The upper troposphere and lower stratosphere (UTLS) form a key region of the Earth's climate system because of a large sensitivity of radiative forcing to greenhouse gas variations in that region, such as water vapour (H2O) and ozone (O3) . The transport and distribution of these trace gases in the UTLS are determined by the stratospheric Brewer–Dobson circulation (BDC), defined as the meridional overturning circulation which transports air masses upward from the tropics, poleward, and then downward in the extratropics through its transition and shallow branches in the UTLS and its deep branch in the middle and upper stratosphere . Any changes in the composition of these radiatively active trace gases in the UTLS region induced by the BDC and its modulation by the modes of climate variability lead to large impacts on surface climate e.g..

Ozone is mainly produced in the lower and middle stratosphere between about 16 and 35 km in altitude, often referred to as the ozone layer . In addition, ozone variability in the tropical lower stratosphere is a good proxy for the tropical upwelling of the BDC . The ozone transport and lifetime in the UTLS region are both modulated by the seasonality in the BDC and the modes of climate variability, such as the quasi-biennial oscillation (QBO) . Lower-stratospheric water vapour and its multi-timescale variations ranging from days to decades are mainly controlled by changes in the tropical cold-point temperatures and its modulations by the natural climate variability . Therefore, the amount of water vapour in the UTLS region is directly linked to the dehydration (i.e. the process of removing water) of the air parcels crossing through the coldest temperatures in the tropical tropopause layer e.g. between 14 and 19 km;.

Mostly driven by gravity waves and equatorially trapped waves, the QBO is a quasi-periodic oscillation between tropical westerly and easterly zonal winds . The QBO is considered a dominant mode of climate variability of the equatorial stratosphere, and it globally impacts the transport and distributions of stratospheric trace gases, including water vapour and ozone. Both alternating QBO easterly and westerly zonal wind regimes modulate the vertical and meridional components of the BDC and affect temperature structure, thereby impacting the water vapour and ozone composition and radiative feedback in the UTLS region .

The quasi-periodic QBO mean cycle of an approximately 28-month period, which alternates between westerly and easterly zonal winds, was subject to two disruptions in the past 5 years. In February 2016 and January 2020, the QBO westerlies in the tropical lower stratosphere were unexpectedly interrupted by anomalous QBO easterlies caused by planetary waves propagating from the mid-latitudes toward the equatorial region combined with equatorial convective gravity waves . Hitherto, there has been no clear understanding of how these QBO disruption events are linked to anomalously warm or cold sea surface temperatures , volcanic aerosols , wildfire smoke , and climate change . However, a recent study based on climate model simulations from Phase 6 of the Coupled Model Intercomparison Project (CMIP6) predicts increased disruption frequencies to the quasi-regular QBO cycle in a changing climate . Previous studies also suggest that the QBO amplitude in the tropical stratosphere is decreasing in the lower stratosphere due to the climate-change-induced strengthening of the tropical upwelling . Thus, in the context of a changing climate, the predictable QBO signal associated with the quasi-regular phase progression and amplitude as well as its potential impacts on UTLS composition faces an uncertain future. Therefore, it is of particular importance to quantify and better understand the different anomalous circulations and the impacts of the QBO disruption events on UTLS water vapour and ozone, which have the potential to locally and globally affect the radiative forcing of the Earth's climate system through their impacts on surface temperatures .

Here, we use satellite observations to quantify the similarities and differences in the strength and depth of perturbed/disrupted QBO impacts in 2016 and 2020 on water vapour and ozone in the lower stratosphere. Also, we analyse the main drivers of the differences in anomalous circulation and UTLS composition changes. Section  describes the satellite observational data sets and the multi-variate hybrid regression model used for the quantification. Section  describes the anomalous BDC and UTLS composition changes following the 2016 and 2020 QBO disruption events. Section  discusses the results of a well-established multi-variate hybrid regression analysis to provide evidence for the impact of the QBO disruption events on lower-stratospheric water vapour and ozone. Finally, we discuss the main reasons for the differences between the 2016 and 2020 impacts of the QBO disruption events on BDC and UTLS composition and the related dynamical processes associated with planetary and gravity wave dissipation, which are likely caused by the anomalous surface conditions associated with the strong El Niño–Southern Oscillation (ENSO) in 2015–2016 and the strong Indian Ocean Dipole (IOD) in 2019–2020. We also discuss the differences in BDC and UTLS composition between 2016 and 2020 in terms of the particularly warm stratosphere in the context of Australian wildfire smoke in 2020.

