Journal of Geophysical Research: Atmospheres

A climatology of polar winter stratopause warmings and associated planetary wave breaking



[1] This work presents a climatology of synoptic-scale disturbances in the upper stratosphere lower mesosphere (USLM) based on 20.5 years of assimilated data analyses from the U. K. Meteorological Office (1991–2012). USLM disturbance criteria are established, based on stratopause warmings at the 2 hPa level, to create climatologies in both hemispheres that delineate their timing, frequency, and geographic location. USLM disturbances occur on average 2.3 times per winter in the Northern Hemisphere (NH) (November through March) and 1.6 times per winter in the Southern Hemisphere (SH) (May through September), persist on average for 8 days in the NH and only 4 days in the SH, occur most frequently in December (July) in the Northern (Southern) Hemisphere, and are predominantly located in the longitude sector between 0oE and 90oE in both hemispheres. This is the first work to show that all major Sudden Stratospheric Warmings (SSWs) over the 20.5 year data record are preceded by USLM disturbances. One third of USLM disturbances evolve into a major SSW; only 22% of minor SSWs evolve into a major SSW. USLM disturbances and minor SSWs illustrate, at times, similar occurrence statistics, but the minor warming criteria seem to include a more diverse range of dynamical conditions. USLM disturbances are more specific in their dynamical construct with strong baroclinicity being a necessary condition. Potential vorticity analysis indicates that all USLM events occur with planetary wave breaking and that subsequent baroclinic instability may lead to the development of USLM disturbances.

1 Introduction

[2] The circulation in the polar winter stratosphere is dominated by a large circumpolar cyclone known as the polar vortex which forms due to Earth's rotation, tilt of its rotation axis, and decreased solar insolation with latitude [e.g., Holton, 2004; Schoeberl et al., 1992]. The resulting westerly circulation is conducive for upward propagation of planetary-scale, quasi-stationary Rossby waves [Charney and Drazin, 1961]—i.e., planetary waves (PW)—which can break and add a great deal of variability and structure to the polar winter stratosphere. PW breaking (hereafter PWB) events are intimately linked to extreme disruptions to the polar vortex and rapid warmings of the lower stratosphere known as Sudden Stratospheric Warmings (SSWs) [e.g., Scherhag, 1952; Labitzke, 1972; Manney et al., 1999, 2005a, 2005b, 2008; Charlton and Polvani, 2007; Sathishkumar et al., 2009, and references therein]. Major SSWs are the most dramatic dynamical events that occur in the polar winter middle atmosphere. Using the established World Meteorological Organization (WMO) definitions, “minor” SSWs are identified when the zonal mean temperatures at 10 hPa are warmer over the pole than at 60°N. “Major” SSWs are identified when minor warming conditions are satisfied and the zonal mean zonal wind at 10 hPa poleward of 60°N reverses from westerly to easterly [Andrews et al., 1985].

[3] However, PWB events are very common, often occur at the vortex edge, can occur but not result in minor or major SSW events, and can occur at various levels within the vertical column of the polar atmosphere [see review by Waugh and Polvani, 2010, and references therein]. An illustration of a disturbed polar vortex due to PW activity in the upper stratosphere is illustrated in Figure 1, where maps of temperature and potential vorticity (PV) are displayed on the 1600 K isentropic surface, near the 2 hPa level. Maps of PV on isentropic surfaces provide useful information on the structure, evolution, and dynamics of the polar vortex. For instance, the PV signature displayed in Figure 1 is characteristic of a PWB event [e.g., McIntyre and Palmer, 1983] where regions of high PV near the vortex edge are pulled off with irreversible deformations and mixed with the low PV midlatitude air. Consequently identification and climatologies of PWB events have been associated with reversals in the latitudinal gradient of PV along a specific longitude sector [e.g., Baldwin and Holton, 1988; Knox and Harvey, 2005; Hitchman and Huesmann, 2007]. In 3-D simulations of the polar vortex by Dritschel and Saravanan [1994], Waugh and Dritschel [1999], and Polvani and Saravanan [2000], PWB events can be organized into two classifications: remote events where PWs propagate up the vortex edge and break in upper levels of the stratosphere or local events where PWB occurs at lower levels of the stratosphere. Abatzoglou and Magnusdottir [2007] used ERA-40 data to illustrate distinctively different climatologies for upper-level (800–1220 K) and lower-level (400–700 K) PWB events. They find upper-level breaking events are characterized by large-amplitude wave number 1 and 2 overturning the PV field. Figure 1 is characteristic of this type of PWB event with its typical “comma”-shaped PV structure. Also, displayed in Figure 1 is a map of temperature on the 1600 K isentropic surface with an anomolously high-temperature maximum on the edge of the vortex between 90° and 135°E longitude. It is the climatology of these stratopause temperature maxima and the dynamical influence of upper-level PWB events that are the focus of this paper.

Figure 1.

Polar stereographic plot on the 1600 K isentropic surface (~2 hPa) for 24 December 2011 of potential vorticity [10−6Km2kg−1s−2 (PVU)] contours (black). The white dashed contour indicates the edge of the Arctic vortex. Shading denotes temperature with superimposed colored contours at 270 K, 275 K, and 280 K to emphasize the location of the warm pool.

