Observed changes in Brewer–Dobson circulation for 1980–2018

Previous work has examined the Brewer–Dobson circulation (BDC) changes for 1980–2009 based on satellite Microwave Sounding Unit (MSU/AMSU) lower-stratospheric temperature (TLS) observations and ERA-Interim reanalysis data. Here we examine the BDC changes for the longer period now available (1980–2018), which also allows analysis of both the ozone depletion (1980–1999) and ozone healing (2000–2018) periods. We provide observational evidence that the annual mean BDC accelerated for 1980–1999 but decelerated for 2000–2018, with the changes largely driven by the Southern Hemisphere (SH), which might be partly contributed by the effects of ozone depletion and healing. We also show that the annual mean BDC has accelerated in the last 40 years (at the 90% confidence level) with a relative strengthening of ∼1.7% per decade. This overall acceleration was driven by both Northern Hemisphere (40%) and SH (60%) cells. Significant SH radiative warming is also identified in September for 2000–2018 after excluding the year 2002 when a very rare SH stratospheric sudden warming occurred, supporting the view that healing of the Antarctic ozone layer has now begun to occur during the month of September.


Introduction
The global residual circulation of the stratospherethe Brewer-Dobson circulation (BDC)-consists of a meridional cell in each hemisphere, with air rising across the tropical tropopause, moving poleward, and sinking into the extratropical troposphere (e.g. Holton et al 1995, Plumb 2002, Butchart 2014. Brewer (1949) and Dobson (1956) first proposed this circulation to explain the observed stratospheric water vapor and ozone distributions. By moving air into and out of the stratosphere, the BDC sets the mean age or the residence time of air in the stratosphere (e.g. Hall and Plumb 1994). The BDC affects the distribution and abundance of stratospheric ozone directly by transporting ozone from the tropics to polar regions, and indirectly through its effect on ozone chemistry via temperature changes and transport of other chemical species (e.g. Dobson 1956, Shepherd 2008. The BDC determines the stratosphere-to-troposphere ozone flux on the global and hemispheric scales (e.g. Holton 1990, Holton et al 1995, Appenzeller et al 1996. The BDC directly modulates the tropical tropopause temperatures, tropical tropopause layer cirrus, and stratospheric water vapor, all of which are associated with climate change processes (Birner 2010, Dessler et al 2013, Fu 2013, Tseng and Fu 2017.
In the last 15 years, there has been a surge of interest in the BDC mainly resulting from the development and continuing improvements of general circulation models (GCMs) and chemistry-climate models (CCMs) with detailed representations of the stratosphere (e.g. Pawson et al 2000, Eyring et al 2005, Gerber et al 2012, Butchart 2014. In agreement with the pioneering work by Rind et al (1990), the stratosphere-resolving GCMs and CCMs consistently predict a strengthening of the BDC in response to greenhouse gas (GHG)-induced climate change (e.g. Butchart et al 2006, Garcia and Randel 2008, Li et al 2008, McLandress and Shepherd 2009, Okamoto et al 2011, Bunzel and Schmidt 2013, Lin and Fu 2013, Original content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Oberlander et al 2013), making it one of the robust manifestations of projected GHG-induced climate change. Several modeling studies also find that the BDC becomes stronger (weaker) in response to the ozone depletion (recovery) (e.g. Shindell and Schmidt 2004, Li et al 2008, Oman et al 2009, McLandress et al 2010, Polvani et al 2011, Lin and Fu 2013, Polvani et al 2018, Polvani et al 2019. It is essential that such model predictions be checked against all relevant observations. The BDC strength and its changes cannot be measured directly, but can be inferred from observations. Changes in the BDC strength have been examined on the decadal time scale from observations of age of stratospheric air (e.g. Engel et al 2009, Stiller et al 2012, Engel et al 2017, Stiller et al 2017, but it has become clear now that changes in mean age of stratospheric air do not constrain residual circulation changes since they represent a combination of changes in mixing and in residual transport (Garny et al 2014, Ray et al 2014, Ploeger et al 2015, Dietmuller et al 2017. On the other