The formations of heavy rainfall/flood (HRF) events in Vietnam are studied from diagnostic analyses of 31 events during the period 1979–2009. HRF events develop from the cold surge vortices formed around the Philippines. These vortices’ speed, size, and rainfall, which evolve into HRF events, are enhanced distinguishably from non-HRF vortices, as they reach Vietnam. The HRF cyclone, the North Pacific anticyclone, and the northwestern Pacific explosive cyclone simultaneously reach their maximum intensities when the HRF event occurs. An HRF cyclone attains its maximum intensity by the in-phase constructive interference of three monsoon (30–60, 12–24, and 5 days) modes identified by the spectral analysis of zonal winds. The rainfall center of an HRF event is formed and maintained by the in-phase constructive interference of rainfall and convergence of water vapor flux anomalies, respectively, from three monsoon modes. Forecast times of regional models are dependent and constrained on the scale of the limited domain. For 5-day forecasts, a global or at least a hemispheric model is necessary. Using the salient features described above, a 5-day forecast advisory is introduced to supplement forecasts of HRF events made by the global model. Non-HRF vortices are filtered by threshold values for the deepening rate of explosive cyclones and basic characteristics of the HRF events predicted by the global model. A necessary condition for an HRF event is the in-phase superposition of the three monsoon modes. One-week forecasts for 12 HRF events issued by the NCEP Global Forecast System are tested. Results demonstrate the feasibility of the forecast advisory to predict the occurrence dates of HRF events.
Summer and winter monsoons characterize the annual variation of the atmospheric circulation in East and Southeast Asia (Ramage 1971). Stretching equatorward from 22° to 9°N, the east coast of Vietnam covers a distance of about 2000 km. The weather and climate systems along this coast are modulated by these two monsoons. Analyzing the annual variation of rainfall measured by World Meteorological Organization (WMO) surface stations along the Vietnam coast, Chen et al. (2012a) noticed that the monsoon rainfall season exhibits three regimes: 1) a summer regime in the north, 2) a fall regime in the central region, and 3) a combination of both regimes (May–November) in the south. The rainfall regime farther south along the eastern coast of Malaysia belongs to a winter regime (Cheang 1987). In central Vietnam, some rain-producing disturbances may develop into heavy rainfall/flood (HRF) events. The cause of these events has been explored from two perspectives: 1) interannual variation of late fall rainfall and 2) the formation/development mechanism of HRF events.
Using correlation coefficient patterns and empirical orthogonal function (EOF) analysis, Nguyen et al. (2007) and Yen et al. (2011) showed the late fall rainfall in central Vietnam undergoes an interannual variation that is out of phase with the El Niño–South Oscillation (ENSO) activity. During the warm (cold) ENSO phase, there is less (more) rainfall produced in central Vietnam. This interannual rainfall variation is related to changes in the circulation pattern in southeastern Asia and the intensity of monsoon westerlies, and the number and type of rain-producing weather systems in central Vietnam (Chen et al. 2012a). The intensity variation of the near-equator trough, which extends from the eastern tropical Indian Ocean across tropical Southeast Asia to the western tropical Pacific during late fall, is out of phase with the SST (Niño-3.4; 5°N–5°S, 170°–120°W) anomalies. This interannual variation of monsoon westerlies is a response to tropical Pacific sea surface temperature anomalies, following the ENSO cycle. Consequently, the interannual variation of the late fall rainfall in central Vietnam is in phase with the low-level westerlies around 15°N and the low-level easterlies around 5°N.
Yokoi and Matsumoto (2008) argued that the coexistence of a cold surge flow and a tropical easterly depression is a necessary condition to produce heavy rainfall in central Vietnam. Chen et al. (2012a) estimated the rainfall produced by a cold surge vortex (CSV), tropical cyclone, and HRF cyclone. Two-thirds of the late fall rainfall in central Vietnam is contributed by HRF events. Over this region, the interannual rainfall variation during late fall in the past three decades (1979–2009) was primarily attributed to the rain produced by HRF events. This interannual variation is not determined by the frequency of HRF events, but is determined by its rain-producing efficiency. During the cold phase of ENSO activity, this rain-producing efficiency during the cold late fall is almost twice that during the warm/normal late fall.
On 30–31 October 2008, an unusual HRF event, classified as a class-2 event (occurs once in more than a century) by the Dartmouth Flood Observatory (DFO 2011), occurred at Hanoi in northern Vietnam. The development mechanism was different from the HRF scenario suggested by Yokoi and Matsumoto (2008). Chen et al. (2012b) explored the formation and development of this HRF event from the perspective of the multiple-scale processes within the context of the midlatitude–tropical interaction. Their findings may be briefly summarized as follows. A westward-propagating CSV formed in the Celebes Sea and developed into a Vietnam HRF event. The extratropical surface low system, related to this vortex formation through the northeastern Asian cold surge flow, formed an explosive cyclone (Sanders and Gyakum 1980) simultaneously with the occurrence of the HRF event. At the same time, the intensity of the explosive cyclone measured by sea level pressure, the North Pacific anticyclone, and the HRF cyclone reached their maxima. Three monsoon (30–60, 12–24, and 5 day) modes were identified by the spectral analysis of zonal winds to the north and south around the HRF event center. The HRF event occurred when the three monsoon modes were in phase, constructively interfering with the water vapor transport and convergence of water vapor flux by these three monsoon modes toward the HRF rainfall.
