Comparison of Wave Packets associated with Extratropical. Transition and Winter Cyclones. Ryan D. Torn. Gregory J. Hakim

Transcription

1 Generated using version of the official AMS L A TEX template 1 Comparison of Wave Packets associated with Extratropical 2 Transition and Winter Cyclones 3 Ryan D. Torn Department of Atmospheric and Environmental Sciences, University at Albany, State University of New York, Albany, NY 4 Gregory J. Hakim University of Washington, Seattle, WA Corresponding author address: Ryan Torn, University at Albany, ES 351, 1400 Washington Avenue, Albany, NY rtorn@albany.edu 1

2 ABSTRACT Differences in the development of wave packets associated with midlatitude cyclones and the extratropical transition (ET) of tropical cyclones in the Western North Pacific (WNP) and Atlantic basins are diagnosed observationally by compositing reanalysis data over a 32-year period and applying the null hypothesis of no difference in the genesis, structure, propagation, or downstream extent. While the development of midlatitude cyclones during the fall and winter amplify a preexisting wave packet moving through the midlatitude storm track, wave packet amplification during ET appears to be primarily an in situ process, although cases involving interaction with a pre-existing midlatitude trough lead to a longer-lasting, more-amplified downstream response. In the WNP, ET wave packets have greater amplitude than winter midlatitude cyclones, but similar to fall midlatitude cyclones; ET events 16 are associated with a statistically significant anomaly farther downstream. By contrast, wave packets associated with fall and winter midlatitude cyclones in the Atlantic basin have greater amplitude and have a statistically significant anomaly further downstream relative to those associated with ET. In comparison to midlatitude cyclones, ET is characterized by small forcing for geopotential tendency due to lower tropospheric differential temperature advection. WNP ET are characterized by larger integrated moisture flux convergence and thus latent heat release relative to their midlatitude counterparts, while Atlantic basin ET has smaller moisture flux convergence compared to midlatitude cyclones, which could explain why Atlantic ET leads to less-amplified wave packets. 1

3 25 1. Introduction Poleward moving tropical cyclones (TCs) often undergo the process of extratropical transition (ET) from a symmetric warm-core to asymmetric cold-core cyclone. During this process, the decaying TC develops surface fronts, begins to derive more energy from baroclinic sources, particularly if a midlatitude trough is located immediately upshear of the TC, and is often associated with the amplification of an upper-tropospheric anticyclone downstream of the TC (e.g., Klein et al. 2000; Harr et al. 2000; Jones et al. 2003; Agusti-Panareda et al. 2004; Harr and Dea 2009; Grams et al. 2013a). The development of this ridge occurs in response to a number of dynamical mechanisms, including the adiabatic interaction of the TC circulation with the midlatitude waveguide (e.g., Ferreira and Schubert 1999; Riemer et al. 2008), and the diabatic outflow from either the TC core or the baroclinic zone that often develops on the downshear side of the cyclone (e.g., Bosart and Dean 1991; Harr and Elsberry 2000; Riemer et al. 2008; Riemer and Jones 2010; Torn 2010). The amplification of this ridge serves as an impulsive disturbance on the midlatitude waveguide (i.e., potential vorticity (PV) gradient), which can give rise to the downstream development of a wave packet (e.g., Simmons and Hoskins 1979; Chang and Orlanski 1993; Orlanski and Sheldon 1995; Hakim 2003). The energy of this impulsive disturbance is transported by the group velocity, which is often faster than the cyclone itself, and contributes to the development of new cyclones downstream. Previous studies have found large case-to-case variability of the downstream response of the midlatitudes during ET. Harr and Dea (2009) use an eddy kinetic energy framework to investigate downstream development following the ET of four Western North Pacific (WNP) 47 TCs during Their results show significant variability in the downstream response depending on the characteristics of the large-scale flow (e.g., background baroclinicity, existence of an upstream trough) and that the baroclinic reintensification of the ex-tc is not necessary to obtain a significant downstream impact. Archambault et al. (2013) found that the amplitude of the downstream meridional wind following the recurvature of a TC is a 2

4 function of the time of the year, latitude, and phasing of the TC with the midlatitude flow, such that TCs occurring during or after September, recurve between N, or reintensify as an extratropical system, are associated with larger-amplitude downstream flow. Riemer and Jones (2010) explore the role of ET on the midlatitude flow by comparing two idealized baroclinic development simulations, one with and one without ET. These simulations show that ET leads to more pronounced jet streak and ridge development through enhanced divergent flow due to the modified lower tropospheric baroclinic zone structure. Moreover, the indirect impact of ET is to produce a more favorable environment for downstream cyclogenesis, which helps to maintain the amplitude of the meridional flow. These wave packets are often associated with forecast errors well downstream of the ET (e.g., Jones et al. 2003; Harr et al. 2008; Anwender et al. 2008), which are in part due to errors and/or uncertainty in the relative phasing of the ET system and an upper-level trough (e.g., Klein et al. 2002; Torn and Hakim 2009; Riemer and Jones 2010; Grams et al. 2013b). Consequently, this motivates a deeper understanding of the development and propagation of these wave packets over many cases. In order to better understand how ET influences the downstream midlatitude state, it is worthwhile to compare the downstream response during ET to other phenomenon that produce impulsive disturbances and subsequent wave packet development along the midlatitude waveguide. Much of the literature on downstream development has focused on the role of midlatitude cyclones. Previous studies of individual cases (e.g., Orlanski and Katzfey 1991; Chang 2000) and larger statistical analyses (e.g., Chang 1993, 1999; Chang and Yu 1999; Hakim 2003) suggest that these wave packets are energy sources whereby upstream eddies can amplify the downstream flow. Similar to ET, other studies have demonstrated how midlatitude forecast errors can develop and propagate similarly to wave packets (e.g., Hakim 2005). For example, forecasts have also been shown to be sensitive to initial condition errors associated with a wave packet (e.g., Langland et al. 2002) and that the impact of assimilating targeted observations spreads downstream as a wave packet (e.g., Szunyogh 3

5 et al. 2000). The goal of this work is to compare wave packets associated with ET with those associated with winter cyclones in the WNP and Atlantic basins. These two basins are chosen because of the overlap between the region of maximum ET and midlatitude cyclogenesis frequency, with midlatitude cyclones most frequent in winter (e.g., Sanders and Gyakum 1980; Klein 84 et al. 2000; Hart and Evans 2001). In particular, this study evaluates whether one can reject the null hypothesis that there is no meaningful difference in the genesis, structure, propagation, or downstream extent of wave packets associated with ET and winter cyclones. This hypothesis is tested by comparing a large sample of wave packets associated with ET and winter cyclones by averaging over many cases and applying the packet diagnostic technique outlined in Hakim (2003) to quantitatively compare packet properties. While most of the previous work on the downstream impact of ET has focused on individual case studies or a small number of cases, this study bridges the ET and wintertime wave packet literature for a large sample. The remainder of the paper proceeds as follows. Section 2 describes the dataset and methods used to compute the wave packets and their properties. Results of the calculations are presented in section 3 followed by a summary and conclusions in section Method Wave packets associated with ET and winter storms in the Western North Pacific (WNP) and Atlantic basins are evaluated by compositing atmospheric fields, similar to the strategy employed by Hakim (2003). Atmospheric fields are taken from version 1 of the National Centers for Environmental Prediction (NCEP) Climate Forecast System (CFS) Reanalysis dataset (Saha et al. 2010) on mandatory constant pressure surfaces, with horizontal and 4