[4] This regionally confined temperature enhancement near 2 hPa in Figure 1 is part of an upper stratosphere and lower mesosphere disturbance that includes an unusually low-altitude stratopause near 42 km (2 hPa), a stratopause temperature in excess of 290 K (50 K above nominal conditions), a ~40 K cooling in the mesosphere near 75 km, and an undisturbed lower stratosphere [e.g., Labitzke, 1972; von Zahn et al., 1998; Meriwether and Gerrard, 2004; Thayer and Livingston, 2008]. This thermal structure has been observed over the years and termed a “stratopause warming” [Duck et al., 2000; Braesicke and Langematz, 2000], a “stratopause temperature enhancement (STE)” [Merriwether and Gerrard, 2004], and upper stratosphere/lower mesosphere (USLM) disturbances [Manney et al., 2008; Thayer et al., 2010]. In this study, USLM will refer to the whole 3-D structure of the disturbance while stratopause warmings near 2 hPa will be used as the observable to relate the USLM disturbance to PW activity. Thayer et al. [2010] used SABER measurements from the TIMED satellite to describe the 3-D characteristics of a USLM disturbance and likened it to a front-like structure in the polar winter middle atmosphere invoking baroclinic instability as playing a possible role in the mechanism for the enhanced temperatures and vertical structure. Based on the improved understanding of USLM characteristics from SABER measurements as presented in Thayer et al. [2010] showing recurring anomalous thermal structure of regional extent, criteria were established to identify these events in 20.5 years of U. K. Meteorological Office (MetO) assimilated data. This paper presents a climatology of USLM disturbances, including seasonal and inter-seasonal variability, geographic location with respect to the polar vortex, seasonal distribution, and life cycle duration.

[5] Major SSW events are the most well-known type of middle atmospheric disturbance, and their climatology has been extensively studied [e.g., Limpasuvan et al., 2004; Charlton and Polvani, 2007]. USLM disturbances tend to be more frequent synoptic-scale features of the polar winter upper stratosphere and lower mesosphere whose climatology and relationship to SSW events has not been documented. This paper will investigate the relationship of USLM disturbances to SSWs and diagnose the dynamical mechanisms responsible for the development of USLM disturbances. Given the climatology, a breaking PW and secondary hydrodynamic instability is proposed as a mechanism for generating USLM disturbances. Drawing on work by Hoskins et al. [1985] and Fairlie et al. [1990], PV is used to diagnose and analyze the dynamics in the upper stratosphere and establish connections between the USLM disturbances, PWB events, and SSWs.

[6] An outline of the paper is as follows. The MetO data and the method for identifying USLM disturbances are described in section 2. Section 3 presents the USLM disturbance climatology. This includes the frequency of occurrence, duration, geographical distribution, a composite lifecycle, and the relationship to SSWs. The discussion in section 4 focuses on dynamical mechanisms responsible for the development of USLM disturbances. Section 5 concludes the paper with a summary of the results.

2 Data and Analysis Methods

2.1 MetO Data

[7] U. K. Meteorological Office (MetO) stratospheric assimilated data are used to identify both SSWs and USLM disturbances from October 1991 through April 2012. The MetO stratospheric data set has also been known as the UK Meteorological Office (UKMO) stratospheric assimilated data. Once daily (12Z) temperature, winds, and geopotential heights at 2.5° latitude by 3.75° longitude resolution are obtained on 22 pressure surfaces extending from the 1000 to 0.3 hPa (26 pressure surfaces up to 0.1 hPa after late 2003) [Swinbank and ONeill, 1994]. The assimilation scheme also incorporates satellite soundings from Advanced TIROS Operational Vertical Sounder data from the NOAA-15 satellite in addition to TIROS Operational Vertical Sounder data from NOAA-14 [Lorenc et al., 2000]. Prior to November 2000, the assimilation used an analysis-correction scheme as described by Lorenc et al. [1991]. Mid-November 2000 and late October 2003 mark major changes in the MetO analyses with the former involving the implementation of 3-D variational assimilation [Lorenc et al., 2000] and the latter a new dynamical core in the Unified Model [Davies et al., 2005]. These periods of change to the MetO assimilation scheme involving 3D var and a new dynamical core were compared to the Northern Annular Mode (NAM) index [e.g., Waugh and Polvani, 2010] to check for consistency. It is seen that in years experiencing NAM values corresponding to a strong polar vortex, the MetO database showed diminished winter disturbances [e.g., Manney et al., 2005a, 2005b]. This indicates the assimilation scheme shows variability on a decadal scale and should be satisfactory for our use of the database. This work is based on MetO temperature, horizontal winds, and geopotential height data on pressure surfaces. The isobaric data are interpolated to potential temperature surfaces ranging from 330 to 2000 K in order to compute isentropic PV (IPV) and to calculate the edge of the polar vortices using the method described by Harvey et al. [2002]. The identification of USLM events, discussed below, is based on localized stratopause warmings near 2 hPa in the MetO polar winter temperature field without any zonal averaging.

2.2 USLM Disturbance Identification Algorithm

[8] Based on the dynamical arguments presented by Thayer et al. [2010] and past observations [e.g., Thayer and Livingston, 2008; von Zahn, 1998], USLM disturbances display the following characteristics: (1) Strong baroclinic conditions near the stratopause, (2) strong positive vertical temperature gradients below the stratopause, (3) stratopause temperatures in excess of 290 K, (4) stratopause height near 42 km +/− 2 km (~2 hPa), (5) separated mesopause located between 65 km and 85 km, (6) regionally concentrated latitudinal and longitudinal extent of synoptic-scale temperature anomalies in the upper stratosphere and mesosphere, and (7) rapid development over several days.

