Experiments using the Weather Research and Forecasting (WRF) Model were conducted to investigate the effects of multiscale motions on the genesis of Typhoon Manyi (2001) in the western North Pacific. The precursor signal associated with this typhoon genesis was identified as a northwest–southeast-oriented synoptic-scale wave train (SWT). The model successfully simulated the genesis of the typhoon in the wake of the SWT. Further experiments were conducted to isolate the effects of the SWT, the intraseasonal oscillation (ISO), and high-frequency (shorter than 3 days) eddies in the typhoon formation.
Removing the SWT in the initial and boundary conditions eliminates the typhoon genesis. This points out the importance of the SWT in the typhoon genesis. It was noted that the SWT strengthened the wake cyclone through southeastward energy dispersion. The strengthening wake cyclone triggered multiple episodes of strong sustained convective updrafts, leading to aggregation of vertical vorticity and formation of a self-amplified mesoscale core vortex through a “bottom up” development process. Removing the ISO flow eliminates the typhoon genesis, as the ISO significantly modulated the strength of the SWT through accumulation of wave activity. In the absence of SWT–ISO-scale interaction, the southeastward energy dispersion was weakened significantly, and thus the strengthening of the wake cyclone did not occur. As a result, the successive strong sustained convective updrafts disappeared. Removing the high-frequency eddies did not eliminate the typhoon genesis but postponed the genesis for about 36 h.
Tropical cyclone (TC) genesis is a complex process in which multiple temporal- and spatial-scale motions are involved. In the tropical western North Pacific (WNP), the most active TC genesis basin in the world, there exist pronounced synoptic-scale (3–10 day) activity and atmospheric intraseasonal (10–90 day) oscillation that impact the genesis (Li and Wang 2005). The TC seasonal mean state in the WNP is characterized by warm sea surface temperature (>27°C), high humidity, monsoon trough, and low-level cyclonic vorticity. While these environmental conditions favor TC formation (Gray 1968), it is synoptic-scale disturbances or intraseasonal oscillation that actually lead to individual TC genesis events (Li 2012). Different from the tropical Atlantic, where the most popular precursor disturbance is African easterly waves (Burpee 1972; Landsea 1993), the following types of pregenesis disturbances were identified in the WNP: Rossby wave train induced by the energy dispersion of a preexisting TC (Holland 1997; Li et al. 2003, 2006; Li and Fu 2006; Ge et al. 2010; Krouse and Sobel 2010), northwest–southeast-oriented synoptic wave train (SWT) or tropical depression (TD)-like disturbances (Lau and Lau 1990; Takayabu and Nitta 1993; Chang et al. 1996; Dickinson and Molinari 2002; Frank and Roundy 2006; Li 2006), Pacific easterly waves (Chang et al. 1970; Ritchie and Holland 1999), and mixed Rossby–gravity waves and equatorial waves (Frank and Roundy 2006). Among these low-level precursor disturbances, SWT has the highest frequency (Fu et al. 2007; Xu et al. 2013). It is suggested that as the northwest–southeast-oriented SWTs move northwestward, their energy disperses southeastward leading to the development of the wake cyclone and TC formation (Tam and Li 2006). Recently, Wang et al. (2012) applied a pouch theory (Dunkerton et al. 2009; Wang et al. 2010a) to the northwestward-propagating SWTs in the WNP and suggested SWTs may provide a favorable environment for TC formation. However, observational analyses (e.g., Li and Fu 2006; Fu et al. 2007) showed that not all SWTs develop into TCs. With this being the case, by what processes do TCs develop in SWTs, and why do some SWTs spawn TCs while others do not?