6 temporal resolution of 2.5 and 6 h, respectively, from Anomalies are defined as deviations from a moving-average climatology, meaning that every day of the year has a unique climatology, defined as the average of all daily fields within ± 14 d of the day of interest during the 32 year period. This method of computing the climatology has the advantage of producing a smooth climatology from one day to the next. Statistically significant anomalies are defined in terms of a two-sided Student s t-test with a threshold of 95%. The sample of winter midlatitude cyclone wave packets is determined by identifying cyclones at the location of highest frequency in each basin. As in Hakim (2003), cyclones are defined as local maxima in 1000 hpa geostrophic relative vorticity exceeding 10 4 s 1 during November March. Several additional checks are employed to ensure that individual cyclone events are identified. Any cyclone that is within 25 of another cyclone is removed from consideration. To ensure that the same event is not identified twice, any cyclone within 15 latitude and longitude of a previously identified cyclone during the previous 24 h is removed from the list of potential candidates. Hereafter, t = 0 refers to the time when the cyclone exceeds the above critical value, which approximates when the cyclone reaches the mature stage. Based on these criteria, the maximum number of cyclone events in the WNP occurs within 35 N-40 N, 145 N-155 N, while in the Atlantic the maximum number occurs within 40 N-45 N, W; the latter region is slightly to the east of the box used in Hakim (2003), likely due to the difference in the time period considered. These boxes contain 281 and 334 cases in the WNP and Atlantic Basins, respectively. Wave packets for ET cases are determined by evaluating all TCs contained in the Joint Typhoon Warning Center (JTWC) WNP best track data and the National Hurricane Center (NHC) Atlantic data from A TC is considered a candidate for ET if the track 1 Although the raw resolution of the CFS reanalysis is 0.5, at this resolution, the mass fields are noisy near TCs, likely due to the method used to relocate the TC from its location in the 6 h forecast to the observed position in the analysis (e.g., Liu et al. 2000); therefore, the lower resolution 2.5 dataset, which shows no evidence of this issue, is employed. Given that this study mainly focuses on synoptic-to-planetary scale features, this choice of resolution should not be a limitation. 5

7 underwent recurvature at some point in its life, meaning it had an easterly component of motion, and the TC moved poleward of 20 N; the latter condition removes any TCs that 127 drifted within the deep tropics, but never moved into the midlatitudes. For each of the remaining TCs, the reanalysis position is determined by finding the minimum in 1000 hpa geopotential height each 6 h; for a majority of times, the best track and reanalysis positions are within 1 of each other. At each time, the asymmetry parameter from the Hart (2003) cyclone phase space 2 is calculated from reanalysis data. This parameter provides a measure of the thermal asymmetry across the cyclone and has been used to objectively identify the onset of transition. As in Hart (2003), the onset of transition is defined as when the asymmetry parameter exceeds 10 m and is hereafter referred to as t = 0. In each basin, the maximum number of ET cases within any 5 10 box is 34, which does not provide a robust composite to calculate wave packet properties on an earth-relative grid due to the small number of cases. Instead, this study considers all cases where the onset of ET occurs between N and E in the WNP and N and W in the Atlantic; these latitude bands contain the largest number of ET onset within each basin. This choice results in 112 and 91 cases in the WNP and Atlantic, respectively. It is worth noting that this study considers all cases of ET within this location, regardless of whether the TC completed its transition into a baroclinic system. Previous studies have suggested that the reintensification as a baroclinic system is not necessary for a midlatitude response to occur (e.g., Harr and Dea 2009; Riemer and Jones 2010; Archambault et al. 2013). Figure 1 shows the number of ET and winter cyclones in each basin as a function of 146 month. While the largest number of ET cases occur between August and October, the winter cases are mainly in February and March in the WNP and December-January in the Atlantic. The difference in timing between ET and winter cyclones results in different background states through which the wave packets develop and propagate; the implications 2 The asymmetry parameter is defined as the difference between the hpa thickness averaged over a semicircle to the right of the track and left of the track. 6

8 of this difference are explored in greater detail in the next section. To overcome some of the differences associated with season, we also compute wave packet properties for midlatitude cyclones that occur during August-October and ET during November-December; details on these experiments are located in the next section. The remainder of this paper will employ composite averaging of ET and midlatitude cyclone cases in each basin over a range of time lags. The relatively large number of midlatitude cyclone cases allow for straightforward averaging in an Earth-relative frame of reference; however, for the ET cases, all fields are shifted to a common longitude, which is defined as the longitude of the cyclone at the onset of ET. As pointed out in Hakim (2003), the ensemble averaging method has the advantage that it is relatively simple and does not integrate away time and amplitude information; however, the signal contained in the ensemble mean will degrade to zero with increasing time lags due to different trajectories and velocities of individual events. Properties of the composite wave packet at each time are analyzed using the methods outlined in Hakim (2003) as summarized here. Wave packets are identified from the 300 (250) hpa meridional wind field for winter cyclones (ET) over a 250 longitude window centered on the maximum in the absolute value at a grid point. The peak of the packet is determined by fitting a sixth order polynomial to the six grid point extrema about the maximum value, while the local extrema and zero crossings are determined from the interpolated polynomial near the packet peak. The leading edge of the packet is determined by linear regression of the interpolated packet to an exponential profile, with the leading edge defined as 2.5 e-folding distance from the peak (8% of the peak value). Group velocity is found by computing the distance between the packet peak 6 h into the future and 6 h in the past and dividing by 12 h. To reduce the inherent noise that comes from discrete estimates of derivatives, a smoother is applied to the packet peak latitude and longitude values prior to computing the group velocity. Finally, the wavelength of the packet is determined by computing the distance between zero crossings and extrema. 7

9 Results 178 a. Western Pacific (WNP) Composites of the upper tropospheric meridional wind 3 during WNP ET and winter cyclone cases reveal important differences in the wave packet evolution through t = 0 (Fig. 2a,c,e). The dominant signal in the ET wave packet at 48 h is a 5 m s 1 positivenegative meridional wind anomaly couplet centered near 155 E with associated weak ridging in the PV field 4 (Fig. 2a). By 24 h, the amplitude of the meridional wind in this region exceeds 10 m s 1, with the corresponding undulation in the PV field suggestive of an amplifying ridge 5 to the east of the 48 h position (Fig. 2c). At the onset of transition (0 h), the ridge amplitude increases further; however, the axis of this ridge has only moved 5 relative to the 24 h position (Fig. 2e). Moreover, there is an indication of a downstream trough centered at 170 W, suggesting a nascent wave packet has developed. Over this same period of time, a lower amplitude positive-negative meridional wind anomaly dipole, indicative of an upstream wave packet, originally near 100 E at 48 h moves eastward and intercepts the ridge near t=0 h; however, the amplitude of this packet never exceeds 3 m s 1, thus the ridge appears to the be the dominant anomaly in the midlatitude flow. Overall, this result suggests that the process of ET is associated with an amplifying, but nearly stationary ridge, which appears to be primarily responsible for the development of a downstream wave packet. In contrast to ET cases, the winter cyclone composite is characterized by the amplification of the predecessor wave packet prior to t = 0. At t = 48 h, there is a relatively weak (3 m s 1 ) wave packet centered to the west of Japan that subsequently amplifies as it moves eastward with time (Fig. 2b). Over the next 48 h, the meridional wind anomalies increase to 18 m s 1, during which time the packet moves 30 to the east (Fig. 2d,f). Similar to Hakim 3 The winter (ET) cyclone composites are performed at 300 (250) hpa because the meridional wind anomalies are maximized and reflect the seasonal cycle of tropopause height. 4 To facilitate comparison with the winter cyclone cases, the geography on the ET figures is oriented such that the cyclone position at t = 0 matches the winter cyclone position at t = 0. 8