Despite the unusual location of the Hanoi HRF event, we investigated whether the formation and development mechanism and salient features of this HRF event are common to other HRF events identified by the DFO during the past three decades (1979–2009). Models alone may not produce sufficiently detailed forecasts, as regional models limit the forecast domain and time scale, while the global models, in general, underforecast the severity of the HRF event. The goal of the present study is to develop a forecast advisory for the HRF events in central Vietnam, using the extended operational model forecasts with more detailed diagnostic analyses for model predictions. To accomplish this goal, this study is organized in the following manner. A synoptic analysis for the development of the North Pacific explosive cyclones following severe cold surges, the HRF cyclones from CSVs, and maximum intensities of the North Pacific anticyclone when HRF events occurred are presented in section 2. The formation mechanisms of the HRF events by the constructive interference of the three monsoon modes are further analyzed, using the zonal wind north and south of all HRF event centers, and substantiated by the streamfunction tendency analysis in section 3. The maintenance of an HRF rainfall center by the convergence of water vapor flux and the constructive interference of the enhanced convergence of water vapor flux by the three monsoon modes are provided in section 4. Salient features from all HRF events revealed by these diagnostic analyses are used to develop a forecast advisory for future events. Some statistical analysis of the available National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) forecasts for more recent HRF events are used to test the feasibility of the forecast advisory in section 5. A summary and the concluding remarks of this study are provided in section 6.
2. Synoptic analysis of heavy rainfall events
a. Development of the 1999 Hue HRF event
In 1999, an HRF event occurred in Hue, the ancient capital of Vietnam located at 16.36°N, 107.28°E. Figure 1 provides an overview of this event’s synoptic conditions. Indicated by the daily rainfall measurements at Hue (Fig. 1a), four rainfall events occurred in central Vietnam. On 2 November, rainfall from the Hue HRF event was close to 1000 mm day−1 (Fig. 1b), twice that of the 2008 Hanoi HRF event. As shown by the 925-hPa streamline chart superimposed with the Tropical Rainfall Measuring Mission (TRMM) rainfall (Fig. 1b), the Hue event, which supplied moisture to this event center, occurred downstream of the southeasterly flow in the vicinity of an HRF cyclone centered over southern Indochina with a high pressure system in East Asia. In the Sea of Okhotsk, the low pressure system northeast of the cold surge flow formed an explosive cyclone. Except for the differences in their latitudinal locations, the 1999 Hue and 2008 Hanoi HRF events have similar synoptic features.
The synoptic development of the Hue HRF event in the tropics and midlatitudes is illustrated by the 925- and 500-hPa streamline charts in Fig. 2. At 1200 UTC 27 October, a cold surge flow crossed Lake Baikal, but did not reach the eastern coast of northeast Asia (Fig. 2a). However, two additional remnant cold surge flows already moved into the Pacific Ocean, following the eastward propagation of the upper troughs (Fig. 2b). The Okhotsk cold surge flow was blocked by the Pacific subtropic high from any interaction with the tropical trade easterlies. The interaction between the Korean cold surge flow and the easterly wave (short red line) is possibly due to their proximity, as shown in Fig. 2a. The two cold surge flows across Korea and in the northwest Pacific are moved farther eastward by the trough (Fig. 2d). On 29 October, the Lake Baikal cold surge flow moved into the Yellow Sea (Fig. 2c). The interaction of this cold surge flow and the easterly wave in the trade easterlies leads to the formation of a cold surge vortex in the Celebes Sea.
During the next 2 days, the size of the newly formed CSV grew to cover the entire tropical South China Sea by 31 October (Fig. 2e). The cyclonic flow is discernible at 500 hPa (Fig. 2f), so the size and depth of this cyclonic system no longer resemble a small-scale vortex; therefore, it is appropriate to refer to this system as an HRF cyclone (Chen et al. 2012b). On 2 November, the Lake Baikal cold surge flow moved northeastward and split into two cold surge flows—one crossed the Yellow Sea and the other entered the northern part of the Sea of Japan and northern Japan (Fig. 2g). The low pressure system, associated with the northern cold surge flow, moved to the Sea of Okhotsk and formed an explosive cyclone, coupled with a deep cutoff low (Fig. 2h). The southern cold surge flow formed a strong easterly flow with the HRF cyclone centered along the coast of southern Vietnam. The strong transport of moist air by easterly flow from the South China Sea maintained the 1999 Hue HRF event.
When the Hue HRF event and the formation of an explosive cyclone occurred simultaneously on 2 November, the synoptic conditions in the western North Pacific (Fig. 3a) were characterized by the maximum intensities of these three weather systems—the HRF cyclone, the North Pacific anticyclone, and the northwestern Pacific explosive cyclone. The deepening of the HRF cyclone was accomplished by two midlatitude–tropics interaction processes. The interaction of the CSV/HRF cyclone with 1) the East Asian cold surge flow and 2) the remnant cold surge flows associated with the previous anticyclone propagating out toward the open North Pacific. These remnant cold surge flows can intensify the tropical trade easterlies and the CSV/HRF cyclone embedded in the near-equator trough. These two processes are indicated by the y–t diagrams of u(850 hPa) at two longitudinal locations: 108°E (the location of the Hue HRF cyclone center) and 150°E (the location of the NE Pacific explosive cyclone center and the maximum isotach of the tropical trade easterlies). The intensification processes of these three systems can be illustrated by the y–t diagrams of u(850 hPa) at two longitudinal locations (thick red lines in Fig. 3a): 1) centers of the HRF cyclone at 108°E (Fig. 3b) and 2) the explosive cyclone at 150°E (Fig. 3c).
