Gravity wave variations during elevated stratopause events using SABER observations

Authors


Abstract

[1] Elevated stratopauses formed at ~80–90 km altitude during the recovery phase of stratospheric sudden warmings in February 2006 and 2009. These likely occurred in response to changes in the downward circulation due to gravity waves (GWs) and/or planetary waves in the mesosphere and the lower thermosphere (MLT). However, the physical mechanisms are not fully understood, due in part to the lack of global GW observations in the MLT. This study presents global-scale GW observations in the MLT during elevated stratopause events using Thermosphere, Ionosphere, Mesosphere Energetics Dynamics (TIMED)-Sounding of the Atmosphere using Broadband Emission Radiometry (SABER) temperature observation, which provide a better insight into the formation of an elevated stratopause. During the downward movement of elevated stratopause events in 2006 and 2009, GWs were suppressed below ~60 km and enhanced above ~60 km at high latitudes compared to non-elevated stratopause years (2005 and 2007). Global SABER GW observations indicate that the regions of GW enhancement propagate from low-mid latitudes to high latitudes in association with the equatorward shift of the polar night jet during elevated stratopause events. Ray-tracing simulations show enhancements of the poleward propagation of GWs during elevated stratopause events as well as continuous propagation of non-orographic GWs within high latitudes. Therefore, our results suggest that meridional propagation of GWs from lower to higher latitudes, which is typically not included in global-scale models, plays an important role in determining GW variations and thus the downward movement of an elevated stratopause, in addition to non-orographic GWs originating at high latitudes.

1 Introduction

[2] A stratospheric sudden warming (SSW) is the most dynamic event to occur in the stratosphere. During SSWs, the temperature increases, and the eastward polar night jet reverses to westward in the winter polar region. The interactions between planetary waves (PW) and the mean flow are believed to be the main driver of SSWs [Matsuno, 1971]. Although SSWs have been considered as events in the winter polar stratosphere, recent studies have shown that SSWs have global impacts from the troposphere to the thermosphere and ionosphere and also affect the summer polar mesopause region [e.g., Baldwin and Dunkerton, 2001; Thompson et al., 2002; Liu and Roble, 2002; Karlsson et al., 2007; Goncharenko and Zhang, 2008]. During the recovery phase of SSWs, enhanced downward transport of chemical species from the mesosphere and the lower thermosphere (MLT) to the stratosphere has been occasionally observed [Randall et al., 2009; Hauchecorne et al., 2007; Orsolini et al., 2010]. Downward transport of NOx is particularly important because it causes ozone depletion in the stratosphere [Hauchecorne et al., 2007; Randall et al., 2006, 2009]. Associated with such downward transport of chemical species, a new stratopause forms in the altitude range of ~80–90 km. The stratopause usually exists at ~50–60 km, and the extreme increase in the stratopause height is often called an “elevated stratopause” [e.g., Manney et al., 2008, 2009; Chandran et al., 2011]. An elevated stratopause likely results from the same downward flow that affects the downward transport of chemical species during the recovery phase of SSWs [Siskind et al., 2007]. Physical mechanisms of anomalies in downward flow have been studied and seem to be related to gravity wave (GW) and/or PW fluxes in the MLT [Siskind et al., 2007, 2010; Chandran et al., 2011; Limpasuvan et al., 2012; Tomikawa et al., 2012; France et al., 2012].

[3] Siskind et al. [2007] proposed a mechanism for the formation of an elevated stratopause related to the filtering of orographic GWs. During the recovery phase of SSWs, zonal winds are still westward in the stratosphere but eastward in the mesosphere because the recovery of the eastward polar night jet occurs in these higher altitudes first. The wind reversal between the troposphere and the mesosphere filters out most orographically forced GWs, which are stationary with respect to the Earth, resulting in suppressions of GWs in the MLT (if only orographic GWs are considered). Due to the lack of GW drag in the MLT, the upwelling vertical circulation becomes weak, resulting in warmer temperature in the MLT compared to temperatures under the normal winter conditions. Based on this theory, Siskind et al. [2007] successfully simulated a realistic structure of an elevated stratopause after the 2006 SSW using the Navy Global Atmospheric Prediction System-Advanced Level Physics-High Altitude (NOGAPS-ALPHA) model with an orographic GW parameterization scheme. However, the magnitude of the elevated stratopause was weaker than that observed by the Sounding of the Atmosphere using Broadband Emission Radiometry (SABER) instrument.

[4] Following Siskind et al. [2007], Siskind et al. [2010] reexamined an elevated stratopause using NOGAPS-ALPHA with both orographic and non-orographic GW parameterizations. The simulations of an elevated stratopause at ~90 km with non-orographic GWs were improved from results only with orographic GWs. Chandran et al. [2011] also simulated an elevated stratopause using the Whole Atmosphere Community Climate Model (WACCM) model with both non-orographic and orographic GW parameterization. Chandran et al. [2011] concluded that non-orographic GW forcing is most likely responsible for the formation of an elevated stratopause. PW contributions to the formation of an elevated stratopause were also discussed by Limpasuvan et al. [2012] and Tomikawa et al. [2012].

[5] In contrast to these various modeling studies of an elevated stratopause, GW observations during elevated stratopause events are still rare. The High Resolution Dynamics Limb Sounder (HIRDLS) on the Aura satellite observed the suppression of GW momentum fluxes in the pressure level range of 100–0.1 hPa (~15–60 km) after the 2006 SSW during a period of an elevated stratopause [Wright et al., 2010]. France et al. [2012] also used HIRDLS and showed altitudinal variations of GW momentum flux from 20 km to 55 km for the 2006 elevated stratopause event. They also showed the suppression of GWs in February in the 20 km to 55 km altitude range. Using the ground-based Rayleigh lidar at Poker Flat Research Range, Alaska (65°N, 147°W), Thurairajah et al. [2010] showed that GW potential energy densities were weaker in the altitude range of 40–50 km during the 2004 elevated stratopause event. In addition to observations, Yamashita et al. [2010a] showed the suppression of GWs in the altitude range of 20–50 km when the 2009 elevated stratopause occurred using the European Centre for Medium-Range Weather Forecasts (ECMWF-T799) high-resolution model that can partially resolve the GW spectrum. Although WACCM and NOGAPS required the non-orographic GW forcing to induce an elevated stratopause [Siskind et al., 2010; Chandran et al., 2011], most observations showed the significant suppression of all GWs in the stratosphere. The main limitation is that the above GW observations were all confined within the stratosphere, missing GW observations in the MLT. The lack of GW observations in the MLT during an elevated stratopause event has prevented us from confirming any of the theories regarding the mechanisms that produce an elevated stratopause.

