The Central Highlands are Vietnam’s main coffee growing region. Unusual wet spells during the early dry season in November and December negatively affect two growing cycles in terms of yield and quality. The meteorological causes of wet spells in this region have not been thoroughly studied to date. Using daily rain gauge measurements at nine stations for the period 1981–2007 in the Central Highlands, four dynamically different early dry-season rainfall cases were investigated in depth: 1) the tail end of a cold front, 2) a tropical depression–type disturbance, 3) multiple tropical wave interactions, and 4) a cold surge with the Borneo vortex.
Cases 1 and 4 are mainly extratropically forced. In case 1, moisture advection ahead of a dissipating cold front over the South China Sea led to high equivalent potential temperature in the southern highland where this air mass stalled and facilitated recurrent outbreaks of afternoon convection. In this case, the low-level northeasterly flow over the South China Sea was diverted around the southern highlands by relatively stable low layers. On the contrary, low-level flow was more orthogonal to the mountain barrier and high Froude numbers and concomitant low stability facilitated the westward extension of the rainfall zone across the mountain barrier in the other cases. In case 3, an eastward-traveling equatorial Kelvin wave might have been a factor in this westward extension, too. The results show a variety of interactions of large-scale wave forcings, synoptic-convective dynamics, and orographic effects on spatiotemporal details of the rainfall patterns.
In about the last 20 yr, Vietnam grew to one of the leading producers and exporters of coffee in the world. In 2013, Vietnam’s contribution to the worldwide Robusta (coffea canephora) production was about 40% with an export share of about 23% (U.S. Department of Agriculture Foreign Agricultural Service 2014), accounting for about 2% of Vietnamese export revenues (General Statistics Office of Vietnam 2016). The main coffee growing region of Vietnam are the Central Highlands, spanning from about 11° to 15.5°N and 107° to 109°E and being the southwestern part of the Southeast Asian Annamese Cordillera1 (Fig. 1). The Central Highlands are aligned parallel to the coast, are subdivided into a northern and southern part exceeding 2000 m in elevation with the Dak Lak Plateau in between (Fig. 1). The dry season in the highlands commences in November, as can be seen in Fig. 1 of Nguyen et al. (2014). Their climate region S2 corresponds to the highland region considered here. November also heralds the start of the coffee bean harvest. In the November–December period, a return of substantial rainfall negatively impacts the yields in two ways: first, the rains lead to flowering of the buds, while the ripening coffee beans of the precedent growing cycle are still on the bush (Alvim 1960; Crisosto et al. 1992). To avoid damage to the flowers, the beans are harvested prior to the optimum time. Second, the subsequent harvest is impacted since the buds are actually in need of a resting period during the dry season. Besides impacts on the cultivation of coffee and other agricultural products, heavy rainfall bears a risk of flooding and landslides.
The large socioeconomic impacts of wet spells over the Central Highlands in the early dry season lead us to thoroughly analyze the synoptic-dynamic causes of such events. While no such study covering the Vietnamese Central Highlands for this season is known to the authors, several studies (Yokoi and Matsumoto 2008; Wu et al. 2011, 2012; Chen et al. 2012a, 2012b, 2015a, 2015b) investigated the causes of extreme rainfall events along Vietnam’s central and northern coast [i.e., climate regions S1 and N4 in Nguyen et al. (2014)]. Contrary to the Vietnamese Central Highlands where an extended rainy season occurs between May and October, the peak of the rainy season in these regions is October–November and daily rainfall totals exceeding several 100 mm are not uncommon (Yokoi and Matsumoto 2008).