10 (2003), this pattern suggests that winter cyclones tend to amplify a weak pre-existing wave packet in the midlatitude wave guide. In addition to differences in how wave packets are generated during ET and winter cyclones, there are also subtle differences in how the packets evolve after t = 0. For ET, the wave packet is associated with an amplifying trough at 160 W at 24 h (Fig. 2g) and finally a nascent ridge with the axis over the west coast of the United States by 72 h (Fig. 2i,k) before vanishing over North America by 96 h (not shown). Although the winter cyclone wave packet exhibits a similar eastward propagation, the southern edge of the packet is characterized by a positive tilt, implying refraction into the tropics starting at 24 h (Fig. 2h), which is not present in the ET composites. Moreover, it appears that the winter cyclone wave packet peak does not reach North America, while it does in ET, which suggests that, on average, ET wave packets propagate farther downstream. This difference is at least partly due to the nature of the waveguide associated with these two types of systems. The ensemble-mean meridional PV gradient during ET is nearly constant across the entire Pacific Ocean, which implies a more consistent waveguide. By contrast, the winter cases are characterized by a larger meridional PV gradient to the west of the Dateline compared to the east, implying a weaker restoring force for Rossby waves and wave breaking (Hoskins et al. 1985, e.g.,). The time evolution of the downstream response to ET and midlatitude cyclones can be compactly summarized via Hovmoller diagram of the meridional wind anomalies averaged between N, which represents the latitude band with the largest anomalies (Fig. 3). Similar to the plan view composites, the largest amplitude meridional wind anomalies for ET prior to 0 h is associated with the ridge just downstream of the ET (Fig. 3a). Moreover, the upstream signal is also apparent, which indicates a wave packet moving from 90 W at 60 h, intercepting the ridge at t=0 h. We suspect that that the upstream anomalies appear when averaging over a latitude range because such averaging allows for case-tocase variability in latitude. Nevertheless, at all times prior to t=0 h the larger amplitude anomalies are associated with the amplifying ridge downstream of ET. Moreover, the positive 9

11 meridional wind anomaly downstream of the ridge becomes statistically significant prior to the upstream packet reaching the ridge. Overall, this figure supports the notion that wave packet initiation is primarily related to the amplification of the downsream ridge rather than an upstream packet. By contrast, the Hovmoller diagram for winter cyclones clearly indicates the amplification of the upstream feature originally near 100 W at t= 60 h as it propagates to the east and forward in time (Fig. 3c). To further understand the role of upstream precursors on the subsequent downstream wave packet, the ET cases are further subdivided depending on whether a 500 hpa trough interacts with the TC near the onset of transition, which has been found to determine if the ET system reintensifies as a baroclinic system (e.g., Klein et al. 2000; Harr et al. 2000). Here, we define an ET system as having a trough interaction if a relative maximum in 500 hpa relative vorticity exceeding s 1 is less than 1000 km upstream of the cyclone position within 24 h of the onset of transition. The value of 1000 km represents the distance between the TC and upper-level trough in the Northwest composite of Harr et al. (2000), and the 24 h time tolerance allows for the possibility that the trough interaction could come after the onset of ET. Of the 112 ET cases used in this study, 54 had an identifiable upstream trough. Figure 4 shows that the ET case composite with troughs has a maximum 0 h meridional wind perturbation of 22 m s 1, while the no-trough composite has a 16 m s 1 maximum. Both composite wave packets subsequently propagate downstream, but the cases with an upstream trough are characterized by larger-amplitude anomalies. At 36 h, both composites show the center of the wave packet east of 180, with the trough composite having larger amplitude (Fig. 4c,d), while at 72 h, only the wave packet associated with the trough cases 250 reach North America (not shown). Similar to the composite of all ET cases, averaging the meridional wind between N reveals an upstream wave packet that intercepts the ridge associated with ET; however, the amplitude of this upstream signal is smaller than the ridge at all times and the downstream trough forms prior to the arrival of the upstream 10

12 packet (not shown). These composites suggest that a pre-existing trough is not necessary for wave-packet formation, but that those ET events with predecessor troughs have an amplified downstream signal. The larger-amplitude signal for cases having an upstream trough is not surprising since this flow configuration is associated with the baroclinic reintensification of the TC (e.g., Harr et al. 2000) and is often associated with the generation of eddy kinetic energy, which acts to amplify the upper-tropospheric flow (e.g., Harr et al. 2000; Harr and Dea 2009; Keller et al. 2014). For the ET composite, it is not clear what is the appropriate definition of t = 0; this choice could lead to apparent differences between ET and winter cyclone wave packets. Previous studies have attempted to determine the downstream response to recurving TCs by compositing based on the recurvature time, which is defined as the time when the TC reaches its westernmost position (e.g., Archambault 2011; Archambault et al. 2013), while this study defines t = 0 to be the onset of transition. These two times can differ by 0 3 days depending on the case. To determine the potential implications of how t = 0 is defined, the compositing procedure is repeated for all TCs that begin the process of ET and undergo recurvature between N (the latitude band where the greatest number of recurvature points are located), but where t = 0 is defined as the recurvature point. Results for this case are comparable to Archambault (2011), in that the composite is characterized by a weak wave packet signal (2 m s 1 ) that moves across Asia into the WNP prior to recurvature. This definition of t = 0 yields a maximum meridional wind perturbation of 10 m s 1 at 24 h (not shown), compared to 19 m s 1 in this study which defines t = 0 as the onset of transition. Further subdiving the ET cases by the difference between recurvature and ET onset times reveal some differences in the midlatitude response. The composite of ET cases where the difference between these two times is less than 24 h (35 cases) yields Hovmoller diagrams qualitatively similar to the trough cases, with an upstream signal that intercepts the ridge at t=0, with the downstream trough forming prior to the arrival of the upstream feature (not shown). By contrast, cases where the recurvature time and ET onset differ by more 11