[9] The unique characteristics of USLM disturbances allow for the development of numerical identification criteria to isolate these events in the MetO data and construct a climatology. Given that the thermal structure is of narrow latitudinal and longitudinal extent, this analysis avoids using zonal means. The numerical algorithm is described as follows. First, the maximum temperature at 2 hPa poleward of 40° latitude in both hemispheres is archived on each day. We then fit a periodic function to the annual cycle of daily polar cap temperature maxima to represent seasonal variations in temperature. Using regression analysis, the following function is fit to the temperature time series using a least squares technique:

display math(1)

where t is time, and b is a vector of coefficients to be fit. This equation accommodates annual and semi-annual temperature variations. Table 1 lists the coefficients that were determined for each hemisphere.

Table 1. Polynomial Fit Coefficients of the Seasonal Function Described in Equation (1)

[10] In equation (1), b1 is the mean of the temperature data (units of K), b2 is the linear trend component (units of K/year), the square root of the sum of the squared b3 and b4 coefficients give the annual amplitude while the square root of the sum of the squared b5 and b6 coefficients give the amplitude of the semi-annual variation. The fit is significant at the 95% confidence level. The annual variation dominates in the Southern Hemisphere (SH), but the semi-annual variation dominates in the Northern Hemisphere (NH) due to the larger temperature variability during the Arctic winter. From the analysis, temperatures in excess of 15 K from the fitted function were flagged and found to occur only in the winter months (November through March for the NH, April through October for the SH). The criteria of 15 K above the fitted temperature was selected due to this difference exceeding one standard deviation of winter temperatures in both hemispheres (σNH = 13.1 K, σSH = 9.4 K), and it is generally robust. If a day meets this requirement, it is designated as a candidate day for a USLM disturbance. If this condition persists for 2 or more days in any 4 day window, it is identified as a USLM event and is included in the climatology. The start and end dates of all USLM events are identified as the days over which these criteria are met. This methodology reduces the effects of abrupt but short-lived temperature enhancements due to transient waves.

[11] Figure 2a shows the 20.5 year time series (1991–2012) of the MetO 2 hPa polar cap maximum temperature and an enlarged view of a single NH season. Figure 2b shows the same set of plots for the SH. In both panels, the solid black line is the daily maximum temperature poleward of 40° latitude. The gray box in the left panels indicates the season shown in the right panels. The gray dashed line in the single season panel is the fitted seasonal function. The mean seasonal cycle amplitude in the NH is ~20 K, and the daily variation in maximum polar temperature exhibits 30 K to 50 K temperature spikes. In the SH, there is a larger mean seasonal cycle amplitude (~30 K) but smaller daily variations in the maximum polar cap temperature. This difference between hemispheres is presumably due to the more stable polar vortex in the SH hemisphere. Shading in the right panels indicates USLM disturbances based on the established criteria.

Figure 2.

(a) Time series of the daily maximum MetO temperature [K] poleward of 40°N for 20.5 years (left) and for the 2005–2006 season (right). The gray dashed line is the fitted seasonal function. The gray box in the left panel indicates the season shown on the right. During the 2005–2006 season, the vertical gray shading indicates periods when USLM conditions meet the established criteria. (b) Time series for the Southern Hemisphere on the left and the 2005 winter season on the right, using the same notation as in the Northern Hemisphere.

[12] The robustness of choosing the 2 hPa level is demonstrated in Figure 3 which shows MetO temperature profiles on the warmest day during NH USLM disturbances. Individual daily profiles are thin black lines, an average of USLM profile is plotted in red plus symbols. For comparison, an average temperature profile at 61°N (the mean latitude of USLM disturbances) for all MetO winter months of December, January, and February (DJF) is shown as a dashed blue line. Temperature profiles that meet both USLM disturbance conditions and major SSW conditions differ from profiles that only meet USLM disturbance conditions in that they show a warmer lower stratosphere. However, profiles that also meet WMO criteria for major SSW conditions are excluded in Figure 3 to accentuate the prominent thermal features associated with the USLM disturbance. The temperature profiles are located at the longitude and latitude of the largest temperature anomaly, which varies from day to day and from event to event. These results indicate that the temperature profiles peak with exceptional reliability at 2 hPa over the 20.5 year period, a conclusion established by von Zahn et al. [1998] but with much less data, and represent stratopause warming events. The DJF temperature profile is not representative of the vertical temperature structure during USLM events. During USLM events, the stratopause is ~50 K warmer and approximately 10 km lower than in the DJF profile. Above ~0.3 hPa (~60 km), the mesosphere is ~20 K colder than in the DJF profile, a feature illustrated by Thayer et al. [2010].

Figure 3.

Northern Hemisphere MetO temperature profiles that intersect the 2 hPa warm anomaly on peak days during USLM disturbances. Individual daily temperature [K] profiles are black, the average profile is indicated by red + symbols for each pressure level. An average temperature profile at 61°N from December, January, and February months for the entire MetO database is plotted for comparison as a dashed blue line.