b. Statistics of synoptic analysis
The dates and locations of HRF events are identified and archived by the DFO, but those of their parent CSVs are identified by the process presented in section 2a for the 1999 Hue case. Locations of these parent CSVs, identified around the Philippines, are marked in Fig. 4a by dark-blue crosses. Translated by the northeasterlies/trade easterlies, these vortices travel about 2–6 days (with an average time of 4.4 days) to reach Vietnam to form the HRF events (purple dots) and red lines depict their tracks. The formation of cold surge vortices is associated with the northeastern Asian cold surge flows coupled with surface lows. Locations of these surface low centers related to the parent CSVs of 31 HRF events at the times these CSVs formed are marked by dark-blue dots in Fig. 4b. These low centers moved northeastward across the east coast of northeast Asia into the northwestern Pacific by strong westerlies associated with the east coast trough and formed explosive cyclones (purple dots). Starting with the time their related CSVs are formed, it also takes 2–6 days (with an average of 4.4 days) (Fig. 4d) for these surface lows to move along trajectories (marked by red lines in Fig. 4b) and become explosive cyclones. Note that 125 cold surge vortices are identified in the vicinity of the Philippines during October–November of 1979–2009 and only 31 of them (~25%) evolved into HRF events. This raises two concerns.
1) Characteristic differences between non-HRF and HRF CSVs
The characteristic differences between non-HRF and HRF CSVs were previously analyzed by Chen et al. (2012a, their Fig. A2). A brief summary of the three basic characteristics (maximum speed, size, and rainfall) is provided below.
Maximum 850-hPa wind speed: The averaged maximum |V(850 hPa)| of non-HRF and HRF CSVs are close to each other around the Philippines. Their difference became distinguishable as they moved west of 120°E. A clear separation between an HRF cyclone and a CSV maximum wind speed along the Vietnam coast is 15 m s−1. Thus, the HRF cyclone is defined, if its speed ≥ 15 m s−1.
Size (east–west dimension): Estimated by the distance between the maximum isotachs at the east and west periphery of all cyclonic disturbances centered at the Vietnam coast, the size is about 16° longitude (≤1800 km) for the HRF event and about 7° longitude (≤800 km) for the CSV.
Rainfall: Average rainfall over an area (rainfall ≥ 3 mm day−1) around the rainfall center of a CSV at the Vietnam coast is 90 mm day−1, but around the rainfall center of an HRF event the average rainfall is 250 mm day−1.
The differences in these three basic characteristics between HRF and non-HRF CSV increase substantially west of the Philippines—by about 100% for an HRF event at the east coast of Vietnam, but less than one-third for a non-HRF CSV.
2) Simultaneous occurrence of the HRF event and the formation of an explosive cyclone
Simultaneous occurrences of an HRF event and the formation of an explosive cyclone are verified by plotting the HRF event dates on the abscissa and the formation dates of explosive cyclones as the ordinate in Fig. 5. All 31 HRF events were identified and archived by the DFO, while the corresponding explosive cyclones in the northwest Pacific were determined using the definition of Sanders and Gyakum (1980) and the NCEP Service Records Retention System (SRRS) 6-h surface charts. Projecting the verified dates of synoptic disturbances on this coordinate system, three groups of synoptic cyclone events are classified by comparing with the formation of explosive cyclones (ECs in Fig. 5):
An HRF event (blue open circle) in Vietnam related to an explosive cyclone (red dot) in the northwest Pacific: All 31 HRF events developed from CSVs. Along the diagonal are simultaneous occurrence dates of these HRF events and the corresponding northwest Pacific explosive cyclones, except for two cases that formed 2 days before the occurrence of related HRF events.
CSVs related to explosive cyclones: Six of 19 CSVs (green dots) that formed in the vicinity of the Philippines became tropical cyclones (green TC symbols) across the South China Sea. These TCs weakened when they reached the Vietnam coast about 1–3 days ahead of the explosive cyclones formation in the northwest Pacific, except for two TCs, which formed 1 day behind.
CSVs and TCs not related to explosive cyclones: The arrival of 66 CSVs and 19 TCs at the Vietnam coast are not relevant to the northwest Pacific explosive cyclones.
It was shown in Fig. 3 that the simultaneous occurrence of the HRF event and explosive cyclone formation is related to the intensification of the North Pacific anticyclone. The occurrence of every HRF event can be substantiated by a time series of u(850 hPa) at the longitude of the maximum westerly of an explosive cyclone and the tropical trade easterly of the North Pacific anticyclone at the same longitude. Interestingly, the formation of every explosive cyclone occurs when both the westerly winds of the explosive cyclone and the tropical trade easterly winds reach their maxima simultaneously. Results obtained for the 31 HRF events by the analysis in this study can be used to validate the simultaneous occurrence dates of the HRF event and the formation of extratropical explosive cyclones shown in Fig. 5. The time series for these 31 HRF events analyzed in this study can be found online (http://eamex.iastate.edu/Vietnam_31HRF_events.pdf). To save space, six HRF events during 2004–09 are shown in Fig. 6, as an example of these events. The date for each explosive cyclone formed is marked by a cross on the u(850 hPa) time series.
3. Development of an HRF cyclone: Analysis of streamfunction tendency
a. Spectral analysis
Chen et al. (2012b) determined the 2008 Hanoi HRF event occurred when the zonal wind anomalies of the three monsoon modes were constructively in phase. Over the period 1979–2009, other HRF events can be checked for the in-phase constructive interference mechanism. Spectra of the u(850 hPa) time series are analyzed over a 6-month period (August 2008–January 2009) at the locations of the maximum wind speeds at both the northern (blue open circle) and southern (green open circle) peripheries of all 31 HRF events (Fig. 7a) for the 1979–2009 period; these spectra are shown in Figs. 7b and 7c, respectively. The periods for the three monsoon modes at both the northern and southern locations are indicated by three light blue strips. Physically, these three modes represent the Madden–Julian oscillation, the dipole of the surface high–low pair juxtaposed with the cold surge flow, and the easterly wave. For reference, the u(850 hPa) spectrum (black line) and its reference spectra with a 99% confidence level (thin dashed black line) for the 1999 Hue HRF event are provided. Despite interannual variations in the latitudinal locations for all HRF cyclone centers, and variations in their maximum easterlies and westerlies, the spectral peaks of the three monsoon modes [30–60, 12–24, and 5-day, denoted by , , and , respectively] emerge coherently for all HRF events.
