In this paper, ECMWF 40-yr reanalysis data have been examined to study the influence of upper-level wave packets propagating across Asia into the Pacific on surface cyclone development over the Pacific. Previous studies have shown that in winter, wave packets propagate across Asia over two branches—a northern branch over Siberia and a southern branch along the subtropical jet across southern Asia. Results presented here show that subsequent to the presence of wave packets on either branch, the frequency of occurrence of deep cyclones (defined as cyclones with central pressure below 960 hPa), as well as explosively deepening cyclones (those with a deepening rate of 1 Bergeron or more), are significantly enhanced. This enhancement also clearly follows the wave packet eastward as it propagates across the Pacific.
Wave packets from the two branches are found to interfere with each other, such that if wave packets of the appropriate configuration are present on both the northern and southern branch, subsequent surface cyclone development over the western Pacific is further enhanced. Examination of the evolution of the anomalies suggests that these interferences can largely be explained by linear superposition of wave packets from the two branches.
Examination of the evolution of the composite structure of wave packets that are followed by the development of a significant surface cyclone indicates that cyclones that develop as the northern packet propagates into the Pacific are phase locked with the upper-level trough and maintain a favorable westward tilt with height throughout their development, consistent with the hypothesis that cyclogenesis is triggered by the approach of the wave packet. In contrast, significant cyclones whose development are influenced by the southern packets initially develop west of the upper-level trough, and propagate eastward with a phase speed that is much faster than that of the upper-level trough, attaining a westward phase tilt with height only at the mature stage, suggesting that cyclogenesis for these cases is probably not triggered by the wave packet.
During the Northern Hemisphere cool seasons, it is well known that cyclone activity peaks over the Pacific and Atlantic storm-track regions (see, e.g., Chang et al. 2002, and references therein). Previous studies (e.g., Whittaker and Horn 1982, 1984) have clearly shown that major peaks in cyclogenesis lie in the western Pacific, over the western Atlantic, and to the east of the Rockies, with the occurrence of the first two peaks related to the maximum in baroclinicity due to the strong ocean–continent temperature contrasts, and the latter peak related to orographic cyclogenesis.
Numerous studies have shown that cyclogenesis over the western Atlantic, just off the east coast of the United States and Canada, is often induced by the approach of an upper-level trough propagating across North America (e.g., Sanders 1986). Some of these precursor troughs are thought to be long-lived coherent structures (Sanders 1988; Hakim 2000; Bosart et al. 1996), while others can be linked to downstream developing wave packets propagating across North America from the eastern end of the Pacific storm track (Chang and Yu 1999; Orlanski and Sheldon 1995; Nielsen-Gammon and Lefevre 1996). Hence, a large number of cyclogenesis events over the Atlantic may be regarded as being seeded by the Pacific storm track.
The situation is a bit less clear over the western Pacific. While there certainly are upper-level waves that propagate across Asia into the Pacific basin, the linkages between these upper waves and surface cyclone development have not been clearly established. Wallace et al. (1988) have suggested that the primary waveguide for upper-tropospheric baroclinic waves over the Pacific is populated by waves propagating southeastward out of Siberia. More recent analyses by Chang and Yu (1999) and Hoskins and Hodges (2002) have shown that upper-tropospheric waves over the Pacific have two sources over Asia, one from the northwest over Siberia, and the second from the southwest along the subtropical jet across southern Asia. However, Hakim (2003) did not find any evidence that waves from the southern branch contribute to Pacific cyclogenesis, and Nakamura and Sampe (2002) have also suggested that waves from the north are coupled more strongly to the surface, while waves propagating out from the subtropical jet are trapped near the tropopause and hence probably decoupled from the surface. Hence it is not clear whether waves propagating out from the subtropical branch affect surface cyclone development over the western Pacific.
In this paper, we examine reanalysis data to study the impact of wave packets on cyclone development over the western and central Pacific. In particular, we examine how the presence of wave packets propagating out from Asia affects the frequency of development of deep and rapidly deepening cyclones. Finally, we examine composites of the vertical structure of these wave packets to shed some lights on the interactions between these upper-level wave packets and surface cyclones. Results show that wave packets propagating out from Asia, from both the northern and the southern branch, exert significant influence on cyclone development over the Pacific.
2. Composite surface cyclone and upper-level wave packets
We examine the 40-yr reanalysis data produced by the European Centre for Medium-Range Weather Forecasts (ECMWF). The main data considered are mean sea level pressure (MSLP) and 300-hPa meridional velocity (V300), for the 24 winter seasons [December–January–February (DJF)] of 1979/80 to 2002/03, on a 2.5° latitude × 2.5° longitude grid. Anomalies are defined as deviations from the seasonal mean. To examine the three-dimensional structure of developing waves, we have also composited meridional velocity and potential vorticity perturbations from 1000 up to 50 hPa (all together, data from 15 vertical levels have been examined: 1000-, 925-, 850-, 775-, 700-, 600-, 500-, 400-, 300-, 250-, 200-, 150-, 100-, 70-, and 50-hPa levels), as well as temperature and geopotential height perturbations. Selected composites of some of these parameters are shown later.
