Interactions Between Typhoon Megi (2010) and a Low-Frequency Monsoon Gyre*

2682 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 72

Interactions between Typhoon Megi (2010) and a Low-Frequency Monsoon Gyre*

MINGYU BI International Laboratory on Climate and Environment Change, and Key Laboratory of Meteorological Disaster of Ministry of Education, Nanjing University of Information Science and Technology, Nanjing, China, and International Paciﬁc Research Center, and Department of Atmospheric Sciences, University of Hawai‘i at Manoa, Honolulu, Hawaii

TIM LI International Paciﬁc Research Center, and Department of Atmospheric Sciences, University of Hawai‘i at Manoa, Honolulu, Hawaii

MELINDA PENG Naval Research Laboratory, Monterey, California

XINYONG SHEN International Laboratory on Climate and Environment Change, and Key Laboratory of Meteorological Disaster of Ministry of Education, Nanjing University of Information Science and Technology, Nanjing, China

(Manuscript received 21 September 2014, in ﬁnal form 3 March 2015)

ABSTRACT

The ARW Model is used to investigate the sharp northward turn of Super Typhoon Megi (2010) after it moved westward and crossed the Philippines. The NCEP analyzed fields during this period are separated into a slowly varying background-flow component, a 10–60-day low-frequency component representing the monsoon gyre, and a 10-day high-pass-filtered component representing Megi and other synoptic-scale motion. It appears that the low-frequency (10–60 day) monsoon gyre interacted with Megi and affected its track. To investigate the effect of the low-frequency mode on Megi, numerical experiments were designed. In the control experiment, the total fields of the analysis are retained in the initial and boundary conditions, and the model is able to simulate Megi’s sharp northward turn. In the second experiment, the 10–60-day monsoon gyre mode is removed from the initial and lateral boundary fields, and Megi moves westward and slightly northwestward without turning north. Tracks of the relative positions between the Megi and the monsoon gyre centers suggest that a Fujiwhara effect may exist between the monsoon gyre and Megi. The northward turning of both Megi and the monsoon gyre occurred when the two centers were close to each other and the beta drift was enhanced. A vorticity budget analysis was conducted. It is noted that the Megi moves toward the maximum wavenumber-1 vorticity tendency. The sharp change of the maximum vorticity tendency direction before and after the track turning point is primarily attributed to the change of the horizontal vorticity advection. A further diagnosis shows that the steering of the vertically integrated low-frequency flow is crucial for the change of the horizontal advection tendency.

1. Introduction * School of Ocean and Earth Science and Technology Contri- bution Number 9310, International Pacific Research Center Con- Tropical cyclones, to a large degree, move with the tribution Number 1110, and Earth System Modelling Center Contribution Number 50. environmental steering flow (Chan and Gray 1982), while the beta effect and tropical cyclone (TC) struc- tures also play a role (Fiorino and Elsberry 1989; Li and Corresponding author address: Tim Li, International Pacific Research Center, and Department of Meteorology, University of Zhu 1991). Despite the prevailing control of the large- Hawai‘i at Manoa, 1680 East-West Road, Honolulu, HI 96822. scale environmental flow and the great improvement E-mail: [email protected] made in track prediction, some cases of large track error

DOI: 10.1175/JAS-D-14-0269.1

Ó 2015 American Meteorological Society Unauthenticated | Downloaded 09/25/21 10:58 PM UTC JULY 2015 B I E T A L . 2683 still occur because of complex interactions of TCs with other scales of motion, such as the low-frequency mode (Carr and Elsberry 1990). It has been noticed that the largest error in the prediction of TC tracks is observed during TC recurvature and sudden turns. Previous studies have examined the relationship between midlatitude waves and TC recurvature. For example, George and Gray (1977) found that, if the 2 upper-level westerlies are greater than 25 m s 1 within 208 poleward of a typhoon, the typhoon may recurve. Hodanish and Gray (1993) compared the sharply and gradually recurving cases and found that typhoons begin to turn when upper-tropospheric westerlies penetrate to within 68 from the typhoon’s center. Holland and Wang (1995) found that typhoons tend to recurve into the FIG. 1. The JTWC best track of Megi from 13 to 24 Oct 2010. midlatitudes when a synoptic-scale trough moves away from East Asia into the subtropical ocean. waves or disturbances (Zhou and Li 2010). On one Some TCs in the tropical western North Pacific (WNP) hand, the ISOs can influence the development of occasionally experienced a sudden northward track synoptic-scale disturbances through barotropic energy change. Megi (2010) is one example wherein most oper- conversion (Maloney and Hartmann 2000). On the ational numerical models failed to predict the sharp turn other hand, the synoptic-scale perturbations may feed at the right time (see a more detailed description of this back to the ISO through the nonlinear rectification of super typhoon in section 2). Using a barotropic model, surface latent heat flux, diabatic heating, and eddy Carr and Elsberry (1995) investigated the sudden north- momentum transport (Hsu and Li 2011; Hsu et al. ward turning of a vortex when it approached a large-scale 2011). In general, the ISO in the WNP exhibits two monsoon gyre. They suggested that Rossby wave energy spectral peaks at periods of 30–60 days and 10–20 days dispersion associated with the monsoon gyre is critical in (Chen and Chen 1993; Chen and Sui 2010; Mao and causing the sudden northward-turning track. Chan 2005). Harr and Elsberry (1991) found that TC TCs in the WNP are usually accompanied with multi- tracks alternate between westward and recurving scale waves, including intraseasonal (10–90 day) oscilla- clusters at the intraseasonal time scale. Kim et al. tions (ISOs) and synoptic-scale (3–10 day) disturbances (2008) revealed a close relationship between land- (Li and Wang 2005; Li 2012). A typical example of low- falling TCs in the WNP and the phase of the MJO. frequency systems in the WNP is the monsoon gyre Thus, it is likely that the ISO flows may affect not only (Lander 1994). Li et al. (2006) demonstrated the effect of TC formation but also TC tracks. the monsoon gyre in promoting TC genesis in a 3D The objective of the current study is to investigate how model. One of the important aspects of the ISOs in the and to what extent Typhoon Megi (2010) interacted with tropical WNP is their interaction with synoptic-scale low-frequency monsoon gyre flow and how such an

5 2 24 FIG. 2. The wavelet power spectrum (10 W m ) of OLR over the region 58–238N, 1138– 1308E from 1 Aug 2010 to 1 Jan 2011. The black contour is the signiﬁcant level. The red line indicates Megi turning time.