The time series of (850 hPa), (850 hPa), and (850 hPa) anomalies, and the combination of these anomalies ()(850 hPa) at locations of maximum easterly (westerly) wind for the 1999 Hue HRF event are shown in Figs. 8a and 8b with lines of different colors (legend in the bottom-right corner of Fig. 8a). Peak values of the (850 hPa), (850 hPa), (850 hPa), and ()(850 hPa) anomalies and u(850 hPa) when the 1999 Hue HRF event occurred on 2 November are clearly shown in Figs. 8a and 8b. The correlation coefficients between the two time series of ()(850 hPa) and u(850 hPa) at locations north and south of the HRF cyclone center exceed 0.9 with a confidence level of 95%. The difference between these two time series is attributed to the exclusion of the annual-mean and annual variation modes of u(850 hPa). The exclusion of these modes does not affect the in-phase constructive interference of the three monsoon modes that result in the peak intensity of an HRF cyclone.
For the past three decades, only 31 of 125 identified CSVs evolved into HRF events. Does this in-phase constructive interference of the three monsoon modes (, , ) occur with the other 79 non-HRF vortices?1 Two three-dimensional scatter diagrams project the days away from the date of the maximum 850-hPa easterly wind anomalies (Fig. 8c) and westerly wind anomalies (Fig. 8d) of (, , ) on the three axes. The occurrence date of an HRF event is designated as day 0 (large red dot), when the maximum easterly (westerly) wind anomalies of the three monsoon modes (, , ) are in phase. All HRF events are represented by only one scatter point, while the other 79 cold surge vortices (small blue dots) are shown away from this scatter. Apparently, the in-phase constructive interference mechanism of the three (, , )(850 hPa) monsoon modes works only for the formation of all 31 HRF events.
b. Analysis of streamfunction tendency
The data resolution is sufficient for a tropical synoptic disturbance to be properly portrayed by the streamfunction,, and streamfunction tendency, , to show the synoptic development of the HRF cyclone. Dynamically, vorticity advection and vortex stretching primarily contribute to the streamfunction tendency (Sanders 1984). The zonal velocity can be measured by the negative north–south gradient of the streamfunction. On 1 November 1999, a newly developed cold surge flow began to interact with the already exiting CSV, while it was crossing the South China Sea (not shown). This interaction led to the intensification of this vortex and the occurrence of the HRF event (Fig. 2g), as indicated by a negative peak value in the time series (Fig. 9a) on 2 November at 15°N, 108°E, the northern periphery of the Hue HRF cyclone.
As shown in Fig. 9a, the minima of the (925 hPa), (925 hPa), (925 hPa), (925 hPa), and (925 hPa) time series coincide on 2 November 1999. The signature of the Hue HRF event is demonstrated by this in-phase constructive interference of streamfunction anomalies of the three monsoon modes. Following Fig. 8c, a scatter diagram of (925 hPa), (925 hPa), and (925 hPa) is shown in Fig. 9c. The in-phase constructive mechanism of the three monsoon modes works for the occurrence of all Vietnam HRF events; negative minima of () anomalies at 925 hPa are coincident, as indicated by the large red dot. In contrast, the () scatters (dark blue dots) of the 79 cold surge vortices are distributed around the scatter of the HRF events. Apparently, the in-phase constructive mechanism does not work for nonflood CSVs. Corresponding to streamfunction anomalies shown in Fig. 9a, the time series of the streamfunction tendency (, ) anomalies at 925 hPa for the Hue HRF event are shown in Fig. 9b. As expected, these time series are individually in quadrature with the time series of their corresponding streamfunction anomalies shown in Fig. 9a. All three monsoon modes have minimum values for their streamfunction tendency and appear coincidently about a day ahead of the occurrence of the Hue HRF event.
Spatially, how was the HRF event developed by three monsoon modes? The streamline charts of wind anomalies at 925 hPa superimposed with the () streamfunction tendency anomalies at 925 hPa, respectively, on 1 November 1999 are shown in Fig. 10. The easterly disturbance depicted by the (, )(925 hPa) streamline chart (Fig. 11a) propagated along the red-dashed track to central Vietnam; negative and positive (925 hPa) values are separated through the center of the vortex (red cross). Compared with the ΔV(925 hPa) streamline chart (Fig. 10e), a vortex depicted by the (925 hPa) streamlines (Fig. 10b) covering Indochina and the tropical South China Sea provides a link to the eastern Asian cold surge. The east–west zone of negative (925 hPa) along the track of the easterly disturbance is coincident with the tropical cyclonic shear zone in the western tropical Pacific–tropical Southeast Asia. As shown from the (925 hPa) streamline chart (Fig. 10c), the 30–60-day mode exhibits a cyclonic shear flow from South Asia, turning northeastward in the Philippines Sea toward Japan and east Siberia. Because the 30–60-day mode is a slowly varying intraseasonal mode, the (925 hPa) tendency develops a favorable environment over southeastern Asia to facilitate the intensification of the HRF cyclone. The in-phase constructive mechanism of the HRF event is confirmed by a pattern resemblance between , (925 hPa) (Fig. 10d) and Δ(V, )(925 hPa) (Fig. 10e).