13 than 24 h do not have any indication of an upstream signal, yet there are minor differences in the downstream amplitude and extent. This result further supports the result that ET can generate a wave packet in situ. One possible explanation for the difference in ET and winter cyclone wave packets is that ET is most frequent in August, September and October (ASO) when the background state can be quite different than in winter. The possible role of the seasonal cycle in these results is explored by constructing a composite of midlatitude cyclones in the WNP basin during ASO that meet the cyclone criteria described in section 2 and are within the same latitude band as the maximum in ET onset; this criteria yields 100 cases, mainly during October (Fig. 1a). As with ET, the fields are shifted to a common longitude, which is defined as the cyclone position at t = 0, due to the relatively smaller number of cases. At 48 h, the ASO cyclone composite is characterized by a nearly stationary trough centered near 150 E, with an upstream precursor wave packet centered near 100 E that moves eastward with time (Fig. 5a,b). By t = 24 h, the upstream packet reaches the pre-existing trough in the Pacific to form a larger packet, which is better seen in the Hovmoller diagram of this composite (Fig. 3e). The wave packet subsequently amplifies to 20 m s 1 by 0 h and moves across the Pacific, with some evidence of refraction into the tropics between h (Fig. 5c-e). By 72 h, a statistically significant signal is near 105 W (Fig. 5f), which becomes statistically insignificant thereafter due to the combination of wave breaking into the tropics or composite smearing of the signal. Overall, this composite suggests that the evolution of wave packets associated with ASO cyclones are quite similar to their winter counterparts, with the exception of less equatorward refraction. In the WNP, ET can also occur during November and December, which overlaps with the period of midlatitude cyclogenesis frequency. As a consequence, it is possible to account for the difference in background state by comparing November and December ET cyclones with winter cyclones. The November and December ET composite is constructed by considering 12

14 all cases where the onset of ET occurs between 22.5 N and 27.5 N 5 and shifting the composite to a common longitude, similar to what is done with other ET cases; this choice yields 23 cases for the 32 year period. Figure 6 shows that these November and December ET cases are qualitatively similar to ASO ET. Prior to ET onset, there is a statistically significant ridge centered near 150 E at t = 48 h, with evidence of an upstream trough near 120 E (Fig. 6a). Similar to ASO ET cyclones, the positive meridional winds associated with the downstream trough become statistically significant prior to the upstream trough reaching the ET cyclone, suggesting in situ generation of the wave packet (Fig. 6b,c). Although the magnitude of the meridional wind anomalies of the November and December ET are similar to the ASO ET composite between h, the November and December ET composite is characterized by statistically significant positive-negative anomalies characteristic of a ridge centered on 150 W at 48 h (Fig. 6e) and subsequently anticyclonic wave breaking over the central United States at 72 h (Fig. 6f). Overall, this suggests that, on average, the midlatitude response to late-season ET is similar to what occurs during ASO, though there is some suggestion of a coherent wave packet further downstream in November and December. As a result of this and the previous composite, we conclude that differences in background state are not the primary reason for the observed difference between winter cyclone and ET wave packets. Although the plane-view plots show evidence of differences between ET and midlatitude cyclone wave packets, an objective, quantitative comparison of the composite wave packets is performed to test the conclusions outlined above. The composite ASO ET and midlatitude cyclone wave packet properties are objectively analyzed using the methods outlined in section 2b to determine their wavelength, amplitude and group velocity 6. In order to improve the robustness of the results, the packet properties are computed from the meridional wind averaged between N. Statistically significant differences between composite packet properties are determined using error estimates based on bootstrap resampling with replacement. The difference between winter and ET cyclones is evaluated by randomly sam- 5 This represents the latitude band with the largest number of cases. 6 November and December ET cases are not analyzed due to the low number of cases within the composite. 13

15 pling 112 winter cases (the number of ET cases), compositing the meridional wind anomalies, and computing the wave packet properties. This process is repeated 2000 times to yield a 95% confidence bounds on the resulting distribution of packet properties. Differences between ET and ASO midlatitude cyclones are assessed using a similar procedure, but where 100 ET cases are randomly sampled. Figure 7a,b shows the composite wave packets associated with ET and winter cyclones at t = 0; these figures are similar to that obtained for t = 24 to +48 h (not shown). For both ET and winter cases, the composite wave packets are similar to those observed in Hakim (2003). In particular, the composite wave packet amplitudes exhibit a good fit to an exponential profile with an abrupt westerly edge, consistent with impulsive disturbances, which from theoretical arguments should have an exponential zonal structure (e.g., Swanson and Pierrehumbert 1994). The composite packet peak amplitude is consistent with many of the aforementioned ideas of the development and propagation of ET and midlatitude cyclone wave packets (Fig. 8a). Although this composite captures the amplitude of the midlatitude flow, it is contaminated by sample smearing of anomalies away from t=0 h, though the same method is applied to all composites, providing a consistent comparison. For ET, the composite peak amplitude increases from 8 m s 1 at 24 h to 14 m s 1 12 h after the onset of ET, then decreases to 10 m s 1 by 48 h. By contrast, the composite winter packet amplitude is 2 m s 1 smaller than the ET wave packet up to 6 h, peaks at 13 m s 1 at 12 h, and decreases in similarly to the ET packet beyond 24 h. Up until 12 h, the ET value is outside the 95% confidence bounds on the winter cyclone signal, thus ET has greater amplitude up until that point. For ASO cyclones, the composite peak amplitude is greater than ET between 0-24 h, but decays more quickly than ET after 24 h. As a consequence, it appears that ET cases have smaller amplitude compared to ASO cyclones, but retain a larger signal for a longer period of time. While this result implies that the winter cyclone composite wave packet has similar or smaller amplitude to the ET wave packet, this is potentially deceiving because the clima- 14

16 tological standard deviation in meridional wind is not constant across the seasonal cycle. Of these three sets of cases, the climatological standard deviation in meridional wind in the WNP is greatest for the ASO cyclones (primarily during October), smallest for ET (primarily in September), with winter in between (primarily February-March). To address this issue, the packet peak amplitude for each case is normalized by the climatological standard deviation in the meridional winds at the location of the packet peak for that day. For ET, the peak amplitude is 0.92 standard deviations above climatology, compared to 0.88 for ASO cyclones; the difference between these two is statistically significant at all lead times (not shown). The winter peak amplitude is 0.80 standard deviations, with the difference between ET and winter cyclones statistically significant up to 36 h. As a consequence, the composite ET wave packets have the largest amplitude relative to climatology at the time of peak wave packet amplitude. The group velocity calculations support the notion that the development of midlatitude cyclones results in the enhancement of an existing wave packet, while for ET the main signal is in situ packet generation. Fig. 9a shows that the composite winter cyclone wave packet group velocity increases from 5 m s 1 at t = 24 h to a maximum of 18 m s 1 at 6 h, 376 which is slower than the background flow of 32 m s 1. The reduction in group velocity with time beyond 6 h likely reflects the decrease in the background flow across the basin, which could act to focus wave packets (e.g., Esler and Haynes 1999; Chang and Yu 1999; Hakim 2003), though it could also be related to smearing of the composite. By contrast, the ET wave packet group velocity is below 10 m s 1 until t = 0 h and becomes nearly identical to winter cyclones at t = 18 h; the difference between the winter cyclone and ET is statistically significant until 12 h. This relatively slower group velocity for ET is consistent with the notion that the packet is strongly forced near a particular location, rather than freely propagating. While the ASO cyclone group velocity is comparable to ET prior to 12 h, this may be due to the fact that the algorithm used to identify the composite packet peak has difficulty distinguishing between the aforementioned slowly moving trough and 15