3 USLM Disturbances

3.1 Climatology

[13] A total of 49 USLM events are identified in the NH, and 31 USLM events are identified in the SH. Their onset dates are listed in Appendix A. On average, 2.3 USLM disturbances occur each NH season, while the frequency in the SH is 1.6 USLM events/year. The onset dates for major and minor SSWs are also catalogued for the data set. SSW dates are in agreement with Charlton and Polvani [2007] and extend their record to the present. In the NH, 16 major SSWs and 72 minor SSWs are identified. The frequency of major SSWs is 0.76 per NH winter season. In the SH, 1 major SSW and 15 minor SSWs are identified. The one major SSW in the SH in 2002 has been examined extensively in the literature [e.g., Krüger et al., 2005; Simmons et al., 2005; Manney et al., 2005]. The seasonal average frequency of minor SSWs is 3.4 in the NH and 0.75 in the SH. Thus, the frequency of SH minor SSWs is similar to the frequency of major SSWs in the NH.

[14] Figure 4a shows the annual cycle of the distribution of days per month per year that meet USLM criteria in each hemisphere. USLM disturbance days occur from November through March in the NH and from May through September in the SH. The MetO data shows a pronounced preference for USLM disturbances during December in the NH and during July in the SH. Figure 4b shows USLM disturbance event duration in both hemispheres. The mean duration of a USLM event in the NH is 8 days, while the mean duration of a USLM event in the SH is only 4 days. The duration of USLM events are not normally distributed. While USLM disturbances can last as long as 3 weeks, events lasting one week or less comprise 57% of events in the NH and 58% in the SH.

Figure 4.

(a) Monthly frequency of USLM disturbance days per year; (b) USLM event duration as function of month. Frequencies in the Northern (Southern) Hemisphere are in dark gray (light gray).

[15] Figure 5 shows the geographic distribution of USLM disturbance frequency (in color) for all events in the NH and SH. Note the different frequency scales between the hemispheres. The black contour indicates the average position of the edge of the polar vortices during USLM days at the 1600 K potential surface (near 2 hPa) as defined by Harvey et al. [2002]. The warm anomaly associated with USLM disturbances preferentially occurs over Northeastern Russia and Scandinavia in the NH and south of Africa in the SH. In both hemispheres, the warm anomaly occurs most frequently between 0oE and 90oE along the Eastern edge of the polar vortices. Thus, ground-based observation sites in Scandinavia and northern Russia in the NH and Davis Station or Dome Fuji Station in the SH have several opportunities per year to observe USLM events. The location of the thermal anomaly being on the east side of the low is also reminiscent of the structure of a developing tropospheric baroclinic wave [Thayer et al., 2010].

Figure 5.

Polar stereographic projections of the geographical distribution of USLM event occurrence frequency in the Northern Hemisphere (left) and Southern Hemisphere (right). Frequency equals the number of days per year that a location satisfies USLM conditions. The average location of the boundary of the polar vortex for USLM days is indicated by the black contour at the 1600 K potential surface (near 2 hPa) as defined by Harvey et al. [2002]. Note the different frequency ranges between hemispheres.

[16] Using the identification algorithm described in section 2, a composite analysis of the identified NH USLM events was created. USLM disturbances are regional phenomena that occur over a range of longitudes and latitudes (as shown by the colored region in Figure 5a). For this composite analysis, we include a subset of all identified USLM disturbances wherein the center of the warm temperature anomaly is within +/−5° latitude of the mean location of all thermal anomalies at 2 hPa in the data set (52.5°E and 62.5°N). The fields of temperature and geopotential height are shifted in longitude so that all maximum thermal anomalies are in phase and located at the mean longitude location (52.5°E). In addition, the three PW2 events were eliminated for clarity. This results in a composite of 16 USLM events in this latitude/longitude region (the onset dates of these events are noted in Appendix A). Figure 6 illustrates the USLM lifecycle over eight days with day 0 identified as the day when the temperature anomaly was warmest between 40°N and the pole at 2 hPa; the other days are noted in relation to day 0. The eight panels show the progression of 2 hPa temperature (filled color contours) and geopotential height (black contours) for days −4 through day +3. During the onset of USLM events (days −4 to −1), the polar vortex (denoted with an “L”) is increasingly displaced from the pole as the Aleutian high (denoted with an “H”) strengthens and moves poleward. This synoptic development is displayed in the growth of planetary wave 1 (PW1) at 60oN. In the days leading up to day 0, the cold and warm anomalies are displaced from the core of the circulation systems such that large horizontal thermal advection occurs. This leads to strong baroclinic conditions and thermal gradients [see Thayer et al., 2010] that results in strong vertical wind shear through the thermal wind relation. A warm temperature anomaly develops on the east side of the polar low in the region of large geopotential height gradients (the polar night jet). The cool temperature anomaly near 90oW moves equatorward and warms over 10 K between day −4 and day 0. On day −2, PW1 amplitudes at the 2 hPa level at 65°N maximize; on day 0, the warm temperature anomaly maximizes. On day +1, the warm temperature anomaly is cooler compared to day 0, but the warm anomaly expands to cover a larger area. As the event continues, the warm temperature anomaly dissipates, conditions become less baroclinic, and PW2 activity increases. After day +2, the variance of the temperature and geopotential height become much larger, and the individual events begin to diverge in their development and structure (not shown). This is likely due to some USLM events dissipating and some events developing into SSWs.

Figure 6.

Northern Hemisphere composite maps of temperature [K] at 2 hPa in color on days surrounding USLM events. All events have been shifted in longitude such that the phase of the temperature maximum is aligned. Solid black lines are geopotential height; contour intervals are 400 m. The geopotential highs and lows are indicated by white “H” and “L” symbols, respectively.

