Effects of Strongly Enhanced Atmospheric Methane Concentrations in a Fully Coupled Chemistry-Climate Model

In a previous study the quasi-instantaneous chemical impacts (rapid adjustments) of strongly enhanced methane (CH4) mixing ratios have been analyzed. However, to quantify the influence of the respective slow climate feedbacks on the chemical composition it is necessary to include the radiation driven temperature feedback. Therefore, we perform sensitivity simulations with doubled and fivefold present-day (year 2010) CH4 mixing ratios with the chemistry-climate model EMAC and include in a novel set-up a mixed layer ocean model to account for tropospheric warming. 5 We find that the slow climate feedbacks counteract the reduction of the hydroxyl radical in the troposphere, which is caused by the strongly enhanced CH4 mixing ratios. Thereby also the resulting prolongation of the tropospheric CH4 lifetime is weakened compared to the quasi-instantaneous response considered previously. Changes in the stratospheric circulation evolve clearly with the warming of the troposphere. The Brewer-Dobson circulation strengthens, affecting the response of trace gases, such as ozone, water vapour and CH4 in the stratosphere, and also causing 10 stratospheric temperature changes. In the middle and upper stratosphere, the increase of stratospheric water vapour is reduced with respect to the quasi-instantaneous response. Weaker increases of the hydroxyl radical cause the chemical depletion of CH4 to be less strongly enhanced and thus the in situ source of stratospheric water vapour as well. However, in the lower stratosphere water vapour increases more strongly when tropospheric warming is accounted for enlarging its overall radiative impact. The response of the stratospheric adjusted temperatures driven by slow climate feedbacks is dominated by these increases of 15 stratospheric water vapour, as well as strongly decreased ozone mixing ratios above the tropical tropopause, which result from enhanced tropical upwelling. While rapid radiative adjustments from ozone and stratospheric water vapour make an essential contribution to the effective CH4 radiative forcing, the radiative impact of the respective slow feedbacks is rather moderate. In line with this, the climate sensitivity from CH4 changes in this chemistry-climate model setup is not significantly different from the climate sensitivity in 20 carbon dioxide-driven simulations, provided that the CH4 effective radiative forcing includes the rapid adjustments from ozone and stratospheric water vapour changes. 1 https://doi.org/10.5194/acp-2020-519 Preprint. Discussion started: 26 June 2020 c © Author(s) 2020. CC BY 4.0 License.


Introduction
Methane (CH 4 ) is the second most important anthropogenically influenced greenhouse gas (GHG). Apart from its direct radia-25 tive impact (RI), CH 4 is chemically active and induces chemical feedbacks relevant for climate and air quality. Through its most important tropospheric sink, the oxidation with the hydroxyl radical (OH), it affects the oxidation capacity of the atmosphere and thus its own lifetime (e.g., Saunois et al., 2016b;Voulgarakis et al., 2013;Winterstein et al., 2019). CH 4 oxidation is further an important source of stratospheric water vapour (SWV) (e.g., Frank et al., 2018) and affects the ozone (O 3 ) concentration in troposphere and stratosphere via secondary feedbacks. Chemical feedbacks from O 3 and SWV contribute significantly to the 30 total RI induced by CH 4 (e.g., Fig. 8.17 in IPCC, 2013;Winterstein et al., 2019). The abundance of CH 4 in the atmosphere is rising rapidly at present (e.g., Nisbet et al., 2019). Furthermore, emissions from natural CH 4 sources can be prone to climate change and have the potential to strongly enhance atmospheric CH 4 concentrations (Dean et al., 2018). Together with its relevance as a GHG, the latter underlines the importance of examining implications of strongly increased CH 4 abundances in the atmosphere. 35 Chemistry-climate models (CCMs) are useful tools for such studies. A CCM is a General Circulation model (GCM) that is interactively coupled to a comprehensive chemistry module. This online two-way coupling is necessary to assess, on the one hand, chemically induced changes of radiatively active gases and their feedback on temperature, and on the other hand feedbacks on chemical processes driven by changes of the climatic state (e.g. temperature, circulation or precipitation). A range of CCM studies analysed the sensitivity of other atmospheric constituents, such as O 3 (Kirner et al., 2015;Morgenstern et al., 40 2018), SWV (Revell et al., 2016) and OH and CH 4 lifetime (Voulgarakis et al., 2013), to different projections of CH 4 mixing ratios. However, these studies did not focus on the climate impact of CH 4 . Other recent studies assessing climate feedbacks and climate sensitivity of CH 4 did not include radiative contributions from chemical feedbacks in their analysis (Modak et al., 2018;Smith et al., 2018;Richardson et al., 2019). Winterstein et al. (2019) assessed chemical feedback processes and their RI in sensitivity simulations forced by 2-fold (2×) 45 and 5-fold (5×) present-day (year 2010) CH 4 mixing ratios. As their simulation set-up prescribed sea surface temperatures (SSTs) and sea ice concentrations (SICs) and thus suppressed surface temperature changes, the parameter changes in their simulations have the character of rapid adjustments (e.g., Forster et al., 2016;Smith et al., 2018). In the effective radiative forcing (ERF) framework, rapid adjustments of radiatively active species are counted as part of the forcing and are to be distinguished from slow climate feedbacks that are coupled to surface temperature changes (Sherwood et al., 2015). Climate sensitivity pa-50 rameters, reflecting the degree of surface temperature change per unit forcing, have been found to be less dependent on the forcing agent with this definition compared to previous definitions of radiative forcing (RF) (e.g., Shine et al., 2003;Hansen et al., 2005;Richardson et al., 2019).
