Collaborative Effects of Cold Surge and Tropical Depression–Type Disturbance on Heavy Rainfall in Central Vietnam

SEPTEMBER 2008 YOKOI AND MATSUMOTO 3275

SATORU YOKOI Center for Climate System Research, University of Tokyo, Chiba, Japan

JUN MATSUMOTO Department of Geography, Tokyo Metropolitan University, Tokyo, and Institute of Observational Research for Global Change, Japan Agency for Marine-Earth Science and Technology, Kanagawa, Japan

(Manuscript received 13 November 2007, in final form 30 January 2008)

ABSTRACT

This paper reveals synoptic-scale atmospheric conditions over the South China Sea (SCS) that cause heavy rainfall in central Vietnam through case study and composite analyses. The heavy rainfall event discussed in this study occurred on 2–3 November 1999. Precipitation in Hue city (central Vietnam) was more than 1800 mm for these 2 days. Two atmospheric disturbances played key roles in this heavy rainfall. First, a cold surge (CS) northerly wind anomaly in the lower troposphere, originating in northern China near 40°N, propagated southward to reach the northern SCS and then lingered there for a couple of days, resulting in stronger-than-usual northeasterly winds continuously blowing into the Indochina Peninsula against the Annam Range. Second, a southerly wind anomaly over the central SCS, associated with a tropical depression–type disturbance (TDD) in southern Vietnam, seemed to prevent the CS from propagating farther southward. Over the northern SCS, the southerly wind anomaly formed a strong low-level convergence in conjunction with the CS northeasterly wind anomaly, and supplied warm and humid tropical air. These conditions induced by the CS and TDD are favorable for the occurrence of the heavy orographic rainfall in central Vietnam. The TDD can be regarded as a result of a Rossby wave response to a large-scale convective anomaly over the Maritime Continent associated with equatorial intraseasonal variability. Using a 24-yr (1979–2002) reanalysis and surface precipitation datasets, the authors confirm that the coexistence of the CS and TDD is important for the occurrence of heavy precipitation in central Vietnam. In addition, it is observed that CSs without a TDD do not lead to much precipitation.

1. Introduction a northern location in September to a southern location in December (Yokoi et al. 2007). Vietnam is located along the eastern coast of the The central part of Vietnam is characterized by a Indochina Peninsula (ICP) and is separated from other narrow lowland plain with a width narrower than 100 parts of the peninsula by the Annam Range, which lies km. The main rainy season in this region is the Octo- along the coast (Fig. 1). The seasonal march of precipi- ber–November period. Climatologically, precipitation tation in Vietnam differs markedly from place to place in Hue city (16.4°N, 107.7°E; see Fig. 1), for example, is because of its substantial latitudinal extent from 8° to more than 1400 mm in this 2-month period, which is 22°N (Matsumoto 1997). In the northern part, the main about half of the annual rainfall. rainy season is September–October, while that in the The October–November period is also known as the southern part is November–December. In other words, transition phase from boreal summer to winter mon- the maximum precipitation area shifts southward from soon seasons. Over the South China Sea (SCS), boreal summer southwesterly monsoon winds retreat in Sep- tember, followed by the appearance of northeasterly Corresponding author address: Satoru Yokoi, Center for Cli- mate System Research, University of Tokyo, 5-1-5, Kashiwanoha, monsoon winds around 20°N in early October (Fig. 2). Kashiwa, Chiba 277-8568, Japan. The northeasterly winds blow into the eastern ICP E-mail: [email protected] against the Annam Range, bringing monsoon rainfall

DOI: 10.1175/2008MWR2456.1

MWR2456 3276 MONTHLY WEATHER REVIEW VOLUME 136

FIG. 1. Topography of the Indochina Peninsula; shading indicates terrain elevation. FIG. 2. Climatological seasonal march of horizontal wind (vector) and its speed (shading) at the 925-hPa level averaged in 110°– 120°E calculated from the JRA-25 dataset. The reference vector through orographic rainfall processes. The northeast- corresponds to a value of 10 m sϪ1. erly wind axis then shifts southward to reach 10°Nin December in concert with the maximum precipitation area in Vietnam. intraseasonal variability (ISV) with a 30–60-day time During the rainy season in central Vietnam, frequent scale. heavy rainfall events often menace the local popula- The CS is known as an episodic southward progres- tion. One of the heaviest rainfall events occurred on 2–3 sion of cold, midlatitude air along the eastern flank of November 1999. Figure 3 shows the precipitation the Tibetan Plateau in the lower troposphere, and is amount for these 2 days obtained from the Tropical observed mainly in the November–March period. Spa- Rainfall Measurement Mission (TRMM) 3B42 dataset tiotemporal structures of the CS have been revealed by (Huffman et al. 2007). Large amounts of precipitation a number of studies. The CS northerly wind signal first exceeding 600 mm (2 days)Ϫ1 were recorded in central appears near 40°N in northern China in association Vietnam. This heavy rainfall was also observed at sur- with a trough–ridge system aloft (Joung and Hitchman face weather stations. For example, at the Hue obser- 1982; Lau and Lau 1984; Wu and Chan 1997; Chen et al. vation site more than 1800 mm of precipitation was 2002) after the intensification of the Siberian High recorded in these 2 days (Fig. 4), which was more than (Ding 1990; Zhang et al. 1997). The CS signal then 60% of the mean annual precipitation and was the larg- propagates southward into the tropics within a few days est amount on record in the previous 50 yr. In addition, (Chu and Park 1984; Sumi 1985; Wu and Chan 1995; precipitation amounts greater than 700 mm (2 days)Ϫ1 Compo et al. 1999). The arrival of the CS is character- were recorded at Da Nang (16.0°N, 108.2°E) and Nong ized by a sharp increase in surface pressure, strong Son (15.7°N, 108.0°E), which are located southeast of northerly winds, and a decrease in temperature (Chang Hue city. This heavy rainfall event caused several et al. 1979; Lau and Lau 1984). After arriving at the rivers to flood, resulting in severe destruction in down- SCS, the CS intensifies convective activity over the stream cities. Hereafter, this event is called NOV99 for SCS, Borneo Island, and sometimes the Maritime Con- brevity. tinent in the Southern Hemisphere (Chang et al. 1979; It is well known that synoptic- and large-scale atmo- Johnson and Priegnitz 1981; Chang et al. 2005; Wu et al. spheric disturbances cause precipitation variability in 2007). the ICP during the summer monsoon season, with a The ISV prevails in convection and tropospheric cir- wide range of time scales from several days to several culation anomalies in the tropics. It is dominated by the weeks (e.g., Yokoi and Satomura 2005; Yokoi et al. Madden–Julian oscillation (MJO; Madden and Julian 2007). It is quite possible that such disturbances also 1994), which is characterized by eastward-propagating play some role in precipitation variability in the eastern convective and zonal wind signals with a zonal wave- coastal area during the monsoon transition phase. As number of 1 or 2. Its characteristics exhibit pronounced will be shown in this paper, the following two atmo- seasonality (e.g., Knutson and Weickmann 1987). Dur- spheric disturbances played important roles in the ing boreal summer, the southwesterly monsoon winds NOV99 event: a cold surge (CS) and a tropical depres- and other seasonal conditions in south and Southeast sion–type disturbance (TDD) associated with tropical Asia modify the MJO structure and behavior (Lau and SEPTEMBER 2008 YOKOI AND MATSUMOTO 3277

