Sections 3.2 and 3.3 showed that AIRS and MLS do capture the expected behavior of temperature and humidity within the instrumental uncertainty. Interannual variations were also identified in the data record. The two primary interannual modes that modulate tropical temperatures and water vapor are the ENSO and QBO, thus the remainder of this work will focus on these two modes of tropical variability. Note that for the rest of this work the 100 hPa temperatures will be used as a cold point proxy to study the effects of the ENSO and QBO on the temperature and water vapor distribution around the tropopause.
3.4.1. Zonal Equatorial Mean Interannual Time Series
 To remove the record mean annual cycle, the mean is computed (for temperature, water vapor, and RH) over all years in bins of ∼16 days (see section 2 for details on bin construction). This is then subtracted from the same 16 day bins for each year; equation (1) illustrates this more clearly. The interannual signal for parameter “X” is calculated in the following manner:
for the ith time record bin (∼16 days/bin) within the jth year. ΔT and ΔRH, the interannual signals for temperature and RH, are computed using the above equation. Δq, the interannual signal for water vapor, is expressed as a percent deviation from the record mean annual cycle by dividing Equation (1) by Σj Xij; Figure 5 shows these quantities.
Figure 5. Interannual variability of the equatorial mean (180°E–180°W,08°S–08°N) of (a) temperature (ΔT), (b) water vapor (Δq), and (c) RH (ΔRH), in units of T, percent, and percent, respectively. (d) The Ocean Niño Index (ONI, red) and the quasi-biennial oscillation U50 index (QBOI, black). Green markers indicate ENSO events (see sections 2.2 and 2.3 for explanation of indices).
Download figure to PowerPoint
 Figure 5a shows the tropopause region ΔT has peak-to-peak amplitudes (∼1.5 K) that are about half the observed maximum tropopause region peak-to-peak temperature anomalies (∼4 K) presented in Figure 4a. Boundary layer and free tropospheric ΔT follow the ONI with warm (red colors in Figure 5a) and cold (blue colors in Figure 5a) periods corresponding to El Niño and La Niña events, respectively, highlighting the dominance of ENSO on the zonal mean temperature structure. In some periods, e.g., around January 2007 during an El Niño, ΔT in the UTLS (∼150–70 hPa) has the same sign as the free tropospheric ΔT, while other periods, for example, boreal winter of 2009 during a La Niña, the free tropospheric and UTLS ΔT have opposite signs. This is because the zonal mean anomalies in the LS down to 100 hPa is mainly determined by the downward phase propagation of the QBO, while the anomalies in the free troposphere and the UT are primarily controlled by ENSO which does not have a regular cycle like the QBO.
 The Δq signal (Figure 5b) is about half (∼0.5 ppmv) the water vapor anomalies (∼1.2 ppmv) shown in Figure 4b for the peak-to-peak amplitudes. The vertically propagating signal (in time), starting around 100 hPa (denoted by the black dashed line), represents the interannual variability of the water vapor tape recorder [Randel et al., 1998; Geller et al., 2002]. The zonal mean Δq is also dominated by the ENSO through the free troposphere, and up to the bottom of the TTL, with moist (red) and dry (blue) regions corresponding to El Niño and La Niña events. This indicates the influence of the lower tropospheric processes on the UTLS water vapor distribution that is associated with ENSO events. Once again, there are periods (see DJF of 2005, 2006, 2007, and 2008) when the prevailing free tropospheric Δq share the same sign as those in the UTLS. This is a result of the joint impact of the ENSO and QBO being in phase and simultaneously dehydrating TTL air more (less) effectively during La Niña (El Niño) years [Zhou et al., 2004]. This simultaneous ENSO and QBO impact will be discussed in more detail below. ΔRH (Figure 5c) in the ULTS almost exactly mimics ΔT with the largest anomalies in the TTL. However, moving further down into the free troposphere ΔRH signal is reduced from UTLS values and does not correlate well with the ONI or QBOI. This is consistent with RH remaining constant with global changes to temperature and water vapor [e.g., Soden et al., 2005].