A well-known cause of rainfall in the South China Sea (SCS)2 area are northeasterly cold surges during the November–April winter monsoon season. They are mainly controlled by planetary-scale dynamics of the Northern Hemispheric midlatitudes, that then penetrate the tropics, lead to a surge in low-level northeasterlies over the SCS, and enhance convection over the Maritime Continent including the near-equatorial SCS. However, tropical influences on cold surges have been shown in the literature, too; Jeong et al. (2005) and Chang et al. (2005) found an interaction between cold surges and the Madden–Julian oscillation (MJO; Madden and Julian 1972), and Zhang et al. (1997) and Chen et al. (2004) showed that cold surges are also influenced by El Niño–Southern Oscillation (ENSO). This prominent type of tropical–extratropical interaction has been studied in detail, with many studies emerging after the First Global Atmospheric Research Program (GARP) Global Experiment (FGGE) in 1978/79 (e.g., Chang et al. 1979; Chang and Lau 1980; Johnson and Chang 2007, and references therein). However, the bulk of the studies concentrated on East Asia, the SCS, the Maritime Continent and the December–February (DJF) period (e.g., Johnson and Zimmerman 1986; Wu and Chan 1995; Chang et al. 2005; Ooi et al. 2011; Park et al. 2011; Koseki et al. 2014). During the DJF period, the northerly wind enhancement associated with SCS cold surges reaches at least the equator and is often associated with the formation of the Borneo vortex (Chang et al. 2005). Juneng and Tangang (2010) found that the Borneo vortex intensified during the DJF 1962–2007 period and that the centers of the vortices moved northwestward closer to the southeastern coast of Vietnam. Ooi et al. (2011) describe a January 2010 case in which the Borneo vortex moved northwestward and developed into a tropical depression affecting southern Vietnam. However, Yokoi and Matsumoto (2008) highlighted differences in cold surges occurring in October–November and January–February. Basically, early season cold surges tend to stall in the central SCS at about 10°N, where at this time of the year the ITCZ is located, whereas winter cold surges reach the equator.
Yokoi and Matsumoto (2008) and Wu et al. (2011) point to a role of westward-propagating tropical wave disturbances for heavy rainfall events along the north-central Vietnamese coast. These low-level disturbances are alternatively termed easterly waves or tropical depression (TD)–type disturbances. They are known to be involved in tropical cyclogenesis in the western Pacific (Frank and Roundy 2006). In the classical wavenumber–frequency diagram based on outgoing longwave radiation (OLR), they correspond to 2–6-day westward-propagating so-called TD-type disturbances (Kiladis et al. 2006), which have wavelengths of 2500–3500 km (Kiladis et al. 2009). The latter notation is used in the present study. Wu et al. (2012) argued that the concurrent occurrence of the convectively active part of the MJO and a TD-type disturbance led to an extreme rainfall event in central Vietnam in early October 2010. Yokoi and Matsumoto (2008) claim that the TD-type disturbances occurred as a result of a Rossby wave response to a large-scale convection anomaly over the Maritime Continent. However, multiple tropical wave interactions of the MJO, convectively coupled equatorial waves (CCEWs; Wheeler and Kiladis 1999), and TD-type disturbances on rainfall events in the Indochina Peninsula have hitherto not been studied.
Therefore, the present paper will employ both classical synoptic-dynamic and tropical large-scale wave analyses to study the evolution of early dry-season rainfall events in the Vietnamese Central Highlands. The study aims at selecting an, in terms of dynamic forcings, as diverse as possible sample of anomalous rainfall events in the period 1981–2007. It shall contribute to an improved understanding of the chain of atmospheric processes that ultimately lead to rainfall in Vietnam’s most important coffee-growing region. In section 2, data and methods are described. Section 3 discusses the four selected rainfall events and section 4 provides a summary and discussion of results.