Two important features of the streamfunction tendency analysis related to the occurrence of the HRF event are 1) The maximum intensity of the HRF cyclone is reached by the spatial coincidence of the three monsoon modes’ low centers and 2) the in-phase occurrence of the maximum deepening rate of the three monsoon lows appears about 1 day ahead of the HRF event. A three-dimensional scatter diagram for () similar to the () diagram in Fig. 9c is shown in Fig. 9d. These two figures show that the streamfunction and the streamfunction tendency are in phase with the constructive interference of the three monsoon modes for all of the HRF CSV events in Vietnam, but are not for the remaining 79 non-HRF CSVs, as substantiated by Figs. 9c and 9d.
4. Maintenance of HRF events
a. Rainfall of HRF events
Using three rainfall datasets, including WMO station measurements, Tropical Rainfall Measuring Mission (TRMM; Simpson et al. 1996) data, and Asian Precipitation–Highly-Resolved Observational Data Integration Toward Evaluation (APHRODITE; Yatagai et al. 2009) daily rainfall, spectral analyses of rainfall at the centers of all HRF events (red dots in Fig. 8a) were performed. To save space, only results from the station rainfall dataset are shown; the three monsoon modes emerge from rainfall spectra (Fig. 11a). Our concern here is with the relationship among the three monsoon modes of rainfall to generate the HRF event. The Hue event is used to illustrate this relationship in terms of the longitude–time diagrams of the anomalies shown in Fig. 12, superimposed with tracks of easterly disturbances.
It is revealed from the longitude–time diagram of rainfall across Hue (17°N, 107°E) during late September–early December (Fig. 12a) that the westward rainfall propagation of the easterly disturbance appears to be modulated by the eastward migration of some low-frequency modes. Rainfall almost disappeared east of 160°W before 2 November and diminished west of 110°E after the occurrence of the Hue HRF events (2 November 1999). The cause for these interesting features of rainfall propagation may be inferred from the longitude–time diagrams of the anomalies. Tracks of easterly disturbances superimposed on the longitude–time diagram of indicate that most of these tracks coincide with the westward propagation of anomalies. The Hue HRF event occurred on 2 November 1999 (the dark blue line along the abscissa in Fig. 12), when the maximum and arrived simultaneously at 107°E (dark blue line along the ordinate in Fig. 12). The maximum then moved farther eastward and the activity was suppressed west of 110°E by the 30–60-day mode. The slow eastward migration of is basically attributed to the modulation of the eastward-propagating mode. The 12–24-day mode, , is related to the East Asian cold surge, coupled with the midlatitude upper trough–ridge system. Thus, anomalies, exhibiting a clear eastward propagation, are modulated by the 30–60-day mode, so that is inactive when is weak east of 150°E during late September and most of October. The time series of (thick solid red lines) at the rainfall center of the Hue HRF event are presented in Figs. 12b, 12c, and 12e. The Hue HRF event occurred when maxima of anomalies appeared coincidently at Hue on 2 November 1999.
One may question whether rainfall maxima of the monsoon modes always occur simultaneously at the center of every HRF event in Vietnam. The time series of (thin black lines) at all HRF event centers in Vietnam are shown in Fig. 11b. The occurrence dates of these HRF events are designated as day 0. The averaged time series of anomalies for all HRF events is presented by thick red lines, superimposed with a ±1 standard deviation (std dev) zone (yellow). Although and are slow-varying modes, maxima of the anomalies appear simultaneously for all HRF events. For non-HRF cases, a three-dimensional scatter diagram of is shown in Fig. 11c, but none is located at day 0; maxima in the rainfall anomalies for the three monsoon modes do not occur simultaneously, as all HRF events do.
b. Maintenance of HRF events
Water vapor flux can be expressed as , where g, po, p, and q are gravity, surface pressure, pressure, and specific humidity, respectively. In addition, Q may be split into rotational () and divergent () components:
where and . The streamfunction and potential function of water vapor flux are expressed as and , respectively. The maintenance of rainfall by the convergence of the water vapor flux may be expressed by the approximated water vapor equation:
Water vapor is transported by a dipole of cells to Hue (inferred from the 925-hPa streamline chart), but the Hue HRF event is maintained by the convergence of water vapor flux illustrated by the (, , P) chart (Chen 1985) in Fig. 13a.
The Hue HRF rainfall center should be primarily formed by (, , ) and maintained by the convergence of the water vapor flux for the three monsoon modes. The centers of rainfall and water vapor flux convergence anomalies of the three monsoon modes appear in central Vietnam (Figs. 13d–f). Apparently, the rainfall center of the Hue HRF event is maintained by these three monsoon modes combined, (, , ) (Fig. 13c). This argument is supported by the close resemblance of pattern and magnitude between Fig. 13c and [, where = yearly mean value of (·)] (Fig. 13b). The slight difference between these two charts is caused by the absence of the annual variation mode (not shown).
Because P is primarily maintained by convergence of the water vapor flux, the propagation properties of (Fig. 12) should be followed by and (): eastward propagation of (, ) and (, ), and westward propagation of (). This inference is supported by the center coincidence of (, , ), (, , ), and (, , ), shown in Figs. 13d–f, respectively. The prominent hydrological features in the x–t diagrams and the three-dimensional scatter diagrams of and (, , ) can be referenced to those of (Figs. 12 and 11c). However, our major concern is whether the in-phase constructive interference mechanism maintaining the Hue HRF event (Fig. 13) is applicable to other events. As shown in Figs. 11b and 11c, positive maximum anomalies always appear simultaneously at the HRF event centers. This in-phase constructive inference of maximum anomalies is maintained by the corresponding maximum () anomalies. It is substantiated by the three-dimensional scatter diagram of (, , ) in time (Fig. 11c) that the maximum water vapor flux, which convergences from the three monsoon modes (), does not occur simultaneously for the remaining 79 cold surge vortices when they arrive at central Vietnam.