First, we want to examine the evolution of the upper-level structure associated with surface cyclones occurring over the western Pacific. We perform time-lagged, one-point composites based on the MSLP anomaly at individual grid points over the western Pacific. A very simple strategy is taken for the compositing. At each grid point, the composite is done by averaging all instances when the MSLP anomaly at the grid point falls within the lowest 5% on day 0 (the results are not qualitatively sensitive to the limit chosen). While this does not guarantee that the surface low is centered at the grid point, with the chosen limit, most of the time the surface low is centered within 5° of the basis grid point. The results for MSLP anomalies, for the composite based on MSLP at 45°N, 170°E,1 are shown in the rhs panels of Fig. 1. At day 0, a deep low can be seen centered around the base point. Going back in time, we can see the low just starting to develop on day −3 near Korea, subsequently propagating eastward and deepening.
The composites for V300 anomalies (V′) based on the day-0 MSLP anomaly at 45°N, 170°E, are shown in the lhs panels of Fig. 1. At day 0, we see a wave train across the Pacific, with a trough centered around 40°N, 160°E, just to the west of the surface low center. Going back in time, we see that the wave train is apparently connected with two upstream wave packets—one originating from northern Asia around 60°N, and the other from the subtropical jet around 25°N. Hence Fig. 1 suggests that cyclone evolution over the western Pacific is preceded by wave packets propagating across both northern and southern Asia. Results based on other grid points over the western Pacific are qualitatively similar.
In Fig. 1, we see that on day −3, at the 300-hPa level (Fig. 1a), over southern Asia, there is a negative V300 center around 25°N, 70°E, flanked by a pair of slightly weaker positive centers at 30°N, 42.5°E and 27.5°N, 92.5°E, corresponding to a wave train propagating across the southern branch of the Asian waveguide (Chang and Yu 1999). Toward the north, there is a positive center around 57.5°N, 82.5°E, and a negative center at 45°N, 112.5°E, corresponding to a ridge over Siberia. To see how these upper-level wave packets are related to cyclone development over the Pacific, we define indices based on the pattern seen in Fig. 1a. A northern branch index is defined as follows:
The factor 1.12 reflects the relative strength between the two centers seen in Fig. 1a after they are normalized by the standard deviations of meridional velocity perturbations at the two locations. A large positive value of NI indicates that there is a ridge over the northern branch similar to the one seen in Fig. 1a. Similarly, a southern branch index is defined as follows:
A large positive value of SI indicates the presence of a wave train over the southern branch similar to that seen in Fig. 1a.
Let us first examine the composite for the northern branch. Figure 2 is generated by compositing all times when NI is within the top decile on day 0. At the 300-hPa level (lhs panels), a wave train can be seen lying across northern Asia, similar to the one shown in Fig. 1a. However, there is no sign of a wave train across southern Asia, indicating that the evolution of the wave trains over the two branches is independent of each other. As time advances, the wave packet propagates east-southeastward into the western and central Pacific, reaching North America on day +3. On the surface (rhs panels), a cyclonic anomaly starts appearing to the west of Japan on day 0, and it moves eastward and deepens a bit until day +3, and then weakens on day +4. The location of the composite cyclone on day +3 (Fig. 2i) is quite close to the position of the cyclone on day 0 in Fig. 1i.
Comparing Fig. 2 to Fig. 1, we see that while the composite upper-level wave train over northern Asia shown in Fig. 2 is much stronger than the one seen in Fig. 1, the composite surface cyclone in Fig. 2 is much weaker than that seen in Fig. 1. Clearly, not all cyclone events over the western Pacific are preceded by wave trains over northern Asia, and not all wave trains over northern Asia are followed by surface cyclone development over the Pacific. In section 3, we examine more closely how these upper-level wave trains affect surface cyclone development.
Another point to note is that the evolution of the wave packet propagating across the Pacific in Figs. 2b–e is clearly not identical to that seen in Figs. 1b–e. This is not surprising, since the two samples are not identical—Fig. 1 consists of situations when there is a significant cyclone near 45°N, 170°E, while Fig. 2 consists of times when there is a significant wave packet over northern Asia. As discussed above, while these two cases do overlap, there are also a lot of instances when the two situations are distinct; hence there is no reason to expect the subsequent evolution of the two cases to be identical.