To quantify the QBO and Australian wildfire smoke impacts, we used the monthly mean, zonal mean ozone and water vapour mixing ratios from Aura Microwave Limb Sounder (MLS) satellite observations covering the 2005–2020 period . The version 4.4 MLS data set used here has a vertical resolution of 2.5–3.5 km ranging from 8 to 35 km and from 60∘ S to 60∘ N. The individual profile measurements of this version 4.4 have a precision and systematic uncertainty of about ±10 %–40 % and ±10 %–25 % for H2O and ±0.02–0.04 ppmv and ±0.02–0.05 ppmv + ±5 %–10 % for O3, respectively, with a spatial representativeness of ∼200–300 km along the orbital-track line of sight . Previous findings show that MLS monthly mean, zonal mean H2O mixing ratios show very good agreement with 13 water vapour products from 11 limb-viewing satellite instruments throughout most of the atmosphere (including the UTLS), with mean deviations from the multi-instrument mean between +2.5 % and +5 %, making these random errors irrelevant for the averaged monthly mean, zonal mean H2O anomalies used in this study e.g..

In addition to the MLS observation data sets, we also utilize the temperature (T) and zonal mean wind (U) for the 2005–2020 time period from the ERA5 reanalysis of the European Centre for Medium-Range Weather Forecasts (ECMWF) . We have also calculated the residual circulation vertical velocity (w∗‾) using the transformed Eulerian mean (TEM; ) and decomposed the wave drag into planetary wave drag (PWD) and gravity wave drag (GWD) contributions to the circulation anomalies . Note that we are using the ERA5 reanalysis data on the original 137 model levels for calculating the TEM budget, but not the coarse conventional pressure-level data, which can cause large uncertainties in the equatorial waves and zonal wind in the tropical stratosphere . For more details about the ERA5 TEM calculations and wave decomposition, please see .

We disentangle the QBO impact on the MLS monthly mean, zonal mean stratospheric water vapour and ozone mixing ratios from the other sources of natural climate variability using a multi-variate hybrid regression model for the 2005–2020 period (Eq. ). In the figures, only the 2013–2020 period is shown to highlight the impact of the two QBO disruption events. The established multi-variate hybrid regression method is appropriate for separating the relative influences of the considered modes of climate variability, including the QBO, on stratospheric water vapour and ozone. Additional details about the multi-variate hybrid regression model and its applications can be found in . Our multi-variate hybrid regression model decomposes the given monthly zonal mean variable, Vari, into a long-term linear trend, a seasonal cycle, modes of climate variability, and a residual (ϵ). For a given variable Vari (herein H2O, O3, w∗‾, T, PWD, and GWD), the multi-variate hybrid regression model yields 1Vari(tmonth,ylat,zalt)=Trend(tmonth,ylat,zalt)+SeasCyc(tmonth,ylat,zalt)+∑n=15bn(ylat,zalt)⋅Proxyn(tmonth-τn(ylat,zalt))+ϵ(tmonth,ylat,zalt), where Proxyn represents the different climate indices used here. Proxy1 is a normalized QBO index (QBOi) from the 5∘ S–5∘ N ERA5 zonally averaged zonal mean winds with full vertical levels then deseasonalized and normalized by the standard deviation to build the QBOi . Proxy2 is the normalized Multivariate ENSO Index (MEI; ), Proxy3 is the IOD , Proxy4 is the Madden–Julian oscillation (MJO, ), and Proxy5 is the aerosol optical depth (AOD) from satellite data . Trend(tmonth,ylat,zalt) is a linear trend. SeasCyc(tmonth,ylat,zalt) is the annual cycle. The coefficients are the amplitude bn and the lag τn(ylat,zalt) associated with the QBO, ENSO, IOD, MJO, and AOD, respectively. The solar forcing is neglected because our data set is relatively short. Finally, we estimate the uncertainty in the multi-variate hybrid regression model using a Student's t-test technique .

In February 2016 and January 2020, unexpected tropical QBO easterlies (negative QBOi) developed into the downward-propagating tropical QBO westerlies between the altitudes of 16 km and 25 km, thereby breaking the quasi-regular QBO cycle of alternating easterly and westerly phases (Figs. a and S1a, b in the Supplement) . Both QBO disruption events have been associated with a combination of extratropical Rossby waves, equatorial planetary waves (Kelvin, Rossby, mixed Rossby gravity, and inertia gravity), and small-scale convective gravity waves, propagating into the deep tropics and depositing their negative momentum forcing . Both QBO disruption events were primarily triggered by mid-latitude Rossby waves propagating from the Northern Hemisphere in 2016 and from the Southern Hemisphere in 2020 into the deep tropical lower stratosphere. In 2016, the equatorial planetary wave forcing may have pre-conditioned mid-latitude Rossby waves to break easily at the Equator e.g., while in 2020, the equatorial planetary and small-scale convective gravity waves propagating into the UTLS predominantly contributed to the disruption . Note that the potential processes and mechanisms triggering the QBO disruption are still under debate. Recent findings from using a one-dimensional theoretical model of the QBO from pointed out the key role of the upwelling and wave dissipation. Details regarding the triggering of QBO disruptions have been the focus of several recent studies (e.g. ). Although similar in many respects, including the causes of the sudden development of tropical QBO easterlies into tropical QBO westerlies between the altitudes of 16 and 25 km, the two QBO disruption events also exhibit differences, particularly in the structure (strength and depth) of their impacts and the level at which the shift started (Fig. a). Here, we mainly focus on the impact of the QBO disruption events on the lower-stratospheric BDC and on the distributions of trace gases like water vapour and ozone.