 Shown in Figure 6a are the 1013 hPa (green) and 100 hPa (blue) ΔT for the entire equatorial region. The 100 hPa time record shows the cross section of anomalies denoted by the black dash-dotted lines in Figure 5. ENSO events alone should produce an anticorrelation between the 1013 hPa and 100 hPa ΔT since warmer (colder) SST should lead to stronger (weaker) convection which would lead to colder (warmer) tropopause temperatures [e.g., Gettelman et al., 2001].There seems to be some anticorrelation in certain periods of ΔT between 1013 hPa and 100 hPa, however, computing a lag correlation over the entire time record results in a low correlation coefficient (R) of −0.13. This value is not surprising as: (1) our data includes land (land surface temperatures do not follow the ONI), (2) oceanic boundary layer ΔT are largely governed by the ENSO, and (3) 100 hPa ΔT are primarily governed by the QBO.
Figure 6. Equatorial mean (180°E–180°W, 08°S–08°N) interannual variability of (a) temperature at 1013 hPa (green) and 100 hPa (blue) and (b) temperature (blue) and water vapor (pink) at 100 hPa. (c) ONI and QBOI are plotted for reference.
Download figure to PowerPoint
 Therefore, one might expect the 100 hPa ΔT to follow the QBOI rather than the ONI. A correlation coefficient of R = 0.86 (for lag 0) results from correlating the QBOI with 100 hPa ΔT with a confidence interval beyond 95%, reaffirming the thermal wind balance between 50 hPa and the lower stratosphere. The descending QBO ΔT is a minimum around 100 hPa, consistent with previous observations of the QBO [e.g., Wallace, 1973]. The 100 hPa anomalies are about 50% greater than those at 1013 hPa. This tropospheric amplification is consistent with work by Santer et al.  and Gettelman and Fu  where they find that UT temperature anomalies are amplified by a factor of ∼1.5 from the surface.
 Figure 6b shows the same 100 hPa ΔT along with 100 hPa Δq (pink). Δq follows ΔT closely though with a lag of about a half a month. The lag correlation between ΔT and Δq yields R = 0.74 for lag +1 (∼16 days) (at lag 0 R = 0.71). The lag correlation between Δq and the QBOI gives R = 0.74 (at lag +4) for lag of about 2.0 months (R = 0.60 for lag 0). The cause of the difference in lag correlation between ΔT and Δq and the QBOI and Δq is unknown. The low correlation between Δq and the ONI of R = 0.50 (lag 0) affirms that the 100 hPa ΔT is predominately determined by the QBO. At 121 hPa, the correlation between ΔT and Δq yields R = 0.78 for lag 0, suggesting the ∼0.5 month lag at 100 hPa is likely a manifestation of MLS averaging (“smearing”) features of the water vapor tape recorder from higher altitudes; MLS has a nominal vertical resolution of 2 km which may not be able to resolve the details of the cold point tropopause (although other transport processes may play a role in the lag as well). The 100 hPa correlations between ΔT and Δq, as well as Δq and QBOI, increases to R = 0.81 and R = 0.80, respectively, when excluding data beyond January 2008, which corresponds to when the ENSO and QBO begin to fall out of phase. This indicates possible different zonal impacts on water vapor depending on the relative phase of the ENSO and QBO. We note that doing correlations with time bins of ∼16 days, on the one hand, yields more statistically significant time means, however, the coarseness of the time resolution prevents any probing of processes on scales finer than about a month. Thus, these lag correlations should only be seen as gross indications of relationships between remotely sensed parameters. Nevertheless, these correlations are statistically significant and confirm the strong influence of the QBO on the zonal temperature and water vapor structure, especially when the ENSO and QBO are in phase.
 Therefore, the zonal mean structure of tropopause temperature and water vapor primarily follow the QBO. However, this mean structure masks out the zonally varying ENSO signal seen, for example, by Randel et al.  and Gettelman et al. . The ENSO and QBO are approximately in phase from August 2004 through boreal summer of 2008 (see Figure 3). After boreal summer of 2008 the ENSO and QBO fall out of phase. Some interesting questions to ask are: (1) how does 100 hPa ΔT depend on the phase of both interannual modes and (2) how is the water vapor distribution in the UTLS related to the relative phase of the ENSO and QBO. These questions will be addressed in more detail in sections 3.4.2–3.4.5.