Turning our attention to the southern branch, Fig. 3 consists of composites in which SI is within the top decile on day 0. In Fig. 3a, we can see a wave train extending all the way across southern Asia, with the three strongest centers similar to those lying across southern Asia in Fig. 1a. Over the next few days, the wave packet propagates eastward into the Pacific. On the surface, we can see the signature of a negative MSLP anomaly starting to develop to the east of Japan on day 0, and the anomaly deepens a bit on days +2 and +3, before weakening slightly on day +4.
Comparing the rhs panels of Figs. 1, 2 and 3, we see that the composite cyclones in Figs. 1 and 2 propagate eastward with a phase speed of over 13° longitude per day, while the cyclonic anomaly in Fig. 3 propagates with a significantly slower phase speed—about 7° longitude per day. This is consistent with the difference in the phase speed of the upper-level waves—the waves shown in the lhs panels of Fig. 3 clearly propagate with a phase speed much slower than those seen in Fig. 2. This raises the question as to whether the cyclonic anomaly seen in Figs. 3f–j is really related to the evolution of synoptic-scale surface cyclones, which, on average, propagate with the faster phase speed seen in Fig. 1. We return to this point later in section 4.
Figures 2 and 3 indicate that the northern and southern branches are not correlated to each other. Nevertheless, it is still of interest to see what happens when wave packets are present over both branches. In Fig. 4, the composites are based on the times when NI and SI are both within the top 10% on day 0. This occurs about 1% of the time. In Fig. 4a, we see that wave packets are present over both branches, similar to the pattern seen in Fig. 1a, except that the magnitude of the wave packets is significantly stronger. During the next few days, we can see a wave packet propagating across the Pacific. The evolution of the wave packet between day 0 and day +3 (Figs. 4a–d) over the Pacific is quite similar to what happens between day –3 and day 0 in Figs. 1a–d. On the surface, a cyclonic anomaly can be seen over the Sea of Japan on day 0, and this anomaly moves eastward and deepens to a value of –13 hPa on day +3. This cyclonic anomaly is significantly deeper than those seen in Figs. 2 and 3. Clearly, the presence of wave packets of the appropriate configuration over both branches is followed by a much more significant cyclonic anomaly over the western Pacific than when wave packets are present only over one of the two branches.
The presence of a ridge over northeast Asia preceding cyclonic development over the western Pacific (Fig. 2) is consistent with the results of Hakim (2003). Hakim (2003) examined cyclogenesis events over the northwest Pacific and found that these events are accompanied by a developing wave packet propagating across the Pacific. He also found that west Pacific cyclogenesis is preceded by the presence of a ridge over northeast Asia, similar to the situation seen in Fig. 2. However, Hakim (2003) did not find any significant signal from the southern branch. The results shown in Figs. 2 and 4 clearly indicate that while wave packets in the northern branch do precede the development of a cyclonic anomaly over the western Pacific, wave packets over the southern branch also matter. Figure 4 shows that the simultaneous presence of a wave packet of the appropriate configuration over the southern branch is followed by the development of a much stronger cyclonic anomaly over the western Pacific. In Fig. 5, we examine what happens when the ridge over the northern branch is accompanied by a wave packet having the opposite phase in the southern branch. The composites in Fig. 5 are formed by the times when NI is within the top decile on day 0, while at the same time SI is within the bottom decile. In Fig. 5a, we can see a ridge over northeast Asia similar to those seen in Figs. 2a and 4a; however, the wave packet across southern Asia has the opposite phase of those shown in Figs. 3a and 4a.
Let us first examine the evolution of the surface anomalies (Figs. 5f–j). On day 0, just east of the upper-level trough located around 120°E (Fig. 5a), we can barely see a cyclonic anomaly over the Sea of Japan (Fig. 5f). This anomaly is not significant at the 95% level, and is weaker than the one seen in Fig. 2f. This anomaly subsequently propagates eastward and deepens slightly on day +1. However, on day +2, we can no longer see its presence. Instead, it is the cyclonic anomaly over the eastern Pacific that deepens on days +2 and +3.
Examining the evolution of the anomalies in the upper troposphere, we see that the trough over the eastern Pacific (Figs. 5c,d) appears to be much better developed than its counterparts seen in Figs. 2 and 4; thus it is not surprising that the surface cyclonic anomaly over the eastern Pacific is much deeper than those seen in Figs. 2 and 4. Over the western Pacific, in Fig. 2, we can see the trough centered near 120°E on day 0 propagating eastward into the western Pacific. At the same time, the surface cyclonic anomaly propagates eastward from the Sea of Japan and deepens. In Fig. 4, in the presence of wave packets over both branches, this trough is clearly stronger, and the surface cyclonic anomaly is deeper. In Fig. 5, we see that when the southern packet has the opposite phase, the evolution of this trough is quite different. The northern part of the trough apparently still propagates eastward, but the southern part stays west of 130°E. With this upper-level trough staying west of the region of enhanced surface baroclinicity, the surface cyclonic anomaly over the western Pacific fails to develop.