The similarities as well as the differences between the two disruption events are also visible in the inter-annual variability of the tropical lower-stratospheric zonal mean zonal wind (a), H2O (b), and O3 (c) anomalies as a percentage change relative to the monthly mean mixing ratio during the 2013–2020 period (Fig. a–c). Both QBO disruption events are expected to impact the tropical upwelling of the BDC through the two-way interactions between the mean flow and wave propagation associated with the QBO phases as well as through its control of the tropical cold-point temperatures . The impacts of the QBO disruption events in 2016 and 2020 on the transport and distribution of lower-stratospheric H2O and O3 mixing ratios are most effective when the anomalous QBO easterlies reach the tropical cold-point temperature altitude (∼17 km), with the associated enhanced tropical upwelling driven by the anomalous wave breaking from June to December in 2016 and from June to August in 2020 (Fig. ) . The zonal mean, zonal wind shows that the westerly jet between the onset and offset time periods and at the altitude of 25 km is stronger and deeper during the QBO disruption event in 2016 than during the QBO disruption event in 2020 (Figs. a and S1c, d in the Supplement). The QBO disruption event in 2020 shows a clear separation of the westerlies into two parts, while the QBO disruption event in 2016 reestablishes the westerlies at the top of the easterlies, e.g. at the altitude of about 25 km (Fig. a). As soon as the downward propagation of tropical QBO easterlies reaches the tropical cold-point temperature altitude (∼17 km) from June to December 2016, the H2O mixing ratios decrease, i.e. turning from positive to negative anomalies. As reported by , the alignment of the strong El Niño event with a westerly QBO in the early boreal winter of 2015–2016 (September 2015–March 2016) substantially increased H2O mixing ratios and decreased O3 mixing ratios by up to about 20 % in the tropical lower stratosphere between the tropopause (∼16 km) and the altitude of about 23 km (Fig. b–d). The sudden occurrence of the QBO disruption event decreased the lower-stratospheric H2O and O3 mixing ratios from late spring to the early following winter by up to about 20 % (Fig. b–d).

Conversely, during the QBO disruption event in 2020, Fig. b–d show clear differences in the tropical lower-stratospheric trace gas anomalies, particularly in the strength and depth of H2O and O3 anomalies, consistent with the structural zonal mean zonal wind changes (Fig. S1c, d). The tropical lower-stratospheric O3 anomalies are purely responding to the enhanced tropical upwelling of the BDC caused in 2016 by a combination of a strong El Niño event, a negative IOD event, and the QBO disruption event in 2016 and in 2020 by a combination of a weak La Niña, a strong positive IOD event, and the QBO disruption events in 2020 (e.g. easterly winds between 16 and 23 km (100–40 hPa)) . The tropical lower-stratospheric O3 anomaly is a good proxy of the tropical upwelling of the BDC as its concentration is modulated by the advection of tropospheric air generally poor in O3 into the stratosphere . The small decrease in the tropical lower-stratospheric O3 anomalies by up to about 10 % in 2020 compared to about 20 % in 2016 between the altitudes of 16 and 23 km suggests a stronger tropical upwelling and its modulations in 2016 than in 2020 (Fig. c, d).