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC 2684 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 72

25 21 FIG. 3. The evolution of 10–60-day bandpass-ﬁltered wind (vectors) and vorticity (shaded; 10 s ) ﬁelds averaged between 850 and 300 hPa. The black dots and red typhoon marks denote the centers of the low-frequency monsoon gyre and the TC center, respectively. interaction may have led to its sudden northward-turning In section 3, we describe the experiment design and track. The rest of the paper is organized as follows. In simulation results. In section 4, we investigate mecha- section 2, we present an overview of the evolution for nisms responsible for the sudden northward-turning Typhoon Megi and the nearby low-frequency circulation. track. Diagnostics of vorticity tendency are given in

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC JULY 2015 B I E T A L . 2685

FIG. 5. The JTWC best track (black) and simulated Megi tracks in the control (red) and NO_MG (blue) experiments. The pink and purple ovals indicate the before-turning and after-turning periods, respectively.

2. Overview of Megi and associated low-frequency flow Megi can be traced back to a pregenesis tropical depression that emerged east of the Philippines (around 1408E) early on 13 October 2011 (Fig. 1). The low pressure system strengthened throughout the day and became a named tropical storm, Megi, by 1200 UTC 13 October, when its central pressure fell to 998 hPa. In the following 3 days, Megi continued to develop while moving northwestward. It was upgraded to typhoon category by the Joint Typhoon Warning Center (JTWC) at 1200 UTC 16 October. It then moved west-southwestward and continued to strengthen, reaching its peak intensity of super typhoon by 1200 UTC 17 October, with a central minimum 2 pressure of 895 hPa and maximum wind speed of 72 m s 1. It caused 11 deaths, 16 injuries, and the evacuation of more than 200 000 people. After passing through the northern Philippines, Megi weakened a little and slowed down as it entered the South China Sea. Beyond 19 October, Megi moved in a northwest direction. By 0000 UTC 20 October, Megi’s track turned straight northward from its original westward and northwestward movement. The angle of pre- and post- turning tracks from before 0600 UTC 19 October to be- FIG. 4. The patterns of (a) the unfiltered initial wind field, yond 20 October was almost 908. Most operational TC (b) the 10-day high-pass-filtered wind field, and (c) the initial wind forecast models failed to predict such a rather sudden field in the NO_MG experiment averaged from 850 to 300 hPa at 0000 UTC 18 Oct 2010. The red sign indicates the typhoon track change, as most models predicted that Megi would location. continue moving westward in the following 24–72 h. As a result, track forecast error for Typhoon Megi is among section 5.Insection 6, additional experiments with dif- the biggest in recent years. Megi eventually made landfall ferent initial conditions are presented to verify our hy- over the southern China coast, and weakened quickly pothesis. The conclusions and discussion are given in the after landfall. By 1800 UTC 23 October, Megi was last section. downgraded to a tropical depression.

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC 2686 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 72

25 21 FIG. 6. The simulated low-frequency wind (vectors) and vorticity (shaded; 10 s ) averaged between 850 and 300 hPa in the control experiment from (a) 18 to (f) 23 Oct. The red marks indicate the typhoon location.

As stated in the introduction, previous studies have previous studies (Li and Wang 2005; Mao and indicated that ISOs may affect TCs in various ways. To Chan 2005). examine the structure and evolution characteristics of a The 10–60-day mode is here termed as the low- low-frequency mode during the period, we first analyzed frequency ISO mode. To investigate the relative impor- the power wavelet spectrum of outgoing longwave ra- tance of different temporal-scale motions in causing the diation (OLR) averaged over 88–238N, 1128–1308E. The sudden northward turning of Megi, all dynamic and ther- data used for this analysis are daily on 2.5832.58 grids modynamic fields are divided into a high-frequency com- for a period of 14 months (1 March 2010–30 April 2011). ponent (with a 10-day high-pass filter), an ISO component Our calculations indicate that there is a significant peak (with a 10–60-day bandpass filter), and a slowly varying at the intraseasonal (10–60 day) period (Fig. 2). The mean flow (with use of a 60-day low-pass filter). A Lanczos power spectrum analysis results are consistent with filter (Duchon 1979) was applied in these calculations.

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC JULY 2015 B I E T A L . 2687