The results shown in Figs. 4 and 5 clearly demonstrate that wave packets from the northern and southern branches interfere with each other. Comparing the rhs panels of Figs. 2, 3 and 4, it appears that the cyclonic anomaly over the western Pacific in Fig. 4 could just be the sum of the anomalies seen in Figs. 2 and 3. To see whether that is the case, in Fig. 6a, we have plotted the sum of the anomalies shown in Figs. 2d and 3d. This should be compared to Fig. 4d. By and large, the two panels resemble each other quite closely, with an anomaly correlation of 0.83. Examining the surface anomaly, again the sum of the anomalies associated with wave packets from the two branches (Fig. 6b) looks very similar to the anomaly for the case when the two wave packets are present simultaneously (Fig. 4i).
For the other case, we have also formed composites based on the times when SI is within the lowest decile (not shown). When we add the signal from such composites to those from NI within the top decile (see Figs. 6c,d), the results again resemble those shown in Figs. 5d and 5i. The results shown in Fig. 6 suggest that much of the interaction between the wave packets from the two branches can be explained by linear superposition.
As discussed above, the results of Hakim (2003) suggest that western Pacific cyclogenesis is mainly associated with wave packets from northern Asia, without a clear contribution from the southern branch. Our results here suggest that both branches contribute. Certainly the case selection criteria here are different from those of Hakim. By varying our sampling criteria, we believe that two other factors may also have contributed to this difference. Hakim (2003) considered cyclogenesis events from November to March, while we only consider cyclones during December to February. When we composite cyclone events using November and March data only in a manner similar to what is done in Fig. 1, we find that the signal in the southern branch is not present during those two months (not shown here), because during the transition seasons the upper-tropospheric waveguide is not split across Asia. Taking the five months together, we find that while the signal in the southern branch is still statistically significant, it becomes a bit weaker than that seen in Fig. 1. Another contributing factor is that in Fig. 1, we made the composites based on significant cyclone events—the criteria chosen means that the central pressure of the cyclones included in the composite are nearly always lower than 975 hPa. If we compare our cyclonic anomaly on day 0 (−28 hPa; see Fig. 1i) to the cyclonic anomaly shown in Fig. 3h of Hakim (−90 m in terms of 1000-hPa height anomaly, equivalent to about –11 hPa in terms of MSLP anomaly), we see that Hakim’s sample clearly consists of many more weaker events. By changing our criterion to compositing all cyclones with central pressure lower than 990 hPa, we found that the signal in both the northern branch and southern branch becomes weaker, with the signal in the southern branch becoming just barely statistically significant. One should also note that a hint of the signal from the southern branch can also be seen in Fig. 3c of Hakim.
3. Influence of wave packets over Asia on cyclone development over the Pacific
While the results shown in Figs. 2 and 3 indicate that wave packets over Asia are followed by the development of a cyclonic anomaly over the Pacific, it is not obvious how these cyclonic anomalies are related to individual cyclogenesis events. While one could examine individual cases to see whether one can find any dynamical linkages, the mere fact that cyclones occur so frequently over the western Pacific (one or more MSLP minima are present over the western Pacific, between 30° and 60°N, 140°E and the date line, about 95% of the time—even if we restrict ourselves to cyclones with central pressure below 990 hPa, one or more can be found within the region about 66% of the time) means that without detailed diagnosis it is difficult to ascertain whether a particular event is influenced by the presence of upper-level wave packets propagating in from upstream. Here, instead of examining cyclone evolution on a case-by-case basis, we employ a statistical approach to determine the extent to which the presence of wave packets over Asia affects subsequent cyclone development over the Pacific.
a. Frequency of deep cyclones over the Pacific
First, we examine whether the existence of wave packets over Asia influences the likelihood of the presence of deep cyclones over the Pacific. Here, we focus on cyclones having a central pressure below 960 hPa, occurring between 30° and 60°N. The longitudinal distribution of all such occurrences over the 24 winters is shown in Fig. 7. Note that in Fig. 7, all such instances are shown; hence a single deep cyclone can be counted more than once. We have decided to use this simple counting technique instead of tracking cyclones and counting each cyclone only once because all statistics based on the criteria used here can be determined entirely objectively and are easily reproducible, whereas statistics based on cyclone tracking in an attempt to count each cyclone only once is heavily influenced by the cyclone tracking routine used (e.g., Gulev et al. 2001), as well as having to face the difficulty of assigning a unique reference time and location to a cyclone whose evolution may span several days across a wide range of longitudes and latitudes.
