatesthat the development in this nonfrontal type of forma-tion is more barotropic. The subtropical high is shiftedto the western boundary of the WNP. The anticycloneat 200 hPa 2 days before formation (Fig. 9e) is similar tothe frontal-type composite; however, this anticycloneremains to the west (Fig. 9f) throughout the formationprocess without the development of a diffluent regionas in the frontal-type formation.

Composites of the 1000-hPa equivalent potentialtemperature also show contrasts between the frontaland the nonfrontal formations. At 24 h before a frontalformation (Fig. 10a), the equivalent potential tempera-ture has a maximum in the northern SCS that is southof a region with a large equivalent potential tempera-ture gradient. Although the situation for the nonfrontalformations (Fig. 10b) is similar at this time, the maxi-mum temperature is further to the south where thedevelopment is occurring. At the formation time for thefrontal type (Fig. 10c), the large equivalent potentialtemperature gradient in the northern SCS is just northof the system center. In the nonfrontal formation case(Fig. 10d), the gradient of equivalent potential tem-perature throughout the SCS is small.

The magnitude of 200–850-hPa vertical wind shear(not shown) is about the same in the composites of

FIG. 6. The best tracks and mean formation position (typhoon symbol) of the (a) frontal-type (18.5°N, 115.6°E) and (b) nonfrontal-type (14.0°N, 114.2°E) TCs in the SCS. The large arrows indicate their respective mean motions, and the insets show the formationpositions of individual TCs.


frontal-type and nonfrontal formations. As the mon-soon shear line shifts northward to about 15°N duringMay and June, low zonal wind shear exists over thenorthern SCS. The major contrast between frontal-typeand nonfrontal formations is in the meridional windshear, with northerly shear over the SCS for frontal-type formations and southerly shear for nonfrontaltype. That is, the frontal-type formations occur underthe upper-level northerlies associated with the eastwardextension of the anticyclone over Indochina (Fig. 8d).The nonfrontal formations occur under weaker upper-level northerlies.

b. Conceptual model

Seven of the 11 frontal-type formation cases involveda disturbance cyclone that originated over land (labeled

A–G in Table 1). Although the origin of the disturbancefor the remaining four cases is uncertain based on theavailable data, they were probably over the ocean. Theseseven typical frontal-type TC formations during Mayand June 1972–2002 have a unique formation mecha-nism in close relationship to the development of themei-yu front, which can be summarized in a conceptualmodel consisting of three essential steps (Fig. 11).

1) In the incipient step, a low-level disturbance overland that originated to the west from cyclogenesis ina weakly baroclinic environment moves eastwardalong the stationary mei-yu front with a semiclosedsurface isobar.

2) During the movement over sea step, the low-levelcirculation center moves over the SCS along the sta-tionary front and with strengthened southwesterlies

FIG. 7. Composite 925-hPa flow for frontal-type formations at (a) �72, (b) �48, (c) �24, and (d) 0 h. The heavy contours indicatewind speed greater than 6 m s�1 (interval: 2 m s�1) and the shaded contours are positive relative vorticity (interval: 0.5 � 10�5 s�1).The typhoon symbols mark the mean formation positions.


now has a relative vorticity maximum and a closedsurface isobar. If the vertical wind shear over thedisturbance decreases, the baroclinicity also weak-ens and vice versa.

3) During the transition step, the northeasterlies alongthe southeast China coast strengthen as the coldhigh over China moves to the Yellow Sea, and thusthe cyclonic shear vorticity continues to increaseover the SCS. Deep convection associated with la-tent heat release in the region reduces the Rossbyradius of deformation of the system, and after de-taching from the stationary front the system be-comes a tropical depression.

As these frontal-type formations are embedded inthe monsoon trough over the SCS, they may belong tothe monsoon shear line formation pattern depicted inRitchie and Holland (1999) that accounts for 42% of allformation cases in the WNP (including the SCS). Thecomposite of the monsoon shear line pattern in Ritchie

and Holland indicates a region of convergence that ex-tends eastward from the formation location and is con-ducive to the development of sustained convection. Forthe frontal-type formations, this convergent region isprovided by the front, which then leads to the convec-tive activity. The upper-level composite in Ritchie andHolland is also similar to the one for the frontal-typeformation (Fig. 8d) with a diffluent region aloft. Be-cause frontal-type formations in the SCS possess signa-ture characteristics that may be easily recognized, theymay be treated as a special case of monsoon-shear lineformation so that forecasters can estimate the potentialof such a TC formation in the region when similar con-ditions appear.