5. Forecast advisory
The late fall HRF event in Vietnam is a local rainfall extreme, but it develops from the formation of a CSV around the Philippines through the midlatitude–tropical interaction of the weather systems over the eastern Asian continent and the western Pacific Ocean. It takes several days for this vortex to reach the Vietnamese coast. Thus, the forecast time required to predict the development of an HRF event from its parent CSV is several days. So far, only an extreme heavy rainfall event at Hue/Quang Nam on 24–26 November 2004 was simulated by Truong et al. (2009), using the Regional Atmospheric Modeling System (RAMS; Pielke et al. 1992; Cotton et al. 2003) over a domain that includes the South China Sea and its surrounding regions. The major purpose was to test the rainfall produced by this event with different parameterization schemes for a 2-day period. The midlatitude–tropical interaction related to the formation of an HRF cyclone/event is not completely covered by the RAMS domain. The period needed to predict the formation of the HRF event from the date the CSV formed is also more than 2 days.
NCEP issues 1-week GFS weather forecasts with 0.5° × 0.5° resolution (Kanamitsu et al. 1991; Yang et al. 2006). Based on the statistics obtained from diagnostic analyses of the synoptic development and dynamic hydrological processes leading to the formation of the HRF events, it seems feasible to develop a forecast advisory for the occurrence of HRF events 5 days in advance. As soon as the parent CSV is identified in the vicinity of the Philippines, these analyses, supplemented with the operational global and regional forecasts, will help improve the HRF forecast. Special features observed in this study will be adopted to generate parameters that can potentially be used for this forecast advisory.
a. Illustration of the forecast advisory
Based on salient features of the Vietnam HRF events analyzed in sections 2–4, a forecast advisory for the occurrence date of an HRF event is illustrated by the flowchart shown in Fig. 14, which consists of four steps.
1) Identification of the parent CSV in the vicinity of the Philippines and the northeastern Asian surface low
The daily 925-hPa streamline charts superimposed with the TRMM rainfall prepared with the GFS initial analysis are used to identify the formation of the CSV around the Philippines and the surface pressure low in northeastern Asia associated with the cold surge flow related to this CSV. The identified CSV should be verified by the NCEP tropical strip surface analysis and observation chart (SRRS 2011), and the surface analysis map of the Thai Meteorological Department (TMD 2011). The northeastern Asian surface pressure low related to the identified CSV should be validated by both charts from the Japan Meteorological Agency (JMA) surface analysis and NCEP SRRS.
2) Potential formation of the Vietnam HRF event and the northeastern Pacific explosive cyclone
Within 2–6 days, the three basic characteristics (speed, size, and rainfall) of any identified parent CSV depicted by the GFS forecasts may reach the threshold values for an HRF event. On this date, the deepening rate of surface pressure at the center of the identified northeastern Asian surface low center may attain its threshold value, too. On this particular date, this GFS CSV may potentially evolve into an HRF event.
3) Simultaneous occurrence date of maximum intensity of the Vietnam HRF event, the North Pacific anticyclone, and the northwest Pacific explosive cyclone
If the requirement is met by the parent CSV identified in step 2, time series of u(850 hPa) at the northern and southern peripheries of the forecast HRF event, the northern and southern rims of the forecast explosive cycle, and the southern rim of the forecast North Pacific anticyclone are prepared to determine whether simultaneous occurrence of the maximum intensities of these three weather systems can take place.
4) Date of the formation and maintenance of the HRF event by the constructive interference mechanism
The formation date of a forecast HRF event verified by step 3 should be confirmed by the following two requirements: 1) maximum anomalies of 850-hPa zonal winds (, , ) at the northern and southern rims of the HRF event coincide and 2) maximum anomalies of rainfall (, , ) and (, , ) anomalies merge at the HRF rainfall center.
If the date of the HRF event predicted by steps 1 and 2 is used to filter out the non-HRF vortex from the forecast, the predicted occurrence date of an HRF event is determined in step 3. Step 4 further confirms the accuracy of this prediction. The preparation of u(850 hPa) and P anomalies of the three monsoon modes in step 4 may have an end effect of the bandpass-filtering process with GFS forecasts. This impediment may be resolved by the following two approaches:
Two-week forecasts can be operationally issued by the NCEP GFS. The forecast data, which cover the period beyond the formation date of a potential HRF event, may be about 8–12 days. These forecast data can be added to the N-day time series, so the end effect of the bandpass-filtering process is reduced.
The N-day time series of the GFS forecast data can be extended by adding the mirror-image data of the previous 8–12 days to reduce the end effect.
b. Feasibility test of the forecast advisory
Since the NCEP GFS 1-week forecasts (Kanamitsu et al. 1991; Yang et al. 2006) became available in 2004, six Vietnam HRF events have been identified by the Dartmouth Flood Observatory. With the consent of the NCEP Environmental Model Center, forecasts of another six HRF events (with rainfall > 400 mm day−1) before 2004 were performed with the NCEP GFS. These HRF events are used to test the feasibility of this forecast advisory. Steps 1 and 2 of the forecast advisory used to select the potential occurrence of HRF events are classified as necessary requirements. Steps 3 and 4 of this advisory were applied to confirm that this selection is classified as sufficient as defined below. The feasibility of the proposed forecast advisory will be tested by these requirements.