[6] Recently, owing to the improvements of GW analysis method, the SABER temperatures from the Thermosphere, Ionosphere, Mesosphere Energetics and Dynamics (TIMED) satellite have begun to provide observations of global GW activities with altitude coverage of 30–100 km [Russell et al., 1999; Preusse et al., 2009]. From SABER temperature observations, GW temperature amplitudes, potential energy densities, and absolute momentum fluxes have been derived along with horizontal and vertical wavelengths of GWs [Preusse et al., 2009; Ern et al., 2011; John and Kumar, 2012]. The SABER GWs have also been validated with GWs obtained from the ECMWF-T799 model [Schroeder et al., 2009], HIRDLS [Wright et al., 2011], the Constellation Observing System for Meteorology Ionosphere and Climate (COSMIC)/Global Positioning System (GPS) [Wright et al., 2011], and ground-based lidar observations [John and Kumar, 2012]. The global distributions of monthly mean SABER GW momentum fluxes in the stratosphere was comparable to previous results from the Cryogenic Infrared Spectrometers and Telescopes for the Atmosphere (CRISTA) and HIRDLS by Ern et al. [2004] and Alexander et al. [2008], respectively [Ern et al., 2011]. Furthermore, SABER GW observations revealed the interesting response of GWs to the quasi-biennual oscillation like QBO, semiannual oscillation, and the 11 year solar cycle [Ern et al., 2011; John and Kumar, 2012; Zhang et al., 2012], demonstrating the capability of SABER observation for GW research.

[7] In this study, we use the global SABER observations from 30 to 100 km to study GW variation and their physical mechanisms during elevated stratopause events in 2006 and 2009 and compare these with the non-elevated stratopause years (2005 and 2007). In order to use SABER GWs for this study, we further validate GWs derived with our SABER analysis method against COSMIC/GPS observations and examine tidal influence in GW analysis. The Gravity-Wave Regional or Global Ray Tracer (GROGRAT) ray-tracing model [Marks and Eckermann, 1995] is also used to understand physical mechanisms of GW behavior associated with an elevated stratopause.

[8] This paper consists of five sections. Section 2 describes the SABER and COSMIC/GPS data and GW analysis method. The validation of SABER GWs, including comparisons with COSMIC/GPS, SABER tidal amplitudes, and the climatology of GWs, are summarized in section 3. Section 4 presents the GW variations associated with an elevated stratopause. The physical mechanisms behind the observed GW variations are investigated using the Modern-Era Retrospective Analysis for Research and Applications (MERRA) reanalysis output [Rienecker et al., 2011] and the GROGRAT ray-tracing model. These are also included in section 4. Finally, the conclusions and outlook are summarized in section 5.

2 Data and Method

2.1 SABER

[9] The SABER instrument on the TIMED satellite has provided atmospheric temperature and constituent data from ~20 km to ~110 km since 2002 [Russell et al., 1999]. In this study, we used version 1.07 level 2A data. Due to the yaw cycle of the SABER instrument, the observational coverage shifts from 50°S–82°N to 82°S–50°N every 60 days. The SABER temperature retrieval is discussed in Remsberg et al. [2008]. SABER is a limb-sounding infrared radiometer, and previous studies have shown that a limb sounder can observe GWs with horizontal and vertical wavelengths (λz) longer than ~100–200 km and ~4 km, respectively [Preusse et al., 2002].

2.2 Gravity Wave Analysis

[10] In this study, GWs are analyzed with the similar method introduced by Preusse et al. [2002, 2009] and Fetzer and Gille [1994]. The method used here is summarized as the following. First, observed daily temperature profiles (T) are binned into 24° × 5° (longitude × latitude) horizontal grids and 1 km vertical grid. Second, the zonal wave numbers 1–5 components of temperature are estimated by least-squares fitting, and the sum of zonal mean temperature (wave number 0) and zonal wave numbers 1–5 components are considered as “background temperatures (inline image).” For the SABER data, the above method is applied separately to ascending and descending nodes in order to separate temperature profiles dependent upon the local time [Preusse et al., 2001, 2009]. By fitting components of zonal wave numbers 0–5 at fixed local time, all tidal components are aliased into wave numbers 0–5 and so can be extracted along with stationary PWs, even though specific types of tides and PWs cannot be distinguished. Therefore, “background temperatures (inline image)” estimated by the above method should contain zonal mean temperature, stationary PWs, and tidal waves. Third, the residual temperature perturbations (T') are calculated by subtracting background temperature from observed temperature (i.e., inline image), and T' are considered as temperature perturbations induced by GWs in this study. Finally, in order to remove noise and isolate wave-like features in our analysis, we perform a spectral analysis on each vertical temperature profile. Using a wavelet analysis, we determine up to three dominant vertical wavelengths (λz) at each altitude. Following Preusse et al. [2002], the spectrum is detrended using a quadratic fit, and only the spectrum peaks whose amplitudes are above twice the standard deviation are selected as dominant λz. Then, waves are reconstructed using three dominant λz by harmonic fitting with a 20 km sliding window. This method is similar to the maximum entropy method and harmonic analysis used in Preusse et al. [2002, 2009]. Here we only consider GWs with λz less than 30 km. GW amplitudes shown in this study are total GW amplitudes of three dominant wave components.

[11] An example of the above method is shown in Figure 1. Figure 1 shows vertical profiles of the observed temperature (T), estimated background temperature (inline image), and residual temperature perturbation (inline image) from SABER. Background temperatures estimated by different zonal wave number components, in ranges of 0–1, 0–2, 0–3, 0–4, 0–5, and 0–6 are shown in Figure 1a along with the observed temperature. All estimated background temperatures capture general structures of the original temperature profile, and estimations of background temperatures improve, as higher order components are included. The estimated background temperatures with wave numbers 0–5 and 0–6 components show no significant difference. Hence, we use zonal wave numbers 0–5 components for background estimation in this study, which accounts for most large-scale features (zonal mean, tides, and PWs). Figure 1b shows a residual temperature perturbation, which is the difference between an observed temperature profile and an estimated background temperature profile (zonal wave numbers 0–5 components). The sum of reconstructed waves based on three dominant λz is shown in red. Although detailed features require the full spectrum of the GW to be captured, most of the features in the residual temperature perturbations are reconstructed with the three large components of GW.

Figure 1.

(a) (black) A raw temperature profile and (colors) estimated background temperatures with different zonal components of large-scale features at location of 44°S and 23°E on 30 January 2009. (b) (black) A temperature perturbation (difference between a raw temperature profile and an estimated background temperature profile with zonal wave numbers 0–5) and (red) reconstructed waves based on the three dominant wavelengths.

[12] We also analyze SABER GWs without separating ascending and descending nodes (i.e., no local time separation) to examine the tidal influence in GW analysis. In the following section, GWs analyzed by separating descending and ascending nodes is called “SABER-LT,” while the nodes are not separated in the analysis called “SABER-no-LT.”