Ϫ1 Ϫ FIG. 3. TRMM 3B42 surface precipitation [mm (2 days) ]on FIG. 4. Daily precipitation [mm (day) 1] at Hue in 2–3 Nov 1999. October–November 1999.

Chan 1986; Lawrence and Webster 2002). The convec- 2 describes the datasets used in this study. The structure tive signal propagates northward in the Asian monsoon and behavior of the atmospheric disturbances during region and western Pacific, as well as eastward along the NOV99 event are presented in section 3, and com- the equator in the Eastern Hemisphere (Yasunari 1979; posite analyses are discussed in section 4. Section 5 Murakami and Nakazawa 1985; Krishnamurti et al. presents a summary and discussion. 1985; Lau and Chan 1986). The convective and low- level zonal wind anomalies in south and Southeast 2. Data Asian monsoon regions form a band structure oriented from a west-northwest to east-southeast direction (Mu- The atmospheric circulation field dataset used in this rakami and Nakazawa 1985; Lau and Chan 1986). Cy- study is the Japanese 25-yr Reanalysis (JRA-25; Onogi clonic circulation anomalies aligning along this band et al. 2007) provided by Japan Meteorological Agency move northwestward because of Rossby wave dynamics (JMA) and Central Research Institute of Electric (Kemball-Cook and Wang 2001). Power Industry (CRIEPI). Its horizontal resolution is Although these two atmospheric disturbances have 1.25° in longitude and latitude, and its temporal interval been studied in many papers as introduced heretofore, is 6 h. There are 12 mandatory levels in the tropical the disturbances that occur during the monsoon transi- troposphere (1000, 925, 850, 700, 600, 500, 400, 300, 250, tion phase have yet to be described in detail. Most of 200, 150, and 100 hPa). This paper will focus primarily the studies dealing with CSs have targeted the mature on the 925-hPa wind field. Note that there are seven winter monsoon season, although CSs occur most fre- levels below the nominal 850-hPa level in the JRA-25 quently in November (Zhang et al. 1997), and the CSs assimilation model, so the 925-hPa level is well re- that occur during the monsoon transition phase exhibit solved. To eliminate the diurnal cycle, a 1–2–2–2–1 tem- features that are different from those during the winter poral filter is applied to the reanalysis data in section 3 monsoon season (Compo et al. 1999). In addition, and a daily average is used in section 4. The analysis is some studies on ISV exclude the monsoon transition performed over the 24-yr period from 1979 to 2002. To phase in order to present its exemplary features. describe synoptic-scale disturbances, we also analyze However, for understanding the causal mechanisms of anomaly data, which are constructed by removing the heavy rainfall events in central Vietnam, it is necessary first three harmonics of the climatological seasonal to study the behavior of disturbances during the mon- cycle. soon transition phase. Therefore, the purpose of this This study also uses daily station rainfall data in the study is to describe the synoptic-scale atmospheric con- eastern ICP. These data were routinely collected by the ditions and disturbances during the NOV99 event in Department of Meteorology and Hydrology of Laos (at detail, and to discuss the mechanisms responsible for 8 stations), the Department of Meteorology of Cambo- heavy rainfall in central Vietnam. To confirm the dia (at 4 stations), and the National Hydrometeorolog- mechanisms, we will also conduct composite analyses ical Service of Vietnam (at 52 stations). using 24 yr (1979–2002) of reanalysis and precipitation In addition, daily mean interpolated outgoing long- data. wave radiation (OLR) data provided by the National The rest of this paper is organized as follows: Section Oceanic and Atmospheric Administration (NOAA; 3278 MONTHLY WEATHER REVIEW VOLUME 136

Liebmann and Smith 1996) and routinely observed upper-air sounding data at Xisha Dao station (16.83°N, 112.33°E; Fig. 1) are utilized as supplements. The horizontal resolution of the OLR data is 2.5° in longitude and latitude. The sounding data have a temporal interval of 12 h and are archived at University of Wyoming (available online at http://weather.uwyo.edu/upperair/ sounding.html).