2. Data and methods
Daily rainfall totals from 15 stations operated by the Vietnamese National Hydrometeorological Service (NHMS; NHMS 2014, unpublished data) in the Central Highlands and adjacent coastland were used (Fig. 1 and Table 1). In addition, the Asian Precipitation–Highly-Resolved Observational Data Integration Towards Evaluation of Water Resources (APHRODITE) Monsoon Asia V1101 gridded rainfall product that is based on station measurements was utilized in the 0.25° × 0.25° latitude–longitude resolution (Yatagai et al. 2012). Station data availability before 1981 and the end year of the APHRODITE product restrict the investigations period to November–December 1981–2007. The 24-h period of daily rainfall in station and APHRODITE data is 1200–1200 UTC (1900–1900 LT). The calendar date is assigned to the date of the end of the 24-h period. The three-dimensional wind components, mean sea level pressure (MSLP) and surface pressure, geopotential, temperature, specific humidity, and potential vorticity at standard pressure levels were obtained from the ERA-Interim reanalysis (Dee et al. 2011) at a horizontal resolution of 0.75° × 0.75° and a temporal resolution of 6 hours. Additional surface charts including fronts were provided by the NHMS. The 6-hourly NCEP–NCAR reanalysis MSLP data (Kalnay et al. 1996) at a 2.5° × 2.5° resolution were used to calculate a long-term time series of the Siberian high (SibH) intensity after Jeong et al. (2011). The SibH intensity is the mean DJF MSLP in the region 40°–65°N, 80°–120°E that is standardized with respect to the mean and standard deviation for 1949/50–2013/14. The corresponding intensity of the Aleutian low was assessed using the North Pacific (NP) index (Trenberth and Hurrell 1994). The NP index is the mean monthly sea level pressure averaged over the region 30°–65°N, 160°E–140°W.
To describe the evolution of deep convection, 3-hourly Gridded Satellite (GridSat)-B1 climate data record intercalibrated IR brightness temperature data in the 11-μm window channel (Knapp et al. 2011) at a resolution of 8 × 8 km2 were employed. Finally, daily NOAA interpolated OLR (Liebmann and Smith 1996) in the latitude belt 0°–15°N was used at a 2.5° × 2.5° resolution to filter for the MJO, Kelvin, and equatorial Rossby (ER) waves with the wavenumber–frequency filter after Wheeler and Kiladis (1999). A 2–10-day Lanczos bandpass filter (Duchon 1979) that was applied to NOAA OLR data is used to determine activity of TD-type disturbances (Wu et al. 2011).
The daily rainfall time series of the nine stations located in the Central Highlands region (Fig. 1) were searched for dates matching the following criteria: 1) measurements were available for at least three out of the nine stations and 2) dates were selected if rainfall amounts greater than or equal to 10 mm day−1 were recorded at three or more stations. If only records from between three and five stations were available, this criterion was relaxed to two stations with more than 10 mm day−1. The 10 mm day−1 threshold has been taken after informal interviews with coffee farmers in the Central Highlands by the third author (Phan et al. 2013). For the period 1981–2007, 90 dates matched these criteria in November, and 19 for the climatologically drier month December. Dates with tropical cyclone activity were excluded using Joint Typhoon Warning Center Best Track Data (accessed via http://weather.unisys.com/hurricane/w_pacific/index.php). Cases were subjectively selected for an in-depth investigation based on two criteria: (i) one of the known dynamic features discussed in the introduction was assumed to be the major forcing of rainfall (i.e., cold fronts, cold surges, TD-type disturbance, or active MJO and CCEW phases); and (ii) subjective synoptic analyses of MSLP, geopotential, wind, streamfunction, velocity potential, and OLR confirmed the suitability of the identified cases. This resulted in four cases named tail end of a cold front (case 1), TD-type disturbance (case 2), multiple tropical wave interactions (case 3), and cold surge with Borneo vortex (case 4). The reasons for the naming are described in section 3.
To study the identified cases, some derived quantities have been calculated from ERA-Interim. These are the vertically integrated humidity fluxes, surface convective available potential energy (SF-CAPE), the 100–400-hPa vertically averaged potential vorticity as in Fröhlich and Knippertz (2008), and the Froude number. The Froude number, defined here for elevation heights higher than 400 m as the ratio of wind speed at 850 hPa and the product of Brunt–Väisälä frequency between 925 and 700 hPa and elevation height, is an approximation if an air parcel will overpass an obstacle or not. In case of high wind speeds, low stability, and/or small obstacle the Froude number is large, and the air parcel will likely overpass the obstacle. On the contrary, in the case of a weak wind, high stability, and/or a tall obstacle the Froude number is smaller than 1, and the air parcel will not easily overpass the obstacle or will even be forced to pass along the obstacle.