2.3 COSMIC

[13] To validate SABER GWs, we also use temperature data from COSMIC/GPS in this study. The COSMIC/FORMOSAT-3 constellation was launched in April 2006 [Rocken et al., 2000]. COSMIC/GPS radio occultation provides global coverage of temperature profiles from near the surface to ~60 km. Because of noise introduced by ionospheric variability, we can only use temperature data below ~35–40 km. For validation purpose, COSMIC data should be analyzed using the same method introduced above for SABER. For SABER data, global temperature maps for the estimation of zonal wave numbers 0–5 components can easily be generated at constant local time through separation of the ascending and descending nodes. This causes the tides to be aliased into the wave numbers 0–5 components, thus allowing their removal. Unlike SABER, COSMIC/GPS data are pseudo-randomly distributed and do not provide sufficient data points to produce the global temperature map at constant local time within 1 day. Therefore, in this COSMIC analysis, background temperatures (zonal wave numbers 0–5) are estimated using all data points in 1 day without local time separation. The difference in analysis methods regarding the local time separation is discussed in the following section.

3 Validations of SABER Gravity Waves

[14] Several different methods have been used to analyze GWs from satellite temperature observations. The most common method is using a high-pass filter to separate the GW temperature perturbation from the large-scale background [e.g., Tsuda et al., 2000]. Basically, a high-pass filter with cutoff λz of ~7–10 km is applied on each temperature profile, and the remaining perturbations with λz shorter than ~7–10 km are considered to be GWs [e.g., Tsuda et al., 2000]. Several problems exist with the high-pass filter method. First, previous studies showed that equatorial Kelvin waves can have λz of ~3–4.5 km, which can incorrectly appear as GWs in this analysis [Holton et al., 2001]. Second, the high-pass filter method eliminates information about GWs with λz longer than ~7–10 km that are likely important in the MLT region. To avoid the above problems, a method of removing large-scale features that uses global temperature observations has been used in various studies [e.g., Alexander et al., 2009; Preusse et al., 2009] and is also implemented here. This method allows us to obtain longer λz and can avoid Kelvin-wave contaminations. However, the method of removing horizontal large-scale feature may be adversely affected by the presence of atmospheric tidal influences in the temperature data set. In this section, we further validate our SABER GW analysis against COSMIC/GPS and SABER tidal amplitudes to examine tidal influences. Also, we validate the climatology of SABER GWs.

3.1 Comparison Between SABER and COSMIC/GPS Gravity Waves

[15] Wright et al. [2011] have compared SABER GWs with COSMIC/GPS and HIRDLS GWs. They focused on the comparisons between co-located temperature profiles within 900 s and 180 km range from different instrument. They showed that individual profiles of GWs show good agreement between SABER, COSMIC, and HIRDLS. Here, we extend their validations to the monthly mean GW at 30 km in July (solstice) and October (equinox) in 2009. Figure 2 shows the global distributions of GW amplitudes. The unit of GW amplitudes is decibel (dB) as calculated by GW Amplitude(dB) = 10 ⋅ log 10(T ' 2), where T' is the GW amplitude in temperature. This unit is chosen for comparisons with the global GW map presented by Preusse et al. [2009]. As discussed in section 2, SABER GWs are analyzed using two different methods, SABER-LT and SABER-no-LT [SABER-LT separates ascending and descending nodes (depending on local time), while SABER-no-LT does not separate ascending and descending nodes (combining 12 h separated local times)]. In July (Figures 2a, 2c, and 2e), large GW amplitudes are observed in the winter high latitudes of ~50°S–80°S and the tropical regions of 10°N–30°N. Such general features were also observed by the Microwave Limb Sounder, CRISTA, and HIRDLS instruments [e.g., Ern et al., 2004; Wu and Eckermann, 2008; Alexander et al., 2008; Preusse et al., 2009]. Similar to July, the general features of GWs in October (Figures 2b, 2d, and 2f) are captured by both SABER and COSMIC including relatively large GW amplitudes at 40°S–50°S and 60°W–60°E and at ~50°N and 80°E–120°E. Magnitudes and global distributions of SABER-LT GWs in Figure 2 compare favorably with the previous SABER GW results from Preusse et al. [2009]. In Figure 2, COSMIC GWs correspond better with SABER-no-LT than SABER-LT in terms of global distributions of GW amplitudes. For example, SABER-LT GWs show no significant activities in the tropical region in October (Figure 2b); however, SABER-no-LT and COSMIC GWs show larger amplitudes in the latitude range of 30°N–30°S.

Figure 2.

Monthly mean GW amplitude at 30 km in (left) July and (right) October 2009. White contour lines indicate continents. SABER-LT and SABER-no-LT indicate different GW analysis method. SABER-LT separates ascending and descending nodes (depends on local time), while SABER-no-LT does not separate ascending and descending nodes (no local time separation). In October, there were no SABER observations in 50°S–80°S due to yaw cycle.

[16] To elucidate differences amongst the three methods, Figure 3 shows the line plots of the zonal-monthly mean GW amplitudes at 30 km obtained from COSMIC (red), SABER-LT (black), and SABER-no-LT (blue) method. In general, GW amplitudes from COSMIC and SABER-no-LT show good agreements. SABER-LT GW amplitudes are comparable in midlatitudes but obviously weaker in low latitudes (~20°S–20°N) compared to SABER-no-LT and COSMIC. In contrast to low-mid latitudes, SABER-LT and SABER-no-LT at high latitudes show larger amplitudes than COSMIC in January and July.

Figure 3.

Zonal-monthly mean GW temperature amplitudes for (a) January, (b) April, (c) July, and (d) October 2009 at 30 km using (black) SABER-LT, (blue) SABER-no-LT, and (red) COSMIC methods.

[17] For a quantitative comparison, correlation coefficients between COSMIC, SABER-LT, and SABER-no-LT GW amplitudes at 30 km are reported in Table 1. To determine those values, monthly GW amplitudes are binned to 20° × 10° (longitude × latitude) grids, where the latitude range of GW amplitudes used to calculate correlation coefficients varies by months due to SABER yaw cycle between 80°S and 80°N. For example, in October, only GW amplitudes in 50°S–80°N are used, but in July, GWs in 80°S–80°N are used to calculate correlation coefficients. The significance levels of all correlation coefficients in Table 1 are >95%. The mean correlation coefficient between COSMIC and SABER-LT from January to December 2009 is 0.76 with a range of 0.52–0.93. The mean correlation coefficient increases to 0.81 between COSMIC and SABER-no-LT.