3.2 PW Activity

[17] The composite lifecycle suggests that the growth phase of a USLM disturbance corresponds with increased PW1 and/or PW2 amplitudes. Figure 7 examines the PW amplitudes and phases of USLM events and compares them to “non-USLM event” winter days. Figure 7a illustrates the relationship between the maximum temperature during USLM events and the PW amplitudes (calculated from the geopotential height distribution at 65oN, 10 hPa) two days prior. For the entire NH winter dataset (November through March, 21 seasons), we compute daily maximum polar cap temperatures at 2 hPa and PW amplitudes (the sum of PW1 and PW2) at 10 hPa from two days prior; PW amplitudes were seen to maximize two days prior to USLM events. The black contours represent the 2-D frequency distribution of all 3354 winter days that do not meet USLM conditions. There is a concentration of days with low maximum polar cap temperatures (~250 K) and low PW amplitudes (~300 m); these are days in which the horizontal thermal structure is minimally baroclinic and a strong polar vortex is nearly pole centered. As a subset of all NH winter days, the red crosses indicate days during USLM disturbance lifecycles (379 days). There is a cluster of USLM disturbance temperatures above ~275 K (red crosses) with PW amplitudes larger than ~800 m, indicating that strong PW amplitudes at 10 hPa are observed two days prior to USLM disturbances at 2 hPa. This suggests that increased PW amplitudes favor USLM event development, but large wave amplitudes do not always lead to USLM events. Figure 7b shows the PW1 and PW2 amplitude results for the SH. Overall, the non-USLM day (black contours) are concentrated in smaller PW amplitude ranges while USLM days represent events with a wide range of PW amplitudes. Despite having fewer USLM days in the SH, it is clear that (like the NH) these events are associated with stronger PW amplitudes.

Figure 7.

(a) Comparison of maximum polar cap temperature [K] at 2 hPa with planetary wave amplitudes [m] (sum of PW1 and PW2) two days prior. All NH winter days that do not meet USLM criteria are indicated by the black contours. USLM days are plotted as red crosses. (b) Same as Figure 7a, but for the SH. (c) Comparison of maximum polar cap temperature [K] at 2 hPa with the PW zonal phase change [degrees] (east-west tilt with altitude) between 10 hPa and 2 hPa two days prior. All NH winter days that do not meet USLM criteria are indicated by the black contours. (d) Same as Figure 7c, but for the SH. All USLM days are plotted as red crosses.

[18] Figure 7c shows the relationship between maximum polar cap temperature at 2 hPa and the differential zonal phase of PW1 geopotential height between the 10 hPa and 2 hPa levels at 65°N. The mean differential phase or vertical tilt for all non-USLM winter days is −19.3° of longitude (indicated by dashed back line). Negative phase changes with height are associated with westward tilting structures. Wintertime westward tilts are a sign of upward propagating PWs; when the phase becomes more vertical, it signifies that the PW is breaking [Salby et al., 2002]. As in Figure 7a, all USLM days are plotted as red crosses. The mean westward tilt on these days is −37.9° (red dashed line), nearly double the value for days when USLM conditions are not met. The larger westward tilt is indicative of strong baroclinic conditions, a necessary condition for the onset of a USLM disturbance, and suggestive of the possibility of baroclinic type instabilities. Figure 7d shows the zonal phase analysis for the SH. For the population of all non-USLM winter days, the average phase difference between 10 hPa and 2 hPa is −17.7°, while the USLM days display a phase difference of −26.2°. This is not as strong as in the NH, but still a significant difference indicating propagating PWs and strong baroclinic conditions in the SH during USLM conditions. These results indicate the need for the propagation of large-amplitude PW to the upper levels of the stratosphere for USLM disturbances to occur.

3.3 Relationship With Minor and Major SSW Events

[19] As regularly occurring weather events in the middle atmosphere, USLM disturbances are associated with SSWs by contributing to the preconditioned state for major SSWs to evolve. Figure 8 shows two Venn diagrams, one for each hemisphere, which illustrates four possible groupings among USLM events, minor SSW events, and major SSW events. All of the USLM events that were identified in the MetO database are contained within the green ovals, while all minor SSW events are contained in the blue ovals. Red circles denote major SSW events. Where the ovals overlap suggests a relationship between events. A relationship is defined to exist if different events occur within 14 days of each other. This time period allows for the extended development of a USLM (see Figure 4b), minor SSW and major SSW; the specific criterion is relatively insensitive to periods of 14 days +/− 2 days. Most notable in Figure 8 is that every major SSW is associated with (and preceded by) a minor SSW and a USLM event (as indicated by the red major SSW circle being fully contained within the green USLM oval and the blue minor SSW oval). For these occurrences, the time progression begins with a USLM event followed by a minor warming and then a major warming. This suggests that the development of USLM disturbances is a necessary precondition for major SSWs and may be a useful tool to forecast major SSWs.

Figure 8.

Venn diagram illustrating the various relationships between USLM events (green ovals), minor SSWs (blue ovals), and major SSWs (red ovals) for each hemisphere.