using a Mercator projection. The SST field from daily real-time global sea surface temperature (RTG_SST) analysis data is used to update the SST every 24 h during the simulation. The model has a single mesh with a horizontal resolution of 18 km. The model physics includes a WSM6 microphysics scheme, a Kain–Fritch convective scheme, a Dudhia shortwave radiation parameterization, and an RRTM longwave radiation parameterization. In the control simulation, the model initial and boundary conditions use the NCEP analysis on 1.083 1.08 grids with the boundary values updated every 6 h. The simulation starts at 0000 UTC 18 October 2010, 36 h before Megi made its sharp northward turning, and is integrated for 5 days. Figure 4a shows the initial wind field in the control simulation. To isolate the role of the FIG. 7. The track of the monsoon gyre center in the control low-frequency monsoon gyre, we conduct a sensitivity experiment (black) and in the experiment where Megi was experiment in which we filter out the 10–60-day mode removed (red). of all the prognostic variables in the initial and lateral boundary fields. This 10–60-day mode represents well As a part of the steering flow, the ISO flow may affect the MG structure in the WNP. This sensitivity experi- TC track. Figure 3 shows the evolution of the vertically ment is named the NO_MG simulation. Figure 4b il- integrated (from 850 to 300 hPa) ISO flow of the Na- lustrates the pattern of the initial low-frequency MG tional Centers for Environmental Prediction Final flow. The initial wind field without the low-frequency Analysis (NCEP FNL analysis; NOAA/National Centers MG flow in the NO_MG experiment is shown in Fig. 4c. for Environmental Prediction 2000) from 14 to 24 Octo- Figure 5 shows the tracks for both the control and ber. It is interesting to note that the vertically integrated sensitivity experiments. In the control experiment, the ISO flow has a wavelike structure, propagating from TC initially moves westward and crosses the Philippines. southeast to northwest in the low latitudes. This wave- It takes a sharp northward turn afterward, at about like structure is mostly evident in the lower troposphere. 1178E, over the South China Sea. The simulated TC A black dot in Fig. 3 denotes the center of a cyclonic track (red line) is very close to the JTWC best track vortex within the 10–60-day wave. This cyclonic vortex (black line). resembles a monsoon gyre (MG) depicted in Lander To diagnose how well the model captures the ob- (1994) and Carr and Elsberry (1995). The most in- served structure and evolution of the low-frequency teresting aspect is the spatial phase relation between mode and its phase relationship with the TC, we plot- Typhoon Megi (denoted by a red typhoon mark) and the ted in Fig. 6 the model-simulated low-frequency wind MG center. On 14 October, Megi was located to the east field in the control simulation. Here, a 5-day running- of the MG center. Megi moved anticlockwise along the mean method (subtracted from a 30-day running mean) MG steering flow, and on 16 October it arrived to the was used to extract the 10–60-day low-frequency wind northeast of the MG center. In the subsequent days, Megi evolution. Since the simulation covers a period from 18 moved toward the MG center. At 1200 UTC 19 October, to 23 October, the NCEP analysis 2 days before and the typhoon center and the MG center nearly overlapped. 2 days after the integration period was used in the above Afterward, they moved together toward the north. calculation. It is noted that major structure and evolu- The analysis above suggests an MG–TC interaction tion characteristics of the low-frequency MG were well scenario. To examine the impact of the MG on the TC captured by the model, as compared with Fig. 3. For motion, we design and conduct numerical experiments example, on 18 October, the MG center was located in as described in the next section. the northern tip of the Philippines, while the simulated TC center is slightly to the east. On 20 October, both the MG center and Megi were located to the west of the 3. Model description and numerical experiments Philippines. After that, both the MG center and simu- The model used in this study is the Advanced Re- lated TC moved toward the north. Thus, both the TC search version of the WRF (ARW) Model. The model movement and the low-frequency MG evolution were domain covers a region of 908–1558E and 58S–408N well simulated.

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC 2688 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 72

FIG. 8. (left) The tracks of the MG and the typhoon centers in (a) the NCEP analysis and (c) the simulation; (right) the relative positions of the MG and typhoon centers in (b) the NCEP analysis and (d) the simulation. The origin (0, 0) in (b) and (d) indicates the middle point between the MG and typhoon center at each time (from 18 to 22 Oct).

In the NO_MG simulation, the TC continues moving that a vortex placed to the east of a large-scale MG may westward after crossing the Philippines, and there is no undergo a sharp northward turning because of steering sharp northward turning. The results suggest that the of the southerly flow caused by MG energy dispersion. presence of the low-frequency MG mode and its in- In this scenario, the vortex has little impact on the teraction with Megi might be essential to cause its MG flow. northward turning. The low-frequency circulation evolutions illustrated in Figs. 3 and 6 seem to suggest a two-way interaction scenario; that is, on the one hand, the MG influences 4. Interactions between Megi and the monsoon the TC track, and, on the other hand, it is influenced by gyre the TC. After the TC cyclonically moved toward the The impact of the MG flow on the TC track has been MG center and they became close to each other on studied by many investigators. For example, Carr and 20 October, they moved together to the north. To Elsberry (1995) demonstrated in a barotropic model demonstrate the effect of the TC on the low-frequency

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC JULY 2015 B I E T A L . 2689

of Megi with the 10–60-day MG mode that leads to its unusual track. Now the question becomes why Megi and the MG moved northward after 19 October. Examining Figs. 3 and 6, we see that the northward movement began when the centers of the MG and Megi became very close, but not before. It appears that the center of Megi and the MG are rotating cyclonically with respect to each other and getting closer. Could a Fujiwhara effect exist between Megi and the MG, even though their sizes and intensities are quite different? The Fujiwhara effect (Fujiwhara 1921, 1923) has long been identiﬁed as a fascinating effect caused by the binary interaction between two tropical cyclones. An early observational study of tropical cyclones shows that, when the distance between two cyclones is within 1400 km, they circle around each other; when the distance is less than 740 km, they could attract each other.

FIG.9.AsinFig. 8b, but starting from 15 Oct and ending on 24 Oct. The effect is also different when the two involved systems have different sizes and intensities. Many research efforts have been made to study the interaction of two flow, we conducted another sensitivity experiment in tropical cyclones through numerical simulations (Chang which we retained the low-frequency flow (with a time 1983; Ritchie and Holland 1993; Wang and Holland scale longer than 10 days) while removing the higher- 1995) and quantitative diagnostics (Wu et al. 2003). frequency eddies that included Megi in both the initial In Fig. 8, the tracks of the MG center and the Megi and lateral boundary conditions. All the experiment center are plotted in the Earth reference frame from 18 to settings are the same as those in the control run, ex- 22 October in Figs. 8a and 8c, and the trajectories of them cept for the different initial and lateral boundary in a reference frame centered at the middle point be- conditions stated above. In this new experiment, we tween the centers for Megi and the MG (centroid points) can examine the movement of the low-frequency MG are plotted in Figs. 8b and 8d. The tracks from the NCEP without the impact of TC Megi and other higher- analysis (Figs. 8a,b) and those from our model simula- frequency perturbations. tions (Figs. 8c,d) are very similar, so we will use the tracks Figure 7 shows how the MG center moves in this ex- from the NCEP analysis for discussion. periment. Instead of the northward movement in the The tracks on the Earth reference (Figs. 8a,c) show control simulation, now the MG center is moving west- that the centers of the MG and Megi are far apart on ward and northwestward. Therefore, this experiment 18 October, with Megi located on the eastern side of the indicates there is an interaction between Megi and the Philippines, while the center of MG is on the western MG during the northward journey of both. side. In the next 24 h, Megi moved rather quickly over It is worth pointing out that the MG track in the the Philippines, while the MG moved slowly northward, above sensitivity experiment resembles the track of and they became much closer by 19 October. Beyond Megi in the NO_MG case. Given that both the 10–60- this point, the two centers’ movement was almost par- daymodeintheformercaseandtheTCvortexinthe allel, with Megi lagging behind for about 1 day. By latter case have the environmental mean-flow steering, 21 October, Megi caught up with the MG and moved it is likely that the seasonal mean flow steers both the ahead of it. TC and the MG in the two experiments. An examina- When the positions of Megi and the MG are plotted tion of the seasonal mean flow (i.e., 60-day low-pass- relative to each other (Figs. 8b,d), it becomes clear that filtered wind) confirms that the mean flow in the region Megi and the MG are attracted to each other and moved is indeed west-northwestward. toward each other in a rotating manner. For the 24 h In an observational data analysis, Carr and Elsberry between 18 and 19 October, the movement was fast, but (2000) showed that 39 out of 69 NOGAPS TC large-error it slowed down after 19 October. The rotating motion cases were related to TC interaction with a surrounding between them made a turn after 1800 UTC 20 October. cyclonic flow or vortex. This provides observational evi- By 21 October, the two centers were very close to each dence to support the argument that it is the interaction other (within computational errors of identifying the