1) Necessary requirements
The non-HRF cold surge vortices are eliminated by the following necessary requirements:
Surface pressure tendency for GFS forecasted explosive cyclone
The observed surface pressure tendencies, , a day before all selected explosive cyclones formed (Fig. 15a) are hPa day−1 (Sanders and Gyakum 1980) within ±1 standard deviation with a 95% confidence interval (Table 1). In contrast, the forecast pressure tendencies for the selected 12 HRF cyclones/events are hPa day−1 within ±1 std dev with a 95% confidence interval (Table 1). The differences between and for the 12 events are consistently hPa day−1, that is, an underforecast bias of the GFS. Thus, hPa day−1 is accepted as the threshold surface pressure tendency for a surface low to become an explosive cyclone in the GFS forecast.
Characteristic maximum speed, size, and rainfall of HRF events
The characteristic maximum speed, size, and rainfall of the 12 HRF events and 13 non-HRF cold surge vortices for 2004–09 are shown in Figs. 15b–d. Statistics of these characteristic variables during this period and the past three decades are presented in Table 1. Magnitudes of these variables for the 12 HRF events are close to the long-term (1979–2009) mean values, but there are no significant changes for cold surge vortices. Magnitudes of characteristic variables in the 12 GFS forecasted HRF events are about 10% smaller in maximum speed and size, and 35% off in rainfall, compared to the corresponding observational cases. Regardless of the differences between the forecasted and observed HRF events, the standard deviations of all forecasted HRF events with 95% confidence intervals are smaller than both the corresponding observed and all HRF events. It is inferred from this statistical contrast between the forecasted and observed HRF events that the feasibility test of the forecast advisory with limited cases is statistically confident, although we hope that forecasts of all 31 HRF events can be used for this test.
The differences in the characteristic variables between the forecasted HRF events and the observed non-HRF cold surge vortices are shown in Figs. 15b and 15c and Table 1, from which a set of threshold enhancement values for the three characteristics of forecasted HRF events can be adopted. For the development of a CSV into an HRF cyclone/event predicted by the GFS, the threshold enhancements of maximum speed, size, and rainfall should be 70%, 100%, and 100%, respectively, on the predicted occurrence date of an HRF event.
2) Sufficient requirements
To warrant the potential occurrence of an HRF event 5 days in advance, using the GFS forecasts, the following two sufficient requirements should be met.
Simultaneous occurrence of maximum intensities in the three weather systems (HRF event, N Pacific anticyclone, and NW Pacific explosive cyclone)
The composite time series of observed uo(850 hPa) at both the northern and southern peripheries of the HRF cyclone center 5 days prior to and 5 days after the occurrence of the 12 HRF events are shown in Fig. 16a. Occurrence dates of these HRF events are designated as day 0. Because the time needed by 12 parent CSVs to become HRF events is 2–5 days, the forecast time series of uM(850 hPa) consists of observed uo(850 hPa) prior to the formation of the parent CSVs. The u(850 hPa) time series for the northwest Pacific explosive cyclone and the North Pacific anticyclone are shown in Fig. 16b. The westerlies of an explosive cyclone overlap the northwest rim of the North Pacific anticyclone, so the u(850 hPa) time series at the southern periphery of the explosive cyclone is used for the northern rim of the North Pacific anticyclone. The time series of tropical trade easterlies are generated at the longitude of the explosive cyclone center. Despite the forecast bias, maximum intensities for the three weather systems surprisingly occur simultaneously in all 12 HRF events. The first sufficient requirement for predicting the formation of these HRF events is met when their parent cold surge vortices formed.
Constructive interference mechanism of HRF events and rainfall maintenance mechanism of HRF rainfall centers
Zonal wind of HRF event
The time series of u(850 hPa) for day-N forecasts over a period of 6 months (August–January) are prepared, where N is the number of days between the formation of a parent CSV and the occurrence of an HRF event. Spectral peaks for the three monsoon modes emerge from the spectra of the uM(850 hPa) time series at the northern (Fig. 17a) and southern (not shown) peripheries of the forecasted cyclones of HRF events. Signals from the three monsoon modes are well preserved by the GFS forecasts on the dates of the HRF events, several days after the parent cold surge vortices formed. Following Fig. 8, the occurrence dates of maximum (, , )(850 hPa) anomalies away from the occurrence dates of HRF events are used to prepare a three-dimensional scatter diagram in Fig. 17b. All 12 HRF events occur when maximum (, , )(850 hPa) anomalies coincide, as required by the constructive interference mechanism. During October–November of 2004–09, 25 cold surge vortices formed in the vicinity of the Philippines: six became HRF events, 13 remained as cold surge vortices, and six grew into tropical cyclones. None of the non-HRF cold surge vortices met the requirements for the constructive inference mechanism.
Spectra peaks for the three monsoon modes are distinct for the N-day forecast rainfall time series at the 12 HRF rainfall centers (Fig. 17c). As shown in Fig. 11b, the composite time series of observed rainfall anomalies for the three monsoon modes at rainfall centers of these HRF events (Fig. 17d) indicate that these rainfall centers are formed by the three monsoon modes. The time series of forecast rainfall anomalies for the three monsoon modes on day N (Fig. 17d) also show that the maximum rainfall anomalies of these monsoon modes coincide with the dates of the HRF events. Evidently, the formation of the 12 HRF events is well predicted by the GFS model forecasts, as indicated by the composite time series of forecast anomalies of the three monsoon modes. Rainfall is maintained by convergence of the water vapor flux. It is inferred from the well-predicted formation mechanism that the maintenance mechanism of the HRF rainfall centers are also well predicted by the GFS.
