3.1 Reactive chlorine

Based on the prescribed changes of the CFC-11 boundary conditions the stratospheric content of reactive chlorine compounds (${\mathrm{ClO}}_x=\mathrm{Cl}+\mathrm{ClO}+\mathrm{OClO}+\mathrm{2}\mathrm{ClOOCl}+\mathrm{2}{\mathrm{Cl}}_{\mathrm{2}}+\mathrm{HOCl}$) is expected to change. In Fig. 2, the simulated change (i.e., SEN-C2-fCFC11_2050 and SEN-C2-fCFC11_2020 minus REF-C2, respectively) of the reactive chlorine mixing ratios with time is shown for the lower stratosphere (LS, near 50 hPa) and the upper stratosphere (US, near 2 hPa). Because all model simulations are operated in a “free-running-mode” (i.e., they do not have the same meteorology), the year-to-year difference (thin red and blue) curves as shown in Fig. 2 (and also in other figures shown afterwards) indicate the possible range of interannual fluctuation. Obviously it takes about 10–15 years (as expected) before the ClO_x values of the REF-C2 and SEN-C2-fCFC11_2050 simulations clearly diverge from each other. At the end of the SEN-C2-fCFC11_2050 simulation (i.e., in 2050), the resulting absolute (mean) difference in EMAC amounts to an approximate $\mathrm{6}\times {\mathrm{10}}^{-\mathrm{12}}$ mol mol⁻¹ increase in the LS and about a $\mathrm{50}\times {\mathrm{10}}^{-\mathrm{12}}$ mol mol⁻¹ increase in the US compared with the REF-C2 simulation. The vertical profile (structure) of the changes of the chlorine mixing ratios qualitatively resembles the reference ClO_x profile in EMAC (i.e., changes are biggest where the ClO_x mixing ratios are biggest), which displays a distinct maximum at about 2–3 hPa (not shown). It turns out that the amount of chemically active chlorine species (ClO_x) in the SEN-C2-fCFC11_2050 simulation is about 17 % larger on average than in the REF-C2 simulation in the US (above 30 km) in the 2040s (i.e., the time period from 2041 to 2050); in the LS (below 30 km) the respective amount of ClO_x is enhanced by about 30 %. As can be expected, the respective ClO_x differences between the SEN-C2-fCFC11_2020 and the REF-C2 simulations in the 2040s are clearly smaller, i.e., about $\mathrm{2}\times {\mathrm{10}}^{-\mathrm{12}}$ mol mol⁻¹ in the LS (about 14 %) and about $\mathrm{25}\times {\mathrm{10}}^{-\mathrm{12}}$ mol mol⁻¹ in the US (about 9 %).

Figure 2Temporal evolution of the annual global mean ClO_x mixing ratio differences (in mol mol⁻¹) at around 2 hPa (US, a) and 50 hPa (LS, b) between the SEN-C2-fCFC11_2050 and REF-C2 simulations (in red) and between the SEN-C2-fCFC11_2020 and REF-C2 simulations (in blue). The 11-year solar cycle (smoothed with a 1-2-1 filter) has been removed from both time series. The thicker curves in red and blue show the 5-year running means. The red and blue lines show the linear regression estimate of the unsmoothed time series.

Download

3.2 Total and partial ozone columns

The impact of the enhanced atmospheric ClO_x content due to the constant CFC-11 surface mixing ratio after 2002 on total column ozone (TCO) is shown in Fig. 3. It illustrates the differences in the mean annual cycle for the last decade depending on the geographical latitude between the REF-C2 and the SEN-C2-fCFC11_2050 simulations. The largest changes in the TCO values are found in both polar regions: in particular in the Northern Hemisphere during winter (December, January, and February) and in the Southern Hemisphere in late winter (August) to early spring (September and October). In the SEN-C2-fCFC11_2050 simulation, the TCO values are clearly reduced by up to about 30 Dobson units (DU) (in the order of 15 % in the Southern Hemisphere) in comparison with the REF-C2 simulation. During other times of the year and in other latitudinal regions the identified TCO changes are much smaller: they are mostly below ±5 DU.