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC 2690 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 72

21 FIG. 10. The radial proﬁle of the tangential wind (m s ) across the center of Megi in the east–west direction at 700 hPa for the control (red) and NO_MG experiment (blue). two centers), but it is clear that they never merged into gradual approach while undergoing an anticyclonic orbit, one. Instead, the two centers started to move away from then there is a capture, at which time a faster cyclonic each other beyond this point. One of the reasons that mutual orbit ensues at relatively constant distance, which Megi and the MG did not merge may be the signiﬁcantly is followed by a merger or an escape. In the Megi–MG different scales and characteristics each possessed. This case, there is an initial slow cyclonic orbit at far separa- is different from the traditional Fujiwhara effect be- tion, then a rapid approach with an anticyclonic orbit tween two tropical cyclones of similar characteristics. (which is consistent with LH), and then the two systems The rotating and attracting motions between Megi and get close enough to be considered merged. While two the MG are further illustrated in a domain covering a TCs that merged could become one entity, Megi and the larger area and longer time frame starting on 15 October MG remained separated even when their centers were (Fig. 9). Beyond the early counterclockwise rotation very close. This may be due to the very different scales motion between Megi and the MG, the attraction be- and characteristic of the two systems. The interesting tween them started around 1200 UTC 16 October, when Fujiwhara effect between Megi and a monsoon gyre the two systems came within 1000 km of each other. By warrants future in-depth investigation and is beyond the 19 October, the two centers were so close that Megi scope of the present study. actually followed the track of the MG to the north. The question remains as to why both Megi and the Lander and Holland (1993, hereafter LH) added more MG moved northward when they were in close prox- details to the typically observed interaction of two imity to each other. Our hypothesis is that the near tropical cyclones (binary interaction) to include phases overlapping of Megi and the MG caused a superposition of approach, capture, and orbit, followed by merger or effect of the beta drift. The effect of a planetary vorticity escape. Megi and the MG did not follow the typical in- gradient (beta) that causes a north-northwest movement teraction observed between two tropical cyclones as of tropical cyclones has been well studied (Rossby 1939, outlined in LH.InLH, the interaction begins with a 1948; Adem and Lezama 1960; Anthes and Hoke 1975;

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC JULY 2015 B I E T A L . 2691

FIG. 11. The wavenumber-1 wind in (a),(c) the control and (b),(d) the NO_MG simulation at (top) 1200 UTC 18 2 Oct and (bottom) 0000 UTC 19 Oct at 500 hPa. The shading indicates the wind speed (m s 1).

Kitade 1981; Holland 1983, 1984; Chan and Williams ventilation flow that corresponds well with the move- 1987; Fiorino and Elsberry 1989). Figure 10 shows the ra- ment. The same wavenumber-1 wind fields at 1200 UTC dial profile of the tangential wind across the center of Megi 19 October and 1200 UTC 20 October are shown in in the east–west direction for the control and NO_MG Fig. 12. During this period, Megi moved north- experiments. On 19 October, when Megi was changing northwestward (1200 UTC 19 October) and northward its course from westward to northwestward, the wind (at 1200 UTC 20 October), and the orientation of the profile was larger overall in the control than the one in wavenumber-1 asymmetry corresponds with the TC the experiment without the presence of the MG. We movement. On the other hand, when the MG is absent, extract the azimuthal wavenumber-1 wind field from the asymmetry is weak and not organized (as indicated the two experiments. The patterns at 500 hPa are dis- within the 800-km-radius circle). played in Fig. 11 for 1200 UTC 18 October and 0000 UTC 19 October. These two time snapshots rep- 5. Diagnosis of maximum vorticity tendencies resent the period while Megi moved mostly westward both in the control and in the NO_MG experiments. To understand in detail the role of the TC– The wavenumber-1 asymmetry displays mostly westward low-frequency MG interaction in Megi’s sudden

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC 2692 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 72

FIG. 12. The wavenumber-1 wind in (a),(c) the control and (b),(d) the NO_MG simulation at (top) 1200 UTC 19 2 Oct and (bottom) 1200 UTC 20 Oct at 500 hPa. The shading indicates the wind speed (m s 1). northward turning, we conducted a vorticity budget The vorticity tendency equation in a pressure co- analysis based on the argument that the TC moves to- ordinate can be written as follows: ward the direction of maximum vorticity tendency ›z ›z ›z ›z › (Holland 1983; Li and Zhu 1991). In both the control and 52 2 y 2 v 2 y f › u › › › › NO_MG experiments, we focus our diagnosis on two |{z}t |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}x y p y A periods: before turning and after turning. The before- B turning period includes 18 time periods at a 1-h interval ›v › ›y ›v › ›y 1 u 2 2 u 1 z 1 from 1900 UTC 18 October to 1200 UTC 19 October, as › › › › › › ( f ) . (1) |fflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflffl}y p p x |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}x y indicated by a pink circle in Fig. 5. The after-turning period C D also includes 18 time periods from 0300 to 2000 UTC 20 October, as indicated by a purple circle in Fig. 5. Here, t is time; p is pressure surface; u and y are zonal Special attention is paid to the vorticity tendency before and meridional wind, respectively; v is vertical p velocity; and after turning in the control simulation and the vor- f is the Coriolis parameter; and z is vorticity. Equation ticity tendency between the control and NO_MG (1) states that the vorticity tendency is determined by experiments. three major terms: the 3D vorticity advection term (B),