hand, an accelerated (decelerated) BDC leads to a cooling (warming) of the tropical lower stratosphere and warming (cooling) in high latitudes, due to the close relationship between variations of temperatures and residual vertical velocities in the lower stratosphere (e.g. Yulaeva et al 1994). Observational evidence of a long-term BDC strengthening has been suggested over both the tropics and high latitudes based on consistent changes in lower stratospheric temperatures (Thompson and Solomon 2005, Johanson and Fu 2007, Hu and Fu 2009, Lin et al 2009, Thompson and Solomon 2009, Fu et al 2010, Young et al 2012, Fu et al 2015, Osso et al 2015. Further, Fu et al (2015) examined the BDC changes for 1980-2009 using satellite Microwave Sounding Unit (MSU/ AMSU) lower-stratospheric temperature (T LS ) observations along with ERA-Interim reanalysis data. They found that the annual mean BDC accelerated (at 90% confidence) with a relative strengthening of ∼2.1% per decade and with most of the change coming from the Southern Hemisphere (SH). Here we provide important new information on BDC changes using the full record for 1980-2018 as well as partitioning it into 1980-1999 (ozone depletion) and 2000-2018 (ozone healing) periods. We find that the annual mean BDC accelerated for 1980-1999 but decelerated for 2000-2018, with almost all the changes coming from the SH. We also show that the annual mean BDC accelerated in the last 40 years (at 90% confidence) with a relative strengthening of ∼1.7% per decade, contributed by both Northern Hemisphere (NH) (40%) and SH (60%) cells. Significant SH radiative warming in September for 2000-2018 after excluding the year 2002 is also shown, further supporting a key role for the healing of the Antarctic ozone layer that has now begun to occur particularly during the month of September (see Solomon et al 2016).

Data and methods
The data and methods used in this study closely follow Fu et al (2015), which are only briefly described here. We used the most recent version of the satellite MSU/ AMSU lower stratospheric temperature (T LS ) dataset (v4.1) compiled by the National Oceanic and Atmospheric Administration team (Zou et al 2018). It is gridded at 2.5°latitude by 2.5°longitude, extending from 82.5°N to 82.5°S. We define the tropics as the region from 20°S to 20°N and high latitudes as latitudes from 40°N(S) to 82.5°N(S) (Fu et al 2010, Fu et al 2015. Similar results were obtained by employing the T LS datasets from the Remote Sensing System team (Mears and Wentz 2017)  The observed high-latitude T LS trends were separated into a dynamical component due to the change of the BDC and a radiative component due to the change of radiatively active trace species (Fu et al 2010, Fu et al 2015 as well as solar activity. A regression of gridded T LS data was performed upon the corresponding eddy heat flux index time series for each month and over each hemisphere. The attribution of the T LS trends to changes in the BDC was derived by multiplying the regression maps with the linear trends obtained in the eddy heat flux index as in Fu et al (2015). The T LS trends due to the BDC changes were averaged over high latitudes in each hemisphere for each month. The radiative component, averaged over 40-82.5°S(N), was derived as the observed total T LS trend minus the dynamical component, averaged over the same regions. The confidence intervals in the derived trends were determined using a Monte Carlo method (Fu et al 2015). The method for separation of observed T LS trends over the tropics into the radiative and BDC change-induced dynamical components will be described when the results are given in the next section. The radiative component of tropical T LS trends derived from the MSU/AMSU observations along with the ERA-Interim reanalysis is independently validated both in terms of its small seasonal dependence and annual mean value (Fu et al 2015). See Lin et al (2009), Fu et al (2010 and Fu et al (2015) for more details on the justification of using the reanalysis eddy heat flux trends.  