The inter-annual variability in large-scale upward advection of the tropical stratospheric H2O anomalies (i.e. tape recorder) is more challenging to interpret because of its regulation by the variability in the tropical cold-point temperatures . The negative tropical lower-stratospheric H2O anomalies induced by the interplay of different modes of natural climate variability, including the QBO, are weaker in 2020 than in 2016 (Figs. b, d and S2a, b in the Supplement). The tropical lower-stratospheric H2O anomalies averaged between the altitudes of 16 and 18 km are up to about 20 % more negative in 2016 than in 2020 (Figs. b, d and S2a, b in the Supplement). In particular, the 2020 tape recorder shows positive H2O anomalies as large as 15 % even after the QBO disruption event that are of opposite sign to the 2016 H2O anomalies (Fig. b, d). This complexity in H2O inter-annual variability lies in its dependency on the interplay of different modes of climate variability, including the QBO , volcanic aerosols , seasons (early or late in the winter), and location (western, central, or eastern Pacific, where the ENSO and IOD maximum occurs; ). Therefore, to elucidate the impact of the two QBO disruption events on the Brewer–Dobson circulation and the respective distributions of lower-stratospheric H2O and O3 anomalies, we performed a regression analysis both without and with explicitly including QBO signals to isolate the QBO impact on these trace gases. The difference between the residual (ϵ in Eq. ) with and without explicit inclusion of the QBO signals provides the QBO-induced impact on stratospheric H2O and O3 anomalies. Also, the impact of 2020 Australian wildfire smoke on stratospheric H2O anomalies is analogously obtained by differencing the residuals of the regression model. This approach of differencing the residuals is similar to direct calculations, projecting the best fits of the regression onto the QBO basis functions, i.e. the QBO predictor time series (see Supplement Figs. 2 and 4 in ). In addition, this differencing approach avoids the need to reconstruct the time series after the regression analysis.

Based on an established multiple regression method applied to Aura MLS observations, we found that both the QBO disruption events in 2016 and 2020 induced similar structural changes in the Brewer–Dobson circulation and respective distributions of the lower-stratospheric H2O and O3 anomalies. Both QBO disruption events induced negative anomalies in H2O and O3, a few months after the sudden shift from the QBO westerly to QBO easterly winds. During the boreal winter of 2015–2016 (September 2015–March 2016), the alignment of the strong El Niño and negative IOD events with the QBO westerlies strongly moistened the lower stratosphere between the tropopause and the altitude of 23 km (positive anomalies of more than 20 %). Analogously, the alignment of the weak La Niña and strong positive IOD events with the strong QBO westerlies and the impact of Australian wildfire smoke strongly moistened the lower stratosphere (positive anomalies of more than 15 %) during the boreal winter of 2019–2020 (September 2019–June 2020). The sudden shift from the QBO westerly to QBO easterly winds reversed the lower-stratospheric moistening, thereby leading to large negative H2O and O3 anomalies of up to about 20 % between 16 and 23 km by the end of summer 2016 and to small negative H2O anomalies of up to about 2 %–3 % and moderate negative O3 anomalies of up to about 10 % in 2020. These decreases in H2O and O3 mixing ratios are due to a strengthening of the tropical upwelling of the BDC, cooling tropical cold-point temperatures and their modulations by the QBO disruption events.

However, differences occur in the strength and depth of the QBO disruption-induced negative H2O and O3 anomalies in the lower stratosphere between 2016 and 2020. We found that the impact of the QBO disruption event on lower-stratospheric trace gases is weaker in 2020 than in 2016, up to about 18 % for H2O anomalies and 10 % for O3 anomalies between 16 km and 23 km, respectively. The differences in the strength and depth of the O3 anomalies and its modulation by the QBO disruption events are due to discrepancies in the anomalous tropical upwelling of the BDC, which was up to about 25 % larger in 2016 than in 2020. The analysis of the wave drag shows that the differences in planetary wave breaking in the tropical lower stratosphere between the tropopause and the altitude of about 23 km and the gravity wave breaking near the equatorward upper flanks of the subtropical jet (e.g. the region between 10 and 30∘ S/N and above the tropopause level) are the main reasons for the differences in the anomalous tropical upwelling of the BDC and secondary circulation between the year 2016 and the year 2020. The main differences in lower-stratospheric H2O anomalies between the year 2016 and the year 2020 are due to discrepancies in the tropical cold-point temperatures induced by the 2020 Australian wildfire smoke. Despite the anomalous planetary waves and gravity wave activities, which are likely related to ENSO and IOD, the 2020 Australian wildfire predominantly raised the cold-point temperatures, thereby leading to less dehydration of the lower stratosphere.

Finally, our results suggest that the interplay of QBO phases with a combination of ENSO and IOD events, and in particular also wildfires and volcanic eruptions, will be crucial for the control of the lower-stratospheric H2O and O3 budget in a changing future climate. In particular, increasing future warming will lead to trends in ENSO and IOD as projected by climate models, and a related potential increase in wildfire frequency combined with a decreasing lower-stratospheric QBO amplitude is expected in future climate projections. The interplay will change with strong El Niño/negative IOD and La Niña/strong positive IOD likely controlling the lower-stratospheric trace gas distributions and variability more strongly in a future changing climate. Clearly, both ENSO and IOD impact the tropopause height and tropical cold-point temperatures. Further analysis is needed using climate model sensitivity simulations to pinpoint the impact of these future changes in lower-stratospheric trace gases and the related radiative feedback.