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC JULY 2015 B I E T A L . 2693

29 22 FIG. 13. The composite wavenumber-1 vorticity tendency fields (shaded; 10 s ) averaged between 850 and 300 hPa during (a),(c) before-turning and (b),(d) after-turning periods in (top) the control and (bottom) the NO_MG experiments. The black vector indicates the direction and magnitude of the maximum vorticity tendency averaged over a 400-km radius. the tilting term (C), and the divergence term (D). The maximum vorticity tendency in the control experiment friction term in the free atmosphere was neglected. For directs to the west before turning and changes toward each term above, a vertical integration (from 850 to the north after turning. This differs from the NO_MG 300 hPa) operator is applied. Besides, we mainly focus experiment, in which the maximum vorticity tendency is on the azimuthal wavenumber-1 component of the always toward the west. Therefore, the maximum vorticity tendency, because other higher-wavenumber wavenumber-1 vorticity tendency direction represents components do not contribute to the TC motion well the TC moving direction in both simulations. (Holland 1983; Li and Zhu 1991; Wu and Wang 2000). Given that the wavenumber-1 vorticity tendency can The left-hand side of Eq. (1) was calculated based on well depict the direction of TC movement, we further the vorticity difference between hour 11andhour analyzed specific physical processes/terms that give rise 0 (current time). to such a vorticity tendency change. Figure 14 shows The simulated wavenumber-1 vorticity tendency fields the wavenumber-1 component of each vorticity budget [corresponding to the left-hand side (LHS) of Eq. (1)] term on the right-hand side of Eq. (1) during the averaged during the before- and after-turning periods in before-turning period in the control simulation. As in the control and NO_MG simulations are shown in Fig. 13, a radial average from 0 to 400 km has been Fig. 13. The black arrows in Fig. 13 represent the di- applied in calculating the direction and amplitude of rection and magnitude of maximum vorticity tendency the maximum vorticity tendency in Figs. 14a–e.Given averaged over a 400-km radius. As one can see, the that major wavenumber-1 vorticity tendency is almost

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC 2694 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 72

29 22 FIG. 14. The composite wavenumber-1 vorticity tendency ﬁelds (shaded; 10 s ) averaged between 850 and 300 hPa during the before-turning period in the control simulation: (a) sum of terms B, C, and D; (b) term B; (c) term C; and (d) term D. (e) The azimuthal distribution of each vorticity tendency term averaged over a 400-km radius, with the red dashed line denoting the direction of TC movement. The black vector in (a)–(d) indicates the direction and magnitude of the maximum wavenumber-1 vorticity tendency averaged over a 400-km radius. concentrated near a radius of 350 km from the TC amplitude as the LHS of the vorticity equation center, such a radial domain selection is reasonable. (Fig. 7a), indicating that our vorticity budget analysis is Note that the sum of three terms in the right-hand side reliable. Before Megi’s turning, the vorticity advection of the vorticity equation has the same direction and term (including the beta effect) has the largest

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC JULY 2015 B I E T A L . 2695

FIG. 15. As in Fig. 14, but for the after-turning period in the control simulation. contribution projecting into the TC moving direction northward vorticity tendency is primarily contributed (red dashed line in Fig. 8e), while the tilting term and by the vorticity advection and the tilting term. Partic- the convergence term are relative small. In particular, ularly, comparing the before- (Fig. 14) and after- the divergence term has an opposite effect on TC turning (Fig. 15) periods, one may ﬁnd that the most movement (Fig. 8e). signiﬁcant change appears in the advection term: it The maximum vorticity tendency after turning changes by more than 908 in vorticity tendency di- exhibits a clear northward direction (Fig. 9a). The rection between the two periods, while the tilting term

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC 2696 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 72

FIG. 16. As in Fig. 14, but for the before-turning period in the NO_MG simulation. only alters by about 208. Again, the divergence term is MG, the advection term continues forcing the vortex to mostly opposed to the direction of TC movement. move toward the west during the two periods. While the The diagnosis of the wavenumber-1 vorticity tendency divergence term always tends to oppose the TC moving di- budget for the NO_MG case shows a consistent result: rection, the contribution from the tilting term is positive but that is, the vorticity advection term plays the biggest role in small in magnitude (Figs. 16 and 17). determining the maximum tendency direction and magni- To sum up, the vorticity budget analysis of both the tude (Figs. 16e and 17e). In the absence of the low-frequency control and NO_MG experiments points out that the

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC JULY 2015 B I E T A L . 2697

FIG. 17. As in Fig. 14, but for the after-turning period in the NO_MG simulation. most important cause of Megi’s sharp northward turning following diagnosis, we focus on the horizontal vorticity is the three-dimensional vorticity advection associated advection. with the atmospheric low-frequency flow. The three- To demonstrate the role of the mean-flow steering dimensional advection term consists of horizontal vor- effect, the wind field at each level is simply divided into ticity advection, the planetary vorticity advection, and two components: an area-mean flow and a perturbation the vertical vorticity advection. Our calculation shows component deviated from the mean. For reanalysis data, that the second and third terms are small. Thus, in the the steering flow is often defined as the average wind

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC 2698 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 72