It has been shown that atmospheric 10–90-day intraseasonal oscillation (ISO) is most pronounced in the WNP in boreal summer and it strongly impacts TC formation (Liebmann et al. 1994; Yamazaki and Murakami 1989; Maloney and Hartmann 2000a,b; Sobel and Maloney 2000; Fu et al. 2007; Aiyyer and Molinari 2008; Hogsett and Zhang 2010). A key question is through what processes the ISO influences TC formation. One possible route is through barotropic energy conversion (e.g., Maloney and Hartmann 2000a; Hsu and Li 2011). Another is through its impact on the strength and structure of the SWT- or TD-type disturbances (Zhou and Li 2010). A recent modeling study by Cao et al. (2014) demonstrated that the ISO can influence TC formation through both a dynamic effect (primarily via low-level convergence and cyclonic vorticity) and a thermodynamic (moisture) effect.
Besides the strong intraseasonal and synoptic variability, the WNP is also the active region of high-frequency (at a period of shorter than 3 days) eddies (HFEs). These eddies have a typical spatial scale of 200 km or less, often including mesoscale convective systems and small-scale convective vortices. Small-scale convective clouds with a length scale of 10 km or so are sometimes called vortical hot towers (VHTs). They are often observed during the TC genesis period (Hendricks et al. 2004; Montgomery et al. 2006). It was noted that during the TC formation period these small-scale VHTs may merge into a mesoscale convective system (MCS) (Houze 1982). A key question related to the role of these HFEs is whether their occurrence is critical for predicting the TC formation. In other words, without the information of high-frequency features in the initial condition, can numerical models generate realistic high-frequency eddies, given accurate large-scale circulation patterns and predict observed TC formation?
In this study, we investigate the effects of multiple-scale motions such as SWT, ISO, and HFE on the formation of Typhoon Manyi in July 2001. As shown by Fu et al. (2007), the formation of this Category-4 typhoon was associated with a typical SWT in the WNP. To illustrate the relative roles of multiscale motions in the TC genesis, various idealized numerical experiments were performed using the Advanced Research version of the Weather Research and Forecasting (WRF-ARW) Model. In the control experiment, all scale motions are considered. In the subsequent sensitivity experiments, we intentionally isolate the roles of the SWT (with a time scale of 3–10 days), the ISO flow (10–90 days), and HFE (shorter than 3 days). The rest of this paper is organized as follows. In section 2, we describe the model setup and various numerical experiments. In section 3, we examine the effect of the SWT on the TC formation. The influences of the ISO and high-frequency eddies are examined in section 4 and 5, respectively. Finally a summary is given in the last section.
2. Model setup and numerical experiments
Typhoon Manyi was identified as a tropical depression with maximum surface wind of 15 knots (kt; 1 kt = 0.51 m s−1) by the Joint Typhoon Warning Center (JTWC) at 0600 UTC 1 August 2001. Just 18 h later, at 0000 UTC 2 August 2001, Tropical Cyclone Manyi formed with maximum sustained surface wind of 35 kt. Manyi rapidly developed into a category 4 typhoon after its genesis. The observed maximum wind speed reached 120 kt on 4 August 2001.
The WRF-ARW modeling system, version 3.3 (Skamarock et al. 2008), is used for the simulation of Typhoon Manyi. The model setup includes quadruple domains, shown in Fig. 1a. Three outer domains are stationary and have grid resolutions and dimensions of 54 km and 163 × 145, 18 km and 301 × 280, and 6 km and 667 × 481, respectively. The finest mesh is movable and has a grid spacing and dimensions of 2 km and 406 × 406. This innermost domain moves following the typhoon center from the southeast (BEGIN) to northwest (END) for a simulation period of 120 h as shown in Fig. 1a. All domains have 35 vertical sigma levels, and the model top is defined at 50 hPa.
