Skip to main content
NCBI home page
NASA Author Manuscripts logoLink to NASA Author Manuscripts
. Author manuscript; available in PMC: 2020 Jul 16.
Published in final edited form as: J Geophys Res Atmos. 2019 Jul 16;124(13):6669–6680. doi: 10.1029/2018JD029516

Large impacts, past and future, of ozone-depleting substances on Brewer-Dobson circulation trends: A multi-model assessment

LM Polvani 1,2,3, L Wang 2,4,5, M Abalos 6, N Butchart 7, MP Chipperfield 8, M Dameris 9, M Deushi 10, SS Dhomse 8, P Jöckel 9, D Kinnison 3, M Michou 11, O Morgenstern 12, LD Oman 13, DA Plummer 14, KA Stone 15,16,*
PMCID: PMC6800672  NIHMSID: NIHMS1538740  PMID: 31632893

Abstract

Substantial increases in the atmospheric concentration of well-mixed greenhouse gases (notably CO2), such as those projected to occur by the end of the 21st century under large radiative forcing scenarios, have long been known to cause an acceleration of the Brewer-Dobson circulation (BDC) in climate models. More recently, however, several single-model studies have proposed that ozone-depleting substances might also be important drivers of BDC trends. As these studies were conducted with different forcings over different periods, it is difficult to combine them to obtain a robust quantitative picture of the relative importance of ozone-depleting substances as drivers of BDC trends. To this end we here analyze – over identical past and future periods – the output from 20 similarly-forced models, gathered from two recent chemistry-climate modeling intercomparison projects. Our multi-model analysis reveals that ozone-depleting substances are responsible for more than half of the modeled BDC trends in the two decades 1980–2000. We also find that, as a consequence of the Montreal Protocol, decreasing concentrations of ozone-depleting substances in coming decades will strongly decelerate the BDC until the year 2080, reducing the age-of-air trends by more than half, and will thus substantially mitigate the impact of increasing CO2. As ozone-depleting substances impact BDC trends, primarily, via the depletion/recovery of stratospheric ozone over the South Pole, they impart seasonal and hemispheric asymmetries to the trends which may offer opportunities for detection in coming decades.

1. Introduction

The meridional circulation in the stratosphere, known as the Brewer-Dobson circulation (Brewer, 1949; Dobson, Harrison, & Lawrence, 1929, hereafter BDC), is fundamental to our understanding of the distribution of ozone and other trace gases in the middle atmosphere, as well as their exchange across the tropopause. Documenting past and future BDC trends, and attributing such trends to specific anthropogenic emissions, is thus an integral part to understanding man-made climate change.

Although BDC trends over the last several decades have proven difficult to establish from observations (see Chapter 4 of WMO, 2014, for a recent review), it has long been known from modeling studies (e.g. Butchart & Scaife, 2001; Garcia & Randel, 2008; Rind, Suozzo, Balachandran, & Prather, 1990) that increasing concentrations of well-mixed greenhouse gases (GHG), primarily CO2, cause an acceleration of the BDC. This result is very robust, as it has been validated across dozens of models in several model intercomparison projects (Butchart et al., 2010, 2006; Hardiman, Butchart, & Calvo, 2014). However, that result was largely established by contrasting the BDC in the late 20th and late 21st centuries, over which period CO2 concentrations increase monotonically and roughly double (for the modeling scenarios analyzed in most BDC trend studies). Over the satellite era (1979-present), however, CO2 concentrations have increased by much smaller amounts (a little over 20%) allowing for other anthropogenic emissions to play an important role.

Halogenated ozone-depleting substances (ODS), in particular, have been highlighted in a number of recent studies as being potentially important drivers of BDC trends. On the observational side, analyzing temperature trends from the microwave sounding units, Fu, Lin, Solomon, and Hartmann (2015) uncovered a major hemispheric asymmetry in the dynamically-driven cooling of the lower stratosphere, with statistically significant trends being present only in the Southern Hemisphere. Since GHG are well mixed, this asymmetry points to ozone depletion, and hence to ODS emissions, as a key anthropogenic cause of BDC trends in the late 20th century. More recently, Polvani, Wang, Aquila, and Waugh (2017) have argued that the lack of cooling in the tropical lower stratosphere since the turn of the century (Seidel et al., 2016) contrasted with the substantial cooling in the last two decades of the 20th century, is a clear indicator that ODS are driving tropical upwelling: recall that both temperature and ozone in the lower stratosphere are closely linked to tropical upwelling through adiabatic cooling and upward advection of low-ozone air, respectively (see, e.g., ?).

On the modeling side too, several studies have reported the key role of ODS as drivers of BDC trends. Oman, Waugh, Pawson, Stolarski, and Newman (2009) found ozone depletion to be the major factor causing decreases in age of air, in their model, in the second half of the 20th century. Oberländer-Hayn, Meul, Langematz, Abalichin, and Haenel (2015), comparing time-slice model integrations at year 1960 and year 2000 conditions, reported the BDC acceleration caused by ODS over that period to be of equal magnitude to the acceleration caused by GHG. A similar result, but focusing on mean age of stratospheric air, was reported by Li, Newman, Pawson, and Perlwitz (2018), using time-varying forcings over the period 1960–2010. The importance of ODS emissions for BDC trends in the last few decades was also noted by Garfinkel, Aquila, Waugh, and Oman (2017). And, while seeking to understand the intermodel spread in ozone simulations in the most recent chemistry-climate model intercomparison project, Morgenstern et al. (2018) found that age of air trends are driven by both GHG and ODS, with increases in the latter driving a decrease in age of air in most models (except over the summer South Pole).