3. Synoptic-scale atmospheric conditions during NOV99 In this section, we will examine the synoptic-scale circulation field in the lower troposphere during the NOV99 event. The horizontal wind field at the 925-hPa level at 1200 UTC 2 November (Fig. 5a) was characterized by a strong northerly wind blowing from the Yellow Sea into the northeastern part of the ICP along the eastern coast of the Eurasian continent. While the wind direction was considerably similar to the climatological northeasterly monsoon wind in early November (Fig. 5b), the wind speed over the northern SCS during the NOV99 event was 22 m sϪ1, which was twice as strong as the climatological wind speed of 11 m sϪ1. The equivalent potential temperature field (Fig. 5c) shows that a warm and humid tropical air mass existed south of the northern SCS and a front-like zone with a large horizontal gradient lay in a west-southwest to east- northeast direction. A part of this zone was also reported as a stationary front in a surface weather chart issued by JMA, which is represented by a thick solid line in Fig. 5a. Because the maximum wind speed was observed very close to the 335-K iso-equivalent potential temperature line, the stronger-than-usual northeasterly wind transported relatively thermally unstable air into the ICP, creating conditions capable of activat- ing orographic rainfall processes over and upstream of the Annam Range. The stronger-than-usual northeasterly wind was associated with a CS from the midlatitudes. A time–latitude cross section of meridional wind at the 925-hPa level averaged over 110°–120°E (Fig. 6a) shows that a FIG. 5. (a) Horizontal wind (vector) and its speed (shading) at Ϫ1 northerly wind signal stronger than Ϫ6ms originated the 925-hPa level at 1200 UTC 2 Nov 1999. Thick solid line indi- near 40°N at 0600 UTC 31 October, which was associ- cates the location of the front analyzed in the surface weather ated with a trough–ridge system aloft (figure not chart issued by JMA. Reference vector corresponds to value of 20 msϪ1. (b) Same as in (a), but for the climatological wind field on shown). The signal propagated rapidly southward to Ϫ 2 Nov. Reference vector corresponds to value of 10 m s 1. (c) 25°N within 18 h, and then slowly reached the northern Equivalent potential temperature at the 925-hPa level at 1200 SCS south of 20°N by 0000 UTC 2 November. During UTC 2 Nov 1999. Contour interval is 5 K, and shading indicates the southward propagation, the northerly wind signal values higher than 340 K. The blacked-out areas are where the was accompanied to its south by a horizontal conver- 925-hPa level is below the ground. gence anomaly (Fig. 6b) and to its north by a negative temperature anomaly (Fig. 6c). These characteristics are consistent with well-known CS features derived by SEPTEMBER 2008 YOKOI AND MATSUMOTO 3279

FIG. 6. (a) Time–latitude cross section of meridional wind at the 925-hPa level averaged in 110°–120°E; contour interval is 2 m sϪ1, and dark (light) shading indicates values higher than ϩ6msϪ1 (lower than Ϫ6msϪ1). (b) Same as (a), except for the horizontal divergence anomaly; contour interval is 5 ϫ 10Ϫ6 sϪ1, and dark (light) shading indicates values lower than Ϫ1 ϫ 10Ϫ5 sϪ1 (0 sϪ1). (c) The same as in (a), but for temperature anomaly; contour interval is 1 K, and dark (light) shading indicates values higher than ϩ1 K (lower than Ϫ1 K). (d) The same as in (a), but for zonal wind; contour interval is 2.5 m sϪ1, and shading indicates values lower than Ϫ10 m sϪ1. many researchers, as given in section 1. It is worth not- anomaly with a 6 m sϪ1 magnitude was found over the ing that over the SCS only a weak convergence central SCS centered at 11.25°N, 112.5°E. This south- anomaly was observed at the 850-hPa level, whereas a erly wind anomaly seemed to prevent the CS from strong divergence anomaly was found to the west of the propagating farther southward, possibly through chang- 925-hPa convergence (figure not shown). This seemed ing the dynamic balance on the meridional propagation to represent the thin vertical structure of the CS (Chen of the CS. Because the southward propagation speed of et al. 2002). the CS in 20°–25°N can be estimated from Fig. 6a at One of the distinctive features of this CS was ob- about 5 m sϪ1, the southerly wind anomaly of 6 m sϪ1 served when the northerly wind signal reached the could block the southward propagation. However, it is northern SCS. While ordinary CS signals propagate far- difficult to discuss the mechanism in more detail, be- ther southward to at least 10°N, and sometimes across cause there is no consensus regarding the dynamics of the equator (Chang et al. 1983; Love 1985; Wu et al. the southward propagation of CS signals in the subtrop- 2007), the signal of this CS stopped propagation and ics. lingered there until 0000 UTC 5 November (Fig. 6a). In addition to the blocking effect mentioned above, Concurrently, an easterly wind began to strengthen at the southerly wind anomaly seemed to contribute to the the same latitude (Fig. 6d). As a result of this lingering heavy rainfall in the following two ways: First, the feature, stronger-than-usual northeasterly winds con- anomaly formed a strong horizontal convergence in the tinued for a couple of days. lower troposphere over the northern SCS (Fig. 6b) in To discuss the reason for this lingering feature, Fig. collaboration with the northeasterly wind of the CS. 7a shows the meridional wind anomaly at the 925-hPa Second, the anomaly tended to transport the warm and level at 1200 UTC 2 November. A southerly wind humid tropical air into the northern SCS, particularly 3280 MONTHLY WEATHER REVIEW VOLUME 136

