Corresponding author: A. Chandran, Geophysical Institute, University of Alaska Fairbanks, Fairbanks, AK, USA. (firstname.lastname@example.org)
 Elevated stratopause (ES) events occurring during Northern Hemisphere winter are identified in four climate simulations of the period 1953–2005 made with the Whole Atmosphere Community Climate Model (WACCM). We find 68 ES events in 212 winters. These events are found in winters when the middle atmosphere is disturbed and there are zonal wind reversals in the stratosphere at high latitudes. These disturbances can be associated with both major and minor stratospheric sudden warming events (SSWs). The ES events occur under conditions where the stratospheric jet, the gravity wave forcing, and the residual circulation remain reversed longer than in those winters where an SSW occurs without an ES. We compare ES events with the type of SSW (vortex splitting and vortex displacement) and find that 68% of ES events form after vortex splitting events. We also present a climatology of ES events based on NASA's Modern-Era Retrospective Analysis for Research and Applications reanalysis data from 1979 to 2012 and compare it to the model results. WACCM composites of major SSW and ES also show enhanced Eliassen-Palm flux divergences in the upper mesosphere after the stratospheric warming, immediately before the formation of an ES. However, the formation of an ES in WACCM is due primarily to adiabatic heating from gravity wave-driven downwelling, which follows the reestablishment of the eastward jet in the upper stratosphere. We find nine winters where an ES forms in the absence of any significant planetary wave activity in the upper mesosphere and illustrate one such event.
 Elevated stratopause (ES) events are episodes during which the wintertime stratopause over the polar cap first descends, then becomes indistinct, and finally reforms at a much higher altitude than normal [e.g., Manney et al., 2008a]. ES events often occur in connection with a stratospheric sudden warming (SSW), a dynamical phenomenon that causes widespread disruption of the temperature and winds of the middle atmosphere and has been shown to play a central role in coupling the different layers of the atmosphere from the troposphere to the ionosphere. The World Meteorological Organization (WMO) defines a major SSW as one that leads to a reversal of the zonal mean wind at 60° latitude and 10 hPa (~ 30 km), and a positive poleward temperature gradient from 60° latitude to the pole, at or below 10 hPa. During a minor warming, there is a positive poleward temperature gradient from 60° latitude to the pole at 10 hPa but the zonal mean wind does not reverse at that level.
 Over the last decade the wintertime Arctic stratosphere has shown significant interannual variability. There have been undisturbed winters when the circulation was characterized by a stable cold stratospheric vortex that yielded significant springtime ozone loss [e.g., Rex et al., 2006; Manney et al., 2011]. There have been disturbed winters when strong planetary wave activity triggered an SSW, with a reversal of the wintertime circulation and little springtime ozone loss [e.g., Manney et al., 2005, 2008a, 2009; Siskind et al., 2010]. This recent variability has attracted interest from researchers attempting to understand trends in the general circulation [e.g., Butchart et al., 2006; Garcia and Randel, 2008; Deckert and Dameris, 2008], the dynamics of the stratospheric vortex [e.g., Charlton and Polvani, 2007; Charlton et al., 2007; Charlton-Perez et al., 2008], Arctic ozone depletion [e.g., Tripathi et al., 2007], the role of wave-forcing in the circulation [e.g., Duck et al., 1998; Gerrard et al., 2002; Hauchecorne et al., 2007; Siskind et al., 2010; Wang and Alexander, 2009; Thurairajah et al., 2010a, 2010b; Chandran et al., 2011; de la Torre et al., 2012], and coupling between the lower atmosphere and the ionosphere [Chau et al., 2010, Goncharenko et al., 2010].
 The polar winter stratopause is a gravity wave-driven feature of the wintertime temperature distribution [Hitchman et al., 1989] and is often warmer and located at a higher altitude than the adjacent midlatitude stratopause. Westward propagating gravity waves travel upward into the mesosphere, break between 60 and 80 km, and drive a poleward and downward circulation below the breaking level. Under normal or undisturbed wintertime conditions the maximum downwelling occurs over the polar cap region and the polar stratopause forms in the corresponding region of adiabatic heating [e.g., de la Torre et al., 2012]. In undisturbed winters the stratopause height increases with latitude from about 50 to 60 km poleward of 60°N. During SSW events, breaking planetary waves displace or split the polar vortex and reverse the eastward zonal-mean jet in the stratosphere. This leads to downwelling and adiabatic warming in the stratosphere, and lowering of the stratopause. At the same time, dissipation of eastward-propagating gravity waves reverses the residual circulation in the mesosphere and drives upwelling and adiabatic cooling in the mesosphere.