Table 1. Correlation Coefficients Between Monthly Mean COSMIC and SABER GWs and Differences (%) in GW Amplitudes Between COSMIC GWs and SABER GWs in 2009
 CorrelationsDifferences in Amplitudes [(SABER-COSMIC)/COSMIC]
SABER-LTSABER-no-LTSABER-LTSABER-no-LT
January0.850.88−1.54%12.02%
February0.720.74−13.74%0.76%
March0.660.71−14.05%−3.01%
April0.520.75−16.00%−0.81%
May0.550.53−16.40%−8.51%
June0.860.87−10.51%4.08%
July0.930.96−9.85%4.86%
August0.850.92−4.35%12.50%
September0.830.89−10.22%3.35%
October0.750.77−13.13%−1.14%
November0.720.79−6.67%5.88%
December0.850.86−11.65%5.51%
Average0.760.81−10.68%2.96%

[18] Differences in GW amplitudes are also calculated and shown in Table 1. Differences, (SABER GW − COSMIC GW)/COSMIC GW, are calculated at each grid point and then averaged over the entire globe at 30 km. Differences are averaged over January–December, and results are 2.96% (range of −8.51% to +12.50%) between COSMIC and SABER-no-LT but −10.68% (range of −16.40% to −1.54%) between COSMIC and SABER-LT, indicating that COSMIC compares well with SABER-no-LT than SABER-LT. The differences are all negative between COSMIC and SABER-LT from January to December.

[19] At low latitudes, tidal amplitudes are relatively large because of the latent heat release and IR heating of water vapor in the tropical region. Because the SABER-no-LT and COSMIC analysis methods do not perform tidal separation via constant local time sampling, the high correlation between COSMIC and SABER-no-LT, as well as their higher GW amplitudes at low latitudes, indicates that COSMIC and SABER-no-LT GWs might contain tidal amplitudes. On the other hand, at high latitudes (55°N–85°N and 55°S–85°S) in January and July, amplitudes of SABER-LT GWs are higher than COSMIC GWs by an average of ~13.5% (not shown in table). It should be noted here that the differences in GW amplitudes at both low and high latitudes between SABER and COSMIC can also be caused by differences in the average direction of the line of sight, filtering for ionospheric contamination, and sampling.

3.2 Tidal Influences on Gravity Wave Analysis

[20] Because tidal amplitudes in the MLT region exceed 10–20 K, it is important to further examine potential tidal effects in GW retrievals in the MLT by comparing actual tidal and GW amplitudes. SABER tidal amplitudes are fitted with a linear least-squares algorithm using a 60 day window sliding every 15 days from 2003 to 2009, allowing for full local time coverage and tidal retrieval [e.g., Forbes et al., 2006]. Diurnal and semidiurnal components with westward and eastward wave numbers from 1 to 4, as well as the wave number 0 standing components, are retrieved simultaneously using this method. GW amplitudes are also averaged over 60 days sliding every 15 days to match with the tidal analysis. Figure 4 shows the maximum correlation coefficients between time series of 60 day zonal mean GW amplitudes and tidal amplitudes from January 2003 to December 2009 at each latitude and altitude. The maximum correlation coefficients are selected from correlation coefficients between GWs and individual tidal components (18 components, diurnal and semidiurnal, and wave numbers 0–4). Amplitudes from 2003 to 2009 every 15 days (175 data points) are used to estimate these correlation coefficients. If GW amplitudes contain tidal influences, correlation coefficients are expected to be high, especially in the MLT where tidal amplitudes are large. Figure 4a shows that correlation coefficients between SABER-LT GWs and tides are less than ~0.5 above ~80 km from low to high latitudes. SABER-LT GWs show relatively independent variations and no significant relationship with tides. On the other hand, correlation coefficients between SABER-no-LT GW and tides are over 0.8 in low-latitude mesosphere, indicating heavy tidal influences in GW amplitudes.

Figure 4.

Maximum correlation coefficients between the time series of tidal amplitudes and (a) SABER-LT GWs and (b) SABER-no-LT GWs from January 2003 to December 2009. Correlations are calculated between amplitudes of GWs and each tidal component; then maximum correlation coefficients at each location are shown here. Numbers (1, 2, and 3) mark the locations of line plots shown in Figure 5.

[21] Although correlation coefficients are small in the MLT, the correlations between tides and SABER-LT GWs are large at 35°S–50°S and 35°N–50°N in the stratosphere. To examine the causes of the high correlation, Figure 5 shows the time series of GW and tidal amplitudes at three locations shown in Figures 4a and 4b. Tidal components of semidiurnal westward propagating wave number 3 (SW3), diurnal eastward propagating wave number 2 (DE2), diurnal eastward propagating wave number 3 (DE3), and diurnal westward propagating wave number 1 (DW1, migrating tides) are shown in Figure 5, because these tidal components show the maximum correlation with GW amplitudes at each location. Figures 5a and 5b show the time series of GW (SABER-LT and SABER-no-LT) and SW3 at 45°S and 35 km. Because SABER-LT and SABER-no-LT GWs show similar variability, the cause of the high correlation appears to be that both GW and tides have the same seasonal variations, maximum in winter and minimum in summer. Such a seasonal variation of GWs has been observed by satellite and ground-based observations at high latitudes. Causes of the above seasonal variations are likely related to generation of GWs and wind filtering in the polar vortex edge, and it is not due to tidal contamination in GWs [e.g., Wu and Eckermann, 2008; Preusse et al., 2009; Yamashita et al., 2009]. Although seasonal variations are similar, the small fluctuations in the GWs do not follow the tidal amplitudes. For example, tidal amplitudes show two peaks in the middle of 2009 in Figures 5a and 5b, but GW amplitudes do not follow this tidal variation. Therefore, the high correlations between tides and SABER-LT GWs are not likely due to tidal contamination in GWs but similar seasonal variations.

Figure 5.

Line plots of the zonal 60 day mean GW amplitudes using (left) SABER-LT and (right) SABER-no-LT methods. (black) Tidal amplitudes at (a, b) 45°S at 35 km, (c, d) 20°N at 90 km, and (e, f) 5°S at 85 km as indicated in Figure 4a. Amplitudes of GWs and tides are normalized to be in the range of 0–1. Tidal amplitudes of SW3 (in Figures 5a and 5b), DE2 (in Figures 5c and 5d), DE3 (in Figure 5e), and DW1 (in Figure 5f) are selected because these components show the maximum correlation with GW amplitudes at each location.

[22] Figures 5c and 5d show amplitudes of GWs and DE2 in the tropical MLT. Although correlation coefficient between SABER-LT GWs and DE2 is ~0.55, SABER GWs do not follow detailed DE2 amplitude variations. Both SABER-LT and SABER-no-LT GWs show larger amplitudes in summer, and so do tides. Similar seasonal variations seem to cause correlation coefficients of ~0.55. In addition, because GW variations between SABER-LT and SABER-no-LT are slightly different, local time separation seems to reduce some tidal influences at this location.