In Fig. 7, we see that most Pacific deep cyclones occur between 160°E and 140°W, with the peak occurrence near 175°W.2 We divide the domain into three regions: 160°E–180° (region A), 180°–160°W (region B), and 160°–140°W (region C). To determine the influence of the northern packet on the occurrence of deep cyclones, all instances when NI is within the top 20%3 are found and used as the reference time. The normalized occurrence (relative to the mean occurrence—i.e., a value of 1 means that the frequency of occurrence is equal to the average frequency) of deep cyclones in each of the three regions as a function of time lag after the reference time is computed.4 The results are shown in Fig. 8a. In the figure, on day 0, NI is within the top 20% range. First consider region A (160°E–180°). Between day –1 and day +1, the frequency of occurrence of deep cyclones over region A is slightly below average. Around day +1, the frequency begins to increase until it reaches a peak of more than twice the average frequency on day +2.75. After that time, the frequency starts to decrease again until it falls below average after day +4.5. Hence we see that corresponding to the cyclonic anomaly seen in Figs. 2i and 3 days after a ridge is seen over northeast Asia, it is almost twice as likely as average to find a deep cyclone over region A.
Turning our attention to region B (180°–160°W), between day –1 and +1.5, the frequency of occurrence of deep cyclones is slightly below average. After day +1.5, the frequency increases until it reaches a peak of over 1.6 times average on day +4, before decreasing again to below-average frequency after day +5. For region C (160°–140°W), the frequency peaks at about 1.4 times average after day +4.
The results shown in Fig. 8a show that the presence of a ridge over northeast Asia is followed by significantly increased probability of finding a deep cyclone over the Pacific as the wave packet propagates across the ocean. In Fig. 8b, we examine what happens following the presence of a wave train over the southern branch. In Fig. 8b, day 0 represents the times when the value of SI is within the top 20%. First, for region A (solid line), the frequency of occurrence of deep cyclones peaks at day +3, at a value of twice the average frequency. Farther east, over region B, the frequency peaks later, around day +4.5, at a relative ratio of 1.4 times average. Still farther east, over region C, the frequency peaks around day +5.5. Again, wave packets over the southern branch clearly significantly modulate the frequency of occurrence of deep cyclones over the entire Pacific as they propagate eastward across the basin.
In Fig. 9, we show what happens if we change the selection criterion to restrict ourselves to more significant wave packets. Since Fig. 8 shows that the influence of the wave packets is strongest over region A, we will restrict ourselves to examining what happens over that region. In Fig. 9a, the thin solid line is reproduced from Fig. 8a, showing what happens in region A when NI is within the top 20% range on day 0. The situation is not changed appreciably when we restrict NI to be within the top decile (long dashed line). Even when we restrict NI to be within the top 5% (thick solid line), the peak frequency around day 2.75 is only slightly enhanced. In Fig. 9b, we show the corresponding statistics for wave packets over the southern branch. The restriction of SI to larger amplitude (top 10% and top 5%, respectively) makes slightly larger impacts than similar restrictions of NI, with the peak frequency now seen to be quite a bit larger than when SI is within the top 20% range.
In the same panels, we show what happens if the polarity of the wave packets over the two branches are of the opposite sign. The dotted line in Fig. 8a shows what happens if NI is within the bottom 20% at day 0 (i.e., a trough instead of a ridge is present over northeast Asia). The occurrence of deep cyclones over region A is clearly suppressed around day 2.75, with the relative frequency dropping to around 0.3 that of average. For the southern branch, if SI is within the bottom 20% at day 0, the frequency of occurrence of deep cyclones over region A can be seen to be below average over the entire period between day –1 to day +6.
In section 2, we found that when wave packets are present over both the northern and southern branches, the subsequent MSLP anomaly over the Pacific is significantly stronger than when only a single wave packet is present over either of the branches. Here, in Fig. 10, we will show what impact this has on the frequency of deep cyclones over region A. The thin solid line is again reproduced from Fig. 8a, showing the change in frequency of deep cyclones after NI is within the top 20% range. The thick dashed line shows what happens when both NI and SI are within the top 20% range on day 0. The frequency of deep cyclones over region A is again peaked on day +2.75, but now the peak value exceeds 4.5 times average, more than twice as frequent as when NI alone is within the top 20% range. When NI and SI are both within the top decile on day 0 (dashed–dotted line in Fig. 9), the peak frequency becomes more than 6 times that of average. Finally, the dotted line in Fig. 9 shows what happens when there is a ridge over the northern branch, but the wave packet in the southern branch is of the opposite phase (i.e., NI is within top 20%, while SI is within bottom 20%). The frequency of deep cyclones is less than that for the case when only NI is within the top 20% (thin solid line), and around day +3, the relative frequency is less than 1, even though a ridge is present over the northern branch on day 0. Hence we see that the phase of the wave packet on the southern branch clearly has strong influence over the frequency of deep cyclones over the western Pacific, similar to what we found by examining Figs. 4 and 5.