 Stratospheric sudden warming events can be followed by the occurrence of a near isothermal upper stratosphere, reformation of the polar cap stratopause near 75–80 km, and subsequent descent of the same over a period of weeks [Manney et al., 2009; Orsolini et al., 2010; Siskind et al., 2010; Thurairajah et al., 2010a; Chandran et al., 2011; Marsh, 2011; Ren et al., 2011, de la Torre et al., 2012]. Chandran et al.  presented a case study of an ES event simulated by the free running version of the Whole Atmosphere Community Climate Model (WACCM), where the meteorology closely resembled recent Arctic disturbed winters, and found that the formation of the ES is due primarily to gravity wave-driven adiabatic heating in the downwelling branch of the circulation, following the reestablishment of the eastward jet in the upper stratosphere due to radiative relaxation. Limpasuvan et al.  also analyzed the dynamics of the stratosphere and mesosphere during a similar ES event simulated by WACCM. They noted the presence of enhanced Eliassen-Palm flux divergence in the mesosphere and lower thermosphere, and attributed the formation of the ES to adiabatic downwelling driven by secondarily generated planetary waves. De la Torre et al.  showed that WACCM can reproduce a realistic climatology of SSW and ES events by presenting a composite of ES events in one 53 year WACCM simulation. In this paper, we extend the studies of Chandran et al.  and de la Torre et al.  by examining in detail the climatology of ES events in an ensemble of four WACCM simulations of the 53 year period 1954–2006. In section 2 we describe WACCM, highlighting the key features pertinent to this study. In section 3 we present model simulations of ES events and the dynamical conditions associated with them in terms of both individual years and composites over a set of 212 Arctic winters. We also present a climatology of ES events based on NASA's Modern-Era Retrospective Analysis for Research and Applications (MERRA) reanalysis data from 1979 to 2012 and compare them with the WACCM climatology. In section 4 we present our discussion and conclusions.
2 Model Description and Definition of SSW and ES Events
 The WACCM is derived from the National Center for Atmospheric Research's Community Atmosphere Model and is a fully coupled chemistry-climate model [see Garcia et al., 2007, and references therein]. The model domain extends from the Earth's surface to about 140 km (4.5 × 10–6 hPa). It has 66 vertical levels with a variable vertical resolution of 1.1–1.4 km in the lower stratosphere, ~ 1.75 km around the stratopause and ~ 3.5 km above 65 km. The standard horizontal resolution is 1.9° × 2.5° (latitude × longitude). While WACCM resolves planetary waves, both orographic and nonorographic small-scale gravity waves are parameterized in the model [Garcia et al., 2007; Richter et al., 2008, 2010].
 Previous WACCM simulations of Northern Hemisphere winters tended to generate an anomalously cold polar middle atmosphere in which the occurrence of SSW was less frequent than observed [Charlton et al., 2007]. The inclusion of a turbulent mountain stress parameterization, which accounts for the effect of unresolved topography, has resulted in a much more realistic representation of Northern Hemisphere major SSW in multiyear simulations [Richter et al., 2010; de la Torre et al., 2012]. The nonorographic gravity wave parameterization is based on the scheme of Lindzen . Gravity waves are assumed to be generated by convection and fronts [Richter et al., 2010]. Based on the source parameterization of Beres et al. , the convectively generated gravity wave source spectrum is activated at the top of deep convection as defined by the parameterization of Zhang and McFarlane . As discussed in Richter et al. , the frontal source is triggered in regions of strong wind deformation and temperature gradient identified by the frontogenesis function of Hoskins  applied at 600 hPa.
 The WACCM simulations used in this study were originally conducted for the second Chemistry-Climate Model Validation Activity of the Stratospheric Process and their Role in Climate [Gillett et al., 2011]. These four simulations (referred to as refb1.1 through refb1.4) cover the period 1953–2006; they include both solar and anthropogenic forcing, observed trends in greenhouse gases, and observed sea surface temperatures. In this study, individual Arctic winters from the different model simulations are referred to as “model year” (MY) to differentiate from observed Arctic winters. The simulations use a time step of 1800 s, with daily model output of dynamical variables at UT 00. We determine the Eliassen-Palm (EP) flux divergence due to planetary-scale Rossby waves that are explicitly resolved by the model from this daily output. The daily output also includes the zonal and meridional components of the residual circulation. However, the output for these runs did not include daily values of gravity wave forcing, which is essential to understanding the evolution of the Arctic stratopause during and after an SSW. Therefore, we infer gravity wave forcing as a residual in the transformed Eulerian mean (TEM) zonal momentum budget [Andrews et al., 1987]. The calculation of gravity wave forcing was validated in Chandran et al. .
 We identify ES events in the four WACCM simulations following the procedure used by de la Torre et al. . We first smooth the daily zonal mean polar cap (75°N–90°N) temperature time series with a 9 day running mean. This eliminates changes in stratopause altitude caused by transient waves that do not result in a persistent modification of the polar cap temperature structure. We define the stratopause as the altitude of maximum temperature between 20 and 100 km. We then identify ES events during the Northern Hemisphere winter (November through March) as abrupt increases, or “jumps”, in the stratopause altitude greater than 15 km. While the detection of ES events does not change appreciably by changing the smoothing window from 9 days to 7 or 5 days, the lowering of the stratopause jump altitude from 15 to 12 km results in an increase in ES event detection by ~ 3 %. We define the central date (day 0) of the ES event as the date of the stratopause jump.
 As in the study of de la Torre et al. , major SSWs are identified using the WMO definition of reversal of the zonal-mean zonal wind at 10 hPa, 60 N and a positive poleward temperature gradient, with T(90°) – T(60°) > 20 K. Day 0 is defined as the day of wind reversal at 10 hPa at 60°N and no other SSW can begin within 20 days of a central date. As discussed in detail below, we find that ES events are essentially a polar cap phenomenon and are not always associated with wind reversals at 10 hPa and 60 N (although there is always a wind reversal over some range of altitude at subpolar latitudes).