[20] The 19 NH events found where the USLM (green) and minor SSW (blue) ovals overlap are events when both minor SSW and USLM criteria are satisfied within a two-week period but did not develop into a major SSW. In the SH, the Venn diagram shows that there are seven events in the overlapping green and blue ovals. Inspection of these occurrences indicates that the minor warming criteria are typically met after the USLM criteria. This is indicative of the later stages of the USLM lifecycle, shown in Figure 6, where the enhanced temperatures are spread over a greater area such that zonally averaged temperatures meet the minor warming criteria at 10 hPa. The proportion of USLM events that are associated with minor SSWs is ~50% in both hemispheres.

[21] The region inside the blue minor SSW oval that does not overlap the green USLM oval represents minor SSWs that occur independently of USLM and major SSW events. This suggests that the WMO criteria for identifying minor SSWs represent a diversity of wintertime middle atmospheric thermal disturbances that do not result in major warmings. For example, Canadian warmings are characterized by an amplification of PW1 resulting in a displacement of the vortex from the pole and can satisfy the minor warming criteria. These events do not evolve into major warmings [Labitzke, 1982; Naujokat et al., 2002] and are not associated with USLM events because they are largely confined to the lower stratosphere. In both hemispheres, ~50% of all minor SSWs occur independently of USLM events and in the NH 78% of minor warmings occur independent of major SSWs.

[22] There are also USLM events that occur and dissipate without developing into a minor or major SSW (14 out of 49 events in the NH, 23 out of 31 events in the SH). Furthermore, 67% of NH USLM events do not evolve into major SSWs. A significant difference between the hemispheres is the relative number of independently occurring USLM events. Because the SH vortex is more stable than in the NH at 2 hPa, there are twice as many independent USLM events in the SH. Investigating the relationship between USLM events and SSWs further is the subject of future work. The next section applies PV concepts to MetO data to interpret the dynamics of USLM disturbances.

4 Discussion

[23] The characteristics of synoptic-scale warming events at 2 hPa, their temporal distribution, event duration, geographic distribution, and organized lifecycle progression establish USLM disturbances as regular and repeatable occurrences with coherent spatial structures that precede all major SSWs. In the examination of a USLM disturbance case study from February 2002, Thayer et al. [2010] posit that baroclinic instability plays a key role in their development. However, additional possibilities include barotropic instability, inertial instability, and both barotropic and baroclinic instability working in concert with the PWB event. Here we employ PV analysis to help understand the dynamical conditions under which these USLM form.

[24] Isentropic maps of Ertel's PV (IPV) are standard diagnostic tools for analyzing PWB. IPV also has the advantage of remaining valid in the vicinity of frontogenic regions at the stratopause [Hoskins et al., 1985], where isentropic surfaces are not coincident with isobaric surfaces and ageostrophic motions occur. Here, IPV is materially conserved assuming adiabatic, frictionless motion. While the hydrostatic approximation is applied to retain the invertibility properties, there are no assumptions made of balanced geostrophic motion.

display math(2)

[25] To illustrate the role of PWB during USLM disturbances, four case studies are shown in Figure 9. Figure 9 displays IPV (black contours) on the 1600 K isentropic surfaces (~2 hPa) superimposed on the temperature field (colored contours). The white dashed contour is the edge of the Arctic vortex. Figure 9a is a USLM event on 5 January 2011 that dissipated without further consequences; Figure 9b is a USLM event on 21 January 2008 that resulted in a minor SSW; Figure 9c is a USLM event on 20 February 2005 that evolved into a displacement type major SSW; and Figure 9d is a USLM event on 19 January 2009 that culminated in a vortex splitting type major SSW. These examples represent different regions in the Venn diagram shown in Figure 8. Here we show individual case studies because the analysis requires the examination of fine scale IPV structures that would be obscured if spatially or temporally averaged. The shape of the Arctic vortex suggests PWB in all cases. PWB is further demonstrated by IPV filaments and nodules that have been irreversibly contorted or separated from the high IPV region inside the vortex [McIntyre and Palmer, 1983]. Figure 9 indicates the presence of strong gradients in PV in the vicinity of the stratopause warmings. High IPV air that has been detrained from the polar vortex will be mixed into the stratospheric surf zone. The spatial size of the detrained IPV filament appears to increase from panel a) to panel b) and from panel b) to panel c). This is revealing in that the ULSM event in panel a) did not evolve into a SSW, the USLM event in panel b) evolved into a minor SSW, and the USLM event in panel c) evolved into a major SSW. A review of all NH USLM disturbances showed that PWB occurred during all events (not shown). However, PWB is a “ubiquitous” process in the winter polar regions [McIntyre and Palmer, 1985; Hitchman and Huesmann, 2007], and not all PWB events are accompanied by USLM disturbances. While USLM disturbances appear to require additional dynamical circumstances (such as strong baroclinic conditions at the stratopause level and differential thermal advection between the levels), PWB likely provides the necessary energy for further development mechanisms of USLM formation. The association between USLM events, SSWs, and PWB is likely complex, however, and requires more investigation that will be the subject of future work.

Figure 9.

Same as Figure 1 but for (a) 5 January 2011, (b) 21 January 2008, (c) 20 February 2005, and (d) 19 January 2009. Colored temperature contours at 270 K, 275 K, 280 K, 285 K, and 290 K emphasize the location of the warm pool(s).