5. Formation versus nonformation

a. The nonformation cases

Examination of weather maps near the time of for-mation of the seven typical cases of frontal-type forma-

FIG. 8. Same as in Fig. 7 but for (left) 500- and (right) 200-hPa flows at (a), (b) �72 and (c), (d) 0 h. Shaded contours are (a), (c)positive relative vorticity (interval: 0.5 � 10�5 s�1) and (b), (d) divergence (interval: 0.2 � 10�5 s�1). The typhoon symbols mark themean formation positions.


tions selected in the last section indicates that they allhad the stationary mei-yu front accompanying the in-cipient surface low. Because the development in theseven selected cases is highly similar, the formation

cases mentioned in this section refer exclusively tothese seven cases.

The question is the following: Will a formation in theSCS always occur whenever an incipient disturbance

FIG. 9. Same as in Fig. 7 but for (a), (b) 925-; (c), (d) 500-; and (e), (f) 200-hPa flows for nonfrontal-type formation at (a), (c), (e)�48 h and (b), (d), (f) �24 h. Shaded contours are (a)–(d) positive relative vorticity (interval: 0.5 � 10�5 s�1) and (e), (f) divergence(interval: 0.2 � 10�5 s�1). The typhoon symbols mark the mean formation positions.


moves from the west to near the mei-yu front? Theanswer is negative because some surface low pressuresystems southwest of the mei-yu front did not eventu-ally develop into tropical depressions. Yeh and Chen(2001) and Yeh et al. (2002) examined a mesoscale vor-tex in the northern SCS that formed along the mei-yufront and lasted for about 4 days. The relative vorticityof the system intensified to 12.5 � 10�5 s�1. The sys-tem’s Rossby radius of deformation decreased to about

1000 km so that latent heat release from the mesoscalesystem should have been very efficient in intensifyingthe background disturbance. However, the systemnever developed into a tropical depression. To identifythese kind of nonformation systems, weather maps dur-ing May and June 1972–2002 are reexamined. A non-formation case is defined as a case where 4 days after asurface low pressure disturbance is identified, the threeconditions in section 4 are satisfied (and that the closed

FIG. 10. Composite 1000-hPa equivalent potential temperature (interval: 3 K) and mean formation position(typhoon symbol) for the (a), (c) frontal-type and (b), (d) nonfrontal-type formations at (a), (b) �24 and (c),(d) 0 h.


surface isobar lasts for at least 24 h but the system doesnot develop into a TD). Finally, the cases that persistalong the SCS are eliminated because they are easilyaffected by topography.

Six nonformation disturbances were identified in the31-yr period according to the definition above (Table2). The lifetime of these cases ranges from 2 days tomore than 4 days. Except for the system in 1975 thatdeepened to 998 hPa, the central pressures were slightlyabove 1000 hPa. The JMA classified four of the systemsas tropical depressions, but no TCFA was issued by theJTWC. Most of the six nonformation systems movedslightly northward to south China, and the average timeperiod they stayed over the sea was about 2.5 days. Thesurface features of these six cases are rather similar tothose of the formation cases. Some of them also havethe high pressure system to the north that moved to theYellow Sea together with the mei-yu front. Therefore,the question is the following: What factors prohibitedthem from intensifying to the tropical depression stage?

b. Composites

Separate composites using the NCEP reanalyses forthe seven (typical frontal type) formation cases and thesix nonformation cases are computed to identify theprohibiting factors in the latter. For the nonformationcases, the time when the closed surface isobar firstformed is used as the zero-time reference, and this ap-plies to the composites of the formation cases as well sothat they do not have a higher intensity compared with

the nonformation composites. For the seven typical for-mation cases, this reference time is, on average, about36 h before the formation time defined in section 4.Because the system identification conditions for theformation and the nonformation systems are similar,the composites several days before the reference timealso have little differences. Focus is thus put on timesafter the reference time.