3.1 Characteristics of Different Types of Arctic Winters in WACCM and Their Relationship to ES Events
 Arctic winters in WACCM exhibit different levels of disturbance, which can be termed dynamically quiet years, minor SSW, and major SSW events. In this study, we have classified ES events as a fourth type of event distinct from (but not necessarily exclusive of) minor and major SSW. We illustrate these levels of disturbance by plotting the polar cap temperature (cosine-weighted average over 75°N–90°N) and zonal mean zonal wind at 60°N as functions of altitude and day number for four winters in Figure 1. We have identified these four winters from a single model decade (the 1990s) in one 53 year simulation (refb1.1). Observations from the Arctic have shown similar winter variability in the decade of 2000–2009. In the four WACCM simulations, which comprise a total of 212 years, the occurrence of all four classes of disturbance within a single decade is seen on only four occasions.
 In Figure 1, day 0 represents the first of January. Contour lines of the TEM vertical velocity (w* in cm s–1) are overplotted on the color-filled temperature contours. Solid lines indicate ascent and dashed lines descent. The Arctic winter of MY 1996/1997 (Figure 1a) is an example of a dynamically quiet or undisturbed winter. During this undisturbed winter, the stratopause altitude remains between 50 and 60 km, the stratospheric jet is eastward throughout the winter, the mesospheric jet is westward, and the zero wind line remains near 80 km. A dynamically quiet Arctic winter without a single minor SSW event is an extremely rare event in WACCM simulations with an annual occurrence frequency of 6%. The Arctic winter of MY 1994/1995 (Figure 1b) is an example of a winter with a minor SSW event, which occurs in mid-December (day –14). During this warming the latitudinal temperature gradient becomes positive poleward of 60°N (not shown) and there is warming and descent of the stratopause coinciding with cooling in the mesosphere. Because this is a minor warming, the associated wind reversal in the stratosphere does not extend down to 10 hPa (~ 30 km altitude).
 The Arctic winter of MY 1992/1993 (Figure 1c) is an example of a winter with a major SSW, which occurs in late January (day 26). At this time, the latitudinal temperature gradient becomes positive poleward of 60°N (not shown). As in the minor SSW in 1994/1995, in this major SSW there is upwelling in the mesosphere, which produces mesospheric cooling, and strong descent in the stratosphere, which results in warming in the stratosphere. The warming and descent of the stratopause during the major SSW and the cooling in the mesosphere are more pronounced than during the minor warming, and the associated wind reversal in the stratosphere extends below 10 hPa (~30 km). During the reversal of the stratospheric winds the mesospheric jet also reverses from westward to eastward. However, there is no abrupt jump in the altitude of the stratopause associated with this major SSW. The Arctic winter of MY 1997/1998 (Figure 1d) is an example of a winter with a major SSW event accompanied by an ES event. This winter has a minor SSW in mid-December (day –14) followed by a major SSW in early January (day 8). During the major SSW the zonal winds remain reversed below 10 hPa for over three weeks, and it is during this time that the stratopause reforms at very high altitude and descends slowly thereafter. In particular, the altitude of the stratopause rises more than 45 km in late January (day 18), from below 40 km to about 85 km. There is a two-week period when the stratopause is poorly defined (there are two temperature maxima between 20 and 100 km), which corresponds to the period when the stratospheric wind reverses below 10 hPa.
 We show the behavior of the polar stratopause during a major SSW (in MY 1992/1993, Figure 1c) and during an ES (in MY 1997/1998, Figure 1d) winters by plotting the stratopause height on a polar stereographic map. We plot the stratopause height 12 days before and after, and during the peak warming (identified by arrows in Figures 1c and 1d). For MY 1992/1993, we plot the stratopause height on days 14, 26, and 38 (Figure 2a). On day 14, 12 days before the peak of the major SSW, an anticyclone is developing over Alaska and the vortex is displaced from the pole (not shown) to the longitude sector that encompasses western Europe and Greenland. The stratopause is found near 60 km, the usual height during undisturbed conditions. At the peak of the major SSW (on day 26) the stratopause height decreases across the polar cap due to the adiabatic warming associated with the circulation driven by breaking planetary waves. The highest stratopause altitude is at 56 km, similar to the typical altitude in an undisturbed year, but it is found off the pole, corresponding to the displacement of the polar vortex. The stratopause is at lowest altitude of about 42 km inside the anticyclone over Russia and the eastern Arctic Ocean. This configuration of stratopause height anomalies in the vortex and the anticyclone are consistent with the findings of France et al. . Twelve days after the warming the location and extent of the stratopause is similar to what is found under undisturbed conditions, but the maximum altitude is about 10 km higher (~ 65 km). This occurs because the downward branch of the residual circulation is located at a higher altitude after the SSW than before it.