3. Rainfall events
a. Case 1: Tail end of a cold front (9–15 November 1982)
In the period between 9 and 15 November 1982, the highest precipitation amounts occurred in the southern parts of the Central Highlands (Fig. 2a). Rainfall anomalies with respect to the 1981–2007 period were also positive in the north, but the central part was drier than normal. Figure 2b shows the time evolution of rainfall during the event for the entire study region, as well as for the northern, central, and southern parts. Two rainfall maxima occurred during this period: the first on 9 and 10 November, and the second from 12 November onward; 11 November was rather dry throughout all parts of the Central Highlands (Fig. 2b). Therefore, this event can be divided in two periods that will be discussed below.
The first period of this event was clearly influenced by a subtropical cold front extending deep into the tropics. The cold front belongs to a low pressure system with its center located over the Yellow Sea at 1800 UTC 9 November 1982 (Fig. 3a). At this time, the cold front was extending equatorward to about 13°N, and the location of the cold front was well reflected in MSLP, wind speed, and horizontal wind shear (Fig. 3a). The passage of the cold front can also be seen in the radiosoundings at Hoang Sa3 (16°50′N; 112°20′E) between 1200 UTC 9 November and 0000 UTC 10 November and at Da Nang (16°04′N; 108°21′E, see Fig. 1) between 1200 UTC 9 November and 1200 UTC 10 November (not shown). Schultz et al. (1997) noted that midlatitude cold fronts frequently lose their frontal character when they reach into the tropics and can better be described as shear lines, because there is no longer a pronounced temperature gradient but strong winds. Yet, both upper-air stations show a considerable drop in low-level temperature on top of a wind shift during the passage of the low-level cold front. Thus, though cold fronts are rarely analyzed in surface charts in the southern SCS, it seems justified in this case.
The 3-hourly surface analyses of the NHMS also showed the cold front from Indochina to the Yellow Sea until 2100 UTC 9 November 1982, whereas at 0000 UTC 10 November 1982 the cold front was no longer drawn (not shown). Figure 3a shows strong 24-h pressure rise over mainland Asia peaking at 8 hPa (24 h)−1 over southern China whereas the pressure fall ahead of the cold front was on the order of 1–2 hPa suggesting frontolysis. The Yellow Sea low was associated with a longwave trough, which is reflected both in 500-hPa geopotential height, and vertically averaged potential vorticity (Fig. 3b). The trough originated from a wave disturbance over central Russia on 6 November 1982 and moved eastward with the basic flow (not shown). At 1800 UTC 9 November 1982, it reached the southernmost position and the trough axis extended southward to about 21°N (Fig. 3b). The trough axis was identified using the zonal 500-hPa geopotential gradient as described in Knippertz (2004).
As a consequence of the strong surface anticyclogenesis over China, the MSLP gradient tightened in the postfrontal area over the northern SCS (Fig. 4a), leading to high wind speeds and arguably leading to large uptakes of moisture by strong air–sea fluxes. Though much weaker than the geostrophic wind, Fig. 4a suggests that the ageostrophic isallobaric wind was the major cause of an equatorward deflection of surface winds over the SCS. It is suggested here, that the convergence of the isallobaric wind was the major cause of triggering deep convection over the central SCS (Fig. 4a) after frontal lifting diminished due to ongoing frontolysis. Our suggestion is corroborated by the fact that outbreaks of deep convection over the SCS occurred well ahead of the surface front in the area of high northeasterly winds and humidity flux convergence (cf. Figs. 3a and 3c). In addition, the convergence of the isallobaric wind was the main contribution to the convergence of the total wind over the SCS (not shown). Note, however, that the small values of the Coriolis parameter render the interpretation of the ageostrophic winds critical; thus, ageostrophic wind vectors are only shown for latitudes north of 10.5°N and interpreted with caution between 10.5° and 12°N. Furthermore, the isallobaric wind in Fig. 4a results from a temporal change of MSLP whereas the geostrophic wind is an instantaneous value.