tology. The differences in the flux at the southern boundary and convergence were primarily due to the advection of the climatological equivalent potential temperature by the southerly wind anomaly, shown in the last row of Table 1. To confirm the unstable condition over central Viet- nam, Fig. 8 shows a time series of convective available potential energy (CAPE) at Xisha Dao station (Fig. 1), calculated from the sounding data. During 29–30 Octo- ber, CAPE fluctuated around 900 J kgϪ1; then CAPE increased dramatically on 31 October to reach a maximum of more than 2000 J kgϪ1, and maintained high values until 2 November. This increase of CAPE rep- resents that the atmosphere became more unstable. On 3 November, CAPE fell to nearly zero and the atmosphere became stable, probably because of the ap- proach of the CS bringing cold air into the lower troposphere. This southerly wind anomaly was a component of a cyclonic circulation anomaly centered at the southeast- ern ICP, which is marked as B in Fig. 7b. This circulation anomaly had a horizontal scale of about 2000 km and appeared to combine with the northeasterly wind FIG. 7. (a) Meridional wind anomaly (positive anomaly shaded) anomaly of the CS in its northeastern sector. The ver- and (b) horizontal wind anomaly vector at the 925-hPa level at tical structure of this circulation anomaly was almost Ϫ1 1200 UTC 2 Nov 1999. Contour interval in (a) is 2 m s , and the standing below the 300-hPa level, with a slight inclina- reference vector in (b) corresponds to value of 10 m sϪ1. The blacked-out areas are the same as in Fig. 5. tion to the west with height (figure not shown). The longitude–time cross section of the meridional wind anomaly (Fig. 9) shows that the southerly wind prior to the arrival of the CS. These effects made the anomaly had appeared at 140°E by 28 October and lower troposphere more unstable and favorable for then propagated westward to reach 110°E on 2 Novem- convection. To confirm the second effect, Table 1 ber. These spatiotemporal structures are considerably shows horizontal fluxes of equivalent potential tem- similar to that of the TDD (Takayabu and Nitta 1993). perature at the 925-hPa level across the boundaries of a In addition to disturbance B, there were two other 10°–20°N, 110°–120°E box in the northern SCS. Be- TDDs: one was over the eastern coast of the Indian cause of the northeasterly monsoon wind, climatologi- subcontinent and the other was southeast of the Phil- cal flux is inward at the northern and eastern bound- ippines (marked as A and C, respectively, in Fig. 7b). aries and outward at the other two boundaries. As for The southerly wind anomaly associated with distur- the NOV99 event, on the other hand, the flux at the bance C had appeared by 1 November and propagated southern boundary was inward and horizontal conver- westward (Fig. 9). Likewise, disturbance A had been gence in the box was stronger than that of the clima- generated in the southern SCS by 28 October and prop-

TABLE 1. Horizontal fluxes of equivalent potential temperature at the 925-hPa level passing across northern, southern, eastern, and western boundaries of, and horizontal convergence integrated over, the 10°–20°N, 110°–120°E box in units of 108 Km2 sϪ1. Values of early November climatology and just before the NOV99 event (31 Oct–1 Nov average) are shown. Positive values indicate inward flux and convergence. The difference between the NOV99 event and climatology, and fluxes of the climatological equivalent potential Ј ␪ temperature by the wind anomaly (u e) and resultant convergence anomaly are also shown.

Northern Southern Eastern Western Convergence Climatology ϩ14.0 Ϫ5.5 ϩ21.5 Ϫ22.4 ϩ7.6 NOV99 ϩ7.5 ϩ9.0 ϩ20.8 Ϫ24.3 ϩ13.0 NOV99 Ϫ climatology Ϫ6.5 ϩ14.5 Ϫ0.7 Ϫ1.9 ϩ5.4 Ј ␪ Ϫ ϩ Ϫ Ϫ ϩ u e 6.5 14.4 0.8 1.7 5.4 SEPTEMBER 2008 YOKOI AND MATSUMOTO 3281

Ϫ1 FIG. 9. Longitude–time cross section of meridional wind FIG. 8. A time series of CAPE (J kg ) at Xisha Dao station. anomaly at the 925-hPa level averaged in 7.5°–12.5°N. Contour interval is 1.5 m sϪ1, and anomalies higher than ϩ3msϪ1 are shaded. agated northwestward along the southern coast of the ICP. These three disturbance centers aligned in a west- northwest to east-southeast direction, accompanied by shown in Fig. 7b can be regarded as a result of an a westerly wind anomaly band to their south (Fig. 7b). off-equatorial Rossby wave response to the equatorial In the equatorial region, the large-scale circulation convective anomaly of the ISV, as discussed in field was characterized by a MJO-type signal. The ve- Kemball-Cook and Wang (2001). locity potential anomaly field at the 150-hPa level (Fig. The collaborative effects of the CS and TDD associ- 10a) exhibited a zonal wavenumber 1 structure. Its ated with the equatorial ISV can be summarized as negative anomaly was found in Southeast Asia and the follows: Because of these two disturbances, the lower western Pacific, with a negative maximum over the troposphere over central Vietnam became thermally Maritime Continent. This implies the presence of a cen- unstable. The stronger-than-usual northeasterly wind ter of large-scale divergence and an upward branch of a caused by the CS blew this unstable air into central tropical circulation anomaly. Accompanying this veloc- Vietnam against the Annam Range, resulting in vigor- ity potential anomaly was a negative OLR anomaly, ous orographic rainfall processes. Because of the lin- which indicates active convection. These negative gering feature of the CS, which was probably caused by anomalies originated over the Indian Ocean in late Sep- the TDD, the northeasterly wind lasted for a couple of tember and propagated eastward at a speed of about 3 days, leading to a large amount of precipitation. msϪ1, with a slight weakening in amplitude just east of 90°E (Fig. 10b). These characteristics are consistent 4. Composite analyses with those of a typical MJO, as reported by many studies (e.g., Murakami and Nakazawa 1985; Krishnamurti To confirm the importance of the collaborative ef- et al. 1985; Weickmann and Khalsa 1990). To the east of fects for the heavy rainfall in central Vietnam, which 150°E, the negative OLR anomaly vanished and the were suggested through the case study in section 3, we negative velocity potential anomaly weakened, while will perform composite analyses of CSs and TDDs the latter propagated farther eastward in the Western found during the monsoon transition phase. Hemisphere at a faster speed of about 13 m sϪ1. The There have been a variety of CS definitions proposed differences in amplitude and propagation speed be- by previous studies, most of which were based on a tween the Eastern and Western Hemispheres are also sharp drop in the surface or lower-tropospheric tem- well-known MJO features (Knutson et al. 1986; Hen- perature, meridional pressure gradient, strong north- don and Salby 1994; Kikuchi and Takayabu 2003). The erly wind, or a combination thereof. Because we focus relationship between the MJO-type convective signal on the CS behavior over the SCS, we set the CS defi- centered over the Maritime Continent and the low- nition as follows: 1) a 3-day mean meridional wind level circulation anomaly field in the subtropics shown anomaly at the 925-hPa level averaged over 110°–120°E in Fig. 7 was in good agreement with that of the ISV along 20°N is negatively larger than Ϫ3.1 m sϪ1 (Ϫ1 ϫ during boreal summer (Knutson and Weickmann 1987; standard deviation) and is a temporal minimum, and 2) Kemball-Cook and Wang 2001). This accordance sug- a 3-day mean temperature anomaly at the 850-hPa level gests that the TDD train in the lower troposphere averaged over 105°–115°E along 25°N records a tem- 3282 MONTHLY WEATHER REVIEW VOLUME 136