[23] In Figure 5f, SABER-no-LT GWs clearly follow amplitudes of DW1 migrating tides at 5°S and 85 km, but in Figure 5e, the maximum correlation is not between SABER-LT GW and DW1 but SABER-LT GW and DE3. These results indicate that migrating tides (DW1) influences are removed from SABER-LT GWs, and also there is no significant correlation between GWs and nonmigrating tides (DE3). DW1 migrating tides are expected to be large in the tropical mesosphere, and our GW analysis method using local time separation is especially useful for removing migrating tidal influences [Preusse et al., 2001].

[24] Thus, correlations between tides and GWs may be caused either by tidal contamination of the GW results or because a process such as filtering or modulation by the global-scale background wind can cause similar annual cycles. The example shown in Figures 5e and 5f demonstrates that the local time separation used here is effective in the removal of both migrating and nonmigrating tides. On the other hand, such local time separation did not have an influence at all (e.g., Figures 5a and 5b); this would indicate that the seasonal variations in each must have a common cause, a conclusion supported by comparison of the size of the respective amplitudes. Comparisons between GWs and tides in Figures 4 and 5 show that SABER-LT GWs do not contain obvious tidal influences. Therefore, the rest of this study uses SABER-LT GWs (hereafter referred to as GWs).

[25] Another possible problem in GW analysis in the MLT region is the influences of the sharp mesopause. The sharp mesopause can resemble wave structure and affect GW analysis in MLT. We compared the altitude variations of GW amplitudes and the height of the mesopause (not shown). GW amplitudes continuously grow with altitudes, and no significant increase in GW amplitudes was observed where a mesopause exists. Furthermore, it should be noted here that this GW analysis method can be still affected by PWs, in particular, quasi 2 day waves as pointed out by Ern et al. [2011]. However, our method removes any perturbations that have λz longer than 30 km and so reduces quasi 2 day wave effects [Ern et al., 2011].

3.3 Climatology of SABER Gravity Waves

[26] Finally, before examining GW variations during an elevated stratopause event, the climatologies of GWs are analyzed and compared with previous studies. Figure 6 shows the monthly composite climatology of GW amplitudes from 2002 to 2011 in January and July. Upper Atmosphere Research Satellite Reference Atmosphere Project (URAP) climatology winds are overlaid in Figure 6. URAP winds are estimated by combining the High Resolution Doppler Imager wind in the mesosphere and the UK Meteorological Office assimilation data in the stratosphere and below. Both total amplitudes and those separated by λz are shown in Figure 6. In Figure 6, both in July and January, GWs are larger in winter high latitudes where the polar night jet exists. This is because, in the winter stratosphere, the wind direction is almost always eastward, allowing westward GWs to propagate upward. On the other hand, the wind direction reverses from the troposphere to the stratosphere in the summer hemisphere. Such a wind reversal in summer filters both westward and eastward GWs, resulting in less GW activity in the middle atmosphere.

Figure 6.

Composite monthly zonal mean GWs in (top) January and (bottom) July from 2003 to 2009. Colored contours and white line contours show (a, d) total GWs, (b, e) GWs with 4 km < λz < 15 km, and (c, f) GWs with 15 km < λz < 30 km. Black line contours show Upper Atmosphere Research Satellite Reference Atmosphere Project (URAP) zonal wind (see text for details). Solid/dashed lines indicate eastward/westward winds, respectively. Red thick lines are the zero-wind line of the zonal mean zonal winds.

[27] The Doppler shift of the vertical wavelength can also cause the large GWs when the background wind speed is large [e.g., Alexander, 1998; Preusse et al., 2006; Wu and Eckermann, 2008; Ern et al., 2011]. GWs with long λz tend to be more easily observed by satellite limb observations, similar to those used here, resulting in increased GW activity observed where the background wind is large. Figures 6c and 6f show that the amplitudes of GWs with λz longer than 15 km are larger than those with shorter λz in the winter polar stratosphere. The climatology of GWs presented here compares well with the previous results [e.g., Alexander, 1998; Preusse et al., 2006, 2009; Wu and Eckermann, 2008; Ern et al., 2011], and it confirms that our SABER GWs capture general features of GWs, thus giving us confidence in using them to investigate GWs during elevated stratopause events.

4 Gravity Wave Variations During Elevated Stratopause Events

4.1 Background Temperature and Wind

[28] After validations of SABER GWs in section 3, we now turn our attention to GW variations associated with an elevated stratopause. Figure 7 shows daily zonal mean SABER temperatures in 2005, 2006, 2007, and 2009 from 15 January to 28 February. In this study, we select 2 years without elevated stratopause event (2005 and 2007) and 2 years with elevated stratopause events (2006 and 2009). In Figures 7d (2006) and 7j (2009), the warm temperature layers (stratopauses) exist at ~40–60 km prior to ~25 January. Then, stratopauses jump up to ~70–100 km around 30 January. The elevated stratopauses then descend to ~60 km by the end of February. In comparison with elevated stratopause events in 2006 and 2009, the temperature structures in 2005 and 2007 (Figures 7a and 7g) are relatively stable without any SSW and elevated stratopause events. The zonal-daily mean MERRA temperature and zonal wind data are also plotted in Figure 7. The differences in zonal winds between non-elevated stratopause and elevated stratopause events are clearly shown in Figures 7c, 7f, 7i, and 7j. Zonal winds in 2006 and 2009 are eastward at the middle of January and then reversed to westward at the end of January. In 2006 and 2009, although recovery of the eastward jet starts in the 50–60 km altitude range at the end of January, zonal winds in the 20–30 km altitude range remain westward until the end of February. On the other hand, zonal winds are almost always eastward in 2005 and 2007 until the end of February.

Figure 7.

The zonal-daily mean of (left) SABER temperature, (middle) MERRA temperature, (right) MERRA zonal wind averaged over the latitude range of 65°N–85°N in (top row) 2005, (second row) 2006, (third row) 2007, and (bottom row) 2009. White lines in left plots represent the upper boundary of MERRA reanalysis data, and white lines in middle and right plots indicate the lower boundary of SABER GW analysis. Red thick lines are the zero-wind line of the zonal mean zonal winds.

