In Part I of this study, the annual cycle of the Northern Hemisphere storm tracks was investigated using feature tracking and Eulerian variance-based diagnostics applied to both the vorticity and meridional wind fields. Results were presented and discussed for the four seasons at both upper- (250 hPa) and lower- (850 hPa) tropospheric levels. Here, using the meridional wind diagnostics, the annual cycles of the North Pacific and North Atlantic storm tracks are examined in detail. This is done using monthly and 20° longitudinal sector averages. Many sectors have been considered, but the focus is on sectors equally spaced in the two main oceanic storm tracks situated at their western, central, and eastern regions, with the western ones being mainly over the upstream continents. The annual cycles of the upper- and lower-tropospheric storm tracks in the central and eastern Pacific, as well as in the western and central Atlantic sectors, all have rather similar structures. In amplitude, each sector at both levels has a summer minimum and a relatively uniform strength from October to April, despite the strong winter maxima in the westerly jets. However, high-intensity storms occur over a much wider latitudinal band in winter. The storm track in each sector moves poleward from May to August and returns equatorward from October to December, and there is a marked asymmetry between spring and autumn. There are many differences between the North Pacific and North Atlantic storm tracks, and some of these seem to have their origin in the behavior over the upstream East Asian and North American continents, suggesting the importance of seeding from these regions. The East Asian storm track near 48°N has marked spring and autumn maxima and weak amplitude in winter and summer. The 33°N track is strong only in the first half of the year. In contrast, the eastern North American storm track is well organized throughout the year, around the baroclinicity that moves latitudinally with the seasons. The signatures associated with these features are found to gradually decrease downstream in each case. In particular, there is very little latitudinal movement in the storm track in the eastern Atlantic.
The winter Northern Hemisphere (NH) storm tracks have been the subject of many previous studies (e.g., Blackmon 1976; Chang et al. 2002; Hoskins and Valdes 1990; Hoskins and Hodges 2002), In his seminal study, Nakamura (1992) presented pictures of the annual cycle of the North Pacific and North Atlantic storm tracks based on the high-pass-filtered variance of the geopotential at 250 hPa and the sea level pressure. His focus was on the winter half of the year, and he contrasted the midwinter minimum in the Pacific with the expected winter maximum in the Atlantic. Subsequent papers, for example, Ren et al. (2010), Penny et al. (2010), Chang and Guo (2012), Ren et al. (2014), and recently Afargan and Kaspi (2017) [and also, very recently, Schemm and Schneider (2018)] have shown annual cycles but have focused on this winter behavior. However, relatively little attention has been given to the NH storm tracks in other seasons or to the details of their annual cycles.
An earlier study whose results might be expected to be relevant to the storm tracks is that of Fleming et al. (1987), who considered the annual cycle of the zonally averaged westerly wind at 500 hPa. The focus of Fleming et al. (1987) was on the asymmetry between spring and autumn, with the latitudes of the jet in spring and autumn being 33° and 46°N, respectively. The largest amplitudes and southernmost latitudes were found to occur in midwinter, with the weakest amplitudes in July and highest latitudes in August.
This paper forms the second part of a study of the annual cycle of the observed NH storm tracks. In Hoskins and Hodges (2019, hereafter Part I) high-pass standard deviation and cyclone tracking metrics were applied to vorticity and meridional wind in the upper troposphere (250 hPa) and lower troposphere (850 hPa) to produce the seasonal average storm tracks that were presented and discussed. Comparisons between the results for different metrics, variables, and levels, and between the Pacific and Atlantic storm tracks, were made. Among the results it was found that the North Pacific upper-tropospheric storm track is indeed weaker in winter than in autumn or spring when viewed using the high-pass standard deviation of both variables, but with tracking measures, and in the lower troposphere this minimum was less marked. The seasonal positions of the storm tracks were in general found to be qualitatively consistent with that of the zonally averaged jet found by Fleming et al. (1987). However, these and other aspects would benefit from a more detailed view of the annual cycle. This is the motivation for this second paper in which the annual cycle of the two major Northern Hemisphere storm tracks, those of the North Pacific and the North Atlantic, are examined in detail using averages over particular, representative longitudinal sectors and for the calendar months. The primary motivation for this paper and its companion is to further our basic understanding of storm tracks and the ability to diagnose the storm tracks in climate models. It is not directly aimed at the impacts associated with storms, though this is of course an important area for study.