a. Control experiment
The initial conditions and the outermost domain lateral boundary conditions are obtained from 6-hourly National Centers for Environmental Prediction (NCEP) final analysis (FNL) at 1.0° resolution. No artificial initial bogus vortex was used. We first conducted several test runs with different initial times and physics schemes to determine what initial time and which physical schemes can best simulate the observed track and evolution of Manyi as verified by JTWC best-track data. The results presented have the following physics packages: Yonsei University (YSU) boundary layer scheme (Noh et al. 2003; Hong et al. 2006), the Rapid Radiative Transfer Model (RRTM) for longwave radiation (Mlawer et al. 1997), the Dudhia shortwave radiation scheme (Dudhia 1989), the WRF single-moment 6-class microphysics scheme (WSM6) (Hong et al. 2004), the modified version of the Kain–Fritsch cumulus parameterization (Kain and Fritsch 1993) for the outer two domains, and no cumulus parameterization for the 6- and 2-km-resolution domains. The model was initialized at 1200 UTC 30 July 2001, 60 h prior to Manyi genesis time (0000 UTC 2 August 2001) when Manyi reached the maximum surface wind of 35 kt as reported by JTWC. At the initial time, the perturbation associated with Manyi was quite weak, with a minimum sea level pressure (MSLP) of about 1008 hPa, and was located to the southwest of the subtropical high (Fig. 1a). This information is from the FNL data.
In comparison to the JTWC best track, the control simulation (CTL) reproduces well the observed TC intensity and track (Figs. 1a and 1b). The model was integrated for 120 h (5 days). During that period, Manyi developed from a weak disturbance with MSLP of about 1008 hPa to a category 4 typhoon with the MSLP about 920 hPa. Figure 1a shows that the simulated storm track is close to the observed one during the 120-h integration. The time when Manyi was first identified from the simulation is at 1200 UTC 30 July, 42 h earlier than the first observed location (denoted by black dot) at 0600 UTC 1 August 2001 from JTWC. The simulated storm track initially moved westward and then turned northwestward. The simulated genesis location (red typhoon mark) was a little north to the real genesis location (black typhoon mark). After the genesis, the simulated track appears to be closer to the observed one. Figure 1b indicates that the model well reproduced the intensity evolution of Manyi. During the 120-h integration period, both simulated and observed storms developed from a tropical disturbance (as shown in Fig. 1b at initial time) to a tropical depression (0600 UTC 31 July–1800 UTC 1 August), tropical storm (0000 UTC 2 August–0000 UTC 3 August), and typhoon (from category 1 at 0600 UTC 3 August to category 4 at 1200 UTC 4 August).
b. Sensitivity experiments
Given that the control experiment successfully simulates the TC formation and successive evolution, further experiments are performed to investigate the relative roles of the SWT, the ISO flow, and HFEs in affecting the TC genesis. The synoptic-scale disturbances, ISO, and HFEs were obtained by applying 3–10-day, 10–90-day, and shorter-than-3-day bandpass filtering (Murakami 1979) to the FNL data, respectively. Figure 2 shows the patterns and evolutions of the synoptic-scale disturbances during the TC genesis period. The SWT had a typical northwest–southeast-oriented structure (Lau and Lau 1990) and propagated northwestward (Fig. 2). The structure of the 10–90-day ISO at the model initial time shown in Fig. 3a is not similar to a typical ISO structure at active phase. Figure 3b shows HFE at the model initial time denoting that the initial high-frequency eddies were quite active.
Table 1 describes each of our numerical experiments. The first sensitivity experiment (referred to as EXP_SWT) is for the role of the SWT in the TC formation. It is the same as the control experiment except that the synoptic-scale disturbances were removed from the initial and lateral boundary conditions (Figs. 2a and 3c,d).
The second and third sensitivity experiments were designed to understand the influences of the 10–90-day intraseasonal oscillation and high-frequency eddies on the TC genesis, respectively. The model setup in these two experiments is similar to that in EXP_SWT except that the ISO (HFE) fields were removed from the initial and lateral boundary conditions in the second (third) sensitivity experiment (Figs. 3e,f). The two experiments are named EXP_ISO and EXP_HFE, respectively (Table 1).
To clearly illustrate the simulated SWT patterns, two additional experiments, referred to as EXP_ISO_SWT and EXP_HFE_SWT, were conducted. In the former we remove both the ISO and SWT, whereas in the latter we remove both the HFE and SWT. The difference between EXP_ISO and EXP_ISO_SWT will show the SWT pattern and evolution in the absence of the ISO flow, whereas the difference between EXP_HFE and EXP_HFE_SWT will illustrate the SWT evolving patterns in the presence of the ISO flow, both without the HFE (Table 1).