A closer look at the near-global mean (60^∘ S–60^∘ N) temporal behavior of the TCO (Fig. 4) indicates increasing ozone values in both simulations (Fig. 4a, for both simulations the solar cycle was removed, see figure caption). The results of the SEN-C2-fCFC11_2050 simulation show slightly smaller TCO values compared with REF-C2 and a slightly flatter slope of the linear trend line (this regression accounts for possible autocorrelation at lag 1, see, e.g., the method described in Tiao et al., 1990). The linear regression based on the results of the REF-C2 simulation (here the black line presented in Fig. 4a) indicates an increase of 1.7 DU per decade for the TCO (annual near-global mean), whereas the linear regression based on the results of the SEN-C2-fCFC11_2050 (red line in Fig. 4a) shows a reduced increase of 1.3 DU per decade. This finding is supported by the TCO difference (Fig. 4b), which indicates a small reduction in the TCO (in SEN-C2-fCFC11_2050) of up to about 2 DU (in the order of less than 1 %) until 2050. The linear regression (again accounting for possible lag 1 autocorrelation) gives −0.5 DU per decade (±0.25 DU per decade; given by two times the estimated standard error, which roughly corresponds to the 95 % confidence interval and will be used throughout this paper as a measure of uncertainty). This effect can be rated as negligible in comparison with the expected annual fluctuations in this region. Therefore, in the following, we focus on the analyses of the polar regions, in particular on the Antarctic region in September. One reason for choosing the month September for further analyses of Antarctic ozone chemistry is that we found the most obvious ozone changes here. Another reason is that this month is less noisy than October (Solomon et al., 2016).

Figure 3Mean annual cycle of total column ozone (TCO) differences (in Dobson units, DU) between SEN-C2-fCFC11_2050 and REF-C2 for the 2040s (i.e., SEN minus REF).

Download

Figure 4(a) Temporal evolution of total column ozone (TCO; in DU) for the annual near-global mean (60^∘ S–60^∘ N) in REF-C2 (black curves) and SEN-C2-fCFC11_2050 (red curves). (b) TCO differences (in DU) between SEN-C2-fCFC11_2050 and REF-C2 (i.e., SEN minus REF). For the absolute TCO time series (a) the 11-year solar cycle (smoothed with a 1-2-1 filter) has been removed. Thicker curves show the 5-year running means, respectively. The corresponding lines (a, b) show the respective linear regression estimates based on the unsmoothed data.

Download

In the Southern Hemisphere polar region (70–90^∘ S) obvious ozone changes can be identified in September (Fig. 5). The mean differences of the TCO between the 2000s and the end of the simulation amount to about 20 DU, indicating that the mean September ozone values in the SEN-C2-fCFC11_2050 simulation are about 10 % lower than in the REF-C2 simulation. The trend estimate gives −4.1 DU per decade (±1.7 DU per decade). This trend estimate and the uncertainties were obtained by multiple linear regression, which accounts for possible lag 1 autocorrelation and uses the difference of the temperature anomalies (at 100 hPa over 70–90^∘ S) from the REF-C2 and SEN-C2-fCFC11_2050 simulations as a second independent variable in addition to the linear trend. It is found that much of the interannual variability can be explained by including the difference of the polar temperature anomalies at 100 hPa in the regression model. This agrees with Langematz et al. (2016), who used temperature anomalies to regress polar TCO. The temporal evolution of TCO differences and the size of the ozone disturbance found in the Northern Hemisphere polar region in January have the same order of magnitude (not shown), but the signal is more noisy because of the stronger dynamic variability.

Figure 5Temporal evolution of TCO differences (in DU) between SEN-C2-fCFC11_2050 and REF-C2 (i.e., SEN minus REF) for the Antarctic region (70–90^∘ S) in September. The thicker curve shows the 5-year running mean. The corresponding line shows the trend estimate between the unsmoothed time series using a multiple linear regression – including differences of temperature anomalies as the dependent variable – which accounts for possible autocorrelation with lag 1 (see text for details).