To put the frequency discussed above further into perspective, over the 24 winters, there are a total of 261 instances of cyclones deeper than 960 hPa within the region 160°E to 180° (region A). Out of these 261 instances, 49 instances (19%) occur 2.75 days after NI and SI are both within the top 20% (which occurs 4% of the time). In addition, 163 instances (62%) occur 2.75 days after either NI or SI are within the top 20% (which occurs 36% of the time), and only 38% occurs during the rest of the time (64%). In another words, for the majority of cases, deep cyclones in region A occur about 3 days after a wave packet of the appropriate configuration is present over either the northern or the southern branch (or both).
b. Frequency of rapidly deepening cyclones
We have also examined how the presence of wave packets over Asia affects the frequency of occurrence of rapidly deepening cyclones (or bombs) over the Pacific. Cyclones that deepen at a rate equivalent to 24 hPa over 24 h at 60° latitude [1 Bergeron (B)]) or faster are called bombs by Sanders and Gyakum (1980). The distribution of bombs in our sample peaks near 160° longitude, and most cases occur between 140°E and 170°W. Again, all such occurrences are considered here; hence a single cyclone may contribute to more than one case. Averaging over all cases, these cyclones move from 37.6°N, 154.2°E to 42.8°N, 167.7°E over the 24 h, deepening from an average pressure of 999.0 to 975.1 hPa, at an average rate of 1.34 B. The distribution found here is consistent with that shown in Gyakum et al. (1989).
By conditionally sampling the frequency of occurrence of bombs depending on the phase of NI or SI, we again find that the frequency of bombs over the Pacific is significantly enhanced subsequent to the presence of wave packets over either the northern or the southern branch, and that the simultaneous presence of wave packets of the appropriate phase (both NI and SI positive and large) is followed by much stronger enhancement of the frequency than that following the presence of a single wave packet on either branch. These results are consistent with those presented in the preceding subsection and details are not shown here. We have also examined the influence of wave packets on rapidly deepening deep cyclones (i.e., cyclones that have MSLP below 960 hPa that have undergone a deepening rate of over 1 Bergeron over the preceding 24 h) and found very similar results.
4. Vertical structure of developing wave packets and cyclones
In this section we examine the vertical structure of the wave packets propagating across the two branches. In section 3, we saw that surface cyclone development is statistically enhanced subsequent to the presence of wave packets over the northern and southern branches. However, it is not clear how these upper-level wave packets interact with the surface cyclones, especially for wave packets from the southern branch, which previous work have suggested to be trapped near the tropopause. While such interactions can only be resolved by detailed dynamical analyses of individual cases, we have found that examination of composites of the vertical structure of wave packet can shed some light on some of these interactions. In section 3, we also saw that surface cyclone development is further enhanced after wave packets of the appropriate phase are present over both the northern and southern branches; hence, in this section, we limit our composites to situations in which significant wave packets of the appropriate phase are present only over one of the two branches to limit the effects of the constructive interference between wave packets from the two branches.
In Fig. 11 Hovmoeller diagrams (longitude–time plot) show the evolution of the upper-level wave packet (V300, in shades) and surface anomalies (1000 hPa V′, contours). The quantity shown in the figure is meridional velocity perturbation averaged between 20° and 60°N. Figure 11a is a composite from all instances in which NI is within the top decile on day 0, while at the same time SI is not within the top 20% range (to limit the interference from the southern wave packet). At the 300-hPa level (shades), we can see a wave packet propagating eastward across the Pacific. On the surface, a cyclonic anomaly can be seen developing just to the east of the upper-level trough starting around day –0.5. The surface perturbation can be seen to propagate eastward with about the same phase speed as the upper-level trough, maintaining a westward phase tilt with height at all times. As the upper-level wave packet propagates into the eastern Pacific, the phase speed decreases. At the same time, the phase speed of the surface anomaly also decreases. Over the eastern Pacific, the phase tilt between the upper- and lower-level anomalies is less than that over the western Pacific, reflecting the weaker baroclinicity over that region.
Vertical cross sections of the V′ perturbations for days –1, +1, and +3 are shown in Figs. 12a–c. On day –1, when the wave packet is over land, the velocity perturbation does not quite make it all the way to 1000 hPa. On day +1, when the wave packet has propagated into the Pacific, the velocity perturbation now exhibits a clear westward tilt with height and extends all the way from the tropopause to the surface. Such a change in packet structure is similar to that found in the modeling studies of Whitaker and Barcilon (1992).