 For MY 1997/1998 we plot the stratopause height on days –4, 8, and 20 (Figure 2b). These correspond to 12 days before the major SSW (day –4), the peak of the major SSW (day 8), and 12 days after the major SSW (day 20). On day –4 the stratopause altitude is already higher (about 70 km) from Greenland to Scandinavia than under normal conditions due to the minor SSW that took place on day –15 (see Figure 1d). The vortex has again been displaced off the pole due to the development of an anticyclone over Alaska and the Canadian Arctic (where the stratopause is at ~36 km). At the peak of the major SSW on day 8, the stratopause altitude is below 46 km across the entire polar cap, and below 40 km over northern Russia. Twelve days after the major SSW the stratopause is elevated uniformly across the polar cap to about 75–80 km. The ES altitude corresponds to the maximum downwelling now occurring in the lower thermosphere (cf. Figure 1d). During this period there are sharp meridional gradients in the stratopause height, which decreases from ~75 km to ~40 km over a latitude range of about 5°.
 In Figure 3, we use the methodology developed by Harvey et al.  to illustrate the three-dimensional structure and evolution of the polar vortex (colored by temperature) and the anticyclones (in black), from the tropopause to the mesopause, during the ES event in MY 1997/1998. We include a modification to the vortex identification scheme to accommodate the mesospheric subtropical jet [Dunkerton and Delisi, 1985] such that, if there are two jets at different latitudes, we choose the poleward-most jet as the edge of the vortex. This figure shows the synoptic response of the atmosphere during both SSW events and the ES event that occurred in MY 1997/98. As shown in Figure 1d, this model year has a minor SSW on day –14 followed by a major SSW on day 8 and the formation of an ES by day 18. We show the vortex on six selected days to highlight its evolution. We plot the vortex before the minor SSW, during the minor SSW, between the two SSWs, during the major SSW, at the time of the formation of the ES, and finally during the descent of the ES. We use a perspective view centered on East Asia (~ 120°E).
 On day –26, prior to the minor SSW, the vortex is undisturbed from the lower stratosphere up to ~75 km. The cyclonic flow in the stratosphere is representative of the eastward polar jet in the stratosphere, while the anticyclonic flow in the mesosphere corresponds to the gravity wave-driven westward jet in the mesosphere. The altitude of reversal from cyclonic to anticyclonic flow is consistent with the altitude of the zero zonal mean zonal wind at ~80 km on day –26 in Figure 1d. This structure of the stratospheric cyclone, centered over the pole, and the mesospheric anticyclone is representative of undisturbed conditions. The stratopause is at a height of ~60 km on this day. On day –16, during the minor SSW, the vortex is displaced from the pole by a large anticyclone that spans the depth of the stratosphere. The cyclone-anticyclone pair tilts westward with altitude, indicating upward planetary wave propagation. The stratopause has descended from about 60 to 40 km in the vortex. The anticyclone is associated with westward flow in the stratosphere, such that the gravity wave-driven forcing above the anticyclone in the mesosphere is eastward and, hence, a cold cyclone (associated with adiabatic cooling due to gravity wave-driven upwelling) is situated above the anticyclone. On day –4, the vortex has returned to undisturbed conditions, though it is weaker than on day –26. The stratopause has reformed near 60 km and an anticyclone is present over the pole in the mesosphere.
 On day 9, we show the vortex structure during the major SSW. Unlike the minor SSW on day –16, the stratospheric polar vortex is split rather than displaced. There is an anticyclone over the pole as well as (at some altitudes) a second anticyclone over Asia. The stratopause has descended to ~40 km and the cold mesospheric cyclone, centered near 80 km, is due to the gravity wave-driven wind reversal in the mesosphere and associated adiabatic upwelling (cf. Figure 1d). On day 20, following the formation of the ES near day 18, the vortex remains split below ~40 km but is beginning to reform over the pole from 45 to 80 km. On this day, the stratopause is at a height of ~70 km. On day 28, the polar vortex has reformed over the pole between ~40 and 70 km and the stratopause is at a height of ~65 km. The anticyclone has also reformed in the mesosphere. The vortex however remains split below 40 km and corresponds to the zonal mean wind reversal between 40 and 20 km, which lasts until day 30, as can be seen in Figure 1d.
3.2 Climatology of Elevated Stratopause Events
De La Torre et al.  presented a classification of SSW in WACCM similar to that of Charlton and Polvani  from the same four realizations (refb1.1 through refb1.4) of the period 1954–2005 examined here. They found the annual occurrence frequency of major SSW to be 57%, which is comparable to the frequencies calculated by Charlton and Polvani  for the ERA40 (64%) and NCEP/NCAR (60%) reanalyses. In the previous section, we showed that ES occur concurrently with SSW in some instances but not in others. Table 1 shows ES events identified by abrupt upward shifts in the altitude of the stratopause, as explained in section 2. According to this criterion, ES events occur in 32% of the winters and their annual frequency of occurrence does not vary appreciably across the four 53 year simulations (28% to 36%). Thus, while ES events simulated with WACCM are not isolated or rare phenomena, many SSW are not accompanied by an ES event. ES events do not always coincide with major SSW events because, by definition, a major SSW occurs when the zonal mean winds reverse at 60°N at 10 hPa (~ 30 km); such a reversal is not always accompanied by a large disturbance of the circulation and temperature of the mesospheric polar cap. On the other hand, while most ES events are accompanied by a reversal of the zonal mean winds at subpolar latitudes, in 7 of 68 ES events (about ~ 10% of the cases) there is no major SSW because the wind does not reverse at 60°N and 10 hPa.