As for the coming decades, Polvani, Abalos, Garcia, Kinnison, and Randel (2018) recently found that age of air trends over the period 2000–2080 will be four times smaller than over the period 1960–2000 as a consequence of the Montreal Protocol, adding to earlier work (Lin & Fu, 2013; McLandress et al., 2010; Oberländer, Langematz, & Meul, 2013). Taken as a whole, these studies offer good evidence that ODS are indeed key forcings of past and future BDC changes. However, these studies comprise a relatively small set of models. Furthermore, since the models were forced in different ways and their output was analyzed over different periods, they cannot be quantitatively compared.

Hence the goal of this paper: to quantify the importance of ODS in driving BDC trends, in both past and future decades, using a large number of models with similar forcing scenarios. We also seek to directly contrast the importance of ODS and GHG, and to provide an estimate of the uncertainty in the ODS/GHG impact on BDC trends, as we quantify the spread across the models.

Our work builds on the study of Lin and Fu (2013) who analyzed BDC trends in 13 models from the Stratospheric Processes and their Role in Climate (SPARC)/Chemistry-Climate Model Validation project, Phase 2 (hereafter CCMVal-2, Eyring et al., 2008), and discussed the role of ozone depletion/recovery on those trends. We expand on their findings in three ways. First, we here analyze a larger set of models, adding those from the recent Chemistry-Climate Model Initiative, Phase 1 (hereafter CCMI, Eyring et al., 2013; Morgenstern et al., 2017) to the CCMVal-2 models. Second, in addition to the scenarios where all forcings vary with time, we examine the sensitivity integrations to identify and contrast the separate impacts of ODS and GHG. And, finally, we here shift the emphasis from ozone to ODS, keeping in mind that ozone is naturally abundant in the atmosphere, and ODS are the anthropogenic emissions that need to be contrasted with GHG.

The paper is laid out as follows. In Section 2 we describe the models and the scenario forcings analyzed here. In Section 3 we contrast past and future BDC trends: as GHG increase monotonically from the mid-20th century to the end of the 21st century, whereas trends in the concentrations the ODS change sign in the late 1990s as a consequence of the Montreal Protocol, any difference in forced trends before and after the year 2000 can be unambiguously attributed to ODS. In Section 4, we exploit the existence of the sensitivity integrations, and quantitatively explore the relative importance of ODS and GHG in affecting BDC trends. A summary and discussion close the paper in Section 5.

2. Methods

We here analyze a total of 20 models for which output was made available, combining both the CCMVal-2 and CCMI projects. These models, listed in Table 1, are extensively documented in Morgenstern et al. (2010) and Morgenstern et al. (2017) for the CCMVal-2 and CCMI projects, respectively. Unless otherwise noted, we consider only one run for each model and scenario, in order to weigh all models equally. In a nutshell, with few exceptions, these are atmospheric general circulation models with a well-resolved stratosphere, i.e. with a model top near or above 0.01 hPa. Typical resolutions are 2–3 degrees in latitude/longitude, and 50–90 levels in the vertical direction. It should be noted that all models used here include an interactive chemistry module for stratospheric ozone and, in most CCMI models, comprehensive tropospheric chemistry as well.

Table 1.

The models analyzed in this study, and their data availability (indicated by an X).

AoA w¯*
model name All Forcings Fixed ODS/GHG All Forcings Fixed ODS/GHG
CCMVal-2
CAM3.5 X X
CCSRNIES X X
CMAM X X X X
GEOSCCM X X
LMDZrepro X X X
MPRI X X X X
NIWA-SOCOL X X
SOCOL X X X X
UMUKCA-METO X X
UMUKCA-UCAM X X
WACCM X X
CCMI
ACCESS CCM X
WACCM X X X X
CMAM X X X X
CNRM-CM5–3 X
EMAC L47 X X
EMAC L90 X X
GEOSCCM X X
HadGEM3-ES X
MRI-ESM1r1 X X
NIWA-UKCA X X X X
UMSLIMCAT X X
Open in a new tab

Three sets of integrations are considered here, each set corresponding to a different forcing scenario. The first set, which we will refer to as the “All Forcings” integrations, simulate the climate system over the period 1960–2100 with models forced by time-varying concentrations of GHG and ODS. The sea surface temperature and sea ice concentrations (SSTs) are typically prescribed from either observations or model output although, in a few instances, the atmospheric model is fully coupled to interactive ocean and sea ice models. The chemistry-climate modeling projects technically refer to this scenario as the “REF-B2” and “REF-C2”, for CCMVal-2 and CCMI respectively (Eyring et al., 2008, 2013).

The anthropogenic forcings used for these All Forcings scenarios are quite similar between CCMVal-2 and CCMI, but not exactly identical. For ODS, both REF-B2 and REF-C2 prescribe them following the same A1 scenario (WMO, 2011), but for GHG the specifications are a little different. The newer REF-C2 models follow the RCP6.0 scenario (Meinshausen et al., 2011) for GHG, whereas the older REF-B2 models follow the SRES A1B scenario (Nakicenovic et al., 2000). Fortunately this is not too problematic, as RCP6.0 and SRES A1B forcings from CO2 and N2O are very similar between 1960 and 2100. As noted in Morgenstern et al. (2017), the main discrepancy is found in CH4 forcing, whose volume mixing ratio peaks at about 2.4 ppmv in the late 21st century for REF-B2, but only at 1.95 ppmv for the REF-C2 forcings. To the degree that CO2 constitutes the bulk of GHG forcing, this 20% discrepancy in methane is a minor issue, as it results in a total radiative difference in GHG forcing of a few percent between CCMVal-2 and CCMI.