Sector averages have been used by a number of authors in order to summarize the storm track behavior in broad regions. For example, Nakamura (1992) used 60° sectors in the Pacific and the Atlantic, and Penny et al. (2010) and Afargan and Kaspi (2017) used 40° sectors. These sector widths were satisfactory for the topics considered, but it is clear from the geographical storm track pictures shown in Part I that the storm tracks in some regions can vary significantly over such longitudinal ranges. In particular averages over a broad range of longitudes are problematic for a southwest–northeast-tilted storm track like the North Atlantic in winter. Here, we will use narrower 20° sectors in order to obtain a more local picture of the storm track behavior. The 37 years of reanalysis data used here are sufficient, results for such narrow sectors based on diagnostics of meridional wind generally showing coherent behavior from month to month in the annual cycle.
Previous studies showing the annual cycle, for example, Nakamura (1992), Nakamura and Sampe (2002), Ren et al. (2010), Chang and Guo (2012), and Afargan and Kaspi (2017), have mostly used a high-pass variance measure for the storm track. However, Penny et al. (2010) showed the results from both variance and feature tracking, but focused on the tracking results. [Very recently, Schemm and Schneider (2018) have tracked surface cyclones and used variance measures in the free atmosphere.] Following Nakamura (1992), the high-pass variance technique has mostly been applied to geopotential at an upper-tropospheric level, but Ren et al. (2010) applied it to meridional wind at an upper-tropospheric level. Nakamura (1992) also presented the annual cycle for the variance of mean sea level pressure but there has been limited discussion of lower-tropospheric metrics. In this paper equal weight will be given to high-pass variance and feature tracking measures. The same metrics applied to an upper-tropospheric level, 250 hPa, and a lower-tropospheric level, 850 hPa, will be discussed for each sector. The main results presented here will be for meridional wind V. Those results obtained using vorticity are generally found to be similar.
2. Data and methodology
The data and basic methodology are the same as in Part I, where full details are given. The basic data source is the European Centre for Medium-Range Weather Forecasts (ECMWF) Interim reanalysis (ERAI; Dee et al. 2011) for the years 1979–2016. The data used in this study are the four times per day meridional wind V, on the 850- and 250-hPa levels.
Two approaches to diagnosing the storm tracks are employed. The first is based on the 2–6-day bandpass-filtered variances (Blackmon 1976), presented in terms of the standard deviation (SD). The second approach uses objective feature tracking as in the NH winter study of Hoskins and Hodges (2002), and is the same as used in Part I, where more details can be found. As in Part I, the diagnostics are performed at 250 hPa, representing the upper troposphere, and 850 hPa, representing the lower troposphere. In Part I the tracking results were presented for extrema in two fields: vorticity and the modulus of meridional wind |V|. In Part I a detailed discussion of using the latter and the relative advantages of the two variables is given. The latter is equivalent to tracking both positive and negative extrema in V and combining the results. Briefly, the justification for using |V| is that the growth of storms depends on both warm air moving poleward and cold air moving equatorward, and in addition tracking |V| provides the best comparison with the bandpass SD of V, which also makes no discrimination between signs. In this paper, SD of V and tracking of |V| will be used instead of vorticity as the finer scales described by vorticity sometimes lead to some lack of continuity between months in sectors as narrow as the ones used here. Tracking has also been performed separately for positive and negative V extrema and these results will be mentioned in cases where this detailed additional information is of interest.