and 1980-2018 (right), which are driven by a combination of the direct radiative effects due to atmospheric composition changes and solar activity, and the BDC-induced temperature changes (Fu et al 2015). Radiative cooling trends prevail for 1980-1999 because of ozone depletion, GHG increases, stratospheric aerosols from the El Chichon volcanic eruption in March 1982 and Pinatubo in June 1991, and solar activity (see time series of stratospheric aerosols and solar activities in figure SM1 is available online at stacks.iop.org/ERL/14/114026/mmedia). There is a particularly strong cooling trend poleward of 60°S during October-December, which was caused by the Antarctic ozone hole (Solomon 1999). A slightly positive trend over the tropics in March indicates a strong dynamical warming that cancels direct radiative cooling there, and is related to strong dynamical cooling in the NH high latitudes in the same month. The dynamical cooling in March in the NH high latitudes could be related to a delay in the final warming (i.e. the polar vortex breakdown). The T LS warming in the SH high-latitude winter/spring seasons and in the NH high-latitude winter is coupled to enhanced cooling in the tropics and is also consistent with BDC changes. Similar T LS trend patterns were observed for 1980-2018 (right panel in figure 1) but with smaller magnitudes (also see figure 2 in Fu et al 2010Fu et al for 1980Fu et al -2008. The behavior for 2000-2018 is strikingly different from the first period. It is known that there is much less volcanic aerosol effect then, and a radiative warming is expected due to the ozone healing  -2018 and 1980-1999 suggests that the BDC changes might have opposite signs for these two periods, a quantitative estimate of the BDC change requires the separation of the dynamical and radiative components in T LS trends, which will be discussed next. The blue lines in figure 3 show the T LS trends due to BDC changes (see Data and Methods section), averaged over high latitudes of 40-82.5°S and 40-82.5°N, versus month, while the red lines show observed T LS trends over the tropics. A strong anti-correlation between observed tropical T LS trends and high-latitude T LS dynamical components indicates that the monthly dependence of observed tropical T LS trend is indeed largely driven by the BDC changes for all three periods considered. Following Fu et al (2015), the observed T LS trends in the tropics are next separated into two parts as T LS (Tropics)=aT LS (High-Lat, Dyn)+[T LS (Tropics) − aT LS (High-Lat, Dyn)]. Here a is a coefficient obtained from orthogonal least squares fitting between the observed tropical T LS trends, i.e. T LS (Tropics), and high-latitude T LS dynamical components, i.e. T LS (High-Lat, Dyn). The observed T LS trends in the tropics are thus separated into the BDC change-induced dynamical components and empirically derived radiative components (i.e. black lines in figure 3), corresponding to the first and second terms on the right-hand side of the above equation, respectively. Figure 3 shows a weakening of the BDC in boreal spring and a strengthening in other seasons for 1980-1999 but opposite behavior for 2000-2018. For 1980-2018, the BDC is strengthened in July-January but shows little change in February-June. The derived annual mean radiative components of tropical T LS are −0.29, −0.15, and −0.19 K/decade, respectively, for 1980-1999, 2000-2018, and 1980-2018. The BDC change-induced T LS annual mean trend and the contributions from the NH and SH cells are   shown in figure 4 over the tropics. The tropical NH and SH contributions to the BDC trends were computed by multiplying the corresponding high-latitude NH and SH dynamical contributions by the coefficient a. The BDC change-induced T LS annual mean trends in the tropics are −0.10, 0.11, and −0.08 K/decades for 1980-1999, 2000-2018, and 1980-2018, respec-tively, indicating a strengthening, weakening, and strengthening of the BDC. The BDC changes in the first and second periods are almost entirely caused by the SH cells ( figure 4). For the overall period 1980-2018, the NH and SH cells contribute 40% and 60% of the total BDC change, respectively. The dynamical cooling in tropical T LS in the past four decades is significant at the 90% confidence interval, as in 1980-2009(Fu et al 2015. However, the NH contribution has now become significant at the 90% confidence interval for 1980-2018 while only the SH contribution was significant at the 90% level for 1980-2009(Fu et al 2015. It should be noted that the trends for the combined period of 1980-2018 cannot be derived from those for the separate periods (e.g. figure 4) because of the evident nonlinear temporal variations and changes in behavior.

