Stratospheric Ozone Response to Sulfate Aerosol and Solar Dimming Climate Interventions based on the G6 Geoengineering Model Intercomparison Project (GeoMIP) Simulations

. This study assesses the impacts of sulfate aerosol intervention (SAI) and solar dimming on stratospheric ozone based on the G6 Geoengineering Model Intercomparison Project (GeoMIP) experiments, called G6sulfur and G6solar. For G6sulfur the stratospheric sulfate aerosol burden is increased to reﬂect some of the incoming solar radiation back into space in order to cool the surface climate, while for G6solar the global solar constant is reduced to achieve the same goal. The high emissions scenario SSP5-8.5 is used as the baseline experiment and surface temperature from the medium emission scenario 5 SSP2-4.5 is the target. Based on three out of six Earth System Models (ESMs) that include interactive stratospheric chemistry, we ﬁnd signiﬁcant differences in the ozone distribution between G6solar and G6sulfur experiments compared to SSP5-8.5 and SSP2-4.5, which differ by both region and season. Both SAI and solar dimming methods reduce incoming solar insolation and result in tropospheric temperatures comparable to SSP2-4.5 conditions. G6sulfur increases the concentration of absorbing sulfate aerosols in the stratosphere, which increases lower 4.5 and SSP5-8.5. G6solar counters zonal wind and tropical upwelling changes between SSP2-4.5 and SSP5-8.5 but does not change stratospheric temperatures. Solar dimming results in little change in TCO compared to SSP5-8.5 and does not counter the effects of the ozone super-recovery. Only in the tropics, G6solar results in an increase of TCO of up to 8 DU compared to SSP2-4.5, which may counter the projected reduction due to climate change in the high forcing future scenario. This work identiﬁes differences in the response of SAI and solar dimming on ozone, which are at least partly due to differences and 25 shortcomings in the complexity of aerosol microphysics, chemistry, and the description of ozone photolysis in the models. It also identiﬁes that solar dimming, if viewed as an analog to SAI using a predominantly scattering aerosol, would, for the most part, not counter the potential harmful increase in TCO beyond historical values induced by increasing greenhouse gases. ﬁnding and on

determine the impacts of SAI on ozone. Only two out of the three models, UKESM1-0-LL and CESM2(WACCM), include interactive aerosol microphysical schemes. The other three models used prescribed ozone fields, which differed only between SSP5-8.5 and SSP2-4.5 (Keeble et al., 2021). The CNRM-ESM2-1 chemistry scheme considers 168 chemical reactions, among which 39 are photolysis reactions, and 9 reactions that represent heterogeneous chemistry. This scheme is applied above 560  Grey horizontal lines indicate changes that are not statistically significant over the period considered using a double sided t-test at 95% confidence levels.
with reduced vertical wind velocity around the tropopause and below the injection altitude. Interestingly, the reduction in w * overcompensates conditions for the target simulation and results in values similar to present-day conditions. Above the sulfur injection location, w * is increased compared to SSP5-8.5. In contrast, w * in G6solar matches the target scenario SSP2-4.5.
This indicates that changes in w * in G6solar are largely driven by tropospheric temperatures.
Besides the commonalities among the three models, some differences exist. For the G6sulfur experiments, the three models showing stronger heating and change in w * than injections in higher altitudes (Tilmes et al., 2017). Furthermore, some non-significant differences in the model results for both G6sulfur and G6solar are obvious for the response already in the first 20 years of the application ( Figure A3). CNRM-ESM2-1 shows a weakening of the polar vortex in the Southern Hemisphere for G6sulfur and a strengthening of the polar vortex in both hemispheres in G6solar compared to the baseline simulation. This, however, is not likely related to the solar or sulfur applications because they had not ramped up 215 before 2040.

Effects of SAI on Surface Area Density
The three ESMs with interactive chemistry applied different strategies to counter the warming between SSP5-8.5 and SSP2-4.5 with stratospheric aerosols (see Section 2). Resulting differences in the changes in SAD as described in this section (Figure 6) have different impacts on heterogeneous chemistry and, therefore, ozone in the stratosphere.