The sectors in which the storm tracks are to be studied in detail are shown in Fig. 1. This shows the December–February (DJF) wintertime cyclone track density and mean intensity for maxima in the modulus of the meridional wind at 850 hPa V850, similar to Fig. 10a in Part I. As discussed in detail there, the North Pacific and North Atlantic storm tracks are clearly delineated at this level. The sectors have been chosen to sample the western/upstream, central, and eastern/downstream regions of the two main storm tracks. They are all 20° in longitude, with three equally spaced (with 30° separation) in the Pacific storm track and three equally spaced (with 10° separation) in the Atlantic storm track. The sectors are the western Pacific (WP, 110°–130°E), central Pacific (CP, 160°E–180°), eastern Pacific (EP, 150°–130°W), western Atlantic (WA, 80°–60°W), central Atlantic (CA, 50°–30°W), and eastern Atlantic (EA, 20°W–0°). The nomenclature—western, central, and eastern—is used here with reference to the relevant storm track and not the ocean basin. For example, in the latitudes of interest, the WP sector is mainly in East Asia, and WA is more in North America than over the Atlantic. To understand the context farther upstream and the continuity between sectors, results from a number of other sectors have been computed and considered. Sectors farther west over Asia (100°–120°E) and over North America (90°–70°W) are shown in the online supplemental material to this paper. Also shown there are the results for a 10°–30°E sector that gives a picture of the seasonal cycle of the Mediterranean storm track. In addition it shows the extension of the North Atlantic storm track into Europe, though the interruption of the 850-hPa surface by the Alps should be noted as giving a region of doubtful validity of the diagnostics between the two storm track regions. The borders of the three sectors shown in the supplemental material are indicated by yellow lines in Fig. 1.
In the following section, the North Pacific storm track will be discussed first, starting with the CP sector, motivated by the previous interest in the midwinter minimum in the Pacific.
a. North Pacific
1) Central Pacific
The CP sector zonal means of SD of the bandpass V at both 250 and 850 hPa are shown as functions of time of year (abscissa) and latitude (ordinate) in Figs. 2a and 2c. The January–December period is continued by a repetition of the 6 months from January to June, so that the annual cycle is clear throughout the year. Overlain on the SD panels are contours of a relevant zonal wind U. At 250 hPa, U is shown on the dynamical tropopause [the PV = 2 surface; see, e.g., Hoskins (2015)] and highlights both the subtropical and polar jets. This is used because of its theoretical importance for developments in the upper troposphere in general (see, e.g., Hoskins and James 2014). However, U at the 250-hPa level (not shown) itself is very similar, showing jets at the same latitudes with only slightly smaller strengths. At 850 hPa, the zonal wind at that level is used. Figures 2b and 2d give the track density (line contours) and mean intensity (filled color contours) for |V| maxima at the same upper- and lower-tropospheric levels.
In the CP, the SDs at both levels (Figs. 2a,c) are dominated by the middle-latitude storm track and show the annual cycle in amplitude and latitude. The midwinter minimum is apparent in both the upper and lower troposphere, though much less so at the lower (850 hPa) level. This contrasts with the winter maximum in U at both levels. A marked midsummer minimum in SD, as well as in U, is apparent at both levels. It is also clear that the autumn maxima are farther poleward than those in spring.
The tracking panels (Figs. 2b,d) give a more complex picture. In the upper troposphere, the track density (Fig. 2b) shows spring and autumn maxima. The mean intensity is large in the autumn storm track, but in midwinter the higher-intensity values are less confined to the region of high track density, with the highest values found at even lower latitudes than the track density maximum. In spring, the maximum intensities are back in the storm track, but there is now a secondary maximum at low latitudes. The midwinter minimum in SD is therefore a reflection of both fewer storms and reduced intensities in the main storm track region. However, there are strong storms with tracks over a wide range of latitudes.
The tracking picture is again simpler in the lower troposphere (Fig. 2d) with the highest intensities in the storm-track region throughout the year. Here, the overall impression is of a single extended maximum from September to May. The weak midwinter minimum in SD is related to slightly weaker mean intensities in the storm track region. However, at that time the region of high intensities also spreads in latitude, particularly on the poleward side.
The almost straight bounding contours on the poleward side of the 850-hPa track density for the main storm track imply that it stays equatorward of the Kamchatka Peninsula all year. At both levels there is also a polar Arctic maximum in track density in the summertime (Serreze and Barrett 2008). At 850 hPa, this leads to a weak maximum in SD (Fig. 2b) there, but the more prominent Arctic maximum in SD is in winter, associated with the larger intensities at that time of year (Zhang et al. 2004)
At both levels and in both measures, the major storm track movement in latitude largely reflects that of the zonal wind, with the upper-tropospheric storm track slightly poleward of the maximum in the westerlies. However, in the upper troposphere in winter the relationship is less definite. The SD maximum does not move as far equatorward as that in U. The track density maximum is north of the jet but the storms with highest mean intensity are slightly south of the jet.