The NCEP GFS forecasts of the 12 HRF events in central Vietnam were applied to test the feasibility of the proposed forecast advisory with these 12 events and show promising forecast advisories. However, to make the proposed 5-day forecast advisory operationally feasible, 2-week NCEP GFS operational forecasts are needed.
6. Concluding remarks
The late fall HRF events often bring flooding disasters in Vietnam. Therefore, it is important to accurately predict the occurrence of these events. The regional numerical weather prediction model limits its forecast domain and time, while the global prediction model likely underforecasts the intensities of the dominant features of the weather systems related to the formation of HRF events. A feasible alternative to supplement and improve the operational weather forecasts of HRF events is to develop a forecast advisory, based on pertinent features of the synoptic conditions and dynamic hydrological processes involved with the formation of HRF events. It has been shown through the synoptic and diagnostic analyses of dynamic and hydrological processes for the formation of the 31 HRF events that the important features related to the formation of the Hanoi event appear in 31 events as identified by the Dartmouth Flood Observatory during the past three decades (1979–2009). Applying five important features, a forecast advisory for these events is illustrated by a flowchart (Fig. 14), which consists of two basic elements: 1) necessary requirements through synoptic analysis and 2) sufficient requirements through dynamic and hydrological analysis.
Since 2004, the 1-week global forecasts have been operationally issued by the NCEP GFS. Six Vietnam HRF events occurred during 2004–09. The GFS forecasts for another six HRF events before 2004 were performed by the NCEP/Environmental Modeling Center. These two groups of forecasts during August–January were used to test the feasibility of the proposed forecast advisory. Based on the results of this test, the forecast advisory for the occurrence date of an HRF event in central Vietnam may be issued, if the following requirements are met.
Necessary requirements: Threshold deepening rate of the forecast explosive cyclone is hPa day−1, and threshold enhancements of three basic characteristics (maximum speed, size, and rainfall) of the HRF event are 70%, 100%, and 100%, respectively.
Sufficient requirements: The three sufficient requirements are met by all 12 HRF events.
This test, completed in this study, demonstrated that the proposed forecast advisory for the HRF events is feasible.
The analyzed initial field and 7-day forecast are operationally issued every 6 h by the NCEP GFS. As soon as the operational data become available, the potential HRF event can be determined by steps 1 and 2 of the necessary requirements. The forecast advisory for the occurrence of an HRF event can then be issued as long as steps 3 and 4 meet the necessary requirements. The outcome of step 1 requires a manual checking, but diagnostic analysis performed in steps 2–4 can be automated, if the software needed for diagnostic analysis is in place. Based on our experienced in testing the proposed forecast advisory of HRF events in central Vietnam, the entire process of preparing this forecast advisory requires less than 3 h and possibly keeps its pace with the GFS forecast cycle every 6 h.
Further improvements to this forecast advisory warrant the following future efforts. Currently, the NCEP GFS forecasts are available only for 12 HRF events in Vietnam. Hopefully, the GFS forecasts make more events available, so that more reliable statistics of the test can be achieved. The current operational global forecasts are limited to 1 week. Difficulty to handle the end effect would occur, when applying the bandpass filtering of forecast data to isolate anomalies of the three monsoon modes. This difficulty could be reduced by extending the forecast period to 2 weeks. The time needed for the parent CSV to become an HRF event in the 31 identified cases is about 2–6 days. The 2-week forecasts would offer a buffer time of 8–12 days for the application of a bandpass filter to forecast data. Such forecasts would be generated by The Observing System Research and Predictability Experiment (THORPEX) Interactive Grand Global Ensemble (TIGGE) project (Bougeault et al. 2010). The end effect of the bandpass-filtering problem will be resolved in the operational application of the forecast advisory introduced.
The cold season HRF event occurs not only in central Vietnam, but also in other Southeast Asian countries. During the past 5 years, HRF events have brought disasters to most countries around the South China Sea:
2007 Jakarta event (2 February 2007), Java, Indonesia (http://news.bbc.co.uk/2/hi/asia-pacific/6328873.stm);
2008 Hanoi event (31 October 2008), northern Vietnam (http://www.thanhniennews.com/2008/Pages/20081115134845 043760.aspx);
2008 Kuantan event (29 November 2008), Malaysian Peninsula (http://reliefweb.int/node/ 334708);
2010 Butuan event (16 January 2010), Mindanao, Philippines (http://www.philstar.com/Article.aspx?articleId=650755&publicationSubCategoryId=200); and
2011 Bangkok event (26 October 2011), Thailand (http://www.bbc.co.uk/news/world-asia-pacific-15471849).
Johnson and Houze (1987) pointed out that the Winter Monsoon Experiment (WMONEX, Greenfield and Krishnamurti 1979) provided little information about the convective system that produces the HRF event. The large-scale environmental circulations in various Southeast Asian regions are different. The formation and development mechanism of HRF events in these regions may not be the same. However, the diagnostic analysis used to explore the formation and development mechanism of HRF events and the forecast advisory used to supplement forecasts for the occurrence of HRF events made by the regional and global forecast models for central Vietnam may be applied to these Southeast Asian regions. Hopefully, this research effort will be of use to mitigate the disasters caused by HRF events in these regions.
This study was partially sponsored by the Cheney Research Fund, NSF Grant ATM-0836200. Ming-Cheng Yen’s effort was supported by NSC Grant NSC100-2111-M-008-014. We thank professor Jun Matsumoto of Tokyo Metropolitan University for motivating us to explore the cause of the Vietnam heavy rainfall/flood events during the late fall. Finally, comments and suggestions offered by the three reviewers were very helpful to the improvement of this paper.
Fifteen CSVs became tropical storms/cyclones.