As a follow-up on Winterstein et al. (2019), we assess the respective SST-driven climate feedbacks, their effect on the quasiinstantaneous response of the chemical composition, and consequently resulting radiative feedbacks. Consistent with Winter-55 stein et al. (2019), we perform sensitivity simulations with 2× and 5× present-day CH 4 mixing ratios with the ECHAM/MESSy Atmospheric Chemistry (EMAC) CCM (Jöckel et al., 2016), but this time coupled to a mixed layer ocean (MLO) model in- Table 1. Overview of the two sets of sensitivity simulations (fSST and MLO) with one reference simulation and two sensitivity simulations.
The simulations with prescribed SSTs and SICs have already been analysed by Winterstein et al. (2019) (Stuber et al., 2001;Dietmüller et al., 2016).  Since this study is one of the first to use the MLOCEAN submodel in MESSy, we have carefully checked whether REF MLO reproduces SSTs and SICs of the climatology that was used to determine the heat flux correction with sufficient accuracy. The spatial pattern of the SST climatology is realistically reproduced in REF MLO (see Fig. S1). The largest differences are found at higher latitudes, where a reduction in sea ice area leads to higher SSTs, as exposed sea water is warmer than sea ice. In the Northern Hemisphere (NH), the monthly climatology of sea ice area is generally well reproduced (see Fig. S2).

Tropospheric temperature response and associated climate feedbacks
The tropospheric temperature response to enhanced CH 4 mixing ratios can freely develop in the MLO sensitivity simulations (see Fig. 1). The temperature change patterns of S2 MLO and S5 MLO show the expected warming of the troposphere and 150 cooling of the stratosphere (e.g., IPCC, 2013). The stratospheric cooling is less pronounced than in carbon dioxide (CO 2 )-driven climate change simulations, since the CH 4 cooling is mainly caused by associated O 3 and H 2 O adjustments (Kirner et al., 2015;Winterstein et al., 2019). Maximum warming in polar regions and in the upper tropical troposphere is also consistent with changes expected from increased levels of GHGs (e.g., Chap. 12 in IPCC, 2013). CH 4 doubling (fivefolding) leads to temperature increases of up to 1 K (3 K) in the Arctic on annual average. Antarctica also warms up particularly strongly in the 155 S5 MLO scenario with a maximum warming of up to 3 K. As a result of the especially strong warming in polar regions, the sea ice area is reduced in both sensitivity simulations with respect to the reference (compare Fig. S2).
The Brewer-Dobson circulation (BDC) is expected to accelerate in a warming climate (Rind et al., 1990;Butchart and Scaife, 2001;Garcia and Randel, 2008;Butchart, 2014;Eichinger et al., 2019). Feedbacks on the chemical composition of the atmosphere, especially of the stratosphere, which result from changes of the BDC are of particular interest in this study, 160 as they will modify the mainly chemically induced changes discussed by Winterstein et al. (2019). The BDC influences the spatial distribution of trace gases, such as O 3 , H 2 O, and CH 4 , in the stratosphere and also their transport from the troposphere into the stratosphere (Butchart, 2014). In Fig. 2  simulations. This is expected, since the main driver of changes in the BDC is tropospheric warming (Butchart, 2014). We note that changes of the residual mean streamfunction below the tropical tropopause in response to CH 4 increase exhibit different patterns in the fSST and MLO simulations (see Fig. 2). A similar feature has been noticed and discussed in CO 2 increase simulations, too (e.g. Bony et al., 2013). However, trying to explain the origin of these tropospheric differences would leave the scope of the present paper, which focuses on stratospheric trace gas feedbacks to CH 4 increase. The latter are influenced 175 by the more distinct strengthening of the BDC in the MLO experiments, as we will show in the next section.