Precipitation amounts were high in the north of the Central Highlands due to the cold front passage that was associated with advection of moist air from the SCS (Fig. 3c) and orographic ascent (Fig. 4b). A rain shadow effect for the Dak Lak Plateau (cf. Fig. 2a) seems likely due to the overall small (i.e., lower than 1) Froude number, indicating high stability in the presence of high northeasterly winds (Fig. 4c). The southern part in turn was wetter than normal because the mountains in the south blocked the flow, represented by low values of the Froude number (Fig. 4c). Therefore, orographic lifting of the low-level flow that recurved from northeasterly to easterly around the southern part of the mountains (Fig. 4b) is proposed as one contributing factor that led to high precipitation amounts in the south. In addition, SF-CAPE was more than 1000 J kg−1 from the southern part of the Central Highlands southward and in the southern portion of the SCS (Fig. 4b), suggesting potential instability of the atmosphere in these regions.
Contrary to the first period of this event, small-scale convection characterized the second period lasting from 12 to 15 November 1982 (Fig. 5). The occurrence of small-scale convection, which was rather constrained to the south (Fig. 2b), was favored by the transport of moisture into the south by the cold front (Fig. 3c). This transport resulted in high equivalent potential temperatures and SF-CAPE especially in the south (Fig. 5a), which indicates instability of the atmosphere in this region. Because of weak winds, the moist and instable situation persisted between 12 and 15 November and orographic lifting by the mountains led to the occurrence of small-scale afternoon convection (e.g., at 0900 UTC 13 November 1982; Fig. 5b).
Interestingly, summed over the whole period, conditions were drier than normal at all coastal stations (Fig. 2a). This is due to the strong northerly winds, blowing parallel to the coast and mountain range during the first period, and weak circulation during the second period of the event. During this second period, convection occurred regionally and most likely due to high instabilities and orographic lifting by the mountains. Local effects such as mountain–valley breezes or thermal lifting at the mountains might have triggered outbreaks of small-scale convection during this period.
b. Case 2: Tropical depression–type disturbance (1–4 December 1986)
The period 1–4 December 1986 was wetter than normal for the whole region (Figs. 6a and 6b) except for the southernmost station (Fig. 6a). The highest rainfall amounts occurred at coastal stations. In the mountains, the largest positive anomalies were observed in the northern part of the Central Highlands (Fig. 6a). This rainfall event was characterized by the passage of a low-level, westward-moving TD-type disturbance (Fig. 7). The 2–10-day bandpass-filtered 850-hPa winds, depicting TD-type disturbance activity, show a cyclonic circulation over the western Pacific on 26 November 1986 that was accompanied by low infrared brightness temperatures, which indicate enhanced convection (Fig. 7b). While moving westward, the pattern of cyclonic circulation and enhanced convection intensifies until 30 November 1986 when it is located over the SCS (Fig. 7b). After having passed the Philippines on 29 November 1986, the westward movement of the pattern slows down, and convective activity is slightly reduced after 30 November 1986 (Fig. 7b). The TD-type disturbance reaches southern and central Vietnam on 1 December 1986 and enhances convection in this region until 3 December 1986 (Figs. 7a and 7b). On 2 December 1986 the center of the cyclonic circulation of the TD-type disturbance is located slightly off the coast of southern Vietnam (Fig. 8), leading to moisture advection and flux convergence in south-central Vietnam especially to the right in the direction of the movement of the cyclonic circulation (Fig. 8a). This area is also affected by widespread deep convection (Fig. 8a). Rainfall amounts are higher at the coast (Fig. 6a) and orographic lifting was stronger (Fig. 8b) when compared with the first case (Fig. 4b), because the lower-tropospheric winds were rather zonally oriented from east to west thus impacting more orthogonal on the Central Highlands (Fig. 8b). The high Froude number in the presence of relatively high winds suggests that low stability was present, facilitating rainfall over and in the lee of the mountain ridge (Fig. 8c).