4.2 Gravity Wave Variations

[29] Figure 8 illustrates zonal-daily mean GW amplitudes averaged over the latitude range of 55°N–75°N in 2005, 2006, 2007, and 2009. MERRA zonal winds are overlaid below 60 km as red line contours. After 30 January, GW amplitudes show a significant difference between non-elevated stratopause (2005 and 2007) years and elevated stratopause (2006 and 2009) years. GWs in 2006 and 2009 in the altitude range of 30–60 km are considerably weaker than those in 2005 and 2007, consistent with the previous observations by lidar and HIRDLS [Wright et al., 2010; Thurairajah et al., 2010; France et al., 2012]. One interesting result here is that GWs in the altitude range of ~70–80 km during mid-late February are larger during elevated stratopause years than those during non-elevated stratopause years. Regions of large GW amplitudes above 80 km on 25 January gradually descend to ~60–70 km on 28 February in 2006 and 2009. According to previous modeling studies [Siskind et al., 2010; Chandran et al., 2011; Limpasuvan et al., 2012; Tomikawa et al., 2012; Chandran et al., 2013], combinations of in situ generated PWs and GWs are important for the onset of an elevated stratopause formation (just after SSWs), while the subsequent downward movement of elevated stratopauses is mainly driven by GWs.

Figure 8.

The zonal-daily mean of GW amplitudes (K) averaged over 55°N–75°N in (a) 2005, (b) 2006, (c) 2007, and (d) 2009. The red line contours show MERRA zonal wind. Solid/dashed lines indicate eastward/westward winds, respectively. A 3 day running mean is applied on GW amplitudes. Blank region indicates missing GW data.

[30] To study GW variations during different phases of elevated stratopause events, the GW variations are separated into two different periods, which are 1–10 February and 10–19 February. Figure 9 shows latitudinal variations of GW amplitudes for 1–10 February and 10–19 February at two altitude levels, 40 km and 80 km. During the onset of an elevated stratopause on 1–10 February, amplitudes of GWs at 40 km are weaker in elevated stratopause years (2006 and 2009) than those in non-elevated stratopause years (2005 and 2007) at 40°N–80°N, but there is no clear difference at 80 km. On 10–19 February at 40°N–80°N, GWs at 40 km at are much lower in amplitude in 2006 and 2009 than those in 2005 and 2007, while GW amplitudes are larger at 80 km in elevated stratopause years than those in non-elevated stratopause years at 40°N–80°N. GWs on 10–19 February at 80 km peak at high latitudes but decrease at low latitudes (0°N–40°N). Using our analysis, there are no significant differences between GWs in elevated stratopause years and non-elevated stratopause years in the Southern Hemisphere.

Figure 9.

The zonal-daily mean GW amplitudes at (top, a, c) 40 km and (bottom, b, d) 80 km averaged over (left) 1–10 February and (right) 10–19 February. Each color represents a different year (black: 2005, blue: 2006, green: 2007, and red: 2009).

[31] Reductions of GW activities in the stratosphere after SSWs (1–10 February at 40 km) have been observed [e.g., Wright et al., 2010; Yamashita et al., 2010a; France et al., 2012], and their causes are likely the filtering of GWs [e.g., Thurairajah et al., 2010]. However, the enhancements of GWs in the MLT during downward movement of elevated stratopauses (10–19 February) are a new observational result, and their physical mechanisms are not yet clear. To gain a better insight into GW variations on 10–19 February, Figure 10 shows the latitudinal variations of GW amplitudes that are averaged over 10–19 February overlaid with zonal mean zonal wind from MERRA below 60 km. In the winter hemisphere, enhancements of GWs in 2005 and 2007 exist at high latitude regions from 30 km to ~80 km, while GWs in 2006 and 2009 show that peaks of GWs propagate from midlatitudes to the high latitudes as shown by the black arrows in Figure 10. Such GW variation corresponds well with the strength of background winds. In 2006 and 2009, the polar night jet shifted toward the equator and tilted meridionally, which was also reflected in the GWs. The meridional tilting of GW enhancements can be explained by a combination of the background wind conditions and critical layer filtering. In 2006 and 2009, MERRA zonal winds remain westward, poleward of 50°N below 40 km, preventing GWs with slow phase speeds from propagating from high latitudes. In contrast, in 2005 and 2007, orographic GWs can propagate from high latitudes without encountering wind reversals.

Figure 10.

Contour plots of GW amplitudes averaged over 10–19 February in (a) 2005, (b) 2006, (c) 2007, and (d) 2009. Red line contours indicate the zonal mean zonal wind from MERRA. The solid/dashed zonal winds represent the westward/eastward direction, respectively. The black arrows indicate potential GW propagation direction (see text for details).

[32] Sato [2000] proposed that GWs excited in low-mid latitudes could propagate to high-latitude MLT regions in addition to GW propagation within high latitudes. It is possible that meridional propagation of GWs during elevated stratopause events differ from those during non-elevated stratopause events. In addition to meridional propagation of GWs, in situ generation of GWs above ~60 km can result in the enhancements of GWs in MLT region [Wang et al., 2006]. However, in situ generation of GWs does not likely explain the meridional tilting of GW enhancements shown in Figure 10. Hence, we hypothesize that the suppressions of GWs below ~60 km and enhancements of GWs above ~60 km at high latitudes may result from the enhancement of poleward propagation of GWs from midlatitudes along with the suppression of propagation of orographic GWs in high latitudes.

4.3 Ray Tracing of Gravity Waves

[33] To test our hypothesis, we conducted the ray tracing of GWs using the Gravity-Wave Regional or Global Ray Tracer (GROGRAT) ray-tracing model for the periods of 1–10 and 10–19 February. The GROGRAT model is a four-dimensional global ray-tracing model, and the detailed descriptions of GROGRAT model can be found in Marks and Eckermann [1995] and Eckermann and Marks [1997]. The GROGRAT model has been used extensively to study GW propagations [e.g., Preusse et al., 2009]. The six-hourly MERRA outputs of temperature, zonal wind, and meridional wind are used as the background conditions in the GROGRAT model.

[34] GWs are launched every 5° in latitude and every 30° in longitude from the latitude range of 2.5°N–87.25°N. GW parameters used for the GROGRAT ray tracing are horizontal wavelengths of 100, 200, 400, 600, 800, 1000, and 1200 km and ground-based horizontal phase speeds of 0 m/s, 20 m/s, and 40 m/s. GW propagation directions are homogeneously distributed between 0° and 360° for every 45°. The above GW parameters are chosen to cover most GW spectrum used in GW parameterization scheme and also observed before [e.g., Liu and Roble, 2002; Yamashita et al., 2010b; Chandran et al., 2011; Ern et al., 2011]. GWs are launched on 1 and 10 February each year and freely propagate until they reach 60 km (upper boundary of MERRA wind) or a critical layer, or dissipate. Figures 11 and 12 show GWs launched on 10 February at high latitudes (45°N–90°N) and low-mid latitudes (0°N–45°N), respectively. Figures 11 and 12 only show ray paths of GWs reaching 60 km in the latitude range of 40°N–90°N. Ray paths are separated based on their input ground-based horizontal phase speeds of 0 m/s (Figures 11a, 11d, 12a, and 12d), 20 m/s (Figures 11b, 11e, 12b, and 12e), 40 m/s (Figures 11c, 11f, 12c, and 12f), and the horizontal wavelengths of GWs.