Table 1. The Occurrence of ES Events (Annual Frequency of Occurrence, in Percent, Given in Parenthesis) in Multiple Realizations of WACCM, As Well As the Occurrences Classified as Vortex Displacement and Splitting Events
Elevated Stratopause (ES) Events
Refb 1.1 (1953–2006)
Refb 1.2 (1953–2006)
Refb 1.3 (1953–2006)
Refb 1.4 (1953–2006)
All WACCM 53 Year Runs
Number of ES winters
6 (11 %)
13 (24 %)
14 (26 %)
46 (22 %)
 Following Charlton and Polvani , SSW events can be classified into vortex splitting and vortex displacement events. De la Torre et al.  discussed the occurrence frequencies of vortex displacement and vortex splitting events during major SSW in WACCM simulations refb1.1 through refb1.4 and found an overall ratio of vortex displacements to vortex splitting events of 1.39, with a range of 1.1 to 2.2 among the simulations. This is similar, although somewhat higher overall, than the observed ratio (1.25 in NCEP data and 1.1 in ERA-40 data). In spite of the preponderance of displacement SSW in WACCM, Table 1 shows that about two of every three ES events are associated with vortex splitting events. Note that in Table 1 we have indicated whether ES events are associated with vortex splits or displacements regardless of whether the latter meet the criterion for major SSW. That is, some of the ES events (10%, as noted above) occurred in conjunction with vortex displacements that did not modify the zonal-mean zonal wind sufficiently to qualify as major SSW.
 In Figure 4, we plot the occurrence of WACCM ES events by decade for five successive decades (i.e., winter 1953/1954 through 1962/1963 … winter 1993/1994 through 2002/2003). There is considerable interdecadal variability in all simulations, ranging from 0 to 6 ES events per decade. There is no apparent trend in the occurrence frequency over the five decades of the simulation, although some decades have appreciably fewer ES events than others (e.g., in the 1980s compared to the 1990s). One implication of Figure 4 is that the relatively large number of major SSW and ES events observed during the decade of 2000–2009 and documented by Manney et al., 2008b is well within the statistical variability of the atmosphere as represented by WACCM.
3.3 Comparison of ES Events and Major SSW Events: A Composite Climatology
 To better understand the differences in dynamics between winters with ES events and winters with major SSW but no ES events, we compare ES composites in Figure 5 (68 ES winters) with composites of major SSW winters in Figure 6 (54 major SSW winters without ES) for all such events in all four WACCM simulations. We show composites of the evolution of (a) zonal mean temperatures over the polar cap, (b) midlatitude zonal-mean zonal wind, (c) EP flux divergence, (d) gravity wave forcing, and (e) TEM vertical velocity, and (f) TEM meridional velocity for years with ES events and years with major SSW events but no ES events. For the composite of the ES events we choose day 0 as the day of the jump in the stratopause altitude and plot the result in Figure 5. For the composite of the SSW events we choose day 0 as 10 days after the day where there is a wind reversal at 10 hPa and plot the result in Figure 6. This has been done to make the figures comparable, because stratopause jumps usually occur about 10 days after the time of maximum zonal wind deceleration.
 From the composite plots in Figure 5, it can be seen that the dynamics of the middle atmosphere during ES events are similar to what is shown in the earlier case study by Chandran et al.  and in the composite of ES events detected in a single 53 year WACCM run by de la Torre et al. [2012, Figure 10]. The ES events are preceded by a reversal of the wind over a broad range of altitude in the stratosphere (Figure 5b), which is triggered by an increase in EP flux divergence (Figure 5c). Because most (~ 90%) of the ES events are associated with major SSW, the composite of the zonal mean wind shows a reversal at 10 hPa (about 30 km). The reversal of the zonal wind changes the filtering of upward-propagating gravity waves, such that the gravity wave-induced acceleration in the mesosphere changes from westward to eastward (Figure 5d). The warming of the stratopause is caused by the downwelling branch of the circulation driven by the enhanced EP flux divergence, while the mesospheric cooling is caused by the upwelling branch thereof. As the composite ES event progresses, strong eastward gravity wave forcing develops above 60 km, which reinforces the upwelling and cooling in the mesosphere. Once the eastward jet is reestablished in the upper stratosphere by radiative relaxation, around day 0, the gravity wave forcing in the mesosphere and lower thermosphere (MLT) becomes westward once again and the stratopause warms adiabatically and descends following the wave-driven downwelling.
 In the major SSW composites shown in Figure 6, it can be seen that the changes in the temperature and dynamics of the region are similar to the changes that occur during ES events. However, comparing the major SSW composites with the ES composites from Figures 5, we see that, for the ES composite, the duration of westward EP flux divergence, which causes the warming, the reversal of the stratospheric jet and associated eastward gravity wave forcing in the mesosphere, and the reversal of the residual circulation are all of longer duration than for the major SSW composite. The magnitudes of EP flux divergence and jet reversal in the stratosphere are stronger but shorter-lived in the SSW composite than in the ES composite. The reversal of the zonal mean winds in the stratosphere (at 35 km altitude) is longer by ~ 7 days in the ES winters (12 days vs. 5 days in SSW winters with no ES). This longer-lasting reversal of the zonal winds in turn results in a longer-lasting reversal in gravity wave forcing and the residual circulation.