29 22 FIG. 18. (a),(d) The total horizontal vorticity advection term (10 s ), (b),(e) the mean flow advection term, and (c),(f) the anomalous flow advection term averaged from 850 to 300 hPa during the after-turning period in the (top) control and (bottom) NO_MG experiments. The black vector indicates the direction and magnitude of the maximum vorticity tendency averaged over a 400-km radius. over a deep layer (850–300 hPa) within the radius of 5–78 and vorticity deviated from the mean. Figure 18 shows from the TC center (Carr and Elsberry 1990). For the layer-averaged (from 850 to 300 hPa) MA and AA higher-resolution model data, because of greater wind terms during the after-turning period for both the con- speed and smaller radius of maximum wind, the average trol and NO_MG cases. Again, only the wavenumber-1 wind in a smaller radius is more reasonable. Thus, in this component is retained. As shown in Fig. 18, the major study, the steering flow is defined as a layer average of difference between the control and NO_MG cases lies in 850–300 hPa in a radius of 400 km from the typhoon the mean-flow advection (one directing toward the center. With such a mean-flow definition, the vorticity of north and northeast, another directing toward the west the mean flow is always zero. Therefore, the horizontal and northwest), while the vorticity tendency due to the advection term (HAV) may be decomposed into two anomalous advection points to the same direction. terms: namely, the mean-flow advection (MA) and the Therefore, our sensitivity numerical experiments clearly anomalous-flow advection (AA): demonstrate the important role of steering of the low- frequency mean flow in causing the sharp northward ›z ›z › z0 › z0 › z0 › z0 2 2 y 52 ( ) 2 y ( ) 2 0 ( ) 2 y0 ( ) turning of Megi. u › › u › › u › › . |fflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflffl}x y |fflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl}x y |fflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}x y It is worth mentioning that there are some errors as- HAV MA AA sociated with the vorticity tendency calculation. For (2) example, the maximum vorticity tendency at the after stage in Fig. 15d is nearly twice as large as that in the In the equation above, u and y indicate the mean zonal before stage (Fig. 15c), even though the westward and meridional velocity at each level, and u0, y0, and z0 speeds of the No_MG experiment in the before stage indicate the perturbation zonal wind, meridional wind, (1900 UTC 18 October–1200 UTC 19 October) and

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC JULY 2015 B I E T A L . 2699

extracted, and it is displayed in Fig. 20. The intensity for Megi in the NCEP analysis is much stronger than that in the ECMWF analysis. To demonstrate that the interaction between Megi and the MG would impact the movement of Megi, we remove Megi in the ECMWF analysis and replace it with the Megi vortex extracted from the NCEP analysis. The track of Megi in this experiment is shown in Fig. 21, and it shows a new perfect track of Megi in the model. This new experiment re- inforces our hypothesis that proper interactions between Megi and the MG are responsible for their movement.

7. Conclusions and discussion Through a series of numerical experiments using the WRF Model, we investigated the role of the low- frequency monsoon gyre (MG) in the sharp northward turn of Super Typhoon Megi (2010) beyond 19 October, after Megi passed over the Philippines. The analysis fields from NCEP and ECMWF are divided into a slowly 21 FIG. 19. (top) Track and (bottom) maximum wind (m s )ofMegi varying background mean state (with a 60-day low-pass simulated using ECMWF interim reanalysis data. The red (black) dot filter), a 10–60-day ISO component representing the and line indicates the simulated (JTWC) track and intensity. MG, and a 10-day high-pass-filtered component representing Megi and other synoptic-scale components. The after stage (0300–2000 UTC 20 October) (shown in NCEP analysis data were first used as the initial and Fig. 5) were nearly the same. Such an error is likely due boundary conditions. In the first experiment (the control to a small TC center position error. Thus, what we rely experiment), complete analysis containing all three on more in the vorticity tendency calculation is its di- component flows is included, and the WRF Model is rection (rather than its amplitude) and its relationship able to capture the observed northward turning of Megi. with the TC moving direction. In the second experiment, the 10–60-day flow containing mainly the nearby MG is removed from the initial and lateral boundary conditions (NO_MG). In the absence 6. Experiments using ECMWF analysis of the low-frequency monsoon gyre, Megi experiences a It is arguable that the MG is part of the large-scale westward and slightly northwestward journey without flow and certainly would impact the movement of Megi turning northward, which is very different from the so that removing the MG would impact Megi. In this control experiment. In the third experiment, the higher- section, we examine the interaction between Megi and the frequency eddies containing Megi were removed (NO_ MG using different analysis fields as the initial and Megi). In this case, the MG center moves westward and boundary conditions to further demonstrate the in- has a similar track to that of the TC vortex in the NO_ teraction between Megi and the MG. In this set of exper- MG experiment. The results suggest the importance of iments, the European Centre for Medium-Range Weather the TC–MG interaction in causing the northward turn- Forecasts (ECMWF) interim reanalysis (ECMWF 2009) ing of both the TC vortex and the MG. is used. The reason for choosing this analysis is that the When tracks of the Megi center and the MG center ECMWF global model was not able to catch the were plotted in a centroid reference frame relative to northward motion of Megi. When the ECMWF analysis each other, a Fujiwhara effect appeared to be in play, is used, our model also failed to produce the northward modulating the motions of both the MG and Megi. In motion of Megi (Fig. 19a), and the track is very similar to the early stage, the two systems rotated counterclock- the track in the operational ECMWF model (figure not wise with each other. When the two systems were shown). Our hypothesis is that the intensity of Megi in roughly within 1000-km distance, attraction between the ECMWF analysis may be too weak (Fig. 19b)to them sped up. In the Earth reference frame, Megi and allow proper interactions between Megi and the MG. the MG moved parallel to each other, with Megi lagging With our filtering technique, the TC vortex associated behind the MG by about a day, after both turned to the with Megi in the NCEP and ECMWF analyses is north. In the centroid reference frame, the two centers

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC 2700 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 72

21 FIG. 21. (top) Track and (bottom) maximum wind (m s )of Megi simulated using ECMWF interim reanalysis data, but the TC is replaced with the vortex extracted from the NCEP FNL reanalysis. The red (black) dot and line indicates the simulation (JTWC) track and intensity.