3. Role of the SWT in Manyi formation
By comparing the control simulations with the three simulations that had no SWT, we will focus on those processes by which the SWT affects TC formation. First, Fig. 4 shows the time–vertical sections of tangential velocity, radial velocity, vertical velocity, relative vorticity, and divergence during the 120-h integration period in CTL. All fields in Fig. 4 are averaged within a radius of 200 km from the storm center. A red dashed line in Fig. 4 denotes the TC genesis time (when the maximum surface wind speed exceeds 17 m s−1). Note that the vortex developed rather slowly in the earlier stage. For example, the tangential velocity increased only about 2 m s−1 during the early 42 h (Fig. 4a). In contrast, the vortex experienced a greater intensification rate in both the tangential wind and vorticity fields a few hours prior to the genesis (Figs. 4a and 4b). This is consistent with the two-stage conceptual model proposed by Wang (2014). Another interesting feature is its oscillatory development. For example, a weakening in the vertical velocity occurred after the occurrence of a strong convective updraft at about 24 h (Fig. 4c). The weakening was accompanied by the slight decrease of tangential wind and vorticity near the surface (Figs. 4a,b) and occurrence of low-level outflow (Fig. 4d) and divergence (Fig. 4e) during 24–30 h. Such an oscillatory developing characteristic has been found in the previous studies (e.g., Li et al. 2006). It was suggested that such low-level downdraft and outflow were associated with stratiform precipitation (Houze 1997). Stratiform precipitation often occurred with a typical profile of convergence in the midlevel and divergence in low levels after the development of a strong convective updraft. The downdraft and low-level divergence tend to decrease the vorticity/tangential velocity and suppress vortex development. To ensure the continuous TC development, a transition from stratiform precipitation to the development of a new convective updraft is needed (Mapes and Houze 1995; Raymond et al. 1998; Tory et al. 2006; Raymond and Sessions 2007; Houze et al. 2009). Figure 4e shows that a new convective updraft and low-level convergence occurred again at about 36 h. A stratiform precipitation regime developed again after 36 h. However, this time low-level divergence did not occur. Thus the sustained convergence in the low level after 36 h (Fig. 4e) is favorable for the continuous growth of tangential wind and vorticity. Wang (2012) suggested that weak low-level divergence was associated with weak low-level cooling in a moist environment.
To demonstrate the structure and evolution of convective and stratiform precipitation and their roles in Manyi genesis, these two components were separated following Braun et al. (2010). As in their analysis, grid points with surface rainfall rates greater than 20 mm h−1 or with rainfall rates twice as large as the average of the nearest 24 neighbors were classified as convective precipitation. If a grid point is designated as convective in such a way, its nearest neighbors (within one grid distance) are also designated as convective. To identify convective columns in which significant precipitation was not yet reaching the surface, columns with upward vertical motions greater than 3 m s−1 or cloud liquid water greater 0.5 g kg−1 were also denoted as convective. All remaining grid columns with surface precipitation greater than 0.1 mm h−1 were classified as stratiform. The vertical velocity and divergence over the convective and stratiform regions in the control simulation were averaged over each region, weighted by the fraction of the total number of grid columns in the convective and stratiform classifications. Figure 5 shows the computed vertical velocity, divergence, and heating profiles for the stratiform and convective precipitation regimes. In a conditionally unstable atmosphere, low-level convergence triggers active convection with vertically “penetrating” ascending motion (Fig. 5a) and divergence in the upper level (Fig. 5b). The maximum heating rate associated with the deep convection occurs in the middle troposphere (Fig. 5e) so that lower tropospheric convective instability decreases as the convection increases. Eventually the convection becomes less active. Then stratiform precipitation appears with downdrafts (updrafts) below (above) the midlevel (Fig. 5c); as a result, convergence occurs in the midlevel, while divergence appears in both low and upper levels (Fig. 5d).