3.4.2. Zonal Structure of the ENSO and QBO at 100 hPa
 In section 3.4.1 the equatorial mean time record was computed (8°S–8°N, 180°W–180°E). It was found that the QBO dominates the equatorial zonal mean ΔT and Δq structure at 100 hPa. However, since the time record in Figures 5 and 6 are tropical means, it is difficult to identify any ENSO signature at 100 hPa. Thus, the longitudinal variations are investigated to explore the regional impacts of the ENSO and QBO in the tropopause region.
 Figures 7a and 7b show longitude Hovmöller diagrams (8°S–8°N) of ΔT and Δq, respectively, at 100 hPa. The 100 hPa ΔT (Figure 7a) varies coherently with the phase of the QBOI indicating the dominating impact of the QBO on 100 hPa temperatures. However, upon closer inspection of 1:E+Q+ and 2:E−Q− (periods when the QBO and ENSO are approximately in phase) some interesting features are observed. The TWP anomalies have the same sign as the coherent QBO ΔT and exhibits warmer (colder) anomalies than other longitudes during El Niño (La Niña) events. This is consistent with larger (smaller) dehydration volumes occurring during La Niña (El Niño) seasons [Zhou et al., 2004]. However, in the TCP, strands of opposite signed anomalies exist. ENSO events that produce anomalies opposite to the prevailing QBO signal are circled with a solid black oval.
Figure 7. Longitude Hovmöller plots of (a) ΔT and (b) Δq, meridionally averaged in the 8°S–8°N band, at 100 hPa (longitude bins are 4°). (c) ONI and QBO are plotted for reference. The TWP and TCP are also marked out by arrows. Ovals mark breaks in the zonal symmetry of the QBO (dashed oval for the TWP). Figure 2 is inserted in Figure 7c for reference.
Download figure to PowerPoint
 From the boreal summer of 2008 to the first quarter of 2010 the ONI and QBOI are out of phase. The oscillating QBO signal produces a prevailing warm ΔT, especially between the boreal summers of 2008 and 2009 (composite 3:E−Q+). The ENSO and QBO temperature signatures are in phase in the TCP and the TWP experiences 100 hPa anomalies that are counter to the prevailing QBO signal. This is seen in the small blue bands circled with a dashed black oval. Figure 7b shows the strong impact of the QBO on the zonal 100 hPa Δq distribution. Colocated with the regional temperatures circled in Figure 7a are areas of positive (negative) Δq that correspond to positive (negative) ΔT. However, the spatial coherence is weaker than with ΔT.
 To statistically quantify the zonal breaks in ΔT and Δq at 100 hPa, the time record for the TCP (Figure 8a) and TWP (Figure 8b) are computed. Blue and pink patches indicate when ΔT and Δq, respectively, are statistically significant from 0 at the 95% confidence level; purple patches indicate when both anomalies are statistically significant.
Figure 8. The 100 hPa ΔT (blue curve) and Δq (pink curve) for (a) TCP and (b) TWP. Blue patches correspond to time periods when ΔT signals are statistically significant at the 95% level or better. Pink patches quantify the same statistics for Δq. Periods with purple patches correspond to instances when ΔT and Δq are simultaneously statistically significant at the 95% level or better. (c) The ONI and QBOI are shown for reference.
Download figure to PowerPoint
 The 100 hPa ΔT in the TWP (blue curve in Figure 8b) show relatively large amplitudes during the years when the ONI and QBOI (shown in Figure 8c) are in phase, with ΔT that range between −2.6 to +2.0 K (between the end of 2004 to the boreal summer of 2008), with statistical significant at the 95% level much of this period. Δq is also statistically significant from 0 with anomalies that range between ±0.8 ppmv. During this period, the TCP experiences weaker anomalies with ΔT and Δq that range between −1.8 to +1.3 K and ±0.5 ppmv, respectively, with many oscillations around 0. When the QBOI and ONI fall out of phase (after the boreal summer of 2008), the TWP ΔT and Δq falls off to smaller magnitudes between −0.8 to +1.8 K and ±0.4 ppmv with frequent oscillations around 0. The TCP now experiences statically significant anomalies (at the 95% level), for longer periods of time, with ΔT and Δq ranging between −2.3 to +2.5 K and −0.4 to 0.5 ppmv, respectively. The TWP Δq, as with the equatorial mean case, correlates well with ΔT giving R = 0.77 (for lag +1), while the TCP shows a lower correlation of R = 0.62 (0 lag). This is consistent with previous work identifying that Δq is a function of the TWP CPT [e.g., Holton and Gettelman, 2001].