maximum wavenumber-1 vorticity tendency directions. The change of the maximum vorticity tendency direction before and after the turning in the control experiment was primarily attributed to the change of horizontal advection. A further diagnosis showed that the mean-flow steering in the presence of the 10–60-day 21 FIG. 20. The wind (m s ) associated with Megi at 500 hPa in mode was crucial for the change of the maximum vor- (a) the NCEP analysis, and (b) the ECMWF interim reanalysis. 25 21 ticity tendency. The shading indicates the vorticity (10 s ). To further demonstrate that interactions between Megi and the MG played a critical role in their movement, we closest to one another around 21 October and then conducted additional experiments using data from the drifted away from each other without merging into one ECMWF interim reanalysis as the initial and boundary system like some pairs of tropical cyclones do with the conditions. The reason for these experiments is that the conventional Fujiwhara effect. This may be because the ECMWF model was not able to predict the northward scale and characteristics of the two systems are very turn of Megi. When the ECMWF reanalysis was used, our different. Idealized simulations will be conducted to model was not able to simulate the northward movement investigate and verify this interaction, which is similar to of Megi either. Examining the difference between the the conventional Fujiwhara effect between two tropical ECMWF and NCEP analyses shows that the vortex cyclones. representing Megi in the former is much weaker than in The northward turning of both Megi and the MG the latter. When the Megi vortex was removed from the was attributed to the enhanced beta drift when the ECMWF analysis and replaced by the Megi vortex two centers are nearly collocated, as shown by the extracted from the NCEP reanalysis, our model suc- wavenumber-1 wind fields before and after the turn- cessfully simulated the northward turning of Megi. ing. A vorticity budget analysis was performed for The results of the current study suggest that an accu- both the control and NO_MG experiments. It was rate TC track forecast requires proper representations found that the directions of TC movement in both the of both low-frequency ISO flows and the TC vortex before- and after-turning periods were consistent with itself. The low-frequency MG flow or the ISO flow, in

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC JULY 2015 B I E T A L . 2701 general, may affect not only TC genesis, but also TC ECMWF, 2009: ERA-Interim Project (updated monthly). Na- track. The failure in predicting Megi’s sharp northward tional Center for Atmospheric Research Computational and turning by many operational models may result from a Information Systems Laboratory Research Data Archive, accessed 6 August 2014, doi:10.5065/D6CR5RD9. lack of accurate representation of the atmospheric low- Fiorino, M., and R. L. Elsberry, 1989: Some aspects of vortex frequency (10–60 day) mode, the TC vortex, or both. structure related to tropical cyclone motion. J. Atmos. Sci., 46, 975–990, doi:10.1175/1520-0469(1989)046,0975: Acknowledgments. This work was supported by SAOVSR.2.0.CO;2. China National 973 Projects 2015CB453200 and Fujiwhara, S., 1921: The mutual tendency toward symmetry of motion and its application as a principle in meteorology. 2013CB430103, NSFC Grants 41475084 and 41375058, Quart. J. Roy. Meteor. Soc., 47, 287–293, doi:10.1002/ NRL Grant N00173-13-1-G902, the International Pacific qj.49704720010. Research Center, which is sponsored by the Japan ——, 1923: On the growth and decay of vortical systems. Quart. Agency for Marine-Earth Science and Technology J. Roy. Meteor. Soc., 49, 75–104, doi:10.1002/qj.49704920602. (JAMSTEC), the Jiangsu Shuang-Chuang Team, and the George, J. E., and W. M. Gray, 1977: Tropical cyclone recurvature and nonrecurvature as related to surrounding Priority Academic Program Development of Jiangsu wind-height fields. J. Appl. Meteor., 16, 34–42, doi:10.1175/ Higher Education Institutions (PAPD). 1520-0450(1977)016,0034:TCRANA.2.0.CO;2. Harr, P. A., and R. L. Elsberry, 1991: Tropical cyclone track characteristics as a function of large-scale circulation REFERENCES anomalies. Mon. Wea. Rev., 119, 1448–1468, doi:10.1175/ Adem, J., and P. Lezama, 1960: On the motion of a cyclone em- 1520-0493(1991)119,1448:TCTCAA.2.0.CO;2. bedded in a uniform flow. Tellus, 12, 255–258, doi:10.1111/ Hodanish, S., and W. M. Gray, 1993: An observational analysis of j.2153-3490.1960.tb01308.x. tropical cyclone recurvature. Mon. Wea. Rev., 121, 2665–2689, Anthes, R. A., and J. E. Hoke, 1975: The effect of horizontal doi:10.1175/1520-0493(1993)121,2665:AOAOTC.2.0.CO;2. divergence and the latitudinal variation of the Coriolis Holland, G. J., 1983: Tropical cyclone motion: Environmental in- parameter on the drift of a model hurricane. Mon. Wea. teraction plus a beta effect. J. Atmos. Sci., 40, 328–342, , . Rev., 103, 757–763, doi:10.1175/1520-0493(1975)103,0757: doi:10.1175/1520-0469(1983)040 0328:TCMEIP 2.0.CO;2. TEOHDA.2.0.CO;2. ——, 1984: Tropical cyclone motion: A comparison of theory Carr, L. E., and R. L. Elsberry, 1990: Observational evidence for and observation. J. Atmos. Sci., 41, 68–75, doi:10.1175/ , . predictions of tropical cyclone propagation relative to envi- 1520-0469(1984)041 0068:TCMACO 2.0.CO;2. ronmental steering. J. Atmos. Sci., 47, 542–546, doi:10.1175/ ——, and Y. Wang, 1995: Baroclinic dynamics of simulated tropical cyclone recurvature. J. Atmos. Sci., 410–425, doi:10.1175/ 1520-0469(1990)047,0542:OEFPOT.2.0.CO;2. 52, , . ——, and ——, 1995: Monsoonal interactions leading to sudden 1520-0469(1995)052 0410:BDOSTC 2.0.CO;2. Hsu, P.-C., and T. Li, 2011: Interactions between boreal summer tropical cyclone track changes. Mon. Wea. Rev., 123, 265–290, intraseasonal oscillations and synoptic-scale disturbances over doi:10.1175/1520-0493(1995)123,0265:MILTST.2.0.CO;2. the western North Pacific. Part II: Apparent heat and moisture ——, and ——, 2000: Dynamical tropical cyclone track forecast sources and eddy momentum transport. J. Climate, 24, 942– errors. Part I: Tropical region error sources. Wea. Fore- 961, doi:10.1175/2010JCLI3834.1. casting, 15, 641–661, doi:10.1175/1520-0434(2000)015,0641: ——, ——, and C.-H. Tsou, 2011: Interactions between boreal sum- DTCTFE.2.0.CO;2. mer intraseasonal oscillations and synoptic-scale disturbances Chan, J. C. L., and W. M. Gray, 1982: Tropical cyclone move- over the western North Pacific. Part I: Energetics diagnosis. ment and surrounding flow relationships. Mon. Wea. Rev., J. Climate, 24, 927–941, doi:10.1175/2010JCLI3833.1. 1354–1374, doi:10.1175/1520-0493(1982)110,1354: 110, Kim, J., C. Ho, H. Kim, C. Sui, and S. K. Park, 2008: Systematic . TCMASF 2.0.CO;2. variation of summertime tropical cyclone activity in the ——, and R. T. Williams, 1987: Analytical and numerical studies of western North Pacific in relation to the Madden–Julian oscil- the beta-effect in tropical cyclone motion. Part I: Zero mean lation. J. Climate, 21, 1171–1191, doi:10.1175/2007JCLI1493.1. flow. J. Atmos. Sci., 44, 1257–1265, doi:10.1175/ Kitade, T., 1981: A numerical study of the vortex motion with , . 1520-0469(1987)044 1257:AANSOT 2.0.CO;2. barotropic models. J. Meteor. Soc. Japan, 59, 801–807. Chang, S. W.-J., 1983: A numerical study of the interaction be- Lander, M., 1994: Description of a monsoon gyre and its effects on tween two tropical cyclones. Mon. Wea. Rev., 111, 1806–1817, the tropical cyclones in the western North Pacific during , . doi:10.1175/1520-0493(1983)111 1806:ANSOTI 2.0.CO;2. August 1991. Wea. Forecasting, 9, 640–654, doi:10.1175/ Chen, G., and C.-H. Sui, 2010: Characteristics and origin of quasi- 1520-0434(1994)009,0640:DOAMGA.2.0.CO;2. biweekly oscillation over the western North Pacific during ——, and G. J. Holland, 1993: On the interaction of tropical- boreal summer. J. Geophys. Res., 115, D14113, doi:10.1029/ cyclone-scale vortices. I: Observations. Quart. J. Roy. Meteor. 2009JD013389. Soc., 119, 1347–1361, doi:10.1002/qj.49711951406. Chen, T.-C., and J.-M. Chen, 1993: The 10–20-day mode of the 1979 Li, T., 2012: Synoptic and climatic aspects of tropical cyclogenesis Indian monsoon: Its relation with the time variation of mon- in western North Pacific. Cyclones: Formations, Triggers and soon rainfall. Mon. Wea. Rev., 121, 2465–2482, doi:10.1175/ Control, K. Oouchi and H. Fudeyasu, Eds., Nova Science 1520-0493(1993)121,2465:TDMOTI.2.0.CO;2. Publishers, 61–94. Duchon, C. E., 1979: Lanczos filtering in one and two di- ——, and Y. Zhu, 1991: Analysis and modeling of tropical cyclone mensions. J. Appl. Meteor., 18, 1016–1022, doi:10.1175/ motion: I: Asymmetric structure and sudden change of tracks. 1520-0450(1979)018,1016:LFIOAT.2.0.CO;2. Sci. China, 34B, 222–233.