The other two sets of model integrations analyzed here are used to quantitatively establish the respective contributions of ODS and GHG. These sets follow scenarios in which either ODS or GHG are kept fixed at year 1960 levels, all else being identical (except for the SSTs which, for consistency, are also kept at ~1960 levels when GHG are fixed, in models with prescribed SSTs). In these paper we refer to these sets of integrations as the single-fixed-forcing runs, because that expression immediately conveys the idea that only one forcing is being altered; a more precise expression would be all-but-one-forcing, but that is unecessarily cumbersome. For maximum clarity these two sets are here referred to as the “Fixed ODS” and “Fixed GHG” integrations. And, for the sake of full reproducibility, we note that these integrations correspond to the SCN-B2b and SCN-B2c scenario runs of the CCMVal-2 project (Eyring et al., 2008), and the SEN-C2-fODS and SEN-C2-fGHG scenario runs of the CCMI project (Eyring et al., 2013), respectively.

The effects of ODS (or GHG) on the BDC are here quantified by computing differences between the All Forcings runs and Fixed ODS (or GHG) runs. Note that in the Fixed ODS runs both the effect of ODS on the ozone layer and the radiative effect of ODS remain constant with time. We are here unable to separately quantify the importance of these two mechanisms for BDC trends, as ozone is not prescribed independently of ODS in these models (it is computed interactively). However, as originally noted in Polvani et al. (2018), and as we will show below for the CCMI models, the effects of ODS on BDC trends exhibit clear seasonal and hemispheric asymmetries that are aligned with ozone changes, indicating that the direct radiative forcing of ODS is relatively unimportant for BDC trends. We have recently confirmed this fact using the WACCM model, as detailed in Abalos et al. (2019).

Unfortunately, the majority of models participating in the CCMVal-2 and CCMI projects did not contribute model output for the single-fixed-forcing scenario integrations. In fact, as shown in Table 1, only a handful of models for each project contributed data for the two variables of interest here: the mean Age of Air (AoA) and the residual vertical velocity (w*). To obtain the strongest possible results, therefore, we decided to combine model output from the two projects. The models themselves are quite similar, basically one generation apart at most, and so are the forcings, as explained above.

Finally, unlike CO2 concentrations, ODS concentrations are not monotonically increasing from 1960 to 2100. Hence, it is crucial to consider separately the pre- and postMontreal Protocol periods, as it makes little sense to compute a single linear trend in model responses over periods during which the forcing trends change sign. Since the ozone-depleting power of ODS – as quantified by equivalent effective stratospheric chlorine (see, e.g. Newman, Daniel, Waugh, & Nash, 2007) – peaked around 1998, we follow Polvani et al. (2018) and simply choose the year 2000 as the break-point. Unlike that study, however, we are here forced to select the shorter 1980–2000 period for the pre-Montreal trends, since the AoA tracer needs at least a decade or two to come into equilibrium, and the model integrations were started around 1960. For the post-Montreal period we follow Polvani et al. (2018) and use 2000–2080, to avoid the reversal of CH4 trends in the forcing scenarios in the final decades of the 21st century. For simplicity we will refer to 19802000 and 2000–2080 as “the past” and “the future”, respectively, in the remainder of the paper.

3. Past and future impact of ODS on BDC trends

The main goal of our paper is to bring out the effect of ODS on the BDC and for this the single-fixed-forcing runs are crucial. However, we postpone analyzing those to the next section, as it is instructive to start by looking at the All Forcings runs. We do this for two reasons. First, we have many more models available for the All Forcings runs than for the single-fixed-forcing runs (see Table 1). Second, the All Forcings runs are the most realistic simulations available, and thus represent the closest model approximation to the actual atmospheric evolution.

As it happens, it is relatively easy to bring out the role of ODS in the All Forcings runs, since only ODS trends change dramatically around the year 2000 whereas GHG trends do not change sign at that time (and, in fact, at no other time over the period of interest). Assuming internal decadal variability to be relatively small in the stratosphere (we will illustrate this later in the paper), any changes in BDC trends before/after the year 2000 can therefore be immediately attributed to changes in ODS concentrations, and thus to the Montreal Protocol. A similar approach was used in Barnes, Barnes, and Polvani (2014) to bring out the ODS impacts on the tropospheric circulation and surface climate in the All Forcings simulations of the models that participated in the Climate Model Intercomparison Project, Phase 5 (CMIP5, Taylor, Stouffer, & Meehl, 2012).

We start, therefore, by simply contrasting the past and future AoA trends, computed using simple linear regression, for the All Forcings integrations of 20 models in Table 1: the multi-model mean values are shown in the left and right panels of Figure 1, for the past and future periods respectively. It is immediately clear from the color difference that future AoA trends are substantially smaller than past trends, and this can be directly attributed to the phase-out of ODS emissions by the Montreal Protocol. Quantitatively speaking, as reported in the top row of Table 2, we find that future AoA trends will be less than half the values of past trends.

Figure 1.

Figure 1.

Open in a new tab

Multi-model mean, annual mean trends (shading) and climatology (contours) of AoA for the All Forcings runs (combined CCMVal-2 REF-B2 and CCMI REF-C2) for (a) 1980–2000 and (b) 2000–2080. The stippling represents the trends significant at the 95% interval by a two-tailed t-test. For the climatology, the coutour interval is 6 months; the zero contour is not shown.

Table 2.

Past and future, multi-model (with the number of models in parenthesis) average of annual and global mean AoA trends, vertically averaged between 1 hPa and 100 hPa, in units of years/decade. All values are statistically significant at the 95% confidence interval by a two-tailed t-test (with bounds indicated after the ± symbols). The seven models in the last three rows are those for which the single-fixed-forcing integrations are also available (see Table 1).