Results
Using the relationship between tropical residual vertical velocity (w * ) at 70 hPa and T LS from Fu et al (2015), a dynamical cooling of −0.08 K/decade in tropical T LS trend corresponds to a mean change in w * of 0.0046 mm/s/decade. Noting that the annual mean w * at 70 hPa over the tropics is 0.288 mm s −1 from the ERA-Interim reanalysis, the relative strengthening of the BDC in terms of tropical residual vertical velocity at 70 hPa is 1.7% in the last 40 years, which is smaller than 2.1% for 1980-2009(Fu et al 2015. This is consistent with a reduced role of ozone due to including its longer recovery era. Figure 5 shows the monthly observed T LS trends (left) and their dynamical (middle) and radiative (right) contributions for 1980-1999 (upper), 2000-2018 (middle), and 1980-2018 (lower) over SH high latitudes (dotted lines), NH high latitudes (dashed lines), and combined high latitudes (solid lines). The radiative contribution is obtained as the residual of the total minus the dynamical contribution. Note that the y-axis scale in figure 5 for 1980-2018 is finer than those for other two periods to more clearly show the results.
For the NH cell during 1980-1999, the BDC weakened in the Spring but strengthened in the Winter (Fu et al 2010, Free 2011, Fu et al 2015. For 2000-2018, the opposite occurred. For the entire period (i.e. 1980-2018), the NH cell displayed small changes in the Spring but accelerated in the Winter. The weakening (strengthening) of the BDC NH cell in the Spring for 1980-1999 (2000-2018) thus seems to be associated with the natural variability. This is supported by the fact that the NH radiative components do not change sign in March. Thus, the longer record The SH cell of the BDC is strengthening, weakening, and strengthening, respectively, for 1980-1999(large in August), 2000-2018(large in October), and 1980-2018. show that the BDC changes mainly occur in the Winter and Spring seasons for both SH and NH cells during all three periods considered.
With the MSU/AMSU data in the last four decades (Santer et al 2019), the response of the BDC NH cell to the GHG increases since 1980 may be emerging. Although large dynamical trends are found in boreal spring over shorter periods, trends over the whole period remain very small for this season (figure 5). The SH cell is still the main contributor to the total BDC change in magnitude, but becomes statistically insignificant (figure 4), unlike that in the period of 1980-2009(Fu et al 2015. This could partly be because BDC weakening in response to ozone healing cancels more of the strengthening during the ozone depletion period when the full record is considered.

Summary and conclusions
This study examines the BDC changes in the past four decades  as well as in the ozone depletion (1980)(1981)(1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999) and ozone healing (2000-2018) periods based on satellite MSU/AMSU T LS observations and ERA-Interim reanalysis data. We show that the annual mean BDC has accelerated over 1980-2018 (at the 90% confidence level), as in 1980-2009(Fu et al 2015. However, the NH contribution is significant at the 90% confidence interval for 1980-2018 while it was not for 1980-2009(Fu et al 2015. This is because the weakening and strengthening of the BDC NH cell in the boreal Spring for the first and second periods, which are largely related to natural variability, cancel each other. At the same time, the SH contribution is significant at the 90% level for 1980-2009(Fu et al 2015 but becomes insignificant for 1980-2018, a period of longer ozone healing. We derived a relative strengthening of ∼1.7% per decade (40% from the NH cell and 60% from the SH cell) in the last four decades, which is smaller than 2.1% per decade for 1980-2009. The smaller relative strengthening in the former period could be related to the ozone healing since the beginning of the 21st century.
The annual mean BDC has accelerated for 1980-1999 but decelerated for 2000-2018. Almost all of the change comes from the SH cell for both periods, which is because the weakening (strengthening) of the NH cell in the boreal Spring largely cancels the strengthening (weakening) of the NH cell in the boreal Winter for 1980-1999 (2000-2018). The enhanced SH radiative warming in September for 2000-2018, which is significant at the 90% confidence level after excluding the year 2002, supports Solomon et al (2016) that healing of the Antarctic ozone layer has now begun to occur during the month of September.
With the MSU/AMSU data in the last four decades, the response of the BDC NH cell to GHG increases in the past four decades may be emerging. The observationally-derived changes of the BDC in the ozone depletion and healing periods also support the GCM and CCM predictions.