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC 2702 JOURNAL OF THE ATMOSPHERIC SCIENCES VOLUME 72

——, and B. Wang, 2005: A review on the western North Pacific Rossby, C. G., 1939: Relation between variations in the intensity of monsoon: Synoptic-to-interannual variabilities. Terr. Atmos. the zonal circulation of the atmosphere and the displacements Oceanic Sci., 16, 285–314. of the semi-permanent centers of action. J. Mar. Res., 2, 38–55, ——, X. Ge, B. Wang, and Y. Zhu, 2006: Tropical cyclogenesis doi:10.1357/002224039806649023. associated with Rossby wave energy dispersion of a preexist- ——, 1948: On displacements and intensity changes of atmospheric ing typhoon. Part II: Numerical simulations. J. Atmos. Sci., 63, vortices. J. Mar. Res., 7, 175–187. 1390–1409, doi:10.1175/JAS3693.1. Wang, Y., and G. J. Holland, 1995: On the interaction of tropical- Maloney, E. D., and D. L. Hartmann, 2000: Modulation of eastern cyclone-scale vortices. IV: Baroclinic vortices. Quart. J. Roy. North Pacific hurricanes by the Madden–Julian oscillation. Meteor. Soc., 121, 95–126, doi:10.1002/qj.49712152106. J. Climate, 13, 1451–1460, doi:10.1175/1520-0442(2000)013,1451: Wu, C.-C., T.-S. Huang, W.-P. Huang, and K.-H. Chou, 2003: A MOENPH.2.0.CO;2. new look at the binary interaction: Potential vorticity di- Mao, J., and J. C. L. Chan, 2005: Intraseasonal variability of the agnosis of the unusual southward movement of Tropical South China Sea summer monsoon. J. Climate, 18, 2388–2402, Storm Bopha (2000) and its interaction with Supertyphoon doi:10.1175/JCLI3395.1. Saomai (2000). Mon. Wea. Rev., 131, 1289–1300, doi:10.1175/ NOAA/National Centers for Environmental Prediction, 2000: 1520-0493(2003)131,1289:ANLATB.2.0.CO;2. NCEP FNL Operational Model Global Tropospheric Analy- Wu, L., and B. Wang, 2000: A potential vorticity tendency di- ses, continuing from July 1999 (updated daily). National agnostic approach for tropical cyclone motion. Mon. Wea. Center for Atmospheric Research Computational and In- Rev., 128, 1899–1911, doi:10.1175/1520-0493(2000)128,1899: formation Systems Laboratory Research Data Archive, ac- APVTDA.2.0.CO;2. cessed 12 February 2012, doi:10.5065/D6M043C6. Zhou, C., and T. Li, 2010: Upscale feedback of tropical synoptic Ritchie, E. A., and G. J. Holland, 1993: On the interaction of tropical variability to intraseasonal oscillations through the nonlinear cyclone-scale vortices. II: Discrete vortex patches. Quart. J. Roy. rectification of the surface latent heat flux. J. Climate, 23, Meteor. Soc., 119, 1363–1379, doi:10.1002/qj.49711951407. 5738–5754, doi:10.1175/2010JCLI3468.1.

Unauthenticated | Downloaded 09/25/21 10:58 PM UTC