Figure 5 shows that the convective maximum vertical velocity (denoted by yellows lines) appears as a precursor of stratiform midlevel convergence (Fig. 5d). This suggests that the midlevel vorticity increase follows the low-level spinup associated with low-level convergence and deep convection. This implies that the pathway of Manyi formation is more like a “bottom up” development process. A comparison of the magnitude of convective and stratiform diabatic heating shows that the main diabatic heating source of the vortex development is attributed to the convective process (Figs. 5e and 5f).
A comparison of the CTL and EXP_SWT simulations indicates that the SWT, with its cyclonic vorticity and high moisture content, does provide a favorable condition for the continuous oscillatory development in CTL. Figure 6 shows the time–height sections of area-averaged tangential wind, vorticity, vertical motion, and divergence in EXP_SWT. Because there is no storm formation in EXP_SWT, the area average here is simply based on the vortex center location in CTL. Note that there were no sustained strong convective updrafts after 28 h (Fig. 6c). In general, the magnitudes of vertical motion and diabatic heating were much weaker in EXP_SWT than in CTL throughout the entire integration period (Fig. 6). Given that the sole difference between CTL and EXP_SWT is the presence of the SWT, it is conceivable that the existence of cyclonic circulation and wet column associated with the SWT is crucial in providing a favorable environment for triggering the multiple episodes of strong sustained convective updrafts. The removal of the SWT in EXP_SWT eliminates the reoccurrence of the new convective updrafts, which is necessary for the genesis of Manyi.
To illustrate how the model captures the observed SWT pattern and its evolution in CTL, we plotted the difference fields between CTL and EXP_SWT. We removed HFEs from the simulation output before deriving the SWT pattern. Figure 7 shows that the simulated SWT pattern is similar to the observed (Fig. 2). Manyi developed in the cyclonic vorticity region of the wave train. The amplitude of the simulated SWT appears a little stronger than the FNL data, possibly owing to the fact that the model spatial resolution is much higher than that in the FNL data.
Figure 7 shows that the TC development was accompanied by the strengthening of the wake cyclone within the SWT. What causes the strengthening of the cyclonic circulation? We hypothesize that the development of the wake cyclone is associated with the Rossby wave energy dispersion of the SWT. To show this, an E vector was calculated to highlight how the wave energy propagates as the wave train itself moves northwestward (Hoskins et al. 1983; Trenberth 1986; Sobel and Bretherton 1999; Li and Fu 2006). The E vector is defined as
where the overbar denotes time averaging and and denote zonal and meridional wind components associated with synoptic-scale disturbances. The E vector indicates energy propagation direction during a specified period. Figures 8a and 8b show the horizontal maps of the calculated E vector at 850 hPa based on the observational data (Fig. 2) and the control simulation (Fig. 7) during the genesis period from 1200 UTC 30 July (the model initial time) to 0000 UTC 2 August (Manyi genesis time). There is clear southeastward energy dispersion in both observations and the control simulation.
Previous studies suggested that the formation of a mesoscale core vortex with closed cyclonic circulation is necessary for TC formation (Simpson et al. 1997; Karyampudi and Pierce 2002). Under a background cyclonic vorticity, cumulus-scale convective clouds or VHTs may merge into a mesoscale vortex system (Hendricks et al. 2004; Montgomery et al. 2006, 2009). To illustrate the relationship between the small-scale convective systems and a mesoscale vortex, time filtering (Murakami 1979) and spatial scale separation (Maddox 1980; Xu 2011) techniques were applied to the model outputs. First, a 3-day high-pass filtering was applied to obtain the high-frequency fields. Then, a spatial-scale-separation technique was further applied to the temporally filtered fields to isolate the mesoscale (>50 km) and smaller-scale (<50 km) motions. Figure 9 illustrates the filtered high-frequency wind fields (top panel) and associated smaller-scale motion (bottom panel) in CTL. Their mesoscale components are shown in Figs. 10a, 10d, and 10i, respectively. These figures indicate that the smaller-scale motion and mesoscale motion were effectively separated from the total high-frequency wind fields.