Download

Next, we look into stratospheric partial columns of ozone (PCO) in more detail with respect to the upper stratosphere (US, above about 30 km, i.e., the 10 hPa pressure level) and the lower stratosphere (LS, between 100 and 10 hPa) for the Antarctic region in September. Figure 6 shows the PCO differences for the US (panel a) and the LS (panel b) between the SEN-C2-fCFC11_2050 and the REF-C2 simulations. Both show the expected negative trend, indicating lower values at the end of the SEN-C2-fCFC11_2050 simulation in the late 2040s. Again the trends are obtained by the same regression model as for the TCO (see the previous paragraph), but for the US the temperatures at 10 hPa were used. The mean PCO changes for the US of about 2 DU are much smaller than those calculated for the LS (about 20 DU). The temporal evolution of the PCO differences in the LS show similar results as those found for the TCO differences (Fig. 5): in the SEN-C2-fCFC11_2050 simulation the TCO is reduced by about 20 DU until the year 2050. The strongest signature of ozone change found in the polar LS points to the importance of heterogeneous chemical processes. Viewing the vertical profile of the differences of net ozone production rates of ClO_x between the REF-C2 and the SEN-C2-fCFC11_2050 simulations in the 2040s clearly indicates an absolute minimum (i.e., less net ozone production in SEN-C2-fCFC11_2050) at around 50 hPa and another relative minimum at about 1.5 hPa in September (Fig. 8b; see explanation in Sect. 3.3). Moreover, looking at the PCO changes in middle- and lower-latitude regions (60^∘ S–60^∘ N, not shown), the partial column differences clearly indicate that the small ozone differences detected in the TCO (Fig. 4b) are affected by ozone reductions of similar magnitudes in the US and the LS, displaying only small contributions to the TCO.

Figure 6Temporal evolution of partial column ozone (PCO) differences (in DU) between SEN-C2-fCFC11_2050 and REF-C2 (i.e., SEN minus REF) for the Antarctic region (70–90^∘ S) in September. Panel (a)  shows PCO values for the US (above 30 km); panel (b) shows PCO values for the LS (below 30 km). Red thicker curves show the respective 5-year running means. The corresponding lines show the trend estimates for the unsmoothed time series using a multiple linear regression – including differences of temperature anomalies as the dependent variable – which accounts for possible autocorrelation with lag 1 (see text for details).

Download

In Fig. 7, we compare the calculated differences between the SEN-C2-fCFC11_2050 and REF-C2 simulations as presented in Figs. 4–6 with the corresponding results derived from SEN-C2-fCFC11_2020 and REF-C2. In Fig. 7a–d, it is obvious that an immediate re-implementation of the Montreal Protocol eventually leads back to the direction of the expected ozone conditions around the end of the 2040s, as they are calculated in our REF-C2 simulation. In the REF-C2 simulation the model-diagnosed CFC-11 emissions are nearly zero after about 2030, whereas in the SEN-C2-fCFC11_2020 simulations they steadily decrease from higher values in 2020 down to zero around the year 2050. This is caused by the prescribed CFC-11 surface mixing ratio as indicated in Fig. 1.

3.3 Stratospheric ozone budget

In the following, a detailed analysis of individual ozone production and loss processes is carried out. This ozone budget analysis is used to investigate the role of separate chemical cycles and reactions, which are responsible for ozone production and loss in the stratosphere. For this analysis the MESSy tool “strato3bud” (cf. Meul et al., 2014, based on Jöckel et al., 2006) is employed. The respective reactions responsible for stratospheric ozone production (attributed to photolysis hυ, HO₂, and CH₃O₂) and loss (attributed to O_x, NO_x, HO_x, ClO_x, and BrO_x) are described by Meul et al. (2014, see their Table 2). In Fig. 8, the results of this budget analysis are shown as changes of the ozone production rate between the SEN-C2-fCFC11_2050 and the REF-C2 simulations $\left(\mathrm{\Delta }{P}_{\mathrm{SEN}-\mathrm{REF}}^{\mathrm{prc}}\right)$ with respect to the total production rate in the REF-C2 simulation. The explicit formula for calculating the changes of the ozone production rate at a certain level is given as