3.4.3. The 100 hPa Composite Maps of ΔT and Δq
 To investigate further this apparent zonal break in the QBO signal, the structures of the 100 hPa ΔT (Figure 9a) and Δq (Figure 9b) are investigated for boreal winter (DJF) in the context of the composites presented in Figure 2: 1:E+Q+, 2:E−Q−, 3:E−Q+, and 4:E +Q* (see section 2.2 for composite details). For 1:E+Q+, the QBO induced, zonally symmetric, warm ΔT shows a break of reduced ΔT centered around 120°W–170°W (Figure 9, dashed circled region). This also occurs in 2:E−Q− with the positive ΔT induced by ENSO breaking the zonal symmetry of the negative QBO ΔT. In both cases the ENSO and QBO are in phase and the observed ΔT around the zonal break forms a symmetrical dumbbell shaped pattern in the TCP since the zonal break has a maximum centered around 140W. These zonal asymmetries can also be interpreted as the QBO restricting (during the westerly regime, 1:E+Q+) or enhancing (during the easterly regime, 2:E−Q−) convection in the TWP [Collimore et al., 1998, 2002].
Figure 9. The 4° × 4° maps of 100 hPa (a) ΔT and (b) Δq for composite events 1:E+Q+ (first row), 2:E−Q− (second row), 3:E−Q+ (third row), and 4:E+Q* (fourth row) discussed in section 2.2. The star after Q indicates a nearly 0 QBOI. Dashed lines mark the 8°CS-8°CN band and the approximate meridional extent of the QBO signal in temperature.
Download figure to PowerPoint
 Transitioning to periods when the ENSO and QBO are out of phase, 3:E−Q+ shows a break in zonal ΔT that now occurs over the TWP (Figure 9, dashed circled region) with the TCP taking on zonal anomalies of the same sign as the prevailing westerly phase of the QBO. During this period the La Niña in itself should increase the dehydration potential of the TWP [e.g., Zhou et al., 2004] via enhanced convection; however, the QBO westerly phase induces subsidence in the tropopause region, increasing tropopause temperatures, thus inhibiting convection from penetrating deeper into the tropopause region [Collimore et al., 1998, 2002].
 The 4:E+Q* ΔT pattern appears, in contrast to the other ENSO events previously discussed, because the QBO is transitioning from the westerly to easterly phase, which takes place around boreal spring (see QBOI in Figure 2). Thus, the 4:E+Q* ΔT map shows the approximate temperature patterns one would expect with little influence from the QBO. The 4:E+Q* ΔT exhibits temperature patterns qualitatively consistent with the ENSO temperature patterns shown in Figure 6b of Kiladis et al. . The narrow band of warm anomalies in the TWP are enveloped by symmetrical cold anomalies around the equator in the TCP. These tropopause patterns are consistent with Rossby and Kelvin circulations induced by tropical equatorial heating [e.g., Gill, 1980; Highwood and Hoskins, 1998]. Figure 10 shows, in contrast to 4:E+Q*, a 3 month mean (August, September, October) in 2008 when the ONI is near zero and the QBO reaches a local maximum. The zonal symmetry in ΔT is robust in the tropics with very little evidence of an ENSO signature on tropopause temperatures. The weaker convection during the fall season and weak ENSO allow the QBO signal to manifest as a zonally symmetric feature in temperature and water vapor. Conceptually, the final signal in ΔT resembles a superposition of the QBO signal shown in Figure 10 and the ENSO signal in 4:E+Q* (Figure 9, fourth row).
 The Δq signals of 1:E+Q+ and 2:E−Q− show dominant prevailing signals that are in phase with the ONI and QBOI with 1:E+Q+ and 2:E−Q− showing positive and negative water vapor anomalies, respectively. This is a consequence of the ENSO and QBO being in phase such that the zonally asymmetric ENSO feature does not show up strongly. Although the ENSO is responsible for the anomaly reduction in the TCP, the anomaly enhancement in the TWP dominates the zonal Δq distribution, consistent with Figures 6 and 8. However, 3:E−Q+ and 4:E+Q* both show the TCP with anomalies of the opposite sign from the rest of the tropics. In these cases, Δq is similar to ΔT in that a colder (warmer) tropopause corresponds to a drier (wetter) tropopause. This contrast in zonal Δq for 3:E−Q+ and 4:E+Q* suggests that when the ENSO and QBO are out of phase, the TCP may play a more prominent role in regulating the tropopause region water vapor distribution, though the source/sink of this water vapor cannot be determined from the composites.