AoA trends
(years/decade)		 past
(1980–2000)		 future
(200–2080)
All Forcings (20) −0.118 ± 0.033 −0.052 ± 0.006
All Forcings (7) −0.117 ± 0.035 −0.041 ± 0.005
Fixed ODS (7) −0.053 ± 0.037 −0.055 ± 0.006
Fixed GHG (7) −0.071 ± 0.037 0.009 ± 0.005
Open in a new tab

To give a sense of the spread across models, we plot in Figure 2 the past and future AoA trends for each of the 20 models analyzed. We show the annual mean, global mean AoA trends, averaged over the 100–1 hPa layer, on the left panel: the solid bars show the past trends, the empty bars the future trends. For clarity, the ratio of past to future trends is shown by the black bars in the right panel. With a few exceptions, the vast majority of the models show a clear reduction in AoA trends, confirming the original finding of Polvani et al. (2018), who used the CCMI-WACCM model. That model appears to show the strongest reduction. It is important, however, to keep in mind that only one run for each model was used in Figure 2 (in order to weigh all the models equally). Hence, one cannot determine the relative amplitude of the forced response in the models from this figure, as internal variability may be large, and needs to be removed by averaging several runs. The role of internal variability will be discussed in Section 4 below.

Figure 2.

Figure 2.

Open in a new tab

Past (1980–2000; filled green bars) and Future (2000–2080; empty green bars) global-mean, annual mean AoA trends (left) and their ratio (right), vertically averaged between 100 hPa and 1 hPa for CCMVal-2 REF-B2 and CCMI REF-C2 (All Forcings) simulations.

In addition to the past/future differences, one can exploit the strong seasonal cycle and hemispheric asymmetry of ozone depletion to bring out the key role of ODS on past and future BDC trends. As noted by Polvani et al. (2018), the impact of ODS is primarily (but not uniquely) mediated by the formation of the ozone hole. Although such ODS impact is sufficiently large to survive averaging over all months of the year, as we have just shown, it can be greatly amplified by considering the hemispheric and seasonal asymmetries. We illustrate these in Figure 3, where the multi-model mean, zonal mean trends for the residual vertical velocity w* are shown (the grey stippling is used to mark regions where trends are statistically significant). The climatology is shown by the black contours, solid indicating upwelling (in the tropics) and dashed downwelling (at high latitudes), separated by the dash-dotted zero contour.

Figure 3.

Figure 3.

Open in a new tab

Past (left) and future (right) multi-model mean, zonal mean w¯* trends (shading) in the annual (top row), DJF (middle row) and JJA (bottom row) means. The black contours show the corresponding climatology, with contour lines at −100 −50 −30 −20 −10 0 10 20 30 50 100 km/year, positive in solid, zero in dash-dotted, and negative in dashed. The grey stippling shows where trends are significant at the 95% interval, using a two-tailed t-test.

Let us start by considering the top left panel, which shows the annual mean w¯* trends for the past period: recall that these trends are computed from the realistic All Forcings simulations. For that period, only the Southern Hemisphere shows large statistically significant trends at high-latitdes. This hemispheric asymmetry points to ODS, not GHG, since the latter are well mixed. This is confirmed by the middle and bottom left panels, which show the seasonal DJF and JJA trends over the same period. It is clear that the large BDC trends – the increased downwelling at high southern latitudes – come from austral summer, the season where ozone depletion is known to mostly affect the atmospheric circulation, as widely documented in the literature (see, e.g., figure 4 of Polvani et al., 2018).

Figure 4.

Figure 4.

Open in a new tab

Left panel: past (1980–2000; filled bars) and future (2000–2080; empty bars) global-mean, zonal-mean, annual mean AoA trends, vertically averaged between 100 hPa and 1 hPa for CCMVal-2 and CCMI simulations for All Forcings (green), Fixed ODS (magenta), and Fixed GHG (orange) scenarios. Right panel: future/past trend ratio for each scenario, color-coded as in the left panel.

The large role of ODS on the BDC is further confirmed by contrasting past and future w¯* trends. Starting from the annual mean in the top row, one can see a very large hemispheric asymmetry in future trends (right panel): whereas the tropics show increased upwelling and the northern high-latitudes increased downwelling, as one expects from increases in GHG, the southern high-latitudes show reduced downwelling. This deceleration of the BDC over the South Pole in the 21st century comes mostly from DJF, as seen the middle right panel. We can easily attribute it, therefore, to the closing of the ozone hole, and thus to reduced ODS concentrations. The future DJF trends are very different from the JJA ones, shown in the bottom right panel: in austral winter, when GHG are the key forcing, we see a BDC acceleration at the southern high-latitudes as well as in the tropics, as expected.

4. The relative importance of ODS and GHG for BDC trends

As we have shown, the large impact of ODS on BDC trends in the All Forcings runs, both for the past and for the future, is readily apparent from the seasonal and hemispheric asymmetries associated with the opening/closing of the ozone hole, which is the primary pathway through which ODS affect the BDC. In those runs, however, it is not possible to quantify the relative importance of the ODS and GHG forcings. To do that we need to analyze the single-fixed-forcing runs, which we explore in this section. Unfortunately, even after combining the CCMVal-2 and CCMI project, AoA output is available for only 7 models (and 8 models for w¯*; see Table 1). While not a huge set, this is sufficient to gain a first impression of the relative importance of ODS and GHG forcings from a multi-model perspective.