To summarize the annual cycle in the main storm track, the latitude (ordinate) of the SD maximum and the amplitude (abscissa) of the maximum in each month are shown in Figs. 3a and 3c, respectively, for the two levels. The months are shown as dots with different colors, and the annual cycle is made clearer by lines joining successive months, starting at January and finishing at December. In the upper troposphere (Fig. 3a), the midwinter minimum is apparent in that the March, April, and May amplitudes, as well as those of October and November, are all slightly higher than those in January and February when the storm track is at relatively low latitudes.
The annual cycle shows a poleward shift and growth from February to March, a sharp decrease in amplitude from May to June, a poleward shift through July into August, a shift equatorward by September, and growth that continues into November followed by a further equatorward shift and reduction in amplitude through December into January. Figure 3c shows that at 850 hPa the midwinter shift equatorward to below 40°N does not occur and the midwinter minimum is less marked, with the amplitude being similar over the extended winter from October to April. As in the upper troposphere, minimum values again occur from June to August and this is the period of the main poleward shift. The main increase in amplitude occurs from then until October and the main decrease in latitude from November to December. The hysteresis-like curves at both levels clearly illustrate the asymmetry between spring and autumn.
It is of interest to obtain similar annual cycle summaries for the main storm track using the tracking of |V|. In each month the track density shows a relatively sharp maximum, but the mean intensities have a broader distribution. Therefore, it has been decided to identify the latitude of the storm track by the track density maximum, and use the mean intensity at that latitude as the measure of strength, with these two acting as the descriptors of the storm track behavior for each month. This gives the annual cycle summary pictures in Figs. 3b and 3d for the two levels. The behavior at both levels is generally similar to that shown for the SD in Figs. 3a and 3c. Again apparent are the January–February equatorward displacement with a minimum in amplitude at both levels, low intensities and a poleward shift during the period June–August, and a latitude difference between spring and autumn, with the spring latitude closer to that of winter and the autumn storm track latitude closer to that of late summer. With this diagnostic, the midwinter minimum is apparent at 850 hPa as well as 250 hPa.
It is of interest to note that in this sector, the large sea surface temperature gradients associated with the extension of the Kuroshio are in the 30°–40°N latitude band, and only the winter storm track can be affected directly by them.
2) Western Pacific
The WP sector has its own intrinsic interest. In addition there has been much discussion (e.g., Hakim 2003; Orlanski 2005; Chang 2005; Robinson et al. 2006; Penny et al. 2010; Chang and Guo 2012) of the seeding from the WP of the storm track in the CP, in particular in its possible importance in the midwinter minimum there. The Kuroshio occurs only in the easternmost part of this 20° sector and then with a south–north orientation. The zonal averages in the WP sector are not designed to diagnose the impact of the Kuroshio and in any case the scale resolved by the data used is probably not suitable for this.
Figure 4 shows the WP sector average SD and tracking results at the two levels. The SD results (Figs. 4a,c) for the main storm track are generally weaker versions of those for the CP (Figs. 2a,c). However, in this sector the midwinter minimum in the upper troposphere is almost as marked as the midsummer minimum. The upper-tropospheric tracking diagnostics (Fig. 4b) are also similar to those for the CP (Fig. 2b), except that the mean intensities are weaker and the midwinter minimum is more evident in both track densities and intensities. The behavior is largely decoupled from that of the jet (Fig. 4a) except that there is an intensity maximum that is coincident with the winter jet maximum.
There are indications of a more complex structure in the lower-tropospheric SD (Fig. 4c). In spring there are signs of separate maxima near 45° and 30°N, and in spring and autumn there are also maxima near 65°N. In the tracking results (Fig. 4d) these features become more apparent. Between 30° and 50°N the track density exhibits a double structure in latitude. The northern maximum, centered on 48°N, is present most of the year, and is prominent from March to June and from August to December. Its importance for the Pacific storm track has been discussed by Nakamura (1992), Hakim (2003), Orlanski (2005), Chang (2005), Penny et al. (2010), and Chang and Guo (2012). The southern maximum, centered on 33°N, is present only in the first half of the year, and is prominent from February to May. Nakamura and Sampe (2002) mentioned the possible importance for the Pacific storm track of waves on this southern branch, and Chang (2005) gave evidence of constructive interference between wave packets on the two tracks.