Because, by the zero reference time, the formationcases (not shown) have almost converted from a weaklybaroclinic to a more barotropic system, the compositesof the seven typical formation cases have continuousstrong southwesterlies to the south of the incipient lowand easterlies to the north from 925 to 500 hPa (notshown). Divergent flow above the system starts to de-velop after the reference time, and the outflow struc-ture at 200 hPa is already clear by 48 h.

The 925-hPa composite of the six nonformation casesat the reference time has a vorticity maximum in theSCS and southwesterlies south of the system (Fig. 12a).However, the magnitude of the southwesterlies did notincrease in the next 24 h (Fig. 12b). Moreover, the coldhigh to the north is not well defined and the accompa-nying easterlies are relatively weak. Although a de-tached high pressure system is moving over the EastChina Sea, it is not as far east as in the frontal-typeformation. A 500-hPa short-wave trough is present inthe SCS at the reference time (Fig. 12c), and a ridge islocated over Indochina such that at 24 h the short-wavetrough becomes northwest–southeast-oriented (Fig.

FIG. 11. Three steps in the conceptual model of TC formation associated with a mei-yu front.


12d). At later times when the subtropical ridge contin-ues to strengthen, southerlies persist in the SCS that arenot favorable for further intensification. The 200-hPaanticyclone (Fig. 12e) is at about the same location as inthe frontal-type formation at the reference time, andthis anticyclone extends farther eastward 24 h later(Fig. 12f). However, the divergence over the SCS is notparticularly high.

The low-level equivalent potential temperature com-posites of the nonformation cases are similar to those ofthe frontal formations (Figs. 10a,c), except the highequivalent potential temperature gradient regions arefarther to the north (around 30°N) and the gradient inthe northern SCS is not particularly large. A maximumis not found as in Fig. 10 because the nonformationcases do not intensify and go through a warming pro-cess.

Aerial averages of several quantities can also distin-guish the environmental conditions in which typicalfrontal-type formation and nonformation are embed-ded. The average 925-hPa relative vorticity within a500-km radius centered on the surface pressure mini-mum for the formation cases maintains a magnitudebetween 2 � 10�5 s�1 and 3 � 10�5 s�1 after the ref-erence time (Fig. 13a). Although the nonformationcases have about the same magnitude of 925-hPa vor-ticity at the reference time, the vorticity then decreasesas a result of the weakening southwesterlies to thesouth and absence of strong easterlies to the north. The200-hPa divergence associated with the formation casecontinues to be high several days after the referencetime, which is favorable for further intensification ofthe tropical depression (Fig. 13b). In contrast, the di-vergence associated with the nonformation case is al-most zero throughout the period. The average 200–850-hPa deep vertical wind shear for the nonformationcases is on average 2–3 m s�1 larger than for the for-mation cases for several days after the reference time(Fig. 13c). Although values of vertical wind shear ofabout 4–8 m s�1 are not large enough to cause dissipa-tion, the probability of development is lower when thevertical wind shear is higher.

6. Summary and conclusions

The TC season in the SCS effectively starts in May asalmost no TC formations occur in the first 4 months ofthe year, and about 16.8% of all TCs form in May andJune. The TC activity in the SCS correlates fairly wellwith the SST and OLR variations in the basin. In Mayand June, areas with small OLR (less than 200 W m�2,i.e., deep convection) and large SST (greater than29°C) move to the northern SCS, which provides favor-able environmental conditions for TC formation in thatregion. The synoptic flow also consists of persistentlow-level monsoonal southwesterlies, the midlevel sub-tropical high retreats eastward such that near souther-lies persist, and an upper-level anticyclone to the westover Indochina puts weak northerlies over the SCS.This kind of synoptic flow provides an environmentwith positive low-level relative vorticity and weak ver-tical wind shear, which are then suitable environmentalconditions for TC formation.