We start by discussing the AoA trends, past and future, shown for the single-fixed-forcing and the All Forcings runs in Figure 4. Let us first consider the past trends (1980–2000), which are illustrated by the solid bars. One can see that, in most models, the AoA trends in the Fixed GHG runs (orange solid bars) are comparable to the trends in the Fixed ODS runs (magenta bars). In the multi-model mean, the past impact of ODS appears to be somewhat larger than the impact of GHG (by about 30%, see Table 2), but that difference is not statistically significant, due to the spread across models. We conclude that ODS and GHG have contributed in roughly equal parts to the past AoA trends: this confirms the recent studies of Oberländer-Hayn et al. (2015) and Li et al. (2018), who used different models and different periods.

Next, we contrast future and past trends, in the single forcing runs. Consider the empty orange bars in the left panel, and note that they are negative for 6 of the 7 models analyzed here. As GHG concentrations are fixed for those runs, it is the ODS reversal as a consequence of the Montreal Protocol that causes AoA to increase over the period 2000–2080 in those runs. This ODS forcing, then, opposes the continuous AoA decrease caused by increasing GHG amounts (empty purple bars). This results in future trends that are considerably smaller than in the past, as one can see by contrasting the empty and solid green bars. This multi-model analysis, therefore, confirms the findings reported by Polvani et al. (2018).

Finally, the right panel of Figure 4 shows the future/past trend ratio. For the All Forcings runs, the multi-model mean ratio is 0.35 (see Table 2), a reduction of trends by nearly two thirds (for this smaller subset of 7 models), as the reversed ODS trends cancel the effect of increasing GHG. Finally, as a sanity check, note from the purple bar that the future/past ratio for the Fixed ODS runs is very close to one in the multi-model mean, as one might expect when GHG are the sole forcing acting on the stratospheric circulation, since GHG concentrations increase almost linearly in the scenarios used to force these model runs.

To appreciate that the results presented here are not fundamentally dependent on our choice of past/future periods, we show the entire time series of AoA for two of the models in Table 1: the Canadian Middle Atmosphere Model (CMAM, Grandpré et al., 2000; McLandress et al., 2010; Scinocca, McFarlane, Lazare, Li, & Plummer, 2008) and the Whole Atmosphere Community Climate Model (WACCM, Garcia, Smith, Kinnison, Cámara, & Murphy, 2017; Marsh et al., 2013; Solomon, Kinnison, Bandoro, & Garcia, 2015). For these two models runs are available for all the scenario forcings and, more importantly, a small ensemble of three runs is available for each scenario. To be clear: each member of these small ensembles is identically forced (according to the appropriate scenario), and the only difference between the three members resides in a very small perturbation in the model’s initial condition, in the spirit of Deser, Phillips, Bourdette, and Teng (2012). This allows us to quantify the importance of internal variability, and determine to what degree it is able to alter the forced response of the BDC to the various forcings.

The complete 1960–2100 time series of AoA are shown in Figure 5, for CMAM and WACCM, on the left and right panels, respectively. The colors indicate the different scenarios, as shown in the legend. First, in the All Forcings runs (green curves) one can see a clear kink around the year 2000: this abrupt change in slope, of course, is due to ODS emission, and is readily apparent in the orange curves showing the runs with GHG fixed at 1960 levels and only ODS concentrations varying in time. Second, in the absence of ODS forcing (red curves) one can see that AoA increases almost linearly from 1960 all the way to 2100, as GHG concentrations increase monotonically over the entire 140-year period. Third, the area between the green and purple lines quantifies the impact of ODS on AoA trends: it is clear that the quantitative impact is model dependent (larger in WACCM than in CMAM), and this very fact motivates the present study. In both cases however, we see that the green curves are steeper than the purple curves before 2000, and less steep after 2000. This is the effect of ODS on the BDC, which is robust across models. Finally, a technical note: Figure 5 shows that AoA does not vary linearly prior to 1980. This is because it takes time for tracer to become equilibrated throughout the entire stratosphere and for the forced response to emerge from internal variability: hence our choice to evaluate the past trends starting at 1980 in these runs.

Figure 5.

Figure 5.

Open in a new tab

Annual and global mean AoA time series averaged over 100–1 hPa for (a) CCMVal-2 CMAM and (b) CCMI WACCM. Three ensemble members are show in thin curves for each scenario, while the 1980–2000 and 2000–2080 linear trends of the ensemble mean are plotted in thick lines. Color-coding as per the caption, consistent with Figure 4.

A second important result can be gathered from in Figure 5. Notice that for both models, and for each forcing scenario, the 3 members of each ensemble track each other quite closely. From this we conclude that for multi-decadal BDC trends the role of internal variability appears to be relatively small in these models. For short periods, needless to say, the actual numerical value of trends can vary considerably across different ensemble members of the same model: for instance, for the 1980–2000 period, the AoA trends for the three WACCM members in Figure 5 are −0.200, −0.096 and −0.079 years/decade, showing a considerable spread. This suggests that a substantal fraction of the intermodel spread seen in Fig. 2 might arise from internal variability, rather than from intrinsic model differences. Nevertheless, Figure 5 clearly suggests that large ensembles may not be needed to robustly quantify AoA trends, as is the case, for instance, for the global mean surface temperature (Deser et al., 2012).

We conclude by illustrating the entire vertical structure w¯* trends in the tropics, to highlight the key role of ODS on tropical upwelling, as suggested by Polvani et al. (2017) on the basis of satellite temperature observations. Trends in w¯* from 100 to 5 hPa, averaged between the turnaround latitudes (i.e. over the region of tropical upwelling), are plotted in Figure 6, for each of the forcing scenarios, and averaged across the 8 models for which output is available for all scenarios (see Table 1). In each panel we plot one curve for the past trends (solid) and one for the future trends (dashed), with the shaded envelope indicating the 95% confidence interval from a two-tailed t-test.

Figure 6.

Figure 6.