[26] McIntyre and Palmer [1985] established that PWB occurs when the IPV contours become irreversibly contorted, cascading from large scale down to smaller scales. As the PW breaks, local conditions may be modified by secondary instabilities: inertial, barotropic, or baroclinic. Barotropic instability is associated with strong horizontal shear in the mean flow, while baroclinic instability is associated with strong vertical shear in the mean flow, or equivalently, with strong horizontal temperature gradients. Both types of shear are present in the vicinity of USLM disturbances. The stability properties of the fluid may be described by a normal modes approach [Pedlosky, 1964]. Based on quasi-geostrophic theory, the Charney-Stern conditions necessary for instability [Charney and Stern, 1962; Pedlosky, 1964] are derived from a normal modes analysis wherein a single Fourier mode is introduced into the flow with a complex phase velocity. A necessary condition for baroclinic instability is that the meridional gradient of the quasi-geostrophic PV (q) must change sign. Except under very specific circumstances (including when isentropic surfaces are coincident with isobaric surfaces, which is not valid during USLM disturbances), q is generally not the same quantity as IPV [Hoskins et al., 1985]. Quasi-geostrophic PV may be calculated in various vertical coordinate systems; here pressure (p) is used to be consistent with MetO. A derivation of isobaric q is given by Holton [1997] in terms of the stream function ψ as:

display math(3)

[27] On the right side of equation (3), term A is the planetary vorticity, term B is barotropic vorticity, and term C is baroclinic vorticity. The variable σ2 acts in a manner similar to the Brunt-Väisälä frequency (N2) in equations where potential temperature is used as the vertical coordinate, here inline image , where p0 and inline image are reference pressures and reference potential temperatures, respectively. For clarity, the stream function in terms of our available variables is expressed as inline image, and relates to the deviation in geopotential from the zonal mean. In spherical coordinates, the meridional gradient of quasi-geostrophic PV is expressed as:

display math(4)

[28] Figure 10 shows the meridional gradient in quasi-geostrophic PV based on the composite lifecycle temperature and geopotential height fields at 2 hPa shown in Figure 6. A term analysis of the barotropic and baroclinic components of the meridional gradient in quasi-geostrophic PV reveals that this diagnostic is dominated by the baroclinic component on USLM disturbance days by an order of magnitude, even though there are also strong horizontal wind gradients. Additionally, there is no negative quasi-geostrophic PV near the warm anomaly (not shown) which rules out inertial instability. The eight panels in Figure 10 show the progression of temperature (black dashed contours), geopotential height (black thin solid contours), and the meridional gradient in quasi-geostrophic PV (in color) for day −4 through day +3. A reversal in the meridional gradient in quasi-geostrophic PV is indicated by the thick white contour. The persistent region of negative dq/dφ near the pole is due to the vortex being displaced off the pole (calculating the gradient in vortex-centric coordinates could avoid this polar condition, but is not necessary for this work). By day −3, a coherent region of negative dq/dφ develops in a location southeast of the polar jet (between 30oE and 100oE) near the location of the stratopause warming. As the USLM event progresses to its peak on day 0, the region of negative dq/dφ strengthens, grows, and then becomes less organized on the days following the peak warming. As the event dissipates, the area of negative dq/dφ breaks up. This progression suggests that a local instability is intensifying the temperature gradient in the region of the flow. That the negative dq/dφ becomes less organized after day 0 indicates that the source mechanism may already be diminishing. As baroclinic instability acts to extract energy from the breaking PW, it diminishes as IPV is irreversibly mixed into the surf zone. The variability among the individual USLM events begins to obscure the average dq/dφ pattern beyond day +3. Results shown here suggest that USLM disturbances require baroclinic conditions embedded in a larger region of PWB. When baroclinic conditions are not present, PWB may manifest as a minor SSW event that is associated with neither a USLM event nor a major SSW event (see Figure 8). In these cases, the vortex is displaced but remains barotropic.

Figure 10.

Same as Figure 6 but colored contours are the meridional gradient in quasi-geostrophic potential vorticity [10−9 PVU/m]; the heavy white line indicates where dq/dφ = 0, thin solid contours are lines of geopotential height (contour intervals at 750 m), dashed contours are isotherms (contour intervals at 10 K).

[29] From the results presented, the progression of USLM events begin with a propagating PW that breaks in the upper stratosphere. A deceleration of the westerlies occurs due to wave drag where momentum is transferred from the wave to the mean flow. This altered flow results in ageostrophic motion as the fluid tries to balance the wave forcing; divergence in this flow produces downward motion. Owing to the strong static stability of the stratosphere, downward motion generates significant adiabatic heating in the upper stratosphere. This is demonstrated by the development of enhanced temperatures and baroclinicity on the east side of the polar low as illustrated by Thayer et al. [2010]. The horizontal temperature gradients produced in this region lead to vertical shear in the horizontal wind based on the thermal wind relation. The increase in vertical shear is the reason the baroclinic term (C) in equation (3) is dominant and, thus, supports the growth of baroclinic instability in the region. Through baroclinic instability, regional growth of the disturbance and amplifying thermal structures result in the distinctive characteristics that constitute a USLM event.