Among the 119 TC formations in the SCS during1972–2002, 20 were in May and June. In addition, 11 ofthese are associated with the weak baroclinic environ-ment of a mei-yu front, while the remaining nine arenonfrontal. Composites using NCEP reanalyses showthat in the frontal-type formations, the low-level distur-bance typically forms over land and moves eastwardalong the mei-yu front to form a vorticity center overthe northern SCS. At 200 hPa, a diffluent region overthe northern SCS provides divergence. After forma-tion, the mean motion of a frontal-type TC is still east-ward because of the southwesterlies south of the mei-yufront. In contrast, a nonfrontal TC formation is morebarotropic, develops faster, and possibly intensifies to astronger TC. As the subtropical high is still dominant,the formation location is usually south of the frontaltype. Moreover, the mean motion after formation isnorthwestward.

For contrast, six nonformation cases in the SCS dur-ing the period 1972–2002 are identified to distinguishthem from the typical frontal-type formation. These arecases with similar low-level disturbances near the west-

TABLE 2. Lifetimes and minimum surface pressure attained by the nonformation cases during May and June 1972–2002. See section5a for a definition of the nonformation cases.

No. Dates Lifetime (h) Min surface pressure (hPa)

1 0000 UTC 16 Jun–0000 UTC 19 Jun 1975 72 9982 1200 UTC 5 Jun–0000 UTC 10 Jun 1983 108 10023 0000 UTC 1 Jun–0000 UTC 3 Jun 1990 48 10044 0600 UTC 26 May–1800 UTC 29 May 1998 84 10025 0000 UTC 14 Jun–1800 UTC 18 Jun 2000 114 10026 0600 UTC 27 May–0600 UTC 29 May 2001 48 1002


ern end of a mei-yu front, but without further develop-ment. Comparison of their respective composites re-veals that the typical frontal-type formation has persis-tent northeasterlies (for as long as 36 h) along the south

China coast because of the high pressure system northof the mei-yu front moving to the East China Sea. Inthe nonformation cases, both these northeasterlies andthe monsoonal southwesterlies are weaker and inter-

FIG. 12. Same as in Fig. 7 but for (a), (b) 925-; (c), (d) 500-; and (e), (f) 200-hPa flows for the nonformation cases at (a), (c), (e) 0and (b), (d), (f) 24 h. Shaded contours are (a)–(d) positive relative vorticity (interval: 0.5 � 10�5 s�1) and (e), (f) divergence (interval:0.2 � 10�5 s�1).


mittent so that the vorticity in the northern SCS issmaller. Because of the influence of a strong subtropi-cal high, deep convection is suppressed in the SCS. Al-though a similar diffluent structure exists at the upperlevel as in the typical frontal-type formation, it appearslater in the development and divergence over the low-level disturbance is weaker. Whereas the typical fron-tal-type formations experience moderate vertical windshear (about 5 m s�1), the nonformation cases experi-ence a large vertical wind shear (about 8 m s�1).

Because of the low spatial and temporal data resolu-tion, some important issues cannot be precisely deter-mined. For instance, the timing of the baroclinic-to-barotropic transition during the weakening of the fron-tal structure is not precisely known. The timing of whenthe Rossby radius of deformation starts to decreasesuch that latent heat can efficiently build up the warmcore also has to be examined further. Some successfulnumerical simulations might provide suitable fields forfurther diagnosis (Lee and Lee 2002). Preliminarysimulations of Typhoon Noguri (in 2002) indicate thestrengthening of northeasterlies associated with thehigh north of the surface front play an important role.

Frontogenesis calculations from this simulation suggestthat the tilting term contributes to the weakening of thefrontal structure, and indicates that an appropriate ver-tical shear environment may determine the formationor nonformation. Simulations of more of these frontal-type TC formations will be diagnosed in more detail ina later report.

Acknowledgments. Comments from Prof. RussellElsberry of the Naval Postgraduate School and the twoanonymous reviewers are much appreciated. The au-thors thank Ms. Erin Huang and Ms. Chia-Mei Huangfor their help in manuscript preparation. This researchis supported by the National Science Council of theRepublic of China (Taiwan) under Grants NSC94-2111-M-002-019-AP1 and NSC93-2119-M-002-006-AP1.

REFERENCES

Bosart, L. F., and J. A. Bartlo, 1991: Tropical storm formation ina baroclinic environment. Mon. Wea. Rev., 119, 1979–2013.