Open in a new tab

Vertical profile of multi-model (CCMVal-2 + CCMI: 8 models) mean seasonal trends of w¯* in DJF (top) and JJA (bottom) for the periods of 1980–2000 (solid) and 2000–2080 (dashed) for three scenarios: All Forcings (left), Fixed ODS (middle), and Fixed GHG (right). Here w¯* is computed by averaging over the tropical upwelling zone, i.e., between the turning points, of the corresponding season.

Let us start with the top row, where the DJF trends are shown. In the All Forcings simulations (left panel) one can see that tropical upwelling in the future is considerably smaller than in the past, over most of the stratospheric column. The large values of past trends are due, to a considerable extent, to ODS: this is seen by the fact that when the ODS forcing is absent (middle top panel), GHG alone are not sufficient to yield a statistically significant trends (Lamarque & Solomon, 2010). Furthermore, the top left panel shows that DJF upwelling trends caused by ODS reverse sign after the year 2000, again over the entire stratospheric column: as a consequence the overall future DJF trends are severely reduced. Finally, the corresponding JJA trends are shown in the bottom row: for the future, statistically significant trends are seen in the All Forcings simulations, and these come almost entirely from GHG (contrast the lower middle and left panels). Interestingly enough, past trends are not statistically significant in that season: the GHG forcing alone, over the relatively short 1980–2000 period (only two decades) is not sufficient to produce a significant response. It is only in DJF, when the ODS contribute a substantial forcing, that a statistically significant trend is seen in these models.

5. Summary and discussion

Using a large set of chemistry-climate models, forced in a consistent manner and analyzed over identical time-periods (for the past and for the future, separately), we have shown that ODS are important drivers of trends in the stratospheric circulation. We have found that ODS have contributed at least 50% of the forced response in the late 20th century, confirming the results of several earlier single-model studies (Li et al., 2018; OberländerHayn et al., 2015; Oman et al., 2009; Polvani et al., 2018). Furthermore, in coming decades, the banning of ODS by the Montreal Protocol will result in much reduced BDC trends, roughly half of the value up to the year 2000 when chlorine levels peaked in the stratosphere.

While the attribution of modeled trends to ODS concentrations in our study is completely unequivocal since we have made use of single-fixed-forcing runs, the precise mechanism via which ODS impact the BDC will need further study. To a large degree, ODS act via the formation of the ozone hole: the resulting large perturbations to the temperature affects atmospheric waves which, upon breaking, drive the BDC. This is why the largest ODS signals are found in DJF and in the Southern Hemisphere. However, it is not immediately obvious how to convert a perturbation in the lower stratospheric temperature gradient (resulting from the formation of the ozone hole) into a change in the wave flux divergence, i.e. we have no theory for computing one from the other a priori: all we can do is to diagnose model output. In the case of increasing GHG, Eichelberger and Hartmann (2005) have argued, from numerical experiments with a simple dynamical-core model, that tropical tropospheric warming caused by higher concentrations of GHG results in a larger meridional temperature gradient at midlatitudes, which increases baroclinicity, which increases wave activity at synoptic and planetary scales, and eventually yields a stronger BDC. More recently, Shepherd and McLandress (2011) have argued that increasing GHG alter the location of critical layers on the tropical lower stratosphere resulting in more upwards wave propagation and thus a stronger BDC. Whether similar mechanisms are at work in the case of ODS forcings remains to be determined. The key difference with GHG is that the temperature gradients caused by ODS are mostly in the lower stratosphere, and not in the troposphere. In addition, it is well known that ODS act as greenhouse gases (see, e.g., Velders, Andersen, Daniel, Fahey, & McFarland, 2007), and their impact on the stratospheric circulation might partly be mediated by SST changes.

Another important question should also be explored in future studies. In coming decades, the degree of trend cancellation between the decreasing ODS and increasing GHG will very likely depend on the scenario used. For example, regarding the poleward shift of the tropospheric eddy-driven jet in the Southern Hemisphere, Barnes et al. (2014) showed that the levels of CO2 will be crucial in determining the extent of the shift in austral summer: in the RCP2.6 scenario recovery of stratospheric ozone will result in a net equatorward shift of the eddy driven jet, whereas the jet will shift poleward in the RCP8.5 scenario. Without running a model with different forcing scenarios, it is impossible to determine the sign of future BDC trends. However, it is not inconceivable that in a weak forcing scenario, such as RCP2.6, the recovery of stratospheric ozone might actually overwhelm the increasing CO2 forcing and result in a net deceleration of the BDC into the late 21st century. We hope future model intercomparison projects with chemistry-climate models will address this issue by exploring different scenarios.

Key Points.

  • Impacts of ozone-depleting substances on Brewer-Dobson circulation trends are analyzed in 20 chemistry-climate models

  • For the period 1980–2000 ODS have contributed more than half (roughly 60%) of the stratospheric age-of-air trends

  • For the period 2000–2080 models show that decreasing ODS levels will substantially decelerate the BDC

Acknowledgments

LMP is most grateful for the hospitality of the Wartner family and the entire staff of the Eurotel Victoria in Les Diablerets, Switzerland, where the bulk of this manuscript was penned. He also acknowledges the continued support by the US National Science Foundation (NSF). LW is supported by grant JIH2308109 from Fudan University and grant 41875047 from National Natural Science Foundation of China (NSFC). MA acknowledges funding from the Program Atraccion de Talento de la Comunidad de Madrid (2016-T2/AMB-1405) and the Spanish National Project STEADY (CGL2017-83198-R). NB was supported by the Met Office Hadley Centre Programme funded by BEIS and Defra and by the European Commission’s Seventh Framework Programme StratoClim project 226520. This research was also supported by the NZ Governments Strategic Science Investment Fund (SSIF) through the NIWA programme CACV, and OM acknowledges funding by the New Zealand Royal Society Marsden Fund (grant 12-NIW-006). Furthermore, Gang Zeng (from NIWA) performed the NIWA-UKCA simulations used in this study. All the data used here are publicly available via the CCMVal-2 and CCMI projects, as detailed in Morgenstern et al. (2010) and Morgenstern et al. (2017).