Associated with the 850-hPa double track in spring and the single track in autumn, there are weak maxima in the westerly winds at the same level (Fig. 4c). In the autumn the northern track is prominent at both levels in both track density and SD (Fig. 4), so that the systems are vertically deep. However, in the spring the upper-tropospheric track is centered between the two lower-tropospheric tracks, so that a linkage with either is possible. This behavior contrasts with that in the CP where the upper- and lower-tropospheric track densities are coherent throughout the year. The two tracks in the WP can be seen in the winter and spring 850-hPa tracking pictures given in Part I for meridional wind (Figs. 10a,b) and positive vorticity (Figs. 7a,b). For winter, the northern track is the southeastward extension of the Siberian track discussed by Wallace et al. (1988) and the southern one, in the region of the subtropical jet, was highlighted by Chang and Yu (1999). Later in the year, the latter corresponds to the spring persistent rains in China discussed by, for example, Tian and Yasunari (1998). The possible importance of the WP behavior for that in the CP will be discussed further in section 4.
Also clearly marked in the 850-hPa tracking results (Fig. 4d) is a maximum in track density near 65°N from March to October with a weak minimum in midsummer. The mean intensities emphasize the spring and autumn periods, making this storm-track feature consistent with the signatures near 65°N commented on above. Looking again at Fig. 4b, similar features are also apparent in the 250-hPa tracking results. The features are even clearer in the 100°–120°E sector shown in Fig. S1 in the online supplemental material. Referring to Figs. 7, 8, 10, and 11 in Part I, the presence of the 850-hPa eastward extension near 65°N from the Siberian storm track to 110°–120°E is clear in both the vorticity and meridional wind, and in both SD and tracking. However, in the upper troposphere the track leading southeast toward the region of the 48°N lower-troposphere track is dominant.
Near 15°–20°N, the May–November maximum in track density (Figs. 4b,d) is more marked at both levels than in the CP, and is the signature of westward-moving western Pacific tropical cyclones. At 850 hPa the track density maximum is accompanied by large mean intensities and so the signature is seen also in the SD (Fig. 4b).
3) Eastern Pacific
Moving eastward from the CP, the EP SD and tracking results at the two levels are shown in Fig. 5. The main storm track in the upper troposphere (Figs. 5a,b) again exhibits a midwinter minimum in SD and in track density. In midwinter the spread of larger SD values to lower latitudes is associated with a secondary maximum in track density accompanied by high intensities. This is the signature of the cutoff lows, the “subtropical cyclones” whose surface features are referred to as kona lows and that have been discussed by, for example, Simpson (1952) and Otkin and Martin (2004). The high intensities at low latitudes seen in CP in the winter (Fig. 2b) may also reflect such features.
In the lower troposphere (Figs. 5c,d) both the SD and track density give a slight emphasis to the first half of winter. The midwinter expansion to lower latitudes is again apparent in the tracking fields (Fig. 5d). At this lower level, the storm track in this sector is bounded near 58°N at all times of the year by the southern coast of the Alaskan Peninsula. At low latitudes the track density signature of westward-moving eastern Pacific tropical cyclones is seen from June to October. In high latitudes the track density again has a summer maximum in the Arctic, off the northern coast of Alaska, but it is the winter intensity maximum that is picked up more strongly by the SD.
At both levels, the westerly winds themselves show a slight winter minimum in this sector. The storm tracks and the winds tend to move together through the annual cycle, with the storm track slightly poleward of the maximum winds except for the upper troposphere from January to April.
Figure 6 shows the EP annual cycle summaries for SD and tracking and for the two levels using the same format as for CP in Fig. 3, and the results are quite similar to those. All four panels in Fig. 6 show a fairly flat maximum from October to April with the actual peak mostly in December. Each then shows a strong reduction in magnitudes compared to those of the summer, being about 50% at the lower level. Each also shows a lower latitude for the storm track in winter and early spring, with a higher latitude in late summer and early autumn. This asymmetry between spring and autumn is again marked.
b. North Atlantic
1) Western Atlantic
As seen in Fig. 1, the WA sector includes eastern North America. The Gulf Stream is present in the eastern portion of the sector, moving from near 30°N to near 40°N across the sector. As with the case of the WP, the zonal average in the sector will not, and is not intended to, capture the details of the atmospheric interaction with the Gulf Stream.