17 the packet that moves off the Asian coast. Nevertheless, this result suggests that external forcing, such as latent heat release associated with the transitioning TC, is important to packet development and that the forcing has near zero zonal velocity. Figure 10 shows the packet peak wavelength for both the ET and winter cyclone cases. As suggested by Fig. 2, the ET wave packet wavelengths are about km longer than ASO or winter cyclones (statistically significant until 36 h) for most lead times. These wavelength differences could be due to the structure of the background flow through which the wave packets are traveling and the spatial extent of the impulsive forcing on the waveguide. We have speculated that the different group velocities for ET and winter wave packets prior to 0 h suggests that different dynamical processes may be responsible for the generation and amplification of the wave packets. As stated earlier, both adiabatic and diabatic processes can contribute to the amplification of the midlatitude flow. In order to gain an estimate of the relative contribution of adiabatic processes to the wave packet amplitude in each set of cases, the forcing for geopotential time tendency due to differential temperature advection (e.g., Holton and Hakim 2013, eq. 6.64) is compared for ET, winter and ASO cyclones at t = 0. The geopotential time tendency is related to the differential temperature advection through an elliptical operator, which can be sensitive to the assumed mean stability profile and boundary conditions. Therefore, we present the vertically integrated differential temperature advection between hpa as a proxy, since differential temperature advection through this layer is typically associated with the amplification of the upper-tropospheric flow during midlatitude cyclogenesis (e.g., Holton and Hakim 2013). This quantity is obtained by computing the differential temperature advection at 50 hpa increments and each horizontal point for each case, computing the vertical integral between hpa, then averaging over all cases. Statistical significance of differences between composites is determined through bootstrap resampling of cases via the same procedure used in the packet properties. 413 Figure 11 shows that midlatitude and ET cyclones are characterized by qualitatively 16

18 different vertically integrated differential temperature advection patterns. While ET is characterized by weak forcing for negative geopotential tendency on its north side, which coincides with the upper-level ridge (Fig. 11a), the winter and ASO cyclone composites contain the familiar forcing for increasing (decreasing) geopotential on the eastern (western) side of the surface cyclone (Fig. 11c,e), which in turn would amplify the upper-level flow. This result suggests that, relative to ET, the midlatitude cyclones are characterized by greater adiabatic forcing for upper-level flow amplification. The reanalysis dataset does not contain an analysis of latent heat in order to compare with adiabatic processes; therefore, a proxy quantity must be used instead. Here we use the column-integrated moisture flux convergence that, from the water continuity equation, should be proportional to the column-integrated difference between condensation and evaporation, which is proportional to latent heat release. The column-integrated moisture flux convergence is computed by integrating moisture-flux convergence over the column for each case, and then averaging over all cases. Fig. 12 indicates that all three composites are are characterized by a region of moisture flux convergence to the northeast of the cyclone; however, the ET maximum is 1130 W m 2, compares to 810 W m 2 for winter cyclones (the difference between the two is statistically significant at 95%) and 1010 W m 2 for ASO 431 cyclones (difference statistically significant to 95%). We interpret this to mean that ET cases are characterized by greater latent heat release and thus greater forcing for height rises compared to either winter or ASO cyclones. 434 b. Atlantic Composites of Atlantic basin ET and winter cyclone wave packets share many qualitative similarities to their WNP counterparts, except with smaller amplitude. Figures 13a,c indicates that Atlantic ET is characterized by a nearly stationary, but amplifying ridge centered near 50 W at t = 48 h (Fig. 13a) with a distinct upstream positive-negative anomaly pair centered near 120 W. The upstream feature subsequently propagates to the east and 17

19 intercepts the ridge by t = 24 h (Fig. 13c,e). At all times, the amplitude of the meridional wind anomalies associated with the ridge are larger than the anomaly associated with the upstream feature, which is also apparent in Hovmoller plots of the latitude-averaged wind (Fig. 3b). By comparison, the winter cyclones are characterized by a weaker pre-existing wave packet centered near 110 W at 48 h that subsequently moves east and amplifies with time (Fig. 13b,d,f and Fig. 3d). Following the onset of ET, the wave packet moves eastward toward Europe; however, the 250 hpa meridional wind anomaly associated with the wave packet does not retain statistical significance beyond 48 h at 0 longitude (Fig. 13g,i,k) or beyond 60 h in the Hovmoller 449 (Fig. 3b). Although there is a statistically significant meridional wind anomaly at 25 E starting at 0 h, this anomaly appears to be unrelated to the ET wave packet for reasons that will be explained below. Composites of ET cases with (46 of 91 cases) and without an upstream trough during the onset of transition show that the meridional wind amplitude is 3 m s 1 greater for the trough cases and the statistically significant anomaly reaches further downstream (Fig. 14). Returning to the negative meridional wind anomaly near 25 E in Fig. 13e, this anomaly is present in the no trough composite prior to the arrival of the ET wave packet (Fig. 14b) and has larger amplitude than the trough composite at the same time (Fig. 14a). Moreover, this anomaly becomes statistically insignificant at 12 h in the trough composite (Fig. 14c), but remains statistically significant in the no trough composite through 48 h (Fig. 14d), thus it appears that this negative meridional wind anomaly is not related to the upstream ET. By contrast, the winter cyclone wave packet has a statistically significant signal, of which a portion is refracted into the tropics, while the remainder reaches Europe by 48 h and decays to zero at 15 E (Fig. 13f,h,j,l). To investigate the role of differences in background state, we compare these results with composites of ASO cyclones as for the WNP composites (Fig. 15). This choice yields 46 cases, with the majority occurring in October (Fig. 1b). Similar to winter cyclones, there is evidence of an eastward propagating wave packet that starts over western North America at t = 48 h, 18

20 amplifying to 22 m s 1 by the time it reaches the western Atlantic at t = 0 (Fig. 15a-c). Over time, the packet moves to the east, exhibits poleward refraction as evidenced by the negative tilt to the anomalies at 48 h (Fig. 15d,e); however, a statistically significant anomaly is still evident over the Baltic sea (Fig. 15f) and in Hovmoller diagrams (Fig. 3f) at 72 h. Overall, this suggests that ASO cyclones in the Atlantic behave much like winter cyclones in that basin and that the differences between ET and winter cyclones in this basin are not solely attributable to seasonal effects. ET and winter cyclone wave packet properties are also compared for the Atlantic basin. For winter cyclones, the composite peak packet amplitude is 14 m s 1 at 0 h, which decays to half that value by 42 h when the packet reaches Europe (Fig. 8b). By comparison, the ET packet amplitude has a maximum value of 12 m s 1 before decreasing to less than 9 m s 1 by 36 h. The amplitude difference is statistically significant at 95% from 0-24 h, suggesting that winter cyclone packets obtain greater amplitude. In addition, the ASO cyclone packet amplitude is roughly 6 m s 1 greater than ET cyclones at the time of peak amplitude, with the difference being statistically significant at most times. Normalizing the amplitude by the climatological standard deviation indicates that the ET wave packet amplitude is about 10% less than winter cyclones (statistically significant at 95% between 0-18 h), while ASO cyclones are roughly 20% greater than ET (statistically significant at all times; not shown). Atlantic wave packet group velocities are similar to the WNP results, which showed that winter cyclones enhance an existing wave packet, whereas ET packets are produced in situ (Fig. 9b). The winter-cyclone wave packets have a group velocity that varies between m s 1 throughout the period. By contrast, the ET wave packet group velocity is slightly negative until 0 h, which is consistent with the idea that the wave packet is mainly formed in situ (the difference in group velocity between winter and ET cyclones is statistically significant at 95% confidence up to 12 h). Beyond that time, the ET packet peak group velocity is quite similar to the winter cyclones, suggesting that propagation properties are similar once the packet is of sufficient amplitude and detached from the forcing. Moreover, the composite 19