We have previously seen that the presence of upper-level wave packets of the appropriate phase is followed by enhanced surface cyclone development. As discussed above, not all upper-level wave packets are followed by the development of a significant surface cyclone over the western Pacific. Hence the composites we just examined contain cases in which surface cyclones are absent. Here, we would like to examine the structure of wave packets for the cases in which significant surface cyclones are present. In Fig. 11b, the composites are from the subset of cases in which a deepening surface cyclone, with surface pressure below 975 hPa, is found in the area 40°–60°N, 160°E–180° on day +3. The cutoff value of 975 hPa is chosen here (instead of a value of 960 hPa used above) to give a larger sample. Comparing Fig. 11b to 11a, we can see a much stronger cyclonic perturbation on the surface over the western Pacific. We can start seeing the signature of the surface cyclone around day –0.5 developing around 120°E, just to the east of the upper-level trough axis, and the upper- and lower-level perturbations stay more or less in a favorable configuration (westward tilt with height) throughout the evolution of the surface cyclone, consistent with the hypothesis that the development of the surface cyclone is initiated by the approach of the upper-level trough.
Examining the upper-level wave packet, only minor differences can be seen between days –1 and 0 over the upstream side of the cyclonic development, suggesting that it is not differences in the upper-level wave packets that have led to the differences in the surface development. As we follow the wave packet into the eastern Pacific, we can see that the amplitude of the perturbations east of the surface cyclone is enhanced, suggesting that the development of a significant surface cyclone leads to downstream enhancement of the wave packet. Similar observations can be made if we examine the vertical cross section shown in Figs. 12d–f. Comparing Figs. 12d–f to 12a–c, we can see that on day −1, the structure of the upstream wave packets are quite similar for the two cases, while on day +3, the amplitude of the wave packet around and downstream of the surface cyclone (located at 170°E) is clearly much enhanced for the cases when a significant surface cyclone develop (Fig. 12f).
Finally, we examine the evolution of the vertical structure of the southern packet in the absence of a significant northern packet (i.e., SI in top decile, while NI not within top 20% at day 0). The Hovmoeller diagram of 300 hPa V′, averaged between 20° and 60°N, is shown in shades in Fig. 13a. We can clearly see the upper-level wave packet propagating eastward from central Asia into the Pacific. The 1000 hPa V′ is shown as contours. At low levels, there are few signs of the wave packet over land, as the wave packet propagates over high terrain just to the south of the Tibetan Plateau. After the wave packet reaches the western Pacific, we can see a surface cyclonic anomaly propagating just to the east of the upper-level trough, with a phase speed similar to that of the upper-level trough.
The temporal evolution of the vertical cross section is shown in Figs. 14a–d. We can see that between 60° and 140°E, the wave packet is mainly trapped within the upper troposphere. The packet can only be seen to extend to the surface east of 140°E. This is in contrast to the northern packet that extends all the way to near the surface even over land (Fig. 12a). Another difference is that the northern packet has maximum amplitude near 300 hPa, while the southern packet has its maximum around 250 hPa, reflecting the fact that the southern packet propagates along the waveguide defined by the subtropical jet, which is at a level higher than the midlatitude tropopause. These results are similar to those found by Nakamura and Sampe (2002).
The evolution of the composite southern wave packet, in the presence of a surface cyclone with central pressure lower than 975 hPa within the region 40°–60°N, 160°E–180° on day +3, is shown in Figs. 13b and 14e–h. Unlike the case for the northern packet, the composite southern packet and surface cyclone do not appear to be phase locked during the development of the cyclone. The surface cyclone appears around day +0.5 near 125°E, just to the east of an upper-level ridge instead of a trough. Such an alignment is not favorable for baroclinic development and differs from the classic Petterssen type B development in which a surface cyclone develops to the east of an upper-level positive isentropic potential vorticity (IPV) anomaly (or trough; see Hoskins et al. 1985). This atypical phase alignment can be clearly seen on Fig. 14f, in which the negative V′ anomaly near 120°E can be seen to tilt eastward instead of westward with height.
Between day +0.5 and day +3, the surface cyclone can be seen to propagate eastward with a phase speed of about 18° longitude per day, while the phase speed of the upper-level wave is less than 10° per day, such that after day +2, the surface cyclone is located east of the upper-level trough, attaining a configuration that is now favorable for baroclinic interaction. As for the case for the northern packet, the existence of a significant surface cyclone clearly affects the subsequent downstream development of the upper-level wave packet, with the packet amplitude over the central Pacific much stronger than that of a typical southern packet, and the phase speed of the upper-level wave also increased significantly over the central and eastern Pacific.5
Based on these composites, it seems probable that the surface cyclone seen in Fig. 13b is not initiated by the southern packet propagating out from Asia. Instead, one can postulate that a surface cyclone started developing (due to an as yet unknown reason), and, as it developed, it just turned out that an upper-level wave packet propagated out from Asia, and the cyclone interacted with the wave packet, with the interaction leading to mutual enhancement. Such development is in contrast to the northern packet case in which the surface cyclone and upper-level trough appear phase locked throughout their development (Fig. 11b), indicating that it is likely that the surface cyclone is initiated by the approach of the upper-level wave packet. Here we have arrived at these conclusions based only on statistical analyses of the relationship between upper-level wave packets and surface cyclone development. How the upper-level waves and surface cyclones interact dynamically can only be illuminated by further diagnostic studies that are beyond the scope of this paper.