The WA sector average annual cycle results are given in Fig. 7. In the annual cycle, the track densities (Figs. 7b,d) have a weak summer maximum but show less variation in amplitude through the year than in the CP and EP. The mean intensities are a maximum in the winter half of the year and lower in the summer. The latter is reflected in the minima in the SD results (Figs. 7a,c) in summer, but these minima are not as marked as in the Pacific. The upper-tropospheric mean intensity (Fig. 7b) again spreads to lower latitudes in winter, but with the low number of tracks this is only weakly reflected in SD (Fig. 7a), and is not as prominent as in the CP and EP. The signature of westward-moving Atlantic tropical cyclones is apparent in all the panels from June to October.
As in the CP, the upper-tropospheric westerly winds are strongest in midwinter (Fig. 7a). However, unlike in the Pacific, the jet continues to move slightly equatorward until March. In the upper troposphere in April–May a second westerly maximum appears some 15° farther north, and it is the northern maximum that continues through the summer, moving slowly poleward. The storm track in all measures is poleward of the strongest westerlies, with the displacement becoming large as the westerly jet continues to move equatorward after January.
Figure S2 shows results for the sector centered 10° farther west. Results have also been obtained for a sector 10° farther east (not shown). The storm-track diagnostics are all very similar but with slightly reduced SD amplitude and track mean intensities to the west, and increased values to the east.
Summaries of the annual cycle in the latitude and strength of the middle-latitude storm track for the WA sector are given in Fig. 8. At 250 hPa (Figs. 8a,b) the storm track is farthest equatorward in February, near 40°N, and farthest poleward in August–September, near 50°N, much as in the Pacific. The amplitude varies little during the period from September to May with a weak maximum in November. The minimum occurs in July and August, but the summer reduction is smaller than in the Pacific. A significant poleward shift does not occur until May or June and the return to lower latitudes does not occur until October or November. The 850-hPa results (Figs. 8c,d) are similar, showing loops from low latitudes and high amplitudes in the period from November to April and high latitudes and low amplitudes in July and August, with the early summer and autumn changes occurring first in amplitude. In the lower troposphere the amplitude range is only slightly less than in the Pacific. In all the diagnostics, the latitude of the storm track changes little from December to April, the period when the upper-tropospheric jet is still moving slowly equatorward. As discussed in Part I, it is only in this period that the storm track is close to the region of strong Gulf Stream SST gradients.
2) Central Atlantic
The CA sector average annual cycle results are presented in Fig. 9. There are again considerable similarities with the WA (Fig. 7), though the amplitudes of the SD (Figs. 9a,c) and the mean intensities (Figs. 9b,d) are slightly larger at both levels, and the latitude ranges in the annual cycles are somewhat smaller. The summer minima in SD at the two levels are even less marked than in the WA. This is associated with a stronger summer maximum in track density at both levels.
The two levels also have secondary maxima in track density in winter. In the lower troposphere (Fig. 9d) this is quite strong, and the definite midwinter maximum in SD is consistent with this result. In the upper troposphere, the winter maxima in track density (Fig. 9b) and in SD (Fig. 9a) are weak and the main impression is of there being little change in SD strength from October to March.
In the upper troposphere the large intensities again extend to lower and higher latitudes, even slightly more than in the WA. The northern side of the 850-hPa storm track (Fig. 9d) is bounded by Greenland throughout the year.
As in the other sectors, it is only during the extended winter period that the storm track is at sufficiently low latitude to be directly affected by the stronger sea surface temperature gradients near 40°N in the extension of the Gulf Stream. Afargan and Kaspi (2017) have recently discussed the existence of a winter minimum in the Atlantic storm track, particularly in strong jet years. The diagnostics presented here for sectors from 90° to 30°W emphasize rather that the intensity of the storm track changes little over the extended winter season.