 As noted previously by de la Torre et al. , the ratio of vortex displacement events to vortex splitting events in major SSW in WACCM is 1.4. Despite the greater frequency of vortex displacement events, ~ 68% of ES events in the model are associated with vortex splitting events (46 of 68 ES events; see Table 1). Therefore, the composites in Figure 5 reflect predominantly the behavior of vortex splitting events, which are also longer lasting, while those in Figure 6 are more influenced by vortex displacement events. The period of enhanced (westward) EP flux divergence in the stratosphere, reversal of the stratospheric jets, and (eastward) gravity wave forcing in the mesosphere is about 7 days longer during vortex splitting events than during vortex displacement events. Generally, the vortex splitting events cause a more thorough overturning of the quasi-geostrophic potential vorticity (PV) gradient in the stratosphere than vortex displacement events. The PV gradient becomes negative poleward of 60°N for both displacement and splitting events (not shown). However, the PV gradient remains reversed for a long period after vortex splitting events, whereas it returns to normal conditions much more rapidly following a displacement event. Our analysis also indicates that vortex displacement events that precede ES events are longer lasting, by 4 days, than vortex displacement events that do not result in an ES.
3.4 Climatology of Elevated Stratopause Events in MERRA Data
 In order to compare ES events simulated with WACCM with observations, we have applied our ES detection algorithm to MERRA reanalysis data for the period 1979–2011 (32 winters). MERRA was developed by NASA's Global Modeling and Assimilation Office focusing on the satellite era, from 1979 to the present. Rienecker and Coauthors  provide an overview of the system, the observations used in the data set, and aspects of its performance, including quality assessment diagnostics and comparisons with other reanalysis data sets. They note that, while the state of the stratosphere are well constrained by observations, the mesosphere is not. However, a recent model study by Liu et al.  has shown that the behavior of the middle atmosphere is strongly constrained when the state of the lower atmosphere is sufficiently well specified. A more significant difficulty is that the MERRA reanalysis has a top level of 0.01 hPa or ~ 80 km. Manney et al., [2008a, Figure 3] have shown that reanalysis models with a top level near this altitude cannot capture the rise of the stratopause to very high altitude (~80–85 km), which occurs at the beginning of an ES event. Instead, the reanalysis locates the elevated stratopause slightly above 65 km. Nonetheless, because the stratopause usually descends to about 40 km before an ES event, the abrupt rise from this altitude to 65 km is reliably identified by our algorithm as an ES event.
 Our algorithm detects 10 ES winters in MERRA data between 1979 and 2011, as shown in Table 2. Among these, six are associated with vortex displacement SSW and four with vortex splits. We note that all of the ES events documented during the last decade (2003/2004, 2005/2006, 2008–2009) by Manney and colleagues [Manney et al., 2005, 2008a, 2008b, 2009] are also identified in our analysis of MERRA data. All of these events were associated with major SSW, in which a prolonged wind reversal occurred at 10 hPa. Over the rest of the period covered by MERRA, the winters of 1983/1984, 1984/1985, 1986/1987, and 2009/2010 also had ES events that were accompanied by persistent wind reversals at 10 hPa. On the other hand, ES events in the winters of 1980/1981, 1989/1990, and 1994/1995 were accompanied by wind reversals that did not reach 10 hPa and would have been classified as minor warmings. Of particular interest is the Arctic winter of 1994/1995, which has been identified as unusually cold [Pawson and Naujokat, 1999; Zurek et al., 1996]. Low temperatures in December and January allowed polar stratospheric clouds to form and significant ozone depletion was observed and reported [Manney et al., 1996]. However, Zurek and colleagues noted that the early cold spell ended in a series of stratospheric warmings in late January and early February [Zurek et al., 1996]. These observations are consistent with the evolution seen during that winter in MERRA data (not shown), where zonal-mean temperatures before January 15 remain under 200 K in the lower stratosphere, but there is a minor SSW followed by an ES event in early February. All of this lends confidence to our identification of ES events in MERRA data, and further emphasizes the point that ES events are not universally associated with major SSW.
Table 2. The Occurrence of ES Events in MERRA Reanalysis Data for 1979–2011. The Annual Frequency of Occurrence, in Percent, is Shown in Parenthesis for Each Case
 The occurrence frequency of ES in MERRA (31%) is almost identical to that found for the WACCM ensemble (32%; see Table 1). However, ES events in MERRA are associated more with vortex displacement events than vortex splitting events, while in the WACCM ensemble ES events are more likely to occur in association with vortex splitting events. The major SSW during December 1987 (a vortex splitting event) is detected as an ES event if the detection criterion for stratopause jump is lowered to 12 km from 15 km. In Figures 7a and 7b, we show a composite of the temperature and zonal mean wind between 30 and 100 km averaged between 75°N–90°N for the 10 ES events detected in MERRA data. The composite has been made using the same procedure as for Figure 5. Comparing the MERRA and WACCM ES composites, it can be seen that they show similar characteristics, with ES forming ~ 6 to 7 days after the peak of the stratospheric wind reversal. At 35 km, where the reversal persists the longest, the wind remains westward for ~ 11 to 12 days. Figure 7c shows the decadal occurrences of ES in MERRA for the three decades from the 1980s to the 2010s; this is the counterpart of Figure 4 for WACCM. The number of ES per decade varies between 1 and 5 in MERRA and between 2 and 4 in the WACCM ensemble, while individual WACCM simulations show variations as large as 2 to 6 ES per decade.