FIG. 10. (a) Velocity potential anomaly at the 150-hPa level (contour) and OLR anomaly (shading) averaged over 11-day period from 28 Oct to 7 Nov 1999. Contour interval is 3 ϫ 106 m2 sϪ1, and dark (light) shading indicates OLR anomalies lower than Ϫ20WmϪ2 (Ϫ10WmϪ2). (b) Longitude–time cross section of 11-day running mean velocity potential anomaly at the 150-hPa level and OLR anomaly averaged in 10°S–10°N. Contour interval and shading are same as in (a).

FIG. 11. Lag–latitude cross section of composite meridional poral minimum in the 5-day period from 2 days before wind anomaly averaged in 110°–120°E of (a) JF-CS and (b) ON- Ϫ1 to 2 days after the meridional wind minimum. The sec- CS at the 925-hPa level. Contour interval is 1 m s , with zero ond condition can exclude strong TDDs and typhoons contour omitted. Dark (light) shading indicates positive (negative) anomalies with statistical significance at a 99% confidence over the western Pacific or Philippines that do not ac- level. company CSs, which can also cause northerly wind anomalies over the northern SCS. In the 24-yr (1979– 2002) October–November periods, we identify 45 CS lation fields) of both CS groups are also qualitatively cases (hereafter ON-CS). Although the CS during the similar to each other. A notable difference between NOV99 event also meets the above definition, we ex- JF-CS and ON-CS is found to the south of 15°N. While clude this case from composite analyses. In addition, we the JF-CS northerly wind signal propagates farther identify 40 CS cases in the 24-yr January–February pe- southward to near the equator, the ON-CS meridional riods (hereafter JF-CS) for evaluating the seasonality of wind anomaly hardly exhibits any statistically signifi- CS features. cant signal south of 10°N. To examine the difference in Figure 11 shows the lag–latitude cross section of the more detail, the composite meridional wind anomaly composite meridional wind anomaly at the 925-hPa averaged over lags from ϩ1toϩ3 days is shown in Figs. level averaged over 110°–120°E. In this figure, a lag of 12a,b. In the JF-CS composite (Fig. 12a), a northerly zero days corresponds to the day when the CSs satisfy wind anomaly with statistical significance is extensively the first condition of the definition. Composite merid- found over the SCS and western Pacific; this is consis- ional wind anomalies of JF-CS and ON-CS over the tent with previous studies, such as that of Zhang et al. Eurasian continent (north of 20°N) are qualitatively (1997). On the other hand, as for the ON-CS composite similar to each other; statistically significant northerly (Fig. 12b), a significant northerly wind anomaly is wind anomalies propagate southward from the midlati- found only over northern SCS and the ICP, whereas the tudes. Other atmospheric conditions (e.g., temperature anomaly over most of the central and southern SCS, and pressure anomalies and upper-tropospheric circu- where we found the southerly wind anomaly during the SEPTEMBER 2008 YOKOI AND MATSUMOTO 3283

FIG. 13. Composite horizontal wind anomaly of CS-SW at the 925-hPa level. Reference vector corresponds to value of 5 m sϪ1. The blacked-out areas are the same as in Fig. 5.