3) Eastern Atlantic
For the EA sector (which includes part of western Europe), the 250-hPa annual cycles in the sector average storm track measures are presented in Fig. 10. In addition to the relatively weak polar jet, the tropopause zonal wind in this sector (Fig. 10a) shows the subtropical jet over the winter half of the year. The polar jet is seen also at 850 hPa (Fig. 10c), and at both levels it is near 50°N in June and, surprisingly, moves slightly north through the summer and stays there until its weakening at the end of the winter. At both levels it is weakest in April and May.
In the extended winter season, September–March, the SD storm tracks at the two levels show little change in latitude or magnitude. At both levels, the tracks have a broad distribution in latitude, centered slightly south of the jet in the upper troposphere and north of the jet in the lower troposphere. There are weak maxima in November and January. The weaker, more compact summer storm track shows a slight poleward progression, keeping its relationship to the jet. In the EA, the SD storm tracks at the two levels show remarkably little change in latitude with time of year, particularly in the extended winter season from September to March. Throughout the year the lower-tropospheric storm track is centered about 5° north of that in the upper troposphere.
The track densities (Figs. 10b,d) also show a broad structure and latitudinal movement is generally small but, in concert with the jet, there is a slight poleward progression through the summer. The 250-hPa track density shows a strong summer maximum as in the CA, but is quite uniform through the rest of the year. The mean intensities have annual cycles similar to the other Atlantic sectors. Again it is the larger winter mean intensities that are reflected in the SD pictures. In the EA, the high intensities at 250 hPa spread deep into the subtropics and into the region of the subtropical jet from October to May. At both levels the higher intensities also spread into the polar region except in the summer.
The highest contour in track density at 850 hPa (Fig. 10d) indicates a rather flat distribution with latitude from September to March, with a hint of a double maximum from November to March. The detailed monthly latitudinal profiles of track density at 850 hPa (not presented) show that for eight months of the year, November–May and also September, there are indeed two weak maxima, near 54° and 62°N. Further investigation using the tracking of positive and negative V separately shows that the northern maximum is predominantly associated with southerly winds and the southern maximum with northerly winds.
The track densities at 850 hPa also show a marked track density maximum near 20°N from May to September, indicative of the African easterly waves in this sector (Thorncroft and Hodges 2001). The similar signature seen in Fig. 9 for the CA sector marks a continuation of these trends, with some becoming the tropical cyclone signature in the WA (Fig. 7).
The focus for this study is on the Pacific and Atlantic storm tracks, and these will be the subject of the discussion in this section. However, some analysis of the findings relevant to the Mediterranean storm track, subtropical easterly waves, and Arctic storm tracks is given in the online supplemental material, in sections SM.2, SM.3, and SM.4, respectively.
For the two major midlatitude storm tracks, the diagnostics presented here show many similarities in behavior from one sector to another over much of the Pacific and the Atlantic, despite differences in detail. The annual cycles for the CP, EP, WA, and CA in both the upper and lower troposphere have significant similarities in behavior. In amplitude (both of SD and mean intensity from tracking), each has a summer minimum and a relatively flat distribution from October to April. The latter occurs despite the strong winter maxima in the westerly jets. The ubiquity of this result gives a more general basis for a theoretical discussion of the relationship between jets and storm tracks in winter than the more usual focus on the midwinter minimum in the Pacific. This discussion will have to take into account stability limitations in a strong jet (e.g., Nakamura 1992; Christoph et al. 1997; Chang 2001; Deng and Mak 2005; Nakamura and Sampe 2002; Chang and Zurita-Gotor 2007) and other processes, such as the varying contribution from diabatic heating (Nakamura 1992; Chang 2001, 2009).
In the winter, the storm track structures are less coherent with the region of large storm intensities extending into latitudes outside the region of large track density. In particular, in the upper troposphere high-intensity storms are found at quite low latitudes in the central and eastern ocean basins. In section 3a(3) these have been identified with cutoff lows, and this identification may be applicable more generally.
Each storm track shows a poleward movement from May to reach near 50°N in August and from October to December a return to near 40°N. Consequently, the autumn storm track is poleward of that in spring, consistent with the asymmetry in the zonally averaged wind discussed by Fleming et al. (1987).