with m CH4 being the mass of CH 4 in kg, k CH4+OH (T ) the temperature dependent reaction rate coefficient of the reaction [mol mol −1 ] in all grid boxes b ∈ B. B is the region, for which the lifetime should be calculated, e.g. all grid boxes below the tropopause for the mean tropospheric lifetime. Please recall that we prescribe the CH 4 mixing ratios at the lower boundary using Newtonian relaxation. It is important to 215 note that the prolongation of the tropospheric CH 4 lifetime causes the corresponding CH 4 fluxes at the lower boundary to not scale equally with the mixing ratio increase, but to increase by a smaller factor. Increasing the CH 4 surface mixing ratio by a factor of 2 (5) corresponds to an increase of the CH 4 surface fluxes by a factor of 1.61 ± 0.01 (2.91 ± 0.01) in the MLO simulations, and by a factor of 1.58 ± 0.00 (2.75 ± 0.01) in the fSST simulations (see Tab. 2). The larger increase factors in the MLO sensitivity simulations are in line with the reduced prolongation of the tropospheric CH 4 lifetime compared to 220 the fSST experiments. The fact that the increase in emission fluxes is less than a factor of 2 or 5 suggests that enhanced CH 4 emissions would likewise scale the mixing ratio by a larger factor than the corresponding increase factor of the emissions. The CH 4 surface fluxes that result from the nudging of the mixing ratio towards zonally averaged CH 4 fields are not realistic in terms of spatial distribution, however. linearly to the increase. However, the CH 4 increase between 50 and 1 hPa has found to be smaller than a strictly linear relation would predict. This indicates enhanced chemical CH 4 depletion in the stratosphere due to changes in the chemical composition.

Non-linearities of CH
Ny · y with the mean values of the S2/S5 and REF fluxes x and y, respectively, interannual standard deviations sx and sy, number of analysed years Nx and Ny, α = 0.05, and the degrees of freedom df = ( Another aspect to note in Fig. 5 is the more than 5×CH 4 increase in the lowermost tropical stratosphere for S5 MLO. This feature indicates enhanced tropical upwelling, which leads to larger CH 4 mixing ratios in the tropical lower stratosphere. This feature is more pronounced in S5 MLO than in S5 fSST, in line with the more pronounced changes of tropical upwelling in 240 the MLO set-up as discussed in Sect. 3.2. Furthermore, strengthening of the BDC transports CH 4 more efficiently to higher altitudes leading to higher CH 4 mixing ratios there as well. This can be one explanation for the weaker deviation from a linear CH 4 increase in the MLO compared to the fSST simulations. Another explanation, as already stated, is that the chemical depletion of CH 4 is less strongly enhanced in the MLO sensitivity simulations compared to fSST. We therefore discuss differences of the response of OH, the most important sink partner of CH 4 , in the next paragraph. Stratospheric OH mixing ratios increase in both simulation set-ups (fSST and MLO) at the order of 30 % for 2×CH 4 and 60 %-80 % for 5×CH 4 . As shown by Winterstein et al. (2019), OH precursors (H 2 O and O 3 ) in the stratosphere are also affected by the CH 4 increase. The OH increase in the stratosphere is weaker in the MLO simulations compared to the fSST simulations (see Fig. 4). The differences are, however, small compared to the total increase of OH and mainly not significant.
The difference between the two 5×CH 4 experiments reaches up to 5 percentage points (p.p.) in the middle stratosphere. The  simulation compared to S2 fSST (S5 fSST). This reduction is significant, but small compared to the relative increase of SWV of around 50 % for both 2×CH 4 , and 250 % for both 5×CH 4 experiments. The amount of tropospheric H 2 O transported into the stratosphere is largely determined by the cold point temperature (CPT) (e.g., Randel and Park, 2019). Furthermore, the oxidation of CH 4 is an important in-situ source of SWV (Hein et al., 2001;Rohs et al., 2006;Frank et al., 2018). The SWV 270 mixing ratio at a given location and time can be approximated as the sum of these two terms (Austin et al., 2007;Revell et al., 2016  What remains to be explained is the reason for the weaker strengthening of the CH 4 oxidation in the MLO setup compared to 285 fSST. Strengthened tropical upwelling as shown in Sect. 3.2 transports CH 4 into the stratosphere more efficiently and would be expected to lead to higher rates of the CH 4 oxidation (Austin et al., 2007). However, as the strengthening of the CH 4 oxidation is weaker in the MLO experiments, CH 4 itself seems not to be the limiting factor here. The abundance of SWV feeds back on OH and therefore also on the efficiency of the CH 4 oxidation. However, the increase of SWV seems to be rather a result of the strengthened CH 4 oxidation here, as the increase of H 2 O entering the stratosphere is higher in the MLO experiments compared concerning the RI, which indicates the radiative flux imbalance between the sensitivity and the reference simulation.