After having passed Vietnam, the circulation and convection starts to weaken over Cambodia and the Gulf of Thailand, and on 5 December there was no longer a cyclonic circulation in the bandpass-filtered wind field (Fig. 7b). An eastward-moving Kelvin wave, having passed Vietnam longitudes about at the same time as the TD-type disturbance enhanced convection rather close to the equator (Fig. 7). The latter is especially evident on the map of 3 December 1986 in Fig. 7b showing low brightness temperatures being confined to latitudes south of 10°N. Thus, a direct influence by the convective envelope of the Kelvin wave on rainfall in the Central Highlands seems unlikely. However, as demonstrated by previous studies (e.g., Roundy 2008; Schreck and Molinari 2011; Schreck 2015) the Kelvin wave might have amplified cyclonic anomalies to its north. Note that there was no convectively active part of an ER wave in the Central Highlands during this case (not shown). Therefore, ER wave-filtered contours were omitted for clarity in Fig. 7a.
c. Case 3: Multiple tropical wave interactions (2–5 November 2007)
Between 2 and 5 November 2007, all parts of central Vietnam were wetter than normal except for two stations in the southwest of the Central Highlands (Figs. 9a and 9b). However, the highest positive deviations relative to the long-term rainfall sum for this period occurred at the coast and decreased inland (Fig. 9b). This event is characterized by the passage of both eastward- and westward-moving equatorial waves (Fig. 10a). Namely, an eastward-moving MJO and Kelvin wave, and a westward-moving TD-type disturbance and ER wave all passed with their convectively active centers being collocated over the southern half of Vietnam on 2 and 3 November 2007 (Figs. 10a and 10b). However, the tropospheric moisture fluxes and their convergences were clearly dominated by the TD-type disturbance on 3 November 2007 (Fig. 11a). The region of maximum moisture flux convergence is characterized by deep convection as indicated by low GridSat infrared brightness temperatures (Fig. 11a). Note that the latter dataset is independent from ERA-Interim. The SF-CAPE pattern in Fig. 11b indicates that high potential instability supported the development of deep convection in the southern part, whereas in north-central coastal regions orographic lifting and less deep convection prevailed (Fig. 11a). It is concluded that rainfall anomalies were highest at the coast, because the strongest convective signal came from the SCS by the TD-type disturbance in combination with orographic lifting (Fig. 11b) and not from the MJO and Kelvin wave that reached Vietnam after having passed the Gulf of Thailand.
The high Froude number is one explanation that explains why rainfall reached leeward of the mountain barrier (Fig. 11c). Apparently, the easterly low-level flow from the SCS was quite unstable. Another cause could be the off-equatorial convective signal of a westward-propagating Kelvin wave that is traceable in the unfiltered GridSat brightness temperature maps in Fig. 10b. However, visual inspection of Fig. 10b suggests the largest impact by the TD-type disturbance followed by the Kelvin wave. Nonetheless, our analyses leave open the question as to the quantitative contribution of the tropical waves to the rainfall events.
d. Case 4: Cold surge with Borneo vortex (11–15 December 2005)
From 11 to 15 December 2005, positive rainfall anomalies occurred over the whole region, except at three stations in the north of the Central Highlands (Fig. 12). As for case 3 described in section 3c, the highest anomalies occurred at the coast and decreased inland. The major reason for high rainfall amounts during this period was a cold surge event. The case satisfied both, the cold surge criteria proposed by Chang et al. (2005) and Yokoi and Matsumoto (2008), the latter being more appropriate for boreal fall cases since the latitude at which the strengths of meridional winds at 925 hPa are evaluated is 20°N instead of 15°N. Moreover, Yokoi and Matsumoto (2008) introduced a temperature criterion making the index more robust in terms of the thermal signal.