220
As described in Section 2, CESM2-WACCM6 injected sulfur at 25 km (≈30 hPa). This resulted in an aerosol distribution that covers a larger altitude range compared to that of other models. Furthermore, the experiment required an initial increase of sulfur emissions of around 2 TgSO 2 /yr in the first three years of the start of the application, which stayed around 2-3 TgSO 2 /yr emissions until 2045 (as shown in Visioni et al. (2021b)). This relatively small injection amount results in a sudden increase of SAD from 2 to 10 µm 2 /cm 3 within the first year of the application ( Figure 6). This happens because the aerosol microphysical 225 scheme first produces smaller particles that grow slowly with increasing injection, resulting in the initial increase in SAD (as also discussed in Tilmes et al. (2021) CNRM-ESM2-1 uses a prescribed stratospheric aerosol distribution that is scaled depending on the requirement to offset the 245 warming between SSP2-4.5 and SSP5-8.5. The aerosol and SAD distribution are generally similar to WACCM6 but smaller and slightly less spread out by the end of the century (Figure 6b). CNRM-ESM2-1 does not apply SAI until after 2040 and applies a linear increase with time after that date mainly because the difference in surface air temperature between SSP8-8.5 and SSP2-4.5 cannot be disentangled from the model internal variability before 2040 (Visioni et al., 2021b). We find a linear increase with time in SAD because of the scaled, fixed aerosol distribution that does not consider potential changes in the aerosol size 250 distribution with injection amount, or changes in the spatial distribution of the aerosol from transport. The resulting SAD is smaller (by almost half) in the Tropics compared to the other two models. A similar but slightly smaller increase after 2060 is found in the SH high latitudes in October, half the increase in SAD in the NH high latitudes in March, a similar SAD in NH mid-latitudes in January compared to UKESM1-0-LL. Finally, change in the O 2 photolysis rate and UV-B radiation due to changes in ozone and aerosols impacts tropospheric ozone. Reductions in column ozone and the resulting increase in UV-B in high latitudes are partly offset by the reduction in UV-B from the aerosol layer (e.g., Tilmes et al., 2012;Pitari et al., 2014). In this study, UKESM1-0-LL is the only model that includes an interactive photolysis scheme that takes the effects of aerosols into account, while all the models include changes 300 due to ozone. The increase in aerosol burden and the resulting reduction of oxygen photolysis likely contribute to the increase in tropospheric ozone in UKESM1-0-LL.

Effects of SAI and solar dimming on total column ozone (TCO)
Total column ozone in SSP5.8.5 and SSP2-4.5 increases in mid-and high latitudes between 2020 and 2100 and reaches above 1960 values for high greenhouse gas forcing scenarios due to the slow reduction in stratospheric halogen loading as the result

310
The three GeoMIP models follow the general behavior outlined above; however, specific differences exist (

Effects of SAI on TCO
In the SH polar region in October, for G6sulfur compared to SSP5-8.5 (Figure 9), WACCM6 shows a significant decline in TCO up to 30 DU for the ensemble mean at the start of the sulfur injection in 2020. After that, TCO declines much slower towards 38 DU by the end of the century. The changes are aligned with changes in SAD (Figure 6d), since chemical changes strongly control the ozone in this region as well as the slow decline in stratospheric halogen content resulting in reduce chemical ozone 320 loss. CNRM-ESM2-1 simulates decreasing TCO between 2040 and 2100, which is also aligned with the increase in SAD and is smaller than what is simulated in CESM2-WACCM6. However, due to the linear increase in SAD, CMRM-ESM2-1 does not show a strong decrease in ozone during the onset of the SAI application. UKESM1-0-LL shows much smaller reductions in TCO in the SH polar region than the other models due to a smaller increase in SAD. Because of differences in timing and magnitude of SAD changes, there is a large spread in the TCO response between the three models in this region. The ensemble 325 mean shows an initial decrease in TCO of 10 DU ozone loss and closer to 20 DU by the end of the century (Figure 11).
Compared to SSP2-4.5, there is no significant change in TCO besides the initial reduction.  Figure A3. Zonal mean U winds changes (m/s) (2030-39) between G6sulfur and SSP5-8.5 (left) and G6solar and SSP5-8.5 (right) for the three GeoMIP models with interactive chemistry that participated in the G6 experiment. Contour lines show the baseline SSP5-8.5 winds.
Grey horizontal lines indicate changes that are not statistically significant over the time period considered using a double-sided t-test at 95% confidence levels