21 ASO cyclone packet group velocity is similar to winter cyclones and is statistically indistinguishable from ET cyclones up to 12 h. Finally, the ET packet wavelength is greater than the winter cyclones for most times prior to 0 h, although the difference is not statistically significant thereafter (Fig. 10b). We infer geopotential height tendencies from vertically integrated differential temperature advection patterns as for the WNP, with qualitatively similar results. In particular, while ET is characterized by weak forcing for negative geopotential time tendency to its north, both ASO and ET cyclones have forcing for positive (negative) geopotential tendency on the west (east) side of the surface cyclone (Fig. 11b,d,f), similar to what is obtained in the WNP. One explanation for the weaker wave packets in the Atlantic relative to the WNP ET events could relate to diabatic forcing. The maximum vertically-integrated moisture flux convergence for Atlantic ET (Fig. 12b) is 70 (58) % of winter (ASO) cyclones in the basin (Fig. 12d,f); both of these quantities are statistically significant at the 95% level. Moreover, this moisture flux value is less than half its WNP counterpart (Fig. 12a); therefore, it appears that on average, Atlantic ET cases may have smaller forcing for height rises due to diabatic heating, which could explain why the amplitude of ET wave packets in the Atlantic is smaller than winter cyclones and WNP ET around the time of peak amplitude. Another potential reason for the lower amplitude in Atlantic ET wave packets as compared to winter cyclones relates to the case-selection criteria. Recall that the Atlantic composite includes all TCs that underwent ET between 35 N and 40 N. For some of these cases, this happened while the TC was over land. By contrast, none of the WNP ET cases made landfall over an appreciable land mass, except for Japan. To determine whether these over-land ET cases bias the ensemble-mean ET wave packet, the composite calculations are repeated for all ET cases that remain over water prior to and during ET. The ensemble-mean wave packet for the over-water Atlantic ET cases is nearly identical to the ensemble-mean wave packet for all (not shown); therefore, land does not appear to be a factor in Atlantic 20

22 521 ET having weaker wave packets Summary and Conclusions This study poses the hypotheses that wave packets associated with the ET of tropical cyclones in the WNP and Atlantic basins have, on average, the same: (1) genesis, (2) structure, (3) propagation, and (4) downstream extent as those associated with midlatitude cyclones. These hypotheses are tested on composites of ET and midlatitude cyclone cases from reanalysis data over a 32-yr period. The properties of the wave packet are then analyzed using the techniques described in Hakim (2003). Of the four aspects of wave packet properties studied here, the null hypothesis of no difference in packet genesis between ET and midlatitude cyclones is most easily rejected. Packet amplification in both winter and ASO midlatitude cyclones appears to involve a preexisting wave packet that can be tracked backward in time through the midlatitude storm 533 track. By contrast, the main source of the downstream wave packet during ET appears to be associated with the in situ amplification of the downstream ridge, which eventually leads to the amplification of a new trough further east. Although an upstream precursor trough is present in 48% of ET events in the WNP and 51% of cases in the Atlantic, the precursor amplitude is weaker than the ridge at all lead times and the downstream trough forms before the precursor intersects the ridge. Moreover, ET events that do not involve the interaction with a precursor trough are still characterized by the downstream ridge-trough pair, which supports the conclusion that wave packet generation during ET is primarily due to the amplification of the downstream ridge. Following the onset of ET and winter cyclone maturity, the null hypothesis cannot be rejected for some wave packet structural and propagation properties, including the group velocity and exponentially-decaying structure, while other properties, such as the amplitude and downstream extent, are characterized by clear differences between midlatitude cyclones 21

23 and ET. In the WNP, the peak amplitude of ET wave packets is indistinguishable from ASO midlatitude cyclones, but greater than winter cyclones. By contrast, Atlantic basin midlatitude cyclones (both ASO and winter) have greater amplitude relative to ET. In addition, while WNP ET packets have a statistically significant signal that propagates further downstream relative to winter or ASO midlatitude cyclones, the opposite is true in the Atlantic basin. As a consequence, it appears that, on average, WNP ET leads to a longerlasting, higher amplitude response in the downstream flow compared to midlatitude cyclones, with the reverse true in the Atlantic. Separating the ET events by the existence of a precursor trough indicates that those cases with a trough are characterized by greater downstream amplitude and the wave packet reaches further downstream, thus it appears that a precursors lead to a greater downstream response. Some of the differences between ET and winter cyclone wave packets are likely related to the dynamics of packet development. We have provided proxy evidence that, in both the Atlantic and WNP basins, winter and ASO cyclones are characterized by larger lowertropospheric adiabatic forcing for geopotential height tendency compared to ET events. In contrast, WNP ET cases have larger vertically-integrated moisture flux convergence relative to the winter cases. Assuming that moisture flux convergence correlates with latent heat release implies that diabatic processes are more important in wave packet amplification in ET cases relative to winter cyclones. The important role of latent heat release in amplifying the midlatitude flow ET is consistent with previous idealized and case studies of ET (e.g., Harr and Dea 2009; Torn 2010; Riemer and Jones 2010; Archambault et al. 2013). In the Atlantic basin, the proxy for adiabatic geopotential height tendency and moisture flux convergence associated with ET is smaller than winter or ASO cyclones, which would lead to less wave packet amplification and could explain why Atlantic-basin ET wave packets are on average weaker than those for winter cyclones. Our conclusions are limited by some aspects of the method. Perhaps the most important is that with composite analysis the signal can be significantly degraded for non-zero time lags, 22

24 particularly if the set of cases making up the composite is characterized by a wide variety of trajectories and velocities. This issue makes assessing the far downstream signal particularly problematic; therefore, this method may not provide the most accurate method of assessing 576 the horizontal extent of these wave packets. On the other hand, the same methodology has been applied to both ET and midlatitude cyclones; therefore, unless one particular set of cases is characterized by a greater variety of wave packet trajectories and velocities, the limitations of the method should not introduce a bias toward one type of phenomenon. Although this work indicates that ET and winter cyclone wave packets have many similar characteristics, the differences in how these packets are produced could have important consequences on downstream predictability during each event. Given the relative dearth of in situ moisture observations over the lower-tropospheric ocean, models may have relatively large moisture analysis errors, which could translate into different latent heating rates and details in the downstream ridge during ET cases (e.g., Torn 2010). Moreover, since the packet derives from the ET process itself, TC track errors could also introduce uncertainty in the timing and amplitude of the subsequent wave packet (e.g., Harr et al. 2008; Anwender et al. 2008; Riemer and Jones 2010; Keller et al. 2014). By comparison, the winter cyclones appear to amplify pre-existing wave packets, and provided the upstream disturbance is resolved by the current observation network, downstream predictability in these cases might not be as sensitive to the details of surface cyclogenesis. Future work will compare forecasts during ET events with those of winter cyclones to evaluate these ideas. 593 Acknowledgments This research benefited from discussions with Heather Archambault, Lance Bosart and Jack Beven and Dan Keyser. We thank two anonymous reviewers for their insightful comments on an earlier version of this manuscript. CFSR data was obtained from the National Oceanic and Atmospheric Administration National Climatic Data Center. This work was sponsored by NSF Grants AGS , and AGS