Figure 11.

GW ray paths obtained from GROGRAT ray-tracing model for (top) 2005 and 2007 and (bottom) 2006 and 2009. Input GW ground-based phase speeds are (a, d) 0 m/s, (b, d) 20 m/s, and (c, f) 40 m/s. GWs are launched from the latitude range of 45°N–90°N. GWs with horizontal wavelength of 100 km, 200 km, 400 km, 600 km, 800 km, 1000 km, and 1200 km are indicated as black, dark blue, blue, green, light green, orange, and red, respectively. Total numbers of waves reaching 60 km in the latitude range of 40°N–90°N are shown in each figure as “Total #waves.” “Each #waves” separates numbers depending on their horizontal wavelength (black: 100 km, dark blue: 200 km, blue: 400 km, green: 600 km, light green: 800 km, orange: 1000 km, and red: 1200 km).

Figure 12.

Same as Figure 11 but for GWs launched from 0°N to 45°N instead of 45°N–90°N.

[35] For GWs with a phase speed of 0 m/s (orographic GWs), the number of waves reaching 60 km shows significant differences between years with and without an elevated stratopause. In 2005 and 2007, a total of 646 waves reach 60 km, but only 42 waves reach that altitude in 2006 and 2009. Total number of orographic GWs reaching 60 km in 2006 and 2009 is 93% less than those in 2005 and 2007. This is because wind reversal between the troposphere and the stratosphere at 55°N–80°N filters orographic GWs in 2006 and 2009 (see wind structure in Figure 10). Total numbers of GWs with phase speeds of 20 and 40 m/s (non-orographic GWs) in 2006 and 2009 reaching 60 km are 47% and 21% less than those in 2005 and 2007, respectively. These results indicate that during elevated stratopause events, most orographic GWs are filtered by the wind reversal between the troposphere and the stratosphere, but non-orographic GWs are able to penetrate to the high-latitude mesosphere, consistent with previous simulation results by Chandran et al. [2011] and Siskind et al. [2010].

[36] Figure 12 shows ray paths of GWs launched on 10 February at low-mid latitudes but reaching the high-latitude mesosphere region (40°N–90°N at 60 km). Although the total number of waves reaching the high-latitude mesosphere from low-mid latitudes is smaller than those launched at high latitudes, GROGRAT indicates that GWs are able to propagate from low-mid latitudes to high latitudes for all four cases. The interesting result is that the numbers of waves reaching high latitudes from the low-mid latitude troposphere in elevated stratopause years are larger than those in non-elevated years by 90%, 41%, and 26% for horizontal phase speeds of 0 m/s, 20 m/s, and 40 m/s, respectively. The ray paths of GWs during elevated stratopause events shown in Figure 12 correspond well with the meridional tilt of GW enhancements and the black arrows in Figure 10. GROGRAT results indicate the enhancements of poleward propagation of GWs during elevated stratopause events. Moreover, meridional propagation of GWs from low-mid to high latitudes explains the increases and decreases in GW amplitudes at 40°N–80°N and at 0°N–40°N, respectively, in 2006 and 2009, shown in Figure 9d.

[37] To study GW propagation during the onset of elevated stratopauses, additional GRORAT ray-tracing simulations are conducted by launching GWs on 1 February. During the onset of an elevated stratopause, the numbers of waves reaching high latitudes from the low-mid latitude troposphere in elevated stratopause years are only changed by −3%, −4%, and −19% for horizontal phase speeds of 0 m/s, 20 m/s, and 40 m/s, respectively. Comparing to the results on 10–19 February, there is no significant change in meridional propagation of GWs during the onset of an elevated stratopause. In contrast, GW propagation within high latitudes change by −82%, −61%, and −37% for horizontal phase speeds of 0 m/s, 20 m/s, and 40 m/s, respectively. Therefore, our GROGRAT simulations indicate that the suppressions of GWs at 40 km on both 1–10 and 10–19 February shown in Figure 9 are likely due to reductions in the GW propagation at high latitudes. Meridional poleward propagation increases for GWs launched on 10 February but not 1 February. Similarly, GWs in the MLT are enhanced on 10–19 February but not on 1–10 February (Figures 9b and 9d). These results suggest that the enhancements of GWs in the MLT on 10–19 February are likely induced by the increases in GW meridional propagation from low-mid to high latitudes.

4.4 Planetary Wave Contribution and Discussion

[38] In addition to GWs, previous modeling studies showed that PWs can contribute to the formation of an elevated stratopause [e.g., Limpasuvan et al., 2012, Tomikawa et al., 2012]. To investigate PW behavior, Eliassen-Palm (EP) flux (Fz, FΦ) and its divergence are calculated based on Andrews et al. [1987] using MERRA reanalysis data. Figure 13 shows the EP flux divergence and EP flux vectors on 23–25 January and 10–19 February. As with the results of Tomikawa et al. [2012], our results also show the positive EP flux divergence in the high-latitude upper stratosphere (~50 km and 60°N–80°N) just after SSWs on 23–25 January, indicating the in situ generation of PWs in the upper stratosphere. On 10–19 February, the positive EP flux divergences are weaker than those on 23–25 January, and thus, in situ generation of PWs does not clearly exist in the stratosphere from MERRA data during the downward movement of elevated stratopauses.

Figure 13.

(color contours) EP flux divergence and (red arrows) EP flux vector averaged over (top) 23–25 January and (bottom) 10–19 February for (a, c) 2006 and (b, d) 2009 obtained from MERRA. Red arrows are normalized to be the same length. The positive/negative numbers represent eastward/westward, respectively. The vertical component of the EP flux is multiplied by a viewing factor of 100.