These results imply that the direction as well as the magnitude of the meridional wind anomaly in the central SCS is different among individual ON-CS cases. In fact, the meridional wind anomaly averaged over 10°– ␷Ј Ϫ 15°N, 110°–115°E (hereafter cscs) ranges from 12 to ϩ8msϪ1. ␷Ј To clarify the relationship between cscs and precipitation over the ICP, we categorize the 45 ON-CS cases ␷Ј into the following three groups: CSs where cscs is negatively larger than Ϫ7msϪ1 [CS with strong northerly ␷Ј wind (CS-NW); 6 cases], CSs where cscs is positively larger than ϩ4msϪ1 [CS with southerly wind (CS-SW); 6 cases], and other CSs (CS-Normal; 33 cases). The composite horizontal wind anomaly over and around the SCS for the CS-SW at the 925-hPa level (Fig. 13) is quite similar to Fig. 7b, showing a TDD centered over the central SCS and a northeasterly wind anomaly blowing into the ICP. Low-level horizontal convergence and the northward transport of the warm and humid tropical air over the northern SCS are also observed as in the NOV99 event. On the other hand, the CS-NW and CS-Normal composites have horizontal wind fields that are qualitatively similar to those of the JF-CS composite over and around the SCS, although FIG. 12. (a) Composite meridional wind anomaly of JF-CS at the 925-hPa level averaged over lags of ϩ1toϩ3 days. Contour the wind speed is different. In addition, horizontal con- interval is 1 m sϪ1, and anomalies with statistical significance at vergence at the 925-hPa level over central Vietnam is a 99% confidence level are shaded. (b) The same as in (a), but not found in these two composites. for ON-CS. (c) Standard deviation of meridional wind anomaly Figures 14a–c show the composite precipitation in of ON-CS cases. Contour interval is 1 m sϪ1, and values greater Ϫ1 eastern ICP of the three above-mentioned groups av- than4ms are shaded. The blacked-out areas are the same as in ϩ ϩ Fig. 5. eraged over lags from 1to 3 days. Composite precipitation of CS-SW (Fig. 14c) is more than 60 mm dayϪ1 in the coastal areas between 13° and 16°N, which NOV99 event, is not statistically significant. In addi- is much higher than that of CS-NW (Fig. 14a) and CS- tion, the standard deviation of the meridional wind Normal (Fig. 14b). In particular, precipitation averaged anomaly of the ON-CS cases (Fig. 12c) exhibits a local over the nine coastal stations shown in Fig. 14f is 67.4 maximum in the central SCS; this is consistent with the mm dayϪ1 for CS-SW, which is about 3 times higher statistical insignificance of the composite anomaly. than that for CS-NW (18.9 mm dayϪ1) and CS-Normal 3284 MONTHLY WEATHER REVIEW VOLUME 136

Ϫ FIG. 14. (a) Composite station precipitation (mm day 1) of CS-NW, averaged over lags of ϩ1toϩ3 days. Legend denoting values for circle size is at the right of this figure. Same as (a), except for (b) CS-Normal, (c) CS-SW, and (d) SW-alone, and averaged over lags of Ϫ1toϩ1 day. (e) Climatological precipitation amount during October–November period (mm dayϪ1). (f) Stations used for quantitative comparison of composite precipitation amount.