3.5 Planetary Wave Activity During Elevated Stratopause Events
 As can be seen in the composite EP flux divergence plots in Figures 5c and 6c, stratospheric wind reversals are triggered by an increase in westward planetary wave forcing in the stratosphere. In Figure 8, we show the amplitude of planetary waves 1 and 2 in geopotential height, averaged between 55°N and 70°N and composited for all 68 ES winters. The composites have been constructed such that day 0 coincides with the day of the stratopause jump. Planetary wave 1 dominates the wave field during much of the composite ES event. Below 60 km, the peak in EP flux divergence is seen between day –25 and day –8 (Figure 5c), with peak stratospheric warming happening around day –10. The geopotential perturbations due to planetary waves 1 and 2 peak during roughly the same period as the EP flux divergence. Such enhancements in planetary waves 1 and 2 during SSW have been found in previous studies [e.g., Liu and Roble, 2002; de la Torre et al., 2012].
 The wave 1 perturbation of the geopotential field is about 3 times larger than wave 2 from day –20 to day –13, preceding the warming. However, following the stratospheric wind reversal, from day –10 to day 0, wave 1 and wave 2 have comparable amplitude. In WACCM, wave 1 is therefore the dominant feature in the stratosphere and lower mesosphere leading to the wind reversal, whereas, after the wind reversal (~ day –11), wave 1 and wave 2 have about equal amplitude. In major SSW events that do not lead to an ES, the amplitude of the wave 2 component is weaker since these events are predominantly vortex displacement events, while ES events are predominantly associated with vortex splitting events.
 The EP flux divergence plots in Figures 5c and 6c also show a suppression of westward forcing in the stratosphere immediately following the wind reversals, together with increased westward EP flux divergence in the MLT, between 80 and 100 km. Similar enhanced EP flux values were observed in this region in the case study of an elevated stratopause event presented in Chandran et al.  (their Figure 1b) and in previous studies [Liu and Roble, 2002; Limpasuvan et al., 2012]. This westward forcing is stronger and longer lasting in the ES composite (Figure 5c) than in the SSW composite (Figure 6c). The enhanced EP flux divergence, between 80 and 100 km, lasts from about day –10 to day 0, before the formation of the ES. During this period, both wave 1 and wave 2 components have comparable amplitudes in the geopotential height.
Limpasuvan et al.  attributed the formation of the elevated stratopause to downwelling driven by westward forcing due to planetary wave EP flux divergence. We find that 59 out of the 68 ES winters (87%) in the WACCM ensemble show enhanced EP flux divergence in the MLT following the stratospheric wind reversal. However, in the other nine ES winters, the EP flux divergence in the MLT is generally much smaller than the composite values shown in Figure 5c. As an example, Figure 9 shows an ES winter (from simulation refb1.4, MY 1965/1966) where there is no enhancement of EP flux divergence in the MLT following the SSW. The SSW occurs on day –13 of MY 1966 (all days in this figure are referred to 1 January 1966) and the ES forms on approximately day 0. Between day –15 and day –2, gravity wave forcing in the MLT is eastward due to the reversal of the stratospheric winds over the same period. Once the wind reversal ends, around day 0, gravity wave forcing in the MLT becomes westward. It is clear that, in events such as this one, the formation of the ES must be due solely to westward gravity wave forcing in the MLT, which induces downwelling, and thus adiabatic heating, in the polar cap. In the majority of the cases, where both planetary wave EP flux divergence and gravity wave forcing are large, as shown in the composites of Figures 5c and 5d, the relative importance of the two processes in the initial formation of the ES remains to be elucidated. Nevertheless, the composites also show that that EP divergence in the MLT weakens rapidly after the formation of the ES, while gravity wave forcing remains strong. Therefore, the maintenance and gradual descent of the stratopause following the initiation of the ES event must be ascribed to adiabatic heating associated with the downwelling induced by gravity wave forcing.
 Among the winters with enhanced EP flux divergence in the MLT we find that, following stratospheric zonal mean wind reversals, there are regions in the upper stratosphere and mesosphere that meet the necessary condition for baroclinic or barotropic instability, that is, a reversal of the zonal mean potential vorticity gradient. In a preliminary analysis of individual Arctic winters (not shown), there is some indication of a source of planetary wave activity associated with such regions, which in turn may be responsible for the enhanced EP flux divergence seen near 80–100 km.
4 Discussion and Conclusions
 Major SSW and ES events commonly occur during Arctic winters in WACCM simulations. A dynamically quiet Arctic winter without even a single minor SSW event is an extremely rare event (6% of all model winters examined). In the past decade, observations have reported major SSW events accompanied by ES events in the Arctic winters of 2003–2004, 2005–2006, and 2008–2009 [e.g., Manney et al., 2008a; Manney et al., 2009; Orsolini et al., 2010; Thurairajah et al., 2010a, 2010b]. In WACCM simulations, ES events occur with an overall annual frequency of 0.32 (about three events per decade), which indicates that the behavior observed over the past decade in Arctic winters is well within the interannual variability simulated by the model.