In this paper, we have examined ECMWF 40-yr reanalysis data to study the influence of upper-level wave packets propagating across Asia into the Pacific on surface cyclone development over the Pacific. Previous studies have shown that in winter, wave packets propagate across Asia over two branches—a northern branch over Siberia, and a southern branch along the subtropical jet across southern Asia. Our results show that subsequent to the presence of wave packets on either branch, the frequency of occurrence of deep cyclones (defined as cyclones with central pressure below 960 hPa), as well as explosively deepening cyclones (those with a deepening rate of 1 B or more), are significantly enhanced. This enhancement also clearly follows the wave packet eastward as it propagates across the Pacific.
Our results also show that wave packets propagating over the two branches can interfere with each other—if both packets are of the appropriate phase, they interfere constructively, resulting in further enhancement of cyclone development, while if the phase of the wave packet over the southern branch is opposite to the favorable phase, even the presence of a “correct” northern packet (namely a ridge near Lake Baikal) is not followed by enhanced cyclone development over western and central Pacific. Our results indicate that the observed interference can be explained largely by linear superposition of the two wave packets.
It should be reiterated here that while these wave packets are found to precede enhanced frequency of significant cyclone development, not all cases of wave packet propagating out from Asia are followed by significant cyclone events over the western Pacific. In our comparisons of composites made from cases followed by significant cyclone development to those computed using all cases, we are unable to identify significant differences in either the amplitude or structure of the precedent upstream wave packet, suggesting that whether a significant cyclone develops or not when a wave packet propagates out from Asia is probably due to some other factors. Other studies have suggested other synoptic conditions that may be important in influencing the rate of cyclone deepening, such as thermodynamic preconditioning and surface heat fluxes from the underlying ocean (Gyakum and Danielson 2000; Bosart 1981), as well as frontogenesis occurring both in the vicinity as well as away from the cyclone (Bullock and Gyakum 1993). How these and other effects interact dynamically with upper-level wave packets to affect cyclone evolution remains to be investigated.
Our results also suggest that significant cyclones that develop when a wave packet propagates out from the northern branch are phase locked with the upper-level trough, suggesting that cyclogenesis may well be triggered by the approach of the upper-level wave packet. On the other hand, significant cyclones that develop when a wave packet propagates into the Pacific from the southern branch initially develop upstream of the upper-level trough. As they develop, these cyclones propagate with a phase speed much faster than that of the upper-level trough, eventually overtaking the trough to attain a westward tilt with height. It is hypothesized that for these cases, the initial cyclogenesis event occurs independently of the upper-level wave packet, but as the cyclone develops, it comes under the influence of the southern packet and its development is enhanced (as compared with other cyclones that develop in the absence of the upper-level wave packet). Whether this is the case, what factors have led to the initial cyclogenesis, and other questions such as what the exact dynamical process that allows the upper-level wave packets to affect surface cyclone development is, and how wave packets and surface cyclones will interact under configurations different from those examined in this study, are topics that are currently under investigation.
The author would like to thank Dr. Greg Hakim and an anonymous reviewer for valuable comments on an earlier version of the manuscript. The author would also like to thank ECMWF for making the reanalysis data available on the ECMWF data server. This research is supported by NSF Grant ATM0296076.
Corresponding author address: Edmund K. M. Chang, ITPA/MSRC, State University of New York at Stony Brook, Stony Brook, NY 11794-5000. Email: firstname.lastname@example.org
For this grid point, the lowest 5% MSLP perturbation is equivalent to an MSLP perturbation of –22 hPa or below.
For the sample here, the region between 160°E and 140°W accounts for 95% of all occurrences of deep cyclones over the Pacific. Since this region also coincides with the region with the minimum climatological MSLP, one might argue that the lower background pressure over the region may exert some influence on the location of the peak. However, even if we define cyclones based on deviations from the seasonal mean (which removes the climatological background), the distribution is again sharply peaked over this region, and this region still accounts for 85% of all occurrences of cyclones deeper than −40 hPa.
That is, a ridge occurs over the northern branch—20% is used here instead of 10% to increase the number of cases to give less noisy statistics.
For reference, the average frequency of deep cyclones is 3.5%, 4.8%, and 2.2% in regions A, B, and C, respectively.
From Fig. 14, it may appear that the southern packet does not undergo further downstream development while over the eastern Pacific. However, inspection of subsequent evolution shows that a positive phase develops around 115 W on day +5, and a negative phase develops around 75°W on day +7. Hence there is evidence that the southern packet also propagates across North America into the Atlantic storm track.