(24.1 mm dayϪ1); this difference is statistically signifi- The NOV99 event described in section 3 is a typical cant at a 99% confidence level. In addition, composite example of such a collaborative effect. precipitation of CS-NW and CS-Normal is only compa- It should be mentioned that one-half of the CS-SW rable to the October–November climatology (Fig. 14e). cases are associated with the equatorial ISV, which has Thus, we can conclude that a CS cannot independently the upward branch and convective anomaly over the cause heavy rainfall in the eastern coast of the ICP. Maritime Continent, as in the NOV99 event. There- To examine the importance of the CS, we compare fore, the equatorial ISV is one of the major candidates, precipitation of CS-SW to that under the condition of a but of course not a necessary condition, for the genera- southerly wind anomaly over the central SCS that does tion and maintenance of the TDD that induces south- not accompany a CS. We identify 62 such cases where erly wind anomalies over central SCS. ␷Ј ϩ Ϫ1 cscs is higher than 4ms and is a temporal maximum, and there is no CS over the SCS and China (here- 5. Summary and discussion after SW-alone). Composite horizontal wind and geo- potential height anomaly maps of SW-alone at the 925- This study revealed synoptic-scale atmospheric dis- hPa level (not shown) indicate that the southerly wind turbances in the East and Southeast Asian regions as- anomaly is a component of a TDD centered over the sociated with the record heavy rainfall event that oc- central SCS. Composite precipitation of SW-alone (Fig. curred in central Vietnam on 2–3 November 1999 14d) is less than that of CS-SW, although it is much (NOV99 event). During NOV99, a lower-tropospheric higher than that of CS-NW and CS-Normal. In fact, northeasterly monsoon wind over the northern SCS precipitation averaged over the nine stations in the was strengthened by a CS and a TDD. The CS north- coastal areas (Fig. 14f) for SW-alone is 40.8 mm dayϪ1, erly wind signal originated near 40°N in northern which is less than that for CS-SW with statistical sig- China, propagated southward along the eastern coast of nificance at a 95% confidence level. the Eurasia continent, and then lingered over the Therefore, we can conclude that the coexistence of northern SCS for a couple of days. A southerly wind the CS and the southerly wind anomaly associated with anomaly in the central SCS associated with the TDD the TDD is a more favorable condition for heavy rain- seemed to play a role in this lingering feature. In addi- fall in the eastern coastal areas of the ICP than the tion, this anomaly formed a strong convergence over existence of either the CS or the southerly anomaly. central Vietnam in association with the CS northeast- SEPTEMBER 2008 YOKOI AND MATSUMOTO 3285 erly wind anomaly, and supplied warm and humid tropical air that destabilized the lower atmosphere. As a consequence, stronger-than-usual northeasterly winds transported the unstable air into central Vietnam for a couple of days, resulting in vigorous orographic rainfall processes over and upstream of the Annam Range. We performed composite analyses to confirm that the coexistence of the CS and TDD is important for heavy rainfall. In the October–November periods of 1979–2002, we identified 6 CSs that accompanied the strong northerly wind anomalies over the central SCS (CS-NW), 6 CSs that accompanied the southerly wind anomalies (CS-SW), and 33 CSs for which meridional FIG. 15. Climatological variance of 3–20-day filtered relative wind anomalies were between the former two extreme vorticity at the 925-hPa level averaged in 5°–15°N, 110°–120°E, cases (CS-Normal). The CS-SW corresponded to the calculated on a monthly basis. The variance has been normalized coexistence cases of the CS and TDD. We also identi- by variance of nonfiltered relative vorticity. fied 62 cases in which a southerly wind anomaly associated with a TDD existed over the central SCS but did not accompany a CS (SW-alone). We showed that com- of sensible and latent heat to the lower troposphere, causing the convective precipitation processes to be posite precipitation of CS-SW was much more than that more vigorous. Quantitative estimates are necessary for of CS-NW, CS-Normal, and SW-alone, thus confirming discussing this subject and will be dealt with in our the significance of the collaborative effects of the CS future study. and TDD. In addition, we found that CSs without TDD As presented in section 4, we found that in the Oc- could not cause much precipitation; composite precipi- tober–November period, meridional wind anomalies tation of CS-NW and CS-Normal was only comparable over the central SCS associated with CSs differ mark- to the October–November climatological precipitation edly among individual cases because of the existence amount. and location of TDDs, whereas most CSs in the Janu- The TDD during the NOV99 event could be re- ary–February period are accompanied by a northerly garded as the result of the Rossby wave response to the wind anomaly over the central SCS and seem to be less large-scale convective anomaly over the Maritime Con- subject to the TDD. This seasonality can be attributed tinent associated with the equatorial ISV. In addition, to the seasonal march in the activity of the TDD over the TDDs in one-half of the CS-SW cases also seemed the central and southern SCS. As a proxy for this ac- to be associated with the ISV convective anomaly. tivity, Fig. 15 shows variance of the 3–20-day filtered Therefore, the ISV is one of the major candidates for relative vorticity, which is calculated with the use of a the generation and maintenance of the TDD that con- Lanczos bandpass filter (Duchon 1979), which has cut- tributes to the collaborative effect. off periods of 3 and 20 days and a filter length of 41 In addition to the collaborative effect, other factors days, at the 925-hPa level averaged in 5°–15°N, 110°– may contribute to the occurrence of heavy rainfall. This 120°E. The variance is highest in October and lowest in may be explained by taking into account the fact that January–March. Because TDDs occur more frequently the amount of precipitation during the NOV99 event in environmental trough zones, the seasonal march in was the highest in the previous 50 yr, as mentioned in the variance is consistent with the southward shift of section 1. In addition, precipitation averaged over the the northeasterly monsoon wind and monsoon trough nine coastal stations shown in Fig. 14f during 2–4 No- shown in Fig. 2. Because the CS occurs frequently dur- Ϫ vember 1999 was 203 mm day 1, which is 3 times higher ing November–March, we can conclude that the col- Ϫ than that for the CS-SW composite (67.4 mm day 1). laborative effect of the CS and TDD studied in this One possible factor is an anomalous sea surface tem- paper is peculiar to the October–November period. perature (SST) in the central SCS. The NOAA Opti- Compo et al. (1999) noted that the composite structure mum Interpolation SST (OISST) dataset (Reynolds et of the CS during the October–November period is dif- al. 2002) reveals that the SST in the northern and cen- ferent from that during December–March. The differ- tral SCS during the NOV99 event was higher (by a ence may possibly be attributable to the diversity in the maximum of 1 K) than the climatological value in early meridional wind anomaly over the central and southern November. The warmer SST supplies higher amounts SCS. To fully understand the behavior of the CS during 3286 MONTHLY WEATHER REVIEW VOLUME 136 the October–November period, we also have to con- Hendon, H. H., and M. L. Salby, 1994: The life cycle of the Mad- sider the effect of TDDs. den–Julian oscillation. J. Atmos. Sci., 51, 2225–2237. The interaction of CSs with tropical atmospheric sys- Huffman, G. J., and Coauthors, 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, tems during the winter monsoon season was also re- combined-sensor precipitation estimates at fine scales. J. Hy- ported by Chang et al. (2005), who emphasized the ef- drometeor., 8, 38–55. fects of the Borneo vortex whose center was located Johnson, R. H., and D. L. Priegnitz, 1981: Winter monsoon con- south of 10°N. They showed that the horizontal distri- vection in the vicinity of north Borneo. Part II: Effects on bution of cloud activity over the southern SCS, Malay large-scale fields. Mon. Wea. Rev., 109, 1615–1628. Peninsula, and Maritime Continent during CS events Joung, C. H., and M. H. Hitchman, 1982: On the role of successive downstream development in East Asian polar air outbreaks. depended on the existence of the Borneo vortex. The Mon. Wea. Rev., 110, 1224–1237. results of Chang et al. (2005) and the present study Kemball-Cook, S., and B. Wang, 2001: Equatorial waves and air– suggest that tropical cyclonic vortices, regardless of sea interaction in the boreal summer intraseasonal oscilla- their latitude, considerably affect the distribution and tion. J. Climate, 14, 2923–2942. strength of convection and precipitation induced by CSs. Kikuchi, K., and Y. N. Takayabu, 2003: Equatorial circumnaviga- tion of moisture signal associated with the Madden–Julian oscillation (MJO) during boreal winter. J. Meteor. Soc. Japan, Acknowledgments. The authors are grateful to the 81, 851–869. Department of Meteorology and Hydrology of Laos, Knutson, T. R., and K. M. Weickmann, 1987: 30–60 day atmo- the Department of Meteorology of Cambodia, and the spheric oscillations: Composite life cycles of convection and National Hydrometeorological Service of Vietnam for circulation anomalies. Mon. Wea. Rev., 115, 1407–1436. providing surface rainfall data. This study was finan- ——, ——, and J. E. Kutzbach, 1986: Global-scale intraseasonal cially supported by the Japan Earth Observation Sys- oscillations of outgoing longwave radiation and 250 mb zonal wind during Northern Hemisphere summer. Mon. Wea. Rev., tem Promotion Program (No. 22-02, leader: Jun Mat- 114, 605–623. sumoto) and a Grant-in-Aid for Young Scientists (B- Krishnamurti, T. N., P. K. Jayakumar, J. Sheng, N. Surgi, and A. 19740284) from the Japanese Ministry of Education, Kumar, 1985: Divergent circulations on the 30 to 50 day time Culture, Sports, Science and Technology, and by Core scale. J. Atmos. Sci., 42, 364–375. Research for Evolutional Science and Technology at Lau, K.-M., and P. H. Chan, 1986: Aspects of the 40-50-day oscillation during the northern summer as inferred from outgoing the Japan Science and Technology Agency (leader: longwave radiation. Mon. Wea. Rev., 114, 1354–1367. Prof. Masakazu Suzuki of the University of Tokyo). All ——, and K.-M. Lau, 1984: The structure and energetics of mid- figures were produced with use of the GFD-DENNOU latitude disturbances accompanying cold-air outbreaks over Library. East Asia. Mon. Wea. Rev., 112, 1309–1327. Lawrence, D. M., and P. J. Webster, 2002: The boreal summer REFERENCES intraseasonal oscillation: Relationship between northward and eastward movement of convection. J. Atmos. Sci., 59, Chang, C.-P., J. E. Erickson, and K. M. Lau, 1979: Northeasterly 1593–1606. cold surges and near-equatorial disturbances over the Winter Liebmann, B., and C. A. Smith, 1996: Description of a complete MONEX area during December 1974. Part I: Synoptic as- (interpolated) outgoing longwave radiation dataset. Bull. pects. Mon. Wea. Rev., 107, 812–829. Amer. Meteor. Soc., 77, 1275–1277. ——, J. E. Millard, and G. T. J. Chen, 1983: Gravitational char- Love, G., 1985: Cross-equatorial influence of winter hemisphere acter of cold surges during winter MONEX. Mon. Wea. Rev., subtropical cold surges. Mon. Wea. Rev., 113, 1487–1498. 111, 293–307. Madden, R. A., and P. R. Julian, 1994: Observations of the 40– ——, P. A. Harr, and H.-J. Chen, 2005: Synoptic disturbances 50-day tropical oscillation—A review. Mon. Wea. Rev., 122, over the equatorial South China Sea and western Maritime 814–837. Continent during boreal winter. Mon. Wea. Rev., 133, 489– 503. Matsumoto, J., 1997: Seasonal transition of summer rainy season Chen, T.-C., M.-C. Yen, W.-R. Huang, and A. G. William Jr., over Indochina and adjacent monsoon region. Adv. Atmos. 2002: An East Asian cold surge: Case study. Mon. Wea. Rev., Sci., 14, 231–245. 130, 2271–2290. Murakami, T., and T. Nakazawa, 1985: Tropical 45 day oscilla- Chu, P.-S., and S.-U. Park, 1984: Regional circulation character- tions during the 1979 Northern Hemisphere summer. J. At- istics associated with a cold surge event over East Asia during mos. Sci., 42, 1107–1122. Winter MONEX. Mon. Wea. Rev., 112, 955–965. Onogi, K., and Coauthors, 2007: The JRA-25 Reanalysis. J. Me- Compo, G. P., G. N. Kiladis, and P. J. Webster, 1999: The hori- teor. Soc. Japan, 85, 369–432. zontal and vertical structure of East Asian winter monsoon Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. pressure surges. Quart. J. Roy. Meteor. Soc., 125, 29–54. Wang, 2002: An improved in situ and satellite SST analysis Ding, Y., 1990: Build-up, air mass transformation and propagation for climate. J. Climate, 15, 1609–1625. of Siberian High and its relation to cold surge in East Asia. Sumi, A., 1985: A study on cold surges around the Tibetan Plateau Meteor. Atmos. Phys., 44, 281–292. by using numerical models. J. Meteor. Soc. Japan, 63, 377– Duchon, C. E., 1979: Lanczos filtering in one and two dimensions. 396. J. Appl. Meteor., 18, 1016–1022. Takayabu, Y. N., and T. Nitta, 1993: 3–5 day-period disturbances SEPTEMBER 2008 YOKOI AND MATSUMOTO 3287