References

  1. Abalos M, Polvani L, Calvo N, Kinnison D, Ploeger F, Randel W, & Solomon S (2019). New insights on the impact of ozone depleting substances on the Brewer-Dobson circulation. Journal of Geophysical Research: Atmospheres, in press. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Barnes E, Barnes N, & Polvani L (2014). Delayed southern hemisphere climate change induced by stratospheric ozone recovery, as projected by the CMIP5 models. Journal of Climate, 27(1), 853–867. [Google Scholar]
  3. Brewer AW (1949). Evidence for a world circulation provided by the measurements of helium and water vapour distribution in the stratosphere. Quarterly Journal of the Royal Meteorological Society, 75(326), 351–363. [Google Scholar]
  4. Butchart N, Cionni I, Eyring V, Shepherd T, Waugh D, Akiyoshi H, … others (2010). Chemistry–climate model simulations of twenty-first century stratospheric climate and circulation changes. Journal of Climate, 23(20), 5349–5374. [Google Scholar]
  5. Butchart N, Scaife A, Bourqui M, De Grandpré J, Hare S, Kettleborough J, … others (2006). Simulations of anthropogenic change in the strength of the Brewer-Dobson circulation. Climate Dynamics, 27(7–8), 727–741. [Google Scholar]
  6. Butchart N, & Scaife AA (2001). Removal of chlorofluorocarbons by increased mass exchange between the stratosphere and troposphere in a changing climate. Nature, 410(6830), 799. [DOI] [PubMed] [Google Scholar]
  7. Deser C, Phillips A, Bourdette V, & Teng H (2012). Uncertainty in climate change projections: the role of internal variability. Climate Dynamics, 38(3), 527–546. [Google Scholar]
  8. Dobson GM, Harrison D, & Lawrence J (1929). Measurements of the amount of ozone in the Earth’s atmosphere and its relation to other geophysical conditions. Part III. Proceedings of the Royal Society of London. Series A, Containing Papers of a Mathematical and Physical Character, 122(790), 456–486. [Google Scholar]
  9. Eichelberger SJ, & Hartmann DL (2005). Changes in the strength of the brewer-dobson circulation in a simple agcm. Geophysical research letters, 32 (15). [Google Scholar]
  10. Eyring V, Chipperfield MP, Giorgetta MA, Kinnison DE, Matthes K, Newman PA, … Waugh DW (2008). Overview of the new CCMval reference and sensitivity simulations in support of upcoming ozone and climate assessments and the planned SPARC CCMVal report. SPARC Newslet, 30, 20–26. [Google Scholar]
  11. Eyring V, Lamarque J-F, Hess P, Arfeuille F, Bowman M, Chipperfield K, Duncan B, … Young P (2013). Overview of IGAC/SPARC Chemistry-Climate Model Initiative (CCMI) community simulations in support of upcoming ozone and climate assessments. SPARC Newslet, 40, 48–66. [Google Scholar]
  12. Fu Q, Lin P, Solomon S, & Hartmann DL (2015). Observational evidence of strengthening of the Brewer-Dobson circulation since 1980. Journal of Geophysical Research: Atmospheres, 120(19), 10,214–10,228. [Google Scholar]
  13. Garcia RR, & Randel WJ (2008). Acceleration of the Brewer-Dobson circulation due to increases in greenhouse gases. Journal of the Atmospheric Sciences, 65(8), 2731–2739. [Google Scholar]
  14. Garcia RR, Smith AK, Kinnison DE, Cámara Á. d. l., & Murphy DJ (2017). Modification of the gravity wave parameterization in the Whole Atmosphere Community Climate Model: Motivation and results. Journal of the Atmospheric Sciences, 74 (1), 275–291. [Google Scholar]
  15. Garfinkel CI, Aquila V, Waugh DW, & Oman LD (2017). Time-varying changes in the simulated structure of the Brewer–Dobson circulation. Atmospheric Chemistry and Physics, 17(2), 1313–1327. [Google Scholar]
  16. Grandpre JD, Beagley S, Fomichev V, Griffioen E, McConnell J, Medvedev A, & Shepherd T (2000). Ozone climatology using interactive chemistry: Results from the Canadian Middle Atmosphere Model. Journal of Geophysical Research: Atmospheres, 105(D21), 26475–26491. [Google Scholar]
  17. Hardiman SC, Butchart N, & Calvo N (2014). The morphology of the Brewer-Dobson circulation and its response to climate change in CMIP5 simulations. Quarterly Journal of the Royal Meteorological Society, 140(683), 1958–1965. [Google Scholar]
  18. Lamarque J-F, & Solomon S (2010). Impact of changes in climate and halocarbons on recent lower stratosphereozone and temperature trends. J. Clim, 23, 2599–2611. [Google Scholar]
  19. Li F, Newman P, Pawson S, & Perlwitz J (2018). Effects of greenhouse gas increase and stratospheric ozone depletion on stratospheric mean age of air in 1960–2010. Journal of Geophysical Research: Atmospheres, 123(4), 2098–2110. [Google Scholar]
  20. Lin P, & Fu Q (2013). Changes in various branches of the Brewer-Dobson circulation from an ensemble of chemistry climate models. Journal of Geophysical Research: Atmospheres, 118(1), 73–84. [Google Scholar]