The evolution of the mesoscale (>50 km) component of wind at 850 hPa is shown in Fig. 10. Note that in the early development stage (e.g., at 18 h), the smaller-scale vorticity was at least one order of magnitude greater than the mesoscale vorticity (see Figs. 9d and 10a). Three weak mesoscale vorticity centers can be identified as A, B, and C in Fig. 10a. In subsequent hours, the mesoscale vorticity center moved toward each other and merged into a larger vorticity center with a nearly closed cyclonic circulation (as highlighted by the red dashed line in Fig. 10d). After that, the mesoscale vortex weakened during 24–30 h owing to low-level divergence and downdraft associated with the stratiform precipitation. Several mesoscale positive vorticity systems denoted by D, E, and F started to develop again after 48 h. They eventually merged into a single mesoscale core vortex with closed cyclonic circulation (Figs. 10f–k). It is the development of this mesoscale core vortex that signifies Manyi formation at 60 h.
The timing of merging of the mesoscale vortices shown in Fig. 10 is consistent with the multiple stratiform and convective episodes illustrated in Fig. 5. A comparison of the CTL and EXP_SWT shows that smaller-scale VHTs also occurred in the early stage in EXP_SWT but they were much weaker than those in CTL owing to the lack of the SWT organization. Mesoscale vorticity perturbations also developed in earlier stage in EXP_SWT; however, they never developed into a strong mesoscale core vortex, as shown in Fig. 10k, owing to the less favorable environment in EXP_SWT. This confirms the important role that the SWT plays in maintaining the successive development of sustained convective updrafts, which is necessary for Manyi genesis.
4. Influences of ISO on the cyclone genesis
In the previous section we emphasize the role of the SWT. However, observational analyses (e.g., Li and Fu 2006; Fu et al. 2007) showed that not all SWTs develop into TCs. Previous studies also suggested that intraseasonal oscillations may greatly modulate the SWT (Zhou and Li 2010). This prompts us to investigate the influence of the ISO on Manyi genesis.
In experiment EXP_ISO, we removed the 10–90-day ISO field from both the initial and lateral boundary conditions. Removing ISO fields eliminates the genesis of Manyi. The MSLP near the TC genesis region is around 1005–1008 hPa during the entire 120-h integration period in EXP_ISO. How does the ISO flow impact the TC genesis? Figure 11 shows the time–vertical cross sections of area-averaged vorticity, vertical motion, and divergence fields in EXP_ISO. A strong convective updraft occurred in the early developing stage (at 24 h, Fig. 11b) with a magnitude similar to that in CTL. During this time, prominent VHTs (not shown) occurred in a similar way as those in CTL (Fig. 9) even though the amplitude is a little weaker. However, in the absence of the ISO fields, the successive occurrence of sustained strong convective updrafts after the stratiform precipitation as observed in the CTL did not happen here. As a result, no mesoscale core vortex developed.
It is likely that the ISO may affect the storm development through both a direct and an indirect effect. The direct effect of the ISO is through the modification of the background low-level cyclonic vorticity and moisture content. As shown in Fig. 3b, a positive 850-hPa vorticity anomaly associated with the ISO appeared in the genesis region. Such a cyclonic vorticity anomaly, on one hand, may promote a stronger vorticity segregation process and favor the development of the mesoscale core vortex and, on the other hand, it may increase background moisture through Ekman pumping–induced boundary layer convergence (Li et al. 2006; Cao et al. 2014).
In addition to the direct effect, the ISO may indirectly affect the TC formation through its impact on the strength of the SWT. As discussed in the previous section, the wake cyclone is reinforced by the Rossby wave energy dispersion. It is interesting to note that the southeastward group velocity was weakened significantly in the absence of the ISO flow (Fig. 8c). This implies that the ISO flow may modulate the amplitude of the SWT and thus impact the TC development.