Chen, G. T.-J., and C.-P. Chang, 1980: The structure and vorticitybudget of an early summer monsoon trough (Mei-Yu) oversoutheastern China and Japan. Mon. Wea. Rev., 108, 942–953.

Cheung, K. K. W., 2004: Large-scale environmental parametersassociated with tropical cyclone formations in the westernNorth Pacific. J. Climate, 17, 466–484.

Davis, C. A., and L. F. Bosart, 2001: Numerical simulations of thegenesis of hurricane Diana (1984). Part I: Control simulation.Mon. Wea. Rev., 129, 1859–1881.

——, and ——, 2002: Numerical simulations of the genesis ofhurricane Diana (1984). Part II: Sensitivity of track and in-tensity prediction. Mon. Wea. Rev., 130, 1100–1124.

——, and ——, 2003: Baroclinically induced tropical cyclogenesis.Mon. Wea. Rev., 131, 2730–2747.

Frank, N. L., and G. Clark, 1977: Atlantic tropical systems of1976. Mon. Wea. Rev., 105, 676–683.

Gray, W. M., 1998: The formation of tropical cyclones. Meteor.Atmos. Phys., 67, 37–69.

Hebert, P. H., and K. O. Poteat, 1975: A satellite classificationtechnique for subtropical cyclones. NOAA Tech. Memo.NWS SR-83, 25 pp.

Ho, C.-H., J.-J. Baik, J.-H. Kim, and D.-Y. Gong, 2004: Interde-cadal changes in summertime typhoon tracks. J. Climate, 17,1767–1776.

Lander, M., 1996: Specific tropical cyclone track types and un-usual tropical cyclone motions associated with a reverse-oriented monsoon trough in the western North Pacific. Wea.Forecasting, 11, 170–187.

Lee, C.-S., and C.-H. Lee, 2002: A numerical simulation of theenvironmental momentum influences on typhoon formation.Proc. Fourth Conf. on East Asia and Western Pacific Meteo-rology and Climate, Taipei, Taiwan, World Scientific, 261–271.

Liu, Y., J. C. L. Chan, J.-G. Mao, and G.-X. Wu, 2002: The role ofBay of Bengal convection in the onset of the 1998 SouthChina Sea summer monsoon. Mon. Wea. Rev., 130, 2731–2744.

Ninomiya, K., and T. Akiyama, 1992: Multiscale feature of Baiu,the summer monsoon over Japan and East Asia. J. Meteor.Soc. Japan, 70B, 467–495.

Ritchie, E. A., and G. J. Holland, 1999: Large-scale patterns as-

FIG. 13. (a) Averaged 925-hPa relative vorticity (10�5 s�1), (b)200-hPa divergence (10�5 s�1), and (c) 200–850-hPa vertical windshear magnitude (m s�1) for the formation (dashed) and nonfor-mation (solid) cases.


sociated with tropical cyclogenesis in the western Pacific.Mon. Wea. Rev., 127, 2027–2043.

Smith, T. M., and R. W. Reynolds, 2003: Extended reconstructionof global sea surface temperatures based on COADS data(1854–1997). J. Climate, 16, 1495–1510.

Tao, S.-Y., and L.-X. Chen, 1987: A review of recent research onthe East Asia summer monsoon in China. Monsoon Meteo-rology, C. P. Chang and T. N. Krishnamurti, Eds., OxfordUniversity Press, 60–92.

Wang, B., and J. C. L. Chan, 2002: How strong ENSO events

affect tropical storm activity over the western North Pacific.J. Climate, 15, 1643–1658.

Yeh, H.-C., and G. T.-J. Chen, 2001: Evolution of a meso-vortexalong the Mei-yu front over South China Sea. Proc. Int. Conf.on Mesoscale Meteorology and Typhoon in East Asia, Taipei,Taiwan, National Science Council, Central Weather Bureau,and National Taiwan University, 359–363.

——, ——, and W.-T. Liu, 2002: Kinematic characteristics of amei-yu front detected by the QuikSCAT oceanic winds. Mon.Wea. Rev., 130, 700–711.


Documents

Tropical Cyclone Formations in the South China Sea Associated …ntur.lib.ntu.edu.tw/bitstream/246246/174877/1/07.pdf · As in other TC formation studies (e.g., Cheung 2004), formation