[39] Figure 14 shows amplitudes of EP flux (|Fz| + |FΦ|), EP flux vectors (propagation directions of PWs), and EP flux divergences. Magnitudes of EP fluxes in the altitude range of ~30–50 km are significantly larger in 2005 and 2007 than those in 2006 and 2009. The magnitudes of EP flux and red arrows in Figure 14 indicate that PWs originating from the troposphere are weaker in the stratosphere and the mesosphere during elevated stratopause events (10–19 February) in 2006 and 2009 than those in non-elevated stratopause years. Divergences of EP flux (line contours) indicate that westward PW forcing in ~30–50 km in non-elevated stratopause years are about 2 times larger than that in elevated stratopause years. Reductions of westward PW forcing in the upper stratosphere can cause downward/upward circulation in the mesosphere/stratosphere, respectively. Such changes in the vertical circulation induce adiabatic warming above ~40–50 km and adiabatic cooling below ~40–50 km at high latitudes. Although such warming above ~40–50 km can contribute to an elevated stratopause, differences in PW westward forcing at 40–55 km and 50°N–80°N between years with and without an elevated stratopause are ~6 ms−1 d−1. It is significantly smaller than GW forcing (~50–100 ms−1 d−1) seen in WACCM simulation during an elevated stratopause event [Chandran et al., 2013]. Therefore, during the downward movement of elevated stratopauses (10–19 February), changes in PWs below ~60 km are too weak to contribute to sustain elevated stratopauses, indicating the importance of changes in GWs for the downward movement of elevated stratopauses. In contrast, our PW and GW results suggest that the onset of an elevated stratopause can be influenced by both in situ generation of PWs and also changes in GWs (an overall reduction for this case). Our results agree with previous modeling studies of in situ PW influences for the onset of elevated stratopauses and GW influences for their downward movement [e.g., Siskind et al., 2010; Limpasuvan et al., 2012; Tomikawa et al., 2012].

Figure 14.

(color contours) EP flux amplitudes (|Fz| + |FΦ|), (black line contours) EP flux divergence (ms−1 d−1) at 3 ms−1 d−1 interval, and (red arrows) EP flux vectors averaged over 10–19 February in (a) 2005, (b) 2006, (c) 2007, and (d) 2009 obtained from MERRA. The solid/dash contour lines represent eastward/westward, respectively. In order for EP flux vectors to be visible over the entire altitude, EP flux vectors are divided by the square root of atmospheric density. The vertical component of the EP flux is multiplied by a viewing factor of 100.

[40] Ern et al. [2011] discussed the possibility that our SABER GW analysis method might be influenced by PW amplitudes. However, regions of strong planetary activities are located around latitude ranges of 20°N–70°N in Figure 14 for all four cases and do not show equatorward tilt in 2006 and 2009, indicating that SABER GW and PW activities show independent spatial distributions. Therefore, meridional tilting of GW enhancements in Figure 10 does not correlate with spatial distributions of PW activities and is likely actual GW variations rather than PW contamination.

5 Conclusions

[41] Elevated stratopauses formed at ~80–90 km altitude during the recovery phase of stratospheric sudden warmings in February 2006 and 2009 in association with downward transport of chemical species. These likely occurred in response to changes in the downward circulation due to GWs and PWs [e.g., Siskind et al., 2007; Limpasuvan et al., 2012]. However, the physical mechanisms are not fully understood, due in part to the lack of global GW observations in the MLT. This study presents the global-scale GW observations in the MLT during elevated stratopause events using TIMED-SABER temperature observations in the altitude range of 30–100 km, which provide insight into the formation of an elevated stratopause.

[42] In order to use SABER GWs in this study, we first validate them against COSMIC/GPS GWs. Then, we also compare them with SABER tidal amplitudes to examine tidal influence in SABER GW analysis. Global distributions of SABER GW amplitudes correlate well with those of COSMIC GWs, though COSMIC GWs are relatively larger than SABER GWs in the tropics. SABER GWs analyzed by separating ascending and descending nodes show independent variations compared to amplitudes of both migrating and nonmigrating tides estimated from SABER. Our results show that GW analysis method used here effectively suppresses tidal influences. The climatology of our GW amplitudes is also comparable with previous research and provides us with the confidence to use SABER GWs for an elevated stratopause study.

[43] During the recovery phase of SSWs, an elevated stratopause formed in February 2006 and 2009. During the onset of elevated stratopauses (1–10 February), GWs in the polar stratosphere during an elevated stratopause (2006 and 2009) are about 2 times weaker than that during non-elevated stratopause years (2005 and 2007), but GWs in the polar mesosphere around 80 km in 2006 and 2009 do not show clear differences compared to those in 2005 and 2007. On 10–19 February (the downward movement of elevated stratopauses), GWs are much lower in amplitudes at 40 km in 2006 and 2009 than those in 2005 and 2007, while GW amplitudes are larger at 80 km during elevated stratopause years than those in non-elevated stratopause years at 40°N–80°N. In 2006 and 2009 on 10–19 February, the enhancements of GWs propagate from low-mid latitudes at ~30 km to high latitudes at ~80 km. Such meridional tilting of the region of GW enhancements is likely related to the equatorward shift of the polar night jet. The GROGRAT ray-tracing simulations demonstrate the suppression of GW propagation within high latitudes from the onset to the downward movement of elevated stratopauses. More interestingly, the GROGRAT ray-tracing simulations show the enhancements of poleward propagation of GWs from low latitudes to high latitudes during the downward movement of elevated stratopause events, but such poleward propagation of GWs do not increase during the onset of elevated stratopause events.

[44] In addition to GWs, there are the in situ generated PWs during the onset of an elevated stratopause event, and these can contribute to the formation of elevated stratopauses along with the suppressions of GWs. This result supports previous modeling studies [Limpasuvan et al., 2012; Tomikawa et al., 2012; Chandran et al., 2013]. During the downward movement of elevated stratopauses, in situ generation and propagation of PWs from the troposphere to the upper stratosphere do not show enough changes to account for the formation of elevated stratopauses. Therefore, increases in the meridional propagation of GWs and suppression of GW propagation within high latitudes might play a key role in sustaining elevated stratopauses and inducing the downward movement of elevated stratopauses.

[45] Previous studies have focused on non-orographic GW propagation within high-latitude regions during an elevated stratopause event, and our results agree with their theory [Siskind et al., 2010; Chandran et al., 2011; Limpasuvan et al., 2012; Tomikawa et al., 2012]. In addition to that, our results of SABER GW variations and ray-tracing simulations indicate that meridional propagation of GWs is likely also important to explain the observed GW variations. Such meridional propagation of GWs is not included in most of GW parameterization schemes, and it deserves further study. Because of insufficient knowledge of GW sources at this point, our ray-tracing simulations assumed a homogenous distribution of GW sources. Better understanding of GW source distributions and temporal variations is necessary to improve our knowledge of elevated stratopause events and the related downward transport of chemical species.

Acknowledgments

[46] We sincerely acknowledge Stephen D. Eckermann for providing the GROGRAT ray-tracing model. C.Y. was supported by Japan Society for the Promotion of Science (JSPS) Postdoctoral Fellowship for Research Abroad. S.L.E. was supported by NASA's Living with a Star program through grant NNX11AQ73G. C.Y. and T.J.I. were supported by the National Science Foundation (NSF) grant AGS-1139165. L.C.C. was supported by grant NSC 101-2111-M-008-021-MY2 from the Taiwan National Science Council.

Ancillary