In Table 3  to 0.72 ± 0.04 W m −2 for 2×CH 4 and from 0.30 ± 0.06 W m −2 to 2.23 ± 0.06 W m −2 for 5×CH 4 ). The RI of stratospheric 325 H 2 O increases as well, which is mostly influenced by the increase in SWV in the lowermost stratosphere due to transport of moist air from the tropical troposphere into the stratosphere (see Fig. 6). In a recent multimodel comparison, the multimodel mean efficacy of CH 4 was found to be smaller than one, however, with a large intermodel spread ranging from 0.56 to 1.15 (Richardson et al., 2019). Modak et al. (2018)  that the CH 4 efficacy is significantly larger or smaller than unity in their framework, as the inter-model spread reported by (Richardson et al., 2019) is so large. Estimating a reasonable climate sensitivity value from our simulations in an interactive chemistry framework, requires that rapid adjustments from SWV and O 3 are included in the effective CH 4 forcing. If this is 345 done, these simulations do not point at a significant climate sensitivity deviation from the CO 2 behavior either. The values after the ± sign are the 95 % confidence intervals of the mean.
For λ the confidence intervals are calculated using Taylor expansion and assuming ∆TMLO and total RIfSST to be uncorrelated as ± t α 2 ,df · x y · s 2 x Nx ·x + s 2 y Ny · y with the mean values of ∆TMLO and total RIfSST x and y, respectively, interannual standard deviations sx and sy, number of analysed years Nx and Ny, α = 0.05, and the degrees of freedom df = ( Ny −1 ) −1 . S5 MLO (see Fig. S11 for simulation S2 MLO and Fig. 9 for simulation S5 MLO). The difference of adjusted stratospheric temperature response between S5 MLO and S5 fSST is shown in Fig. 10 (for S2 see Fig. S12). This difference is small for CH 4 and tropospheric O 3 (see Fig. 10 (b) and (g)). Figure 10 (d) confirms the stratospheric radiative cooling effect of increased humidity in the troposphere in S5 MLO, although the effect is quantitatively small. The adjusted stratospheric temperature response pattern induced by SWV in S5 MLO is similar to S5 fSST. However, the stronger increases of SWV in S5 MLO 360 result in more pronounced cooling in the lowermost stratosphere, whereas the reduced increases above consistently result in reduced cooling (see Fig. 10 (e)). The stronger decrease of O 3 in the tropical lower stratosphere in S5 MLO (see Fig. 7) leads to stronger cooling in this region as shown in Fig. 10 (h). These results also apply qualitatively to the comparison of S2 MLO and S2 fSST (see Fig. S12), but the magnitude of the differences is smaller. The effects from SWV and stratospheric O 3 dominate the differences of stratospheric adjusted temperature between S5 MLO and S5 fSST (compare Fig. 10 (a)). In addition, the

Radiatively and dynamically driven atmospheric temperature response
with X being either 2 or 5. A similar approach was, for example, used by Rosier and Shine (2000) and Schnadt et al. (2002) to distinguish between the radiative impact of trace gases and dynamical contributions to the total temperature response. Fig. 11 shows the annual mean of ∆T dyn. for all four sensitivity simulations. It is mostly not significant for S2 fSST and S5 fSST in the stratosphere suggesting that dynamical effects play a minor role in the temperature response in these simulations 375 as already indicated by Winterstein et al. (2019). However, immediately above the tropical tropopause centered at the equator ∆T dyn. indicates warming for both, S2 fSST and S5 fSST. In austral winter (JJA), ∆T dyn. shows significant cooling in the southern polar stratosphere for S2 fSST and S5 fSST. The cooling extends into austral spring (SON), but gradually weakens as time proceeds (see Fig. S15 and Fig. S16). These temperature changes can be associated to the strengthening of the SH stratospheric winter polar vortex (see Fig. S18), which leads to enhanced isolation of airmasses and stronger cooling. The 380 stratospheric polar vortex in boreal winter DJF accelerates in both fSST sensitivity simulations as well (see Fig. S17).