Figure 12 shows the evolution of the SWT in EXP_ISO. Here the SWT patterns were derived based on the following method. The high-frequency eddies were first filtered out from the EXP_ISO and EXP_ISO_SWT outputs, and then the synoptic-scale flows were obtained by subtracting the high-pass filtered wind in EXP_ISO_SWT from its counterpart in EXP_ISO. At the initial time, the SWT pattern (Fig. 12a) is quite similar to that in CTL (Fig. 7a). However, because of the absence of the ISO, the SWT development differs markedly from that in CTL. The most notable difference is the intensity, structure, and propagation of the anticyclone to the northwest of the wake cyclone. It developed much faster and in a larger horizontal extent in CTL (Figs. 7c–e) than in EXP_ISO (Figs. 12c–e). This implies that the background ISO flow contributes to the greater development of the SWT. The enhanced wave train may disperse more energy southeastward, strengthening the wake cyclonic vorticity and leading to the successive development of sustained strong convective updrafts.
The strengthening of the SWT was possibly caused by the accumulation of wave activity associated with the ISO flow. A calculation of wave activity divergence based on 10–90-day wind field shows that there was indeed ISO wave activity convergence in the anticyclonic flow of the wave train (10°–20°N, 140°–150°E; Fig. 13). As the wave train moved northwestward, the anticyclone appeared in the region of the ISO energy accumulation. As a result, it amplified. The strengthened anticyclone dispersed wave energy southeastward, leading to further growth of the wake cyclone (Figs. 7, 8b,c, 12). Removing ISO eliminates the energy accumulation process. The sensitivity experiments support the notion that the full development of the prestorm vortex is necessary for TC genesis and that the ISO flow plays a critical role in Manyi genesis through its impact on the SWT.
5. Impacts of high-frequency eddies on Manyi genesis
Figure 3b shows that high-frequency eddies were quite active over the TC genesis region. The amplitude of these mesoscale vorticity perturbations is about 2 × 10−5 s−1, comparable to that of the SWT. How important were these initial high-frequency eddies in TC formation?
To address this question, an experiment EXP_HFE was carried out in which high-frequency eddies were removed from the initial and boundary conditions. The result indicates that removing the HFE did not eliminate the cyclone genesis but postponed the genesis time for about 36 h (see blue curves in Fig. 1b). The simulated TC track in EXP_HFE is quite similar to that in CTL (Fig. 3f).
The simulated SWT patterns appear to be not affected by removing HFE. Figure 14 shows the evolution of the SWT in EXP_HFE. Here the SWT pattern was derived by subtracting EXP_HFE simulation from EXP_HFE_SWT simulation. The structure and amplitude of the SWT in Fig. 14 were similar to those in CTL (Fig. 7), indicating that the influence of the high-frequency eddies on the structure and evolution of the SWT is small. The result above is further supported by calculated E vectors, which show the same southeastward energy propagation characteristic in EXP_HFE (Fig. 8d) as that in CTL (Fig. 8b).
The time–vertical sections of area-averaged vorticity, vertical motion, and divergence fields in EXP_HFE (Fig. 15) show similar evolution features as in the CTL except for a delayed genesis time. There were four sustained strong convective events at 24, 42, 70, and 84 h, respectively, prior to the genesis at 96 h (Figs. 15b,c). These convective updrafts coincided well with strong low-level convergence (Figs. 15e,f). They were accompanied by the subsequent occurrence of stratiform ascending motion at 10-km height, midlevel convergence, and low-level downdraft/divergence (Figs. 15d,h).
The major difference between EXP_HFE and CTL lies in the high-frequency eddy activities during the initial developing stage. These high-frequency eddies such as small-scale VHTs can quickly develop under a favorable large-scale environment, but the TC genesis is primarily controlled by synoptic-scale and lower-frequency systems (Karyampudi and Pierce 2002; Braun et al. 2010; Wang et al. 2010b). This experiment result suggests that the high-frequency eddies may be regarded as the byproduct of the large-scale activity and are not essential in determining the TC genesis. On the other hand, the inclusion of the initial high-frequency eddies is still important for accurately predicting the timing of the storm formation.