25 REFERENCES Agusti-Panareda, A., C. D. Thorncroft, G. C. Craig, and S. L. Gray, 2004: The extratropical transition of hurricane Irene (1999): A potential-vorticity perspective. Quart. J. Roy. Met. Soc., 130, Anwender, D., P. A. Harr, and S. C. Jones, 2008: Predictability associated with the down- stream impacts of the extratropical transition of tropical cyclones: Case studies. Mon. Wea. Rev., 136, Archambault, H. M., 2011: The downstream extratropical flow response to recurving Western North Pacific tropical cyclones. Ph.D. thesis, University at Albany, 200 pp Archambault, H. M., D. Keyser, L. F. Bosart, and J. M. Cordeira, 2013: A climatological analysis of the extratropical flow response to recurving western North Pacific tropical cyclones. Mon. Wea. Rev., 141, Bosart, L. F. and D. B. Dean, 1991: The agnes rainstorm of June 1972: Surface feature evolution culminating in inland storm redevelopment. Wea. Forecasting, 6, Chang, E. K. M., 1993: Downstream development of baroclinic waves as inferred from 615 regression analysis. J. Atmos. Sci., 50, Chang, E. K. M., 1999: Characteristics of wave packets in the upper troposphere. Part II: Seasonal and hemispheric variations. J. Atmos. Sci., 56, Chang, E. K. M., 2000: Wave packets and life cycles of troughs in the upper troposphere: Examples from the Southern Hemisphere summer season of 1984/85. Mon. Wea. Rev., 128,

26 Chang, E. K. M. and I. Orlanski, 1993: On the dynamics of a storm track. J. Atmos. Sci., 50, Chang, E. K. M. and D. B. Yu, 1999: Characteristics of wave packets in the upper tropo- sphere. Part I: Northern Hemisphere winter. J. Atmos. Sci., 56, Esler, J. G. and P. H. Haynes, 1999: Mechanisms for wave packet formation and maintenance in a quasigeostrophic two-layer model. J. Atmos. Sci., 56, Ferreira, R. N. and W. H. Schubert, 1999: The role of tropical cyclones in the formation of tropical upper-tropospheric troughs. J. Atmos. Sci., 56, Grams, C. M., S. C. Jones, C. A. Davis, P. A. Harr, and M. Weissmann, 2013a: The impact of Typhoon Jangmi (2008) on the midlatitude flow. Part I: Upper-level ridgebuilding and modification of the jet. Quart. J. Roy. Met. Soc., doi: /qj Grams, C. M., S. C. Jones, C. A. Davis, P. A. Harr, and M. Weissmann, 2013b: The impact of Typhoon Jangmi (2008) on the midlatitude flow. Part II: Downstream evolution. Quart. J. Roy. Met. Soc., doi: /qj Hakim, G. J., 2003: Developing wave packets in the North Pacific storm track. Mon. Wea. Rev., 131, Hakim, G. J., 2005: Vertical structure of midlatitude analysis and forecast errors. Mon. Wea. Rev., 133, Harr, P. A., D. Anwender, and S. C. Jones, 2008: Predictability associated with the down- stream impacts of the extratropical transition of tropical cyclones: Methodology and a case study of Typhoon Nabi (2005). Mon. Wea. Rev., 128, Harr, P. A. and J. M. Dea, 2009: Downstream development associated with the extratropical transition of tropical cyclones over the Western North Pacific. Mon. Wea. Rev., 137,

27 Harr, P. A. and R. L. Elsberry, 2000: Extratropical transition of tropical cyclones over the western north Pacific. Part I: Evolution of structural characteristics during the transition process. Mon. Wea. Rev., 128, Harr, P. A., R. L. Elsberry, and T. F. Hogan, 2000: Extratropical transition of tropical cyclones over the western north Pacific. Part II: The impact of midlatitude circulation characteristics. Mon. Wea. Rev., 128, Hart, R. E., 2003: A cyclone phase space derived from thermal wind and thermal asymmetry. Mon. Wea. Rev., 131, Hart, R. E. and J. L. Evans, 2001: A climatology of the extratropical transition of Atlantic tropical cyclones. J. Climate, 14, Holton, J. R. and G. J. Hakim, 2013: An Introduction to Dynamic Meteorology. Elsevier Academic Press, 532 pp Hoskins, B. J., M. E. McIntyre, and A. W. Robertson, 1985: On the use and significance of isentropic potential vorticity maps. Quart. J. Roy. Met. Soc., 111, Jones, S. C., et al., 2003: The extratropical transition of tropical cyclones: Forecast chal- lenges, current understanding, and future directions. Wea. Forecasting, 18, Keller, J. H., S. J. Jones, and P. A. Harr, 2014: An eddy kinetic energy view of physical and dynamical processes in distinct forecast scenarios for the extratropical transition of two tropical cyclones. Mon. Wea. Rev., 142, In Press Klein, P. M., P. A. Harr, and R. L. Elsberry, 2000: Extratropical transition of western North Pacific tropical cyclones: An overview and conceptual model of the transformation stage. Wea. Forecasting, 15, Klein, P. M., P. A. Harr, and R. L. Elsberry, 2002: Extratropical transition of Western North 26

28 Pacific tropical cyclones: Midlatitude and tropical cyclone contributions to reintensifica- tion. Mon. Wea. Rev., 130, Langland, R. H., M. A. Shapiro, and R. Gelaro, 2002: Initial condition sensitivity and error growth in forecasts of the 25 January 2000 east coast snowstorm. Mon. Wea. Rev., 130, Liu, Q., T. Marchok, H.-L. Pan, M. Bender, and S. J. Lord, 2000: Improvements in hurricane initialization and forecasting at NCEP with global and regional (GFDL) models. Tech. rep., NOAA Tech. Procedures Bull. 472, 7 pp., Camp Springs, MD Orlanski, I. and J. J. Katzfey, 1991: The life cycle of a cyclone wave in the southern hemi- sphere. Part I: Eddy energy budget. J. Atmos. Sci., 48, Orlanski, I. and J. P. Sheldon, 1995: Stages in the energetics of baroclinic systems. Tellus, 47, Riemer, M. and S. C. Jones, 2010: The downstream impact of tropical cyclones on a de- veloping baroclinic wave in idealized scenerios of extratropical transition. Quart. J. Roy. Met. Soc., 136, Riemer, M., S. C. Jones, and C. A. Davis, 2008: The impact of extratropical transition on the downstream flow: An idealized modelling study with a straight jet. Quart. J. Roy. Met. Soc., 134, Saha, S., et al., 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, Sanders, F. and J. R. Gyakum, 1980: Synoptic-dynamic climatology of the Bomb. Mon. Wea. Rev., 108, Simmons, A. J. and B. J. Hoskins, 1979: Downstream and upstream development of unstable baroclinic waves. J. Atmos. Sci., 36,

29 Swanson, K. and R. T. Pierrehumbert, 1994: Nonlinear wave packet evolution on a baro- clinically unstable jet. J. Atmos. Sci., 51, Szunyogh, I., Z. Toth, R. E. Morss, S. J. Majumdar, B. J. Etherton, and B. C. H., 2000: The effect of targeted dropsonde observations during the 1999 winter storm reconnaissance program. Mon. Wea. Rev., 128, Torn, R. D., 2010: Diagnosing the factors that influence downstream ridging during extrat- ropical transition. J. Atmos. Sci., 67, Torn, R. D. and G. J. Hakim, 2009: Initial condition sensitivity of western-pacific extrat- ropical transitions determined using ensemble-based sensitivity analysis. Mon. Wea. Rev., 137,