However, there are also significant quantitative differences between the two storm tracks. A hypothesis raised from the present work is that the upstream seeding from East Asia and North America is the origin of many of these differences. The importance of upstream seeding for the Pacific storm track has been the subject of many earlier studies (e.g., Nakamura 1992; Hakim 2003; Orlanski 2005; Chang 2005; Robinson et al. 2006; Penny et al. 2010; Chang and Guo 2012). Over East Asia the southwest extension of the upper-tropospheric Siberian storm track to near 48°N, where there is underlying baroclinicity downstream of the northern side of the Tibetan Plateau, leads to storm track activity with maxima in the spring and autumn seasons, which are not dominated by winter or summer monsoons. The dominance of spring and autumn maxima in the storm tracks weakens downstream, with winter values becoming more comparable, but which can still be detected in the eastern Pacific. The summer minimum is marked along the whole storm track.
The possible importance of the East Asian maximum in storm activity near 33°N between 100° and 130°E has been discussed by Hakim (2003) and Chang (2005). In this paper it has been seen that the lower-tropospheric maximum in track density is present only in the first half of the year, and is likely to be related to the spring rains in central China. It may be linked to the subtropical jet, though the upper-tropospheric maximum that is very clear in vorticity tracking (Part I, Fig. 1) is situated slightly north of it. The signature of this maximum disappears by about 150°E, though as discussed by Chang (2005) there may be interaction with storms on the more northerly track.
Turning to the Atlantic, in the sectors over eastern North America, the summer amplitude minima are less pronounced than in the Pacific sectors. In fact the track density shows a marked peak in summer. Also in the upstream sectors, the strength of the storm track, as given, for example, by SD, is quite flat from October to April. In addition, there is a marked latitudinal movement in the annual cycle. These features link closely to the annual cycle in the latitude and magnitude of the baroclinicity over the North American continent. The relative weakness of the summer minimum, with a maximum in track density at that time, and the flatness of the strength in the extended winter period are found downstream along the Atlantic storm track. The magnitude of the latitudinal cycle decreases to near zero at the eastern end of the track, implying an annual cycle in the meridional tilt of the storm track, from south-southwest to north-northeast in the winter to almost zonal in the summer.
The track density at 250 hPa in the Pacific has a minimum in midwinter, whereas that at 850 hPa has a weak maximum then. This trend with height is consistent with the very recent paper of Schemm and Schneider (2018), where it was found that surface cyclones have a track density maximum in the midwinter. In the Atlantic at 250 hPa, there is little change in the track density over the extended winter, but 850 hPa shows a weak midwinter maximum as in the Pacific.
The annual cycle of the pdfs of the lifetimes of the tracked features (not shown) have similar patterns of behavior in the Pacific and Atlantic, and at 250 and 850 hPa. In each case there is a 3-day peak that is largest in winter and smallest in summer. In contrast, for lifetimes longer than 6 days, the summer is dominant and the winter weakest. This is consistent with the longer periods in summer found for the Pacific by Chang and Yu (1999) and also for surface cyclones by Schemm and Schneider (2018). It is also consistent with the more marked summer minima generally seen here in the 2–6-day bandpass SD pictures than in the tracking pictures.
There are many aspects of the rich Northern Hemisphere storm track behavior that have been shown in Part I and here that are not well understood, and it is hoped that these papers will provide a stimulus for further diagnoses of observations and theoretical understanding of storm tracks. An important extension of the present study will be to consider the interannual variability of the seasonal cycles of the storm tracks and the relationship with low-frequency variability, in particular that associated with El Niño–Southern Oscillation (ENSO), and the North Atlantic Oscillation (NAO), Pacific decadal oscillation (PDO), and Atlantic multidecadal oscillation (AMO). It is also hoped that this pair of papers provides a basis for more detailed evaluations of the performance of climate models in today’s climate and diagnosis of their projections of storm tracks in a changing climate.
We thank Tim Woollings for provoking us into extending our previous analysis of NH storm tracks into other seasons. We also thank Paul Berrisford and the ECMWF Reanalysis team for the provision of the basic data used in this paper. Finally, we thank the reviewers for their helpful comments.
Supplemental information related to this paper is available at the Journals Online website: https://doi.org/10.1175/JCLI-D-17-0871.s1.
The original article that was the subject of this comment/reply can be found at http://journals.ametsoc.org/doi/abs/10.1175/JCLI-D-17-0870.1.