  21. Marsh DR, Mills MJ, Kinnison DE, Lamarque J-F, Calvo N, & Polvani LM (2013). Climate change from 1850 to 2005 simulated in CESM1 (WACCM). Journal of Climate, 26(19), 7372–7391. [Google Scholar]
  22. McLandress C, Jonsson AI, Plummer DA, Reader MC, Scinocca JF, & Shepherd TG (2010). Separating the dynamical effects of climate change and ozone depletion. Part I: Southern Hemisphere stratosphere. Journal of Climate, 23(18), 5002–5020. [Google Scholar]
  23. Meinshausen M, Smith SJ, Calvin K, Daniel JS, Kainuma M, Lamarque J, … others (2011). The RCP greenhouse gas concentrations and their extensions from 1765 to 2300. Climatic change, 109(1–2), 213–241. [Google Scholar]
  24. Morgenstern O, Giorgetta M, Shibata K, Eyring V, Waugh D, Shepherd T, … others (2010). Review of the formulation of present-generation stratospheric chemistry-climate models and associated external forcings. Journal of Geophysical Research: Atmospheres, 115 (D3). [Google Scholar]
  25. Morgenstern O, Hegglin MI, Rozanov E, O’Connor FM, Abraham NL, Akiyoshi H, … others (2017). Review of the global models used within Phase 1 of the Chemistry-Climate Model Initiative (CCMI). Geoscientific Model Development, 10(2), 639–671. [Google Scholar]
  26. Morgenstern O, Stone KA, Schofield R, Akiyoshi H, Yamashita Y, Kinnison DE, … others (2018). Ozone sensitivity to varying greenhouse gases and ozone-depleting substances in ccmi-1 simulations. Atmospheric Chemistry and Physics, 18 (2), 1091–1114. [Google Scholar]
  27. Nakicenovic N, Alcamo J, Grubler A, Riahi K, Roehrl R, Rogner H-H, & Victor N (2000). Special report on emissions scenarios (sres), a special report of working group iii of the intergovernmental panel on climate change. Cambridge University Press. [Google Scholar]
  28. Newman P, Daniel J, Waugh D, & Nash. (2007). A new formulation of equivalent effective stratospheric chlorine (EESC). Atmos. Chem. Phys, 7, 4537–4552. [Google Scholar]
  29. Oberländer S, Langematz U, & Meul S (2013). Unraveling impact factors for future changes in the brewer-dobson circulation. Journal of Geophysical Research: Atmospheres, 118(18), 10–296. [Google Scholar]
  30. Oberländer-Hayn S, Meul S, Langematz U, Abalichin J, & Haenel F (2015). A chemistry-climate model study of past changes in the Brewer-Dobson circulation. Journal of Geophysical Research: Atmospheres, 120(14), 6742–6757. [Google Scholar]
  31. Oman L, Waugh DW, Pawson S, Stolarski RS, & Newman PA (2009). On the influence of anthropogenic forcings on changes in the stratospheric mean age. Journal of Geophysical Research: Atmospheres, 114 (D3). [Google Scholar]
  32. Polvani LM, Abalos M, Garcia R, Kinnison D, & Randel WJ (2018). Significant weakening of Brewer-Dobson circulation trends over the 21st Century as a consequence of the Montreal Protocol. Geophysical Research Letters, 45(1), 401–409. [Google Scholar]
  33. Polvani LM, Wang L, Aquila V, & Waugh DW (2017). The impact of ozone-depleting substances on tropical upwelling, as revealed by the absence of lower-stratospheric cooling since the late 1990s. Journal of Climate, 30(7), 2523–2534. [Google Scholar]
  34. Rind D, Suozzo R, Balachandran N, & Prather M (1990). Climate change and the middle atmosphere. Part I: The doubled CO2 climate. Journal of the Atmospheric Sciences, 47(4), 475–494. [Google Scholar]
  35. Scinocca J, McFarlane N, Lazare M, Li J, & Plummer D (2008). The CCCma third generation AGCM and its extension into the middle atmosphere. Atmospheric Chemistry and Physics, 8(23), 7055–7074. [Google Scholar]
  36. Seidel DJ, Li J, Mears C, Moradi I, Nash J, Randel WJ, … Zou C-Z (2016). Stratospheric temperature changes during the satellite era. Journal of Geophysical Research, 121, 664–681. [Google Scholar]
  37. Shepherd TG, & McLandress C (2011). A robust mechanism for strengthening of the Brewer–Dobson circulation in response to climate change: Critical-layer control of subtropical wave breaking. Journal of the Atmospheric Sciences, 68 (4), 784–797. [Google Scholar]
  38. Solomon S, Kinnison D, Bandoro J, & Garcia R (2015). Simulation of polar ozone depletion: An update. Journal of Geophysical Research: Atmospheres, 120(15), 7958–7974. [Google Scholar]
  39. Taylor KE, Stouffer RJ, & Meehl GA (2012). An overview of cmip5 and the experiment design. Bulletin of the American Meteorological Society, 93(4), 485–498. [Google Scholar]
  40. Velders G, Andersen S, Daniel J, Fahey D, & McFarland M (2007). The importance of the montreal protocol in protecting climate. Proc. Nat. Acad. Sci, 104, 4814–1819. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. WMO. (2011). Scientific Assessment of Ozone Depletion: 2010. Global Ozone Research and Monitoring Project-Report No. 52, . Geneva, Switzerland. [Google Scholar]
  42. WMO. (2014). Scientific Assessment of Ozone Depletion: 2014. Global Ozone Research and Monitoring Project-Report No. 55. Geneva, Switzerland. [Google Scholar]

RESOURCES