A strong SibH and strong Aleutian low, both known to be important factors for the occurrence of cold surges (Park et al. 2011), favored high northeasterly winds from the East China Sea down to the southern SCS during case 4 (Figs. 13a and 13b). The DJF 2005/06 SibH intensity index, as defined in section 2, shows one of the most intense SibHs in the period 1949/50–2013/14 (Fig. 13b). Figure 13b also documents that December 2005 had the lowest NP index value (cf. section 2), which is a measure of the intensity of the Aleutian low, for any December in the period 1949–2013. Additionally, the Aleutian low was located exceptionally far west with its center located above the Sea of Okhotsk and the North Pacific (Fig. 13a). After having been almost stationary since 10 December 2005 over the SCS north of Borneo, a Borneo vortex started to move westward on 14 December 2005 (not shown), and reached the Vietnamese coast on this date. This resulted in moisture flux convergence, deep convection (Fig. 14a), and an increase of rainfall anomalies particularly in the south and center of the Central Highlands (Fig. 12b). As in cases 2 and 3, the Froude number gives a clue as to why the rains extended leeward of the mountain range (Fig. 14c) though lower Froude numbers (not shown) in the northern Central Highlands caused a rain shadow effect, resulting in near-normal conditions at three leeward stations. Altogether, Figs. 14b and 14c suggest lifting in stable environments in the northern part of the Annamese Cordillera, thus deep convection was restricted to instable areas south of 15°N (Fig. 14a).
4. Summary and discussion
Synoptic-dynamic causes of early dry-season (November–December) rainfall events in the Vietnamese Central Highlands, the major coffee-growing region in Vietnam, were analyzed in this study. The 109 rainfall events that have been considered for an in-depth study lasted for several days and led to positive rainfall anomalies relative to the long-term mean throughout large parts of the region. The final selection was then motivated by capturing the diversity of weather patterns that cause anomalous rainfall in the study region in the period of 1981–2007. Altogether four cases were chosen: a tail end of a cold front (case 1), a TD-type disturbance (case 2), multiple tropical wave interactions (case 3), and a cold surge with Borneo vortex case (case 4). To study the four selected cases, a variety of data sources have been used, ranging from station surface and upper-air observations, hand-analyzed weather maps from the national weather service, satellite data, gridded station-based data products to NCEP–NCAR, and ECMWF reanalyses. In addition, both classical synoptic and tropical large-scale wave diagnostics were employed to obtain a thorough description of the synoptic dynamics of the rainfall events.
The tail end of a cold front case in November 1982 (case 1) describes a situation in which the cold front characteristics were maintained deep into the tropics down to about 13°N. Deep convection develops over the SCS ahead of the cold front where convergence of low-level ageostrophic isallobaric winds have likely contributed to triggering of convection in the prefrontal moist and instable atmosphere. The enhanced low-level northeasterly winds transported high equivalent potential temperature air from the SCS toward the southern Vietnamese Central Highlands where this air mass stalled and caused a multiday period of afternoon convective outbreaks. Rainfall in the northern highlands occurred in a relatively stable situation and was restricted to the time of the cold front arrival. This blocking effect due to low Froude numbers is known to have an effect on frontal systems (Houze 2012).
A westward-traveling TD-type disturbance, was instrumental in causing rainfall during case 2. Though the low-level winds blew almost orthogonal to the coastline and mountain range, the rain shadow effect was decreased by an instable lower troposphere, as indicated by high Froude numbers. In case 3, four tropical waves were involved in the rainfall events: a TD-type disturbance, and active phases of the MJO, Kelvin, and ER waves. While the TD-type disturbance has the clearest signature in the deep convection, the relative contribution of each wave type could not be quantified though unfiltered GridSat brightness temperature suggests off-equatorial convection affecting the study region in association with an eastward-moving Kelvin wave (Fig. 10b). Case 4 is a cold surge case, satisfying the Chang et al. (2005) and Yokoi and Matsumoto (2008) cold surge criteria. A related Borneo-type vortex started to move westward, further enhancing the northeasterly flow on the windward side of the mountains. As in cases 2 and 3, the Froude number indicated that rainfall could spread across the mountain range due to low stability, especially in the southern part of the highlands.