Here, P denotes the temporal mean and zonally summed ozone production rate (molecules s⁻¹). The subscript denotes the respective simulations, and the “prc” superscript denotes which process (hυ, HO₂, CH₃O₂, O_x, NO_x, HO_x, ClO_x, and BrO_x) is analyzed. Loss cycles are regarded as negative production rates. Further, the superscript “tot” denotes the sum of all positive production rates (namely of hυ, HO₂, and CH₃O₂) and the summation goes over all latitudes, which lie in the respective latitudinal band R. Here we show profiles of ΔP_SEN−REF for the annual global mean and the Southern Hemisphere polar region (70–90^∘ S) during September 2041–2050. In the Antarctic spring season, (Fig. 8b) it is obvious that the enhanced content of reactive chlorine in the SEN-C2-fCFC11_2050 simulation is responsible for the intensified ozone loss in the LS (around 50 hPa) and the US (around 1.5 hPa). In the LS, ozone loss through reactive bromine compounds (${\mathrm{BrO}}_x=\mathrm{Br}+\mathrm{BrO}+\mathrm{HOBr}+\mathrm{BrCl}+\mathrm{2}{\mathrm{Br}}_{\mathrm{2}}$) is also enhanced, which is probably related to the enhanced chlorine loading as a result of ozone loss due to the reaction of BrO with ClO (see Table 2 in Meul et al., 2014, which specifies the reactions that are considered in our analysis of ozone production and loss cycles). Conversely, as a consequence of more available chlorine, other loss cycles or production processes show a tendency to compensate for the enhanced ozone destruction by chlorine. For instance, the catalytic NO_x cycle shows some balancing in the altitude region between about 19 km (50 hPa) and the stratopause (about 0.7 hPa). This means that the ozone depletion by NO_x is clearly reduced (i.e., a relative ozone production) in the SEN-C2-fCFC11_2050 compared with the REF-C2 simulation above about 40 hPa (see the red line in Fig. 8, which indicates the change of the net ozone production rate) in the 2040s. It is, however, not straightforward to further disentangle the underlying processes of the most relevant chemical cycles, as the underlying kinetic system is highly nonlinear. The system in the SEN-C2-fCFC11_2050 simulation is heading towards a different chemical equilibrium, due to the distribution of educts and the temperature change. The results with respect to the global annual means of ozone production and loss in the 2040s shown in Fig. 8a indicate that below about 50 hPa no obvious changes are detected. Above 20 km the ozone loss is strongly affected by reactive chlorine, and, again, some compensation effects due to other competing ozone loss cycles are clearly identified in the US. The positive values with respect to the photolysis rates indicate a slight downward shift of the ozone layer (ozone maximum) to lower altitudes. This is probably due to enhanced ozone loss via chlorine at higher altitudes, which allows more UV radiation to reach lower altitudes, where this additional radiation then causes higher photolysis rates. Overall, the red line in the Fig. 8a indicates that the change in the net ozone production rate is nearly zero at all altitudes due to these compensating effects.

Figure 7Different temporal evolution of column ozone differences (in DU) between the individual sensitivity simulations and the reference simulation (in DU): SEN-C2-fCFC11_2050 minus REF-C2 values are indicated in red, and SEN-C2-fCFC11_2020 minus REF-C2 values are shown in blue. (a) TCO for the annual near-global mean (60^∘ S–60^∘ N); (b) TCO for the Antarctic (70–90^∘ S) in September; (c) PCO for the Antarctic (70–90^∘ S) in the US in September; (d) PCO for the Antarctic (70–90^∘ S) in the LS in September. The red and blue lines show the trend estimates for the unsmoothed time series using a multiple linear regression – including differences in temperature anomalies as the dependent variable – which accounts for possible autocorrelation with lag 1 (see text for details).

Download

Figure 8The relative change in ozone production rates (in %), which are normalized to the total column production (through photolysis hυ, HO₂, and CH₃O₂) in the REF-C2 simulation. For the individual ozone production and loss processes, mean differences are shown that have been derived from the REF-C2 and SEN-C2-fCFC11_2050 (i.e., SEN minus REF) simulations for the 2040s (from 2041 to 2050). The change in the net ozone production rate, which refers to the sum of all changes, is indicated by the red line. Panel (a) shows the mean annual global mean profiles, and panel (b) shows the values for the south polar region (70–90^∘ S) in September. Negative values indicate an intensified ozone loss or a decreased ozone production in the SEN-C2-fCFC11_2050 simulation, whereas higher values indicate more ozone production or less loss via a specific process. Thin horizontal lines indicate the nearest pressure levels to the model grid boxes.