The pattern of ∆T dyn. for S5 MLO (Fig. 11 (d)) displays a near-symmetrical behavior around the equator. It comprises of two warming patches in the lower stratosphere -unlike S5 fSST not centered at the equator, but at around 30 • S or 30 • N -, as well as cooling in the tropics and warming in the extratropics in the middle stratosphere. The warming patches in the lower stratosphere are present in all seasons, whereas the pattern of cooling in the tropics and warming in the extratropics above is 385 shifted to the respective winter hemisphere (compare Fig. S13 and Fig. S15). For S2 MLO, the warming patches in the lower stratosphere are also present in the pattern of ∆T dyn. . Apart from that, the annual mean ∆T dyn. is mostly not significant for S2 MLO. However, the pattern of cooling in the tropics and warming in the extratropics is indicated in boreal autumn (SON) and winter (DJF) for S2 MLO as well. discussed in Sect. 3.2. Strengthened downwelling in the subtropical and extratropical lower stratosphere results in adiabatic warming in this region in both hemispheres throughout the year. These temperature changes can therefore be associated with the intensification of the shallow branch of the BDC (Plumb, 2002;Birner and Bönisch, 2011). The patterns are present in S2 MLO and S5 MLO. Adiabatic cooling in the tropical middle and upper stratosphere, as well as a respective adiabatic warming in the extratropical and polar winter stratosphere indicate the strengthening of the deep branch of the BDC, more 395 pronounced in S5 MLO than in S2 MLO. The strengthening of the BDC would be expected to result in adiabatic cooling directly above the tropopause from increased tropical upwelling. This effect seems to be masked by other processes in Fig. 11. These could be advection or mixing of warm air from the troposphere, or increased longwave (LW) radiation from the warmer troposphere and potentially more LW absorption in the lowest stratosphere. Lin et al. (2017) found the latter effect to cause strong warming in the tropical tropopause layer. This radiative effect is not accounted for in ∆T addst (SX*-REF*), which is the 400 sum of the individual contributions of radiatively active gases to the adjusted stratospheric temperatures. Furthermore, mixing with air out of the upper tropical troposphere could also contribute to the warming patches in the subtropical and extratropical lower stratosphere. This region is particularly affected by mixing (Dietmüller et al., 2018;Eichinger et al., 2019) and mixing itself can also be influenced by climate change (Eichinger et al., 2019).
The deep branch of the residual mean circulation is closely linked to the strength of the winter stratospheric polar vortex.

405
An increase in the poleward flow and in downwelling at higher latitudes is accompanied with a slow down of the stratospheric polar vortex (Kidston et al., 2015, and references therein). The S5 MLO response of zonal mean winds shows indeed an easterly change of the stratospheric polar vortex in boreal winter (DJF) (see Fig. S17). The respective response for S2 MLO is not significant, but decelerating, too. The SH stratospheric polar vortex strengthens for S2 MLO, but less than in S2 fSST.
Nevertheless, the response of stratospheric zonal winds in both MLO experiments is substantially different from fSST in the 410 SH as well.
The easterly change of polar stratospheric zonal winds in the NH during DJF is consistent with the response of the stratospheric polar vortex in CMIP5 global warming simulations (Manzini et al., 2014;Karpechko and Manzini, 2017). Moreover, differences between the fSST and MLO response signals of stratospheric zonal winds during DJF are qualitatively consistent with the results of Karpechko and Manzini (2017). They identified, on the one hand, a deceleration of the stratospheric polar 415 vortex and associated warming in the polar stratosphere in simulations driven by higher SSTs (comparable to the MLO experiments), and, on the other hand, a strengthened and cooled stratospheric polar vortex in simulations driven by CO 2 increase and suppressed tropospheric warming (comparable to the fSST experiments). Karpechko and Manzini (2017) suggested that tropospheric warming and associated strengthening of subtropical winds lead to enhanced wave activity. In S5 MLO subtropical winds strengthen indicating that similar processes might act in our simulations. However, a detailed analysis of wave activity 420 is beyond the scope of this study.
In summary, SST-driven climate feedbacks affect the chemical composition. The differences in stratospheric temperature adjustment between MLO and fSST (see Fig. 10) reflect radiative impacts of these composition changes on stratospheric temperature. Additionally, the patterns of ∆T dyn. suggest that dynamical effects have changed significantly in the MLO simulations with respect to fSST. The dynamical temperature response effect for S5 MLO is consistent with the strengthening of the BDC.