While case 1 shows some similarities to the cold surge event of 2–3 November 1999 described in Yokoi and Matsumoto (2008), noteworthy differences exist. First, case 1 does not fulfill the cold surge criteria of Yokoi and Matsumoto (2008) and no cold front was analyzed over the SCS and Indochina Peninsula in their study. Figure 14c of Yokoi and Matsumoto (2008) shows that in their cold surge-southerly wind (CS-SW) composite case in which a TD-type vortex in the central SCS causes southerlies over the SCS, rainfall is observed in the Central Highlands area. However, during case 1 no TD is present, thus it is not a CS-SW case.
The four cases presented reveal complex large-scale, synoptic, and local orographic interactions that ultimately determine the spatiotemporal characteristics of rainfall events. Common to all cases is that the synoptic forcing removed the climatological orographic effect of windward rains from relatively warm clouds and downstream dryness over the mountains by providing the moisture, instability, and vertical lifting that lead to outbreaks of deep convection over the mountains and partially the coast. The MJO and types of CCEWs are involved in cases 2 and 3, but direct influences by the waves’ convective envelopes are not discernible in cases 1 and 4 that are more related to midlatitude dynamic forcing. However, the MJO and Kelvin wave are known to remotely influence rainfall by modifying the large-scale circulation (e.g., Roundy 2008; Zhang et al. 2009; Schreck and Molinari 2011; Schreck 2015). These remote influences might also have impacted the evolution of rainfall in cases 1 and 4. This study primarily aimed at identifying, categorizing, and understanding rainfall events over the central Vietnamese mountains. Clearly, determinations of the frequency of certain events, their predictability on weekly time scales, and the future change of occurrence and intensity are left for future research. Because of the association with large-scale extratropical and tropical wave forcing, 1–2-week predictability of these events in a probabilistic sense could be explored and measures to at least dampen the impact on cultivation of coffee could be developed.
Given the data paucity over the Central Highlands and the relatively coarse resolution of ERA-Interim, our conclusions will require further verifications and extensions. Ideally, to explore the transition mechanisms from the heavy rainfall coastal region to the dry highland during late fall–early winter over a distance of about 50–100 km, a field campaign could provide the necessary surface and upper-air data. This should be complemented by modeling studies at convection-resolving resolutions. For example, our proposed mechanisms might not exclusively explain rainfall dynamics over highlands. Second in composite-like approaches, the frequency and climatological relevance of the cases shall be explored. This includes the investigation of a potential linkage to remote indirect influences like equatorial waves and ENSO. Note that cases 1 and 2 occurred during strong and developing El Niño events, respectively. On the contrary, case 3 occurred during a La Niña event.
Nguyen et al. (2014) have shown a statistical significant increase of dry-season rainfall in the Central Highlands region. The recent recovery of the SibH, as documented by Jeong et al. (2011) until 2009/10, is still ongoing until 2013/14 (cf. Fig. 13b). A strong SibH favors the occurrence of cold surges, which are potentially leading to an enhancement of rainfall in the Central Highlands. In addition, Juneng and Tangang (2010) demonstrated that the Borneo vortex moved closer to Vietnam and showed a stronger zonal wind component. By revealing important dynamic causes of rainfall, the present study might help in assessing past and future variability of early dry-season rainfall events in Vietnam’s major coffee growing region.
The first and second authors acknowledge partial support for their research leading to these results by the EWATEC-COAST (BMBF Grant 02WCL1217C) project. The two last authors would like to acknowledge the Vietnam National University Ho Chi Minh City (VNU-HCM) for partial support under Grant NDT2012-24-01/HD-KHCN. We are also thankful to three anonymous reviewers whose comments helped to greatly improve the manuscript.
Truong Son (in Vietnamese).
The Chinese name of this station is Xisha Dao (WMO station ID 59981).