 The resulting patterns of 100 hPa ΔT are interesting and have a simple explanation. During La Niña years (composite 2:E−Q− in Figure 9), deep convection is particularly strong in the TWP. This is associated with large positive temperature anomalies in the free troposphere due to convective heating (see Figure 3a). Associated with this strong heating is a thinner layer of anomalously cold temperatures around the tropopause region. In the TCP a warmer tropopause region is associated with colder free tropospheric temperatures. If the ENSO and QBO are in phase, then the cold ENSO anomalies will be in phase with the zonally symmetric cold QBO anomalies in the TWP, leading to an enhancement of cold anomalies, thus leading to greater dehydration in water vapor [Zhou et al., 2004]. These same cold anomalies will work against the warm anomalies that are seen in the TCP tropopause temperatures, resulting in reduced anomalies. During the El Niño years that are in phase with the QBO (composite 1:E+Q+ in Figure 9), one expects the TCP to have colder tropopause temperatures. If the ENSO and QBO are in phase, once again, the zonally symmetric QBO signal, now positive (warm), will counteract the cold TCP tropopause temperatures while enhancing the warm tropopause temperatures in the TWP, leading to moistening/reduced dehydration in the TWP [Zhou et al., 2004]. In these cases, the zonal structure of water vapor is primarily driven by the TWP CPT [Holton and Gettelman, 2001].
 During the periods when the ENSO and QBO are out of phase (3:E−Q+ and 4:E+Q* in Figure 9), the westerly (easterly) anomalies will enhance the TCP warm (cold) tropopause anomalies associated with La Niña (El Niño) events, while reducing the cold (warm) anomalies in the TWP. Revisiting 3:E−Q+, we find that a prominent water vapor maximum in TCP that is not present when the ENSO and QBO are in phase. This maximum is consistent with the La Niña warm top over the TCP being in phase with the warm QBO westerly, although the source of the water vapor maximum is not known.
3.4.4. Isolating the ENSO Signature
 In section 3.4.2, it was found that the ENSO and QBO either enhance or reduce temperature and water vapor anomalies at 100 hPa depending on the phase and location of observation. The anomalies are magnified in the TWP when the ENSO and QBO are in phase, leading to reduced anomalies in the TCP. When the ENSO and QBO are out of phase, the TCP becomes the region of enhanced anomalies while the TWP anomalies counter the prevailing QBO signal. Although the statistics seem to support this hypothesis, the ENSO effect still has not been quantified. In order to isolate the ENSO signal, the approximate zonal symmetry nature of the QBO is exploited. This is done by subtracting the tropical zonal mean time record (Figure 5) from the corresponding TCP and TWP time record.
 In the TWP, the correlation between ΔTE and ΔqE (ΔT and Δq with their zonal mean subtracted out) at 100 hPa only yields R = 0.48. This is in contrast to the high correlation (R = 0.77, lag +1) for the case when the zonal mean is included. Within the TCP, the 100 hPa correlation between ΔTE and ΔqE yields R = 0.82 (lag 0). This high value suggests that the local temperatures in the TCP do have a strong influence on the local water vapor distribution. During El Niño (La Niña) years, the TCP tropopause region cools (warms) leading to a dryer (moister) tropopause region. Recall, a weaker correlation was computed between ΔT and Δq with the zonal mean included (R = 0.62, lag 0). This, in part, may explain why when the ENSO and QBO are out of phase the TCP water vapor maxima shows up (e.g., 3:E−Q+). When the QBO and ENSO are out of phase, the ENSO impact on tropopause ΔT and Δq is now supported by the QBO.
3.4.5. Horizontal and Vertical Structure of the ENSO and QBO
 We now investigate the vertical and horizontal spatial distribution (Figure 11) of the interannual variability of temperature and water vapor to corroborate the results shown in sections 3.4.2–3.4.4. The same time period composites as shown in Figure 9 are selected for analyses, however, in order to highlight the impacts of ENSO, specifically El Niño, we compute the difference [1:E+Q+] − [2:E−Q−] (herein EL01, brackets inserted for readability) and [4:E+Q*] − [3:E−Q+] (herein EL02).