Download

The analogous ozone budget analysis is carried out for changes in the ozone production rate between the SEN-C2-fCFC11_2020 and the REF-C2 simulations. As expected, the vertical dependence of the ozone production rates in the 2040s looks similar to those in the analysis of the SEN-C2-fCFC11_2050 simulation (not shown), namely (1) the global annual means of the ozone production and loss below about 20 km (50 hPa) do not indicate obvious changes. Higher up ozone loss is strongly affected by reactive chlorine, and, again, some compensation effects in the US due to other competing ozone loss cycles are clearly identified; however, all ozone production rate changes are about half as strong as those found between the SEN-C2-fCFC11_2050 and REF-C2 simulations (see Fig. 8a). (2) In the Antarctic region during September in the 2040s the intensified ozone loss in the LS (around 50 hPa) via the ClO_x and BrO_x cycles are again obvious, but the ozone destruction rates are only one-third of the magnitude of that found between the SEN-C2-fCFC11_2050 and REF-C2 simulations (see Fig. 8b).

Finally, to check the possible impact of temperature changes due to enhanced CFC-11 concentrations on ozone chemistry, we analyze the overall temperature trends in the US (near 1 hPa) and LS (near 50 hPa) as well as the differences between REF-C2 and SEN-C2-fCFC11_2050. The global annual mean long-term temperature behavior in the REF-C2 simulation indicates a cooling of about 1 K in the LS and of about 3 K in the US from the early 2000s until to the year 2050. The temperature difference between REF-C2 and SEN-C2-fCFC11_2050 amounts to an additional cooling of about 0.3 K in the US in the SEN-C2-fCFC11_2050 simulation, whereas no obvious change in the long-term behavior can be identified in the LS (not shown). It is difficult to separate the individual contributions of the additional cooling in a coupled CCM simulation, i.e., radiative cooling due to enhanced CFC-11 concentrations and due to less ozone in the stratosphere caused by the enhanced chlorine loading without additional diagnostics. We assume that both processes will contribute to the calculated additional cooling in the SEN-C2-fCFC11_2050 simulation.

Taking a closer look at the Southern Hemisphere polar region in spring, the REF-C2 simulation indicates a clear cooling trend of about 4 K in the US (near 1 hPa) until 2050, whereas no obvious trend can be identified in the LS (not shown). With respect to the temperature differences of SEN-C2-fCFC11_2050 and REF-C2, the US does not show a clear change, whereas in the LS the SEN-C2-fCFC11 simulation suggests some additional cooling from the early 2000s until 2050 by about 2 K. However, this difference is superposed by large interannual fluctuations.

With this in mind, we can try to evaluate the calculated ozone differences in the sensitivity simulations in comparison to REF-C2. With respect to the global mean in the US, on the one hand, enhanced chlorine mixing ratios lead to enhanced ozone depletion via the catalytic ozone destruction cycle; on the other hand, the extra cooling is known to create a reduction of the ozone depletion rates by gas-phase chemistry (e.g., Haigh and Pyle, 1982). It was found that the net effect here is slightly negative, i.e., there are indications that ozone differences between REF-C2 and the sensitivity simulations in the US are dominated by the enhanced chlorine content. With respect to the global mean in the LS, where no clear cooling is simulated, the smaller ozone values are mainly caused by the enhanced chlorine content.

Looking closer at the south polar region in spring, it is obvious that the enhanced chlorine content is again mostly responsible for the slightly reduced PCO in the SEN-C2-fCFC11_2050 simulation in the US (Fig. 6a). In the LS, where heterogeneous chemical processes are the most important drivers of ozone changes, the enhanced chlorine mixing ratios intensify the ozone destruction. This leads to significantly reduced PCO over the time (Fig. 6b), which eventually leads to the indicated (slight) extra cooling of the polar lower stratosphere in spring. However, a first analysis of polar stratospheric cloud (PSC) statistics for the REF-C2 and SEN-C2-fCFC11_2050 simulations did not display a considerable trend in the PSC surface area (not shown). Therefore, we cannot identify any hint of enhanced chlorine activation.