[30] USLM disturbances have consequences for the redistribution of stratospheric and mesospheric air through vertical ageostrophic circulations. Descent results in adiabatic warming near the stratopause at 2 hPa while upward moving air adiabatically cools the mesosphere near 0.01 hPa (~80 km). In order to maintain hydrostatic balance and quasi-geostrophy, horizontal ageostrophic flow is required where the air is divergent or convergent; a set of closed ageostrophic circulation cells is required between the stratopause and lower mesosphere [see Figures 1 and 6 in Thayer et al., 2010]. This ageostrophic circulation can redistribute constituents, and these motions need to be accounted for in studies of chemical tracers. Finally, ageostrophic motions are important because they may provide in situ sources of gravity waves [Fairlie et al., 1990; Gerrard et al., 2011; Yamashita et al., 2010].

5 Conclusions

[31] Assimilated data from MetO has been used to identify synoptic-scale disturbances in the USLM by observing stratopause warmings and constructing a climatology of these disturbances in the polar winter middle atmosphere from a 20.5 year record (1991–2012). USLM events show remarkable consistency in the perturbation structure of temperature, geopotential height, and PV over the course of individual events. The vertical thermal structure through the warm temperature anomaly shows impressive repeatability of a temperature maximum (i.e., stratopause) located near 2 hPa. These unique characteristics of USLM disturbances allow for the development of criteria to isolate these events in the MetO data. USLM disturbances constitute a significant and regular wintertime disturbance with a total of 49 NH and 31 SH USLM events identified in the 20.5 year MetO database, or 2.2 times per winter in the NH (November through March) and 1.6 times per winter in the SH (May through September). An examination of the relationships between criteria used to define USLM disturbances, minor SSWs, and major SSWs indicates that using the USLM criteria provides additional information to identify preconditioning of the atmosphere for major SSW development. All major SSWs are preceded by a USLM event and provide a less dynamically diverse population of events than when using the minor SSW criteria. In the NH, about 33% of USLM disturbances evolve into a major SSW, while 22% of all minor warmings evolve into major SSWs, and USLM events not satisfying the minor warming definition occur more frequently in the SH. The lifecycle of a NH USLM event, using a composite analysis of 16 events, illustrates how the geopotential heights and temperature evolve in an organized manner over an average period of eight days. The evolution, strong baroclinicity, and preferential formation of the warm temperature anomaly in the upper stratosphere on the eastern edge of the polar vortex (between 0oE and 90oE in both hemispheres) are indicative of PWB that may support the development of baroclinic instability.

[32] Large PW1 and PW2 amplitudes occur prior to and during USLM disturbances. During USLM events, PWs exhibit a significant westward phase tilt between the 10 hPa and 2 hPa levels indicative of upward propagation. It is found from PV fields that all USLM events occur simultaneously with breaking PWs in the upper levels of the stratosphere. As the wave breaking cascades to smaller scales and decelerates the Westerlies, a hydrodynamic instability may grow as ageostrophic vertical motion causes adiabatic heating near the 2 hPa level, which increases the horizontal thermal gradient and vertical wind shear. The meridional gradient in quasi-geostrophic PV in the composite analysis of USLM disturbances is dominated by the baroclinic component of the calculation and shows that the Charney-Stern criterion for baroclinic instability is met for the days leading up to the peak in the disturbance, implicating this mechanism as playing a role in the growth of USLM disturbances. USLM events represent regularly occurring disturbances in the polar winter middle atmosphere that may play a key role in the evolution of major SSWs.


Table A1. Start Dates of USLM Events Identified in the MetO Stratospheric Assimilated Data
USLM Event Dates
  • a

    indicates events that evolved into a minor SSW event.

  • b

    Indicates events that evolved into major SSW events.

  • c

    Indicates events used in the composite lifecycle analysis (Figures 6 and 10).

NH    SH 
15-Dec-1991  c 26-Jul-1992 
8-Jan-1992a   3-Sep-1992 
10-Mar-1992ab  28-Sep-1992 
13-Dec-1992    3-Jul-1996 
17-Feb-1993a c 29-Jul-1996 
28-Dec-1994    9-Aug-1996 
24-Jan-1995a   22-Aug-1997 
16-Feb-1996a   16-Aug-2001a
2-Feb-1998a   17-May-2002 
10-Nov-1998    11-Jun-2002 
1-Dec-1998abc 26-Jun-2002 
18-Feb-1999ab  8-Jul-2002a
9-Mar-2000abc 20-Aug-2002a
25-Nov-2000a   13-Sep-2002a
2-Dec-2000a   24-Sep-2002a
26-Jan-2001ab  5-Jun-2004 
16-Feb-2001ab  18-Jul-2004 
18-Dec-2001abc 20-Aug-2004 
18-Jan-2002a   8-Sep-2004 
12-Feb-2002abc 27-Sep-2004a
18-Dec-2002a   5-Jun-2005 
21-Mar-2003a   17-Jul-2005 
3-Dec-2003a   1-Aug-2005 
1-Jan-2004abc 17-Sep-2005a
20-Feb-2005abc 5-Jul-2007 
26-Nov-2005    12-Jul-2007 
6-Dec-2005  c 16-Sep-2007a
19-Dec-2005    25-Sep-2008 
31-Dec-2005abc 28-Jun-2010 
25-Jan-2006    18-Jul-2010 
13-Dec-2006  c 13-Sep-2010 
7-Jan-2007a c   
18-Nov-2009a c   
14-Jan-2010a c   


[33] This work was supported by NSF CEDAR grant AGS-0940174. VLH was supported by the NASA LWS grant NNX10AQ54G, NSF CEDAR AGS grant 0940124, and NSF grant 1107498. We appreciate the BADC for access to the UK MeO Stratospheric assimilated data.