coupled with convection over the Tropical Pacific Ocean. J. tion of repeated torrential rains over Java Island. SOLA, 3, Meteor. Soc. Japan, 71, 221–246. 93–96. Weickmann, K. M., and S. J. S. Khalsa, 1990: The shift of convec- Yasunari, T., 1979: Cloudiness fluctuations associated with the tion from the Indian Ocean to the western Pacific Ocean Northern Hemisphere summer monsoon. J. Meteor. Soc. Ja- during a 30–60 day oscillation. Mon. Wea. Rev., 118, 964–978. pan, 57, 227–242. Yokoi, S., and T. Satomura, 2005: An observational study of in- Wu, M. C., and J. C. L. Chan, 1995: Surface features of winter traseasonal variations over Southeast Asia during the 1998 monsoon surges over South China. Mon. Wea. Rev., 123, 662– rainy season. Mon. Wea. Rev., 133, 2091–2104. 680. ——, ——, and J. Matsumoto, 2007: Climatological characteristics ——, and ——, 1997: Upper-level features associated with winter of the intraseasonal variation of precipitation over the In- monsoon surges over South China. Mon. Wea. Rev., 125, 317– dochina Peninsula. J. Climate, 20, 5301–5315. 340. Zhang, Y., K. R. Sperber, and J. S. Boyle, 1997: Climatology and Wu, P., M. Hara, H. Fudeyasu, M. D. Yamanaka, J. Matsumoto, interannual variation of the East Asian winter monsoon: Re- F. Syamsudin, R. Sulistyowati, and Y. S. Djajadihardja, 2007: sults from the 1979–95 NCEP/NCAR reanalysis. Mon. Wea. The impact of trans-equatorial monsoon flow on the forma- Rev., 125, 2605–2619.