 In WACCM, vortex displacement SSW occur more frequently than vortex splitting events, but an ES forms much more frequently after vortex splitting events. For the ensemble of four 53 year WACCM simulations analyzed here, 68% of ES events form after vortex splitting events. The ES are formed mostly under conditions where the stratospheric wind reversal, the gravity wave forcing in the MLT region, and the residual circulations remain reversed for a longer period compared to SSW that do not produce ES. The PV gradient, which is reversed in the zonal mean, returns to normal much sooner in vortex displacement events compared to vortex splitting events possibly because there are some sectors where the PV has not reversed in displacement events. The vortex splitting events also produce a positive meridional temperature gradient over the entire polar cap along with associated reversals in zonal mean winds and residual circulation, as detailed in Chandran et al. . Some vortex displacement events do cause ES events, and we find that such events are longer lasting, by about 4 days, than vortex displacement events that do not result in ES events.
 Applying the ES detection technique to MERRA reanalysis data from 1979 to 2011 shows occurrence frequencies of ES events and vortex displacement and splitting events within the range of occurrence frequencies seen in WACCM. The composites of temperature and zonal mean wind associated with ES events in MERRA are also similar to WACCM composites, with comparable warming of the stratosphere and duration of stratospheric wind reversals. The elevated stratopause in both MERRA and WACCM forms ~ 6 to 7 days after the peak of the wind reversal in the stratosphere. In MERRA composites of major SSW (not shown) the stratospheric wind reversals in non-ES events are briefer than wind reversals accompanying ES events. The reversals last for ~ 7 days during major SSW warmings without an ES compared to ~ 12 days for SSW that accompany ES events.
 Some ES winters in WACCM simulations, such as MY 1997/1998 from refb 1.1 (shown in Figure 1d), are preceded by a minor SSW event. There have been a few studies [Labitzke1972; von Zahn et al., 1998; Braesicke and Langematz, 2000] that viewed these minor warmings as weakening the stable polar vortex and preconditioning the stratosphere for a major SSW later in the winter. We have investigated the occurrences of minor warmings (defined as those events which have a positive polar temperature gradient and that result in a reversal of the zonal mean wind at 50 km at 60 N) at least 30 days prior to a major SSW that results in an ES event. We find that 28 of 68 (~ 41%) ES events are associated with a minor SSW before the ES event, while the rest are not. The timing of these minor SSW events varies with respect to the occurrence of the ES events and thus their presence is not apparent in composite plots such as Figures 5 and 7. In any event, it does not appear that preconditioning by minor SSW plays a crucial role in the occurrence of ES events.
 During vortex displacement events, the local meridional temperature gradient in the polar cap can be either negative or positive at different longitudes because the dynamical behavior of the stratosphere and mesosphere is dominated by the presence of planetary wave 1, which produces roughly antiphase behavior on opposite sides of the globe. The polar vortex can be displaced such that there are certain longitude ranges in the polar cap that lie within the vortex, which has been displaced from the pole but otherwise retains its integrity; there, the meridional temperature gradient is generally positive. Other longitude ranges lie within an anticyclonic circulation and display a negative meridional temperature. Thus, vortex displacement (wave 1) events can produce strong longitudinal asymmetry in the meridional temperature and wind distributions, such that ground based observations will document markedly different behavior, ranging from a structure typical of an ES accompanied by wind reversals at 10 hPa to a structure typical of a minor SSW, depending on their location relative to the displaced polar vortex. This has obvious implications for the interpretation of ground-based observations of SSW events.
 Although planetary wave 1 is the dominant feature responsible for SSW in WACCM, wave 2 becomes significant after an SSW and before the formation of an ES. Planetary waves 1 and 2 amplitudes are also enhanced in the MLT after a stratospheric warming and before the formation of an ES. In agreement with the study of Limpasuvan et al. , we find in our analysis that ~ 90% of ES events are accompanied by enhanced EP flux divergence in the MLT. The westward planetary wave forcing in the MLT region acts against the eastward gravity wave forcing and reduces the wave-driven upwelling and mesospheric cooling and helps reestablish the westward zonal jets in the mesosphere more quickly. In these events, it is difficult to separate contributions to downwelling and adiabatic heating due to planetary wave dissipation from those due to gravity wave breaking. However, while gravity wave driving is always present during the formation of the ES, this is not true of planetary waves. We have illustrated one ES event in which there is no enhancement of EP flux divergence in the MLT, such that the ES forms solely as a result of gravity wave-driven downwelling and adiabatic heating.
 The authors thank H. L. Liu and V. Yudin for helpful discussions. The authors acknowledge support from the United States National Science Foundation under grant ARC 1107498 and grant ARC 0632387 as part of the United States International Polar Year Program. AC conducted this work as a participant in the visiting scientist program at the National Center for Atmospheric Research. The National Center for Atmospheric Research is sponsored by the National Science Foundation. VLH was supported by NASA grant NNX10AQ54G and NSF grant AGS 0940124. L. de la Torre has been partially supported by research project CGL2011-24826, financed by the Ministry of Economy and Competitiveness of Spain and the FEDER fund.