30 List of Figures 1 Number of winter midlatitude cyclones (black), ASO midlatitude cyclones (white) and extratropical transition cases (gray) in the (a) WNP and (b) Atlantic basin used in this study as a function of month Ensemble-mean time evolution of the Western North Pacific extratropical transition (left column; 250 hpa) and winter cyclone (right column; 300 hpa) wave packets. The shading denotes the meridional wind anomaly (m s 1 ), where stippled regions are statistically significant at the 95% confidence level. Thin lines denote the 330 K potential vorticity every 1 PVU [1 PVU = 10 6 m 2 K (kg s) 1 ]. The underlying map in the left column is oriented such that the composite center longitude at t = 0 matches the winter cyclone box longitude Meridional wind anomalies averaged between N for WNP (a) ET, (c) winter cyclones, and (e) ASO cyclones as a function of longitude and time (shading, m s 1 ), where stippled regions denote where the value is statistically significant at the 95% level. For ET and ASO cyclones, the longitude is shifted to the winter cyclone location at t = 0. (b), (d), and (f) as in (a), (c), and (e), but for the Atlantic basin As in Fig. 2e, but for ET cases with (left column) and without (right column) an preceding upper-level trough As in Fig. 2a,c,e,g,i,k, but for midlatitude cyclones that occur in the WNP basin between N during August, September and October As in Fig. 2a,c,e,g,i,k, but for ET cyclones that occur in the WNP during November and December

31 WNP (a) ET and (b) winter cyclone wave-packet analysis at t=0. Solid lines denote the anomaly meridional wind (m s 1 ) as a function of longitude, while the dashed lines are a linear fit to an exponential profile from the packet peak to 2.5 e-folding distances from the peak Amplitude (m s 1 ) of the ET (solid gray), winter cyclone (solid black), and ASO cyclone (dashed) packet peak (m s 1 ) as a function of lag (h) for the (a) WNP and (b) Atlantic Basins. The error bars denote the 95% confidence bounds determined through bootstrap resampling Group velocity (m s 1 ) of the wave-packet peak as a function of lag (h) for ET (solid gray), winter cyclone (solid black), and ASO cyclone (dashed) in the (a) WNP and (b) Atlantic basins. The error bars denote the 95% confidence bounds determined through bootstrap resampling Wavelength (km) of the ET (solid gray), winter cyclone (solid black), and ASO cyclone (dashed) wave packet as a function of lag (h) in the (a) WNP and (b) Atlantic basins. The error bars denote the 95% confidence bounds determined through bootstrap resampling Ensemble-mean differential temperature advection integrated between hpa for (a) ET, (b) winter, and (c) ASO cyclones in the WNP (shading, J s 1 m 4). The heavy lines denote the ensemble-mean 900 hpa geopotential height deviations every 20 m, with negative values dashed. The underlying map in the left column is oriented such that the composite center longitude at t = 0 matches the winter cyclone longitude at t = 0. (b), (d), and (f) as in (a), (c) and (e), but for the Atlantic As in Fig. 11, but where the shading is the column-integrated horizontal moisture flux convergence (units: W m 2 ) As in Fig. 2, but for the Atlantic Basin

32 As in Fig. 4, but for ET cases with (top row) and without (bottom row) an upstream upper-level trough As in Fig. 2a,c,e,g,i,k, but for midlatitude cyclones that occur in the Atlantic basin between N during August, September and October

33 Fig. 1. Number of winter midlatitude cyclones (black), ASO midlatitude cyclones (white) and extratropical transition cases (gray) in the (a) WNP and (b) Atlantic basin used in this study as a function of month. 32

34 Fig. 2. Ensemble-mean time evolution of the Western North Pacific extratropical transition (left column; 250 hpa) and winter cyclone (right column; 300 hpa) wave packets. The shading denotes the meridional wind anomaly (m s 1 ), where stippled regions are statistically significant at the 95% confidence level. Thin lines denote the 330 K potential vorticity every 1 PVU [1 PVU = 10 6 m2 K (kg s) 1 ]. The underlying map in the left column is oriented such that the composite center longitude at t = 0 matches the winter cyclone box longitude. 33

35 Fig. 3. Meridional wind anomalies averaged between N for WNP (a) ET, (c) winter cyclones, and (e) ASO cyclones as a function of longitude and time (shading, m s 1 ), where stippled regions denote where the value is statistically significant at the 95% level. For ET and ASO cyclones, the longitude is shifted to the winter cyclone location at t = 0. (b), (d), and (f) as in (a), (c), and (e), but for the Atlantic basin. 34

36 Fig. 4. As in Fig. 2e, but for ET cases with (left column) and without (right column) an preceding upper-level trough. 35

37 Fig. 5. As in Fig. 2a,c,e,g,i,k, but for midlatitude cyclones that occur in the WNP basin between N during August, September and October. 36

38 Fig. 6. As in Fig. 2a,c,e,g,i,k, but for ET cyclones that occur in the WNP during November and December. 37

39 Fig. 7. WNP (a) ET and (b) winter cyclone wave-packet analysis at t=0. Solid lines denote the anomaly meridional wind (m s 1 ) as a function of longitude, while the dashed lines are a linear fit to an exponential profile from the packet peak to 2.5 e-folding distances from the peak. 38

40 Fig. 8. Amplitude (m s 1 ) of the ET (solid gray), winter cyclone (solid black), and ASO cyclone (dashed) packet peak (m s 1 ) as a function of lag (h) for the (a) WNP and (b) Atlantic Basins. The error bars denote the 95% confidence bounds determined through bootstrap resampling. 39

41 Fig. 9. Group velocity (m s 1 ) of the wave-packet peak as a function of lag (h) for ET (solid gray), winter cyclone (solid black), and ASO cyclone (dashed) in the (a) WNP and (b) Atlantic basins. The error bars denote the 95% confidence bounds determined through bootstrap resampling. 40

42 Fig. 10. Wavelength (km) of the ET (solid gray), winter cyclone (solid black), and ASO cyclone (dashed) wave packet as a function of lag (h) in the (a) WNP and (b) Atlantic basins. The error bars denote the 95% confidence bounds determined through bootstrap resampling. 41

43 Fig. 11. Ensemble-mean differential temperature advection integrated between hpa for (a) ET, (b) winter, and (c) ASO cyclones in the WNP (shading, J s 1 m 4). The heavy lines denote the ensemble-mean 900 hpa geopotential height deviations every 20 m, with negative values dashed. The underlying map in the left column is oriented such that the composite center longitude at t = 0 matches the winter cyclone longitude at t = 0. (b), (d), and (f) as in (a), (c) and (e), but for the42atlantic.

44 Fig. 12. As in Fig. 11, but where the shading is the column-integrated horizontal moisture flux convergence (units: W m 2 ). 43

45 Fig. 13. As in Fig. 2, but for the Atlantic Basin. 44

46 Fig. 14. As in Fig. 4, but for ET cases with (top row) and without (bottom row) an upstream upper-level trough. 45

47 Fig. 15. As in Fig. 2a,c,e,g,i,k, but for midlatitude cyclones that occur in the Atlantic basin between N during August, September and October. 46