Figure 11. Vertical cross sections of: (a and b) ΔT, (c and d) ΔT with the zonal mean removed (ΔTE), (e and f) Δq, and (g and h) Δq with the zonal mean removed (ΔqE) for composite differences between [1:E+Q+] − [2:E−Q−] (EL01, Figures 11a, 11c, 11e, and 11g) and [4:E+Q*] − [3:E−Q+] (EL02, Figures 11b, 11d, 11f, and 11h) (composite events shown in Figure 9). Ordinate is pressure in hPa and abscissa longitude (4° bins); meridional mean is taken from 8°S–8°N. Overlaid solid black horizontal line marks 100 hPa. Overlaid blue curve is the interannual mean rain rate anomalies (Δrr), in mm/h, derived from TRMM with ordinate axes on the right; dash-dotted line marks zero Δrr.
Download figure to PowerPoint
 The EL01 ΔT composite (Figure 11a) shows a strong thick band of warm QBO anomalies from 50 hPa down into the TTL. The TWP TTL shows especially warm anomalies as compared to the rest of the tropics contrasting the thin vertical band of cold anomalies over the TCP TTL. The tropical troposphere is also warm. As previously discussed (section 3.4.2), the ENSO dominates the free tropospheric and boundary layer zonal mean ΔT. Subtracting the zonal mean ΔT produces ΔTE (Figure 11c), revealing the asymmetrical ENSO anomalies (with possible other effects as well, for example, tropospheric biennial oscillations which impacts aspects of the rainfall associated with the Indian Monsoon [Chang and Li, 2000]). A quadrupole temperature anomaly structure is revealed in ΔTE, with a cold top over a warm free troposphere in the TCP and colder free troposphere in the TWP with a warm top. A quadrupole pattern in ΔqE is roughly colocated with the ΔTE structure. However, there are a few differences in the water vapor structure. First, the tropospheric anomalies in ΔqE extend into the bottom of the TTL where one finds ΔTE with opposite sign to ΔqE. Secondly, the layer of TTL ΔqE anomalies primarily lie above 100 hPa in the TCP, whereas ΔTE is approximately symmetric around 100 hPa. Thus, although the ENSO impacts on temperature are symmetric around the tropopause, its impact on water vapor is skewed into the stratospheric region.
 The ΔTE signal for EL01 clearly shows the cold TCP TTL is playing a role in reducing the warm QBO anomalies while the warm TWP TTL (Figure 11c) enhances the QBO signal. Associated with these warm anomalies is hydration of the TTL region over the TWP (in a La Niña case, i.e., [2:E−Q−] − [1:E+Q+], there would be enhanced dehydration, highlighting composite 2:E−Q−). Upon further inspection, one finds that the mean Δq, i.e., the mean Δq difference [1:E+Q+] − [2:E−Q−], in the TCP is ∼2 % (∼0.1 ppmv), while the TWP mean is ∼12 % (∼0.4 ppmv). EL02 shows that deep convection in the TCP leads to cold TTL ΔTE that enhances the QBO cold anomalies. The overlaid (light blue curves in Figure 11) interannual rain rate anomalies (Δrr, in mm/hr), computed from Tropical Rainfall Measuring Mission [Iguchi et al., 2000] data corroborates that convection follows the ENSO cycle; high Δrr follows the tropospheric moistening. For EL02, the TCP cold anomalies associated with deep convection are now allowed to constructively work with the QBO to dehydrate TTL air over the TCP, or in the case of 3:E−Q+ (a La Niña), moisten the TTL TCP. In the case of [3:E−Q+] − [4:E+Q*] (highlighting La Niña), the mean TCP and TWP differences in Δq are ∼+8 % (∼+0.3 ppmv) and ∼−13 % (∼−0.4 ppmv), respectively, with the TCP TTL anomalies now switching signs. Although 3:E−Q+ shows a moist anomaly over the TCP, presumably from the QBO westerly anomaly and La Niña warm top having the same sign, the source of this water vapor cannot be determined from this analysis.