Figure 16 depicts the vorticity growth features in low level in both CTL and EXP_HFE. Here the vorticity is calculated based on an area average (within a radius of 200 km from the storm center) at the height of 1.5 km, which is near the 850-hPa level as shown in Figs. 2 and 3 from the FNL analyses. The vorticity growth experiences two stages—a slower development and a faster growth—consistent with previous studies (Nolan 2007; Wang 2014). The growth rates (denoted by black dashed lines) in the two stages are quite similar between the two experiments. The difference lies in that the removal of the initial HFE extended the period of the first slowly growing stage and thus postponed the genesis.
6. Summary and discussion
The northwest–southeast-oriented synoptic wave train (SWT) is a primary type of the precursor synoptic-scale disturbances associated with TC genesis in the WNP (Fu et al. 2007; Xu et al. 2013). To understand how a TC forms in the SWT, Typhoon Manyi in 2001 is selected as a study case. The WRF-ARW with a high-resolution nested grid configuration was employed. A number of numerical experiments were designed to understand the roles of various time-scale motions in the TC formation. In the control simulation, the total fields from the FNL data were included. The model simulated successfully the genesis of Typhoon Manyi in the wake cyclone of the SWT. A number of experiments were further performed to isolate the effects of the SWT, the ISO flow, and the HFE.
For Typhoon Manyi, removing the SWT eliminates the cyclone genesis. A diagnosis of the control simulation shows that Manyi indeed formed in the cyclonic vorticity region of the SWT. As the phase of the SWT moved northwestward, it emitted energy southeastward, strengthening the wake cyclone. The strengthened wake cyclone provided a favorable environment for Manyi formation. It triggers multiple episodes of strong sustained convective updrafts, leading to aggregation of mesoscale vertical vorticity and formation of a self-amplified mesoscale core vortex through a “bottom-up” development process. It is noted that the TC genesis underwent an oscillatory development with convective precipitation preceding stratiform precipitation. As in previous studies, the SWT serves as precursor for Manyi. Removing the SWT during the pregenesis period resulted in no genesis for the storm.
Removing the intraseasonal oscillation (ISO) flow also eliminates the cyclone genesis even though the initial SWT is the same one that produced a category 4 typhoon in the control simulation. The diagnosis of the model output indicates that the ISO exerted both a direct and an indirect impact. The direct impact is through the increase of background low-level cyclonic vorticity and moisture. The indirect impact is through the modulation of the SWT. It is noted that wave activity convergence associated with the ISO flow strengthened the anticyclone within the wave train. In the absence of SWT-ISO-scale interaction, the anticyclone did not develop so that the southwestward energy dispersion was weakened significantly and the wake cyclone could not develop. The nondeveloping wake cyclone would not be able to maintain the successive development of sustained convective updrafts, and as a result the TC would not form.
Removing the high-frequency eddies in the initial state did not eliminate the cyclone genesis but postponed the genesis for about 36 h. The vorticity and divergence evolution characteristics in this experiment are in general similar to those in the control simulation except with time lagging. It appears that large-scale forcing is capable of generating high-frequency vortices (e.g., VHTs and mesoscale vortices) necessary for TC genesis.
In summary, the cyclonic circulation within the SWT is crucial for Manyi’s formation. The ISO flow, on the other hand, can significantly modulate the SWT structure, propagation, and strength. While small-scale vortical hot towers (VHTs) grow quickly under large-scale cyclonic flow, the development of a mesoscale core vortex takes a much longer period. In general, more than one sustained deep convective episode is needed, in order for mesoscale vorticity anomalies to grow into an intense self-amplified vortex. The strengthening of the background cyclone appears necessary before a self-amplified vortex is set up. Whether or not a successor of sustained convective updrafts could occur depends on the development of the environmental cyclone.
The authors thank the three anonymous reviewers for their helpful comments. This work was supported by the National Science Foundation of China under Grant 41475047 and the National Basic Research Program of China (2013CB430104) and ONR Grants N00014-0810256 and N00014-1210450 and NRL Grant N00173-13-1-G902 and by the International Pacific Research Center, which is sponsored by the Japan Agency for Marine-Earth Science and Technology (JAMSTEC).
International Pacific Research Center Contribution Number 1069