425
Dynamic heating counteracts the radiative cooling in the extratropical middle and upper stratosphere and in the subtropical lower stratosphere in S5 MLO. This results in reduced cooling in these regions in S5 MLO in Fig. 8, which is not significant on annual average, but in the respective winter hemispheres (not shown). ∆T dyn. for S2 MLO indicates strengthening of mainly the shallow branch of the BDC.  While it has been long-since acknowledged that the net RF of CH 4 includes substantial contributions from O 3 and SWV (e.g., IPCC, 2013, Fig. 8.17), it is still common to consider climate feedbacks and climate sensitivity of CH 4 in comparison to CO 2 without accounting for these additional radiative components (Modak et al., 2018;Smith et al., 2018;Richardson et al., 2019).
Our study provides a quantification of SST-driven slow radiative feedbacks from CH 4 , O 3 and associated SWV changes in The strongly enhanced CH 4 mixing ratios cause enhanced depletion of OH in the troposphere. Tropospheric warming, in contrast, results in enhanced OH precursors and causes the reduction of OH in the troposphere to be weaker than in the prescribed SST simulations analysed by Winterstein et al. (2019). Additionally, the acceleration of the CH 4 oxidation at higher temperatures leads to a more efficient depletion of CH 4 in a warming troposphere. This so called climate offset results in a 440 reduced prolongation of the tropospheric CH 4 lifetime and is consistent with previous CCM studies ( Voulgarakis et al., 2013).
The prolonged tropospheric CH 4 lifetime has the effect that the corresponding CH 4 surface fluxes increase by a smaller factor than the mixing ratio.
Changes in the stratospheric circulation can be clearly identified in the sensitivity simulations that include SST-driven climate feedbacks, on top of the quasi-instantaneous response analysed by Winterstein et al. (2019). Tropospheric warming leads to 445 the acceleration of the BDC in our sensitivity simulations as expected from climate change scenario calculations (Butchart, 2014). In the lower tropical stratosphere, both the decrease of O 3 and the associated cooling, and the increase in CH 4 become more distinct, which reflects the more pronounced acceleration of tropical upwelling induced by a warming troposphere.
The strengthening of the BDC also manifests in the temperature response. Whereas the stratospheric polar vortices in both winter hemispheres strengthen in the experiments with prescribed SSTs and SICs, polar stratospheric zonal winds decelerate in 450 northern winter in the sensitivity simulation that include tropospheric warming consistent with the response in CMIP5 global warming simulations (Manzini et al., 2014;Karpechko and Manzini, 2017).
As a result of tropical upper troposphere moistening, increased tropical upwelling and more pronounced warming of the cold point, the transport of tropospheric H 2 O into the lower stratosphere is more strongly enhanced in the sensitivity simulations that include SST-driven climate feedbacks, resulting in a stronger increase of SWV in the lower extratropical stratosphere. In 455 the middle and upper stratosphere, where CH 4 oxidation makes an important contribution to SWV, the increase of SWV is weakened in the present sensitivity simulations compared to the quasi-instantaneous response. Less pronounced increases of stratospheric OH in response of the slow adjustments in comparison to the quasi-instantaneous response cause the depletion of CH 4 to be weaker, and thus the in situ source of SWV as well.
The contribution of SST-driven climate feedbacks to the total CH 4 induced O 3 response shows remarkable similarities to 460 the O 3 response to climate feedbacks in CO 2 -forced climate change simulations (Dietmüller et al., 2014;Nowack et al., 2018;Chiodo and Polvani, 2019). The consistency between the O 3 feedbacks resulting from these different forcing agents encourages the separation of the O 3 response patterns into rapid adjustments and climate feedbacks in future studies. Rapid adjustments are specific to the forcing, whereas climate feedbacks are driven by surface temperature changes and are therefore expected to be less dependent on the forcing agent (Sherwood et al., 2015).

465
The doubled and fivefold CH 4 mixing ratios result in global mean surface temperature changes of 0.42 ± 0.05 K and 1.28 ± 0.04 K, respectively. We estimate the corresponding climate sensitivity parameters λ using these temperature changes and the respective RIs from CH 4 and the respective chemical adjustments, as determined by Winterstein et al. (2019), that can well be interpreted as the corresponding ERFs. The respective estimate of λ for 5×CH 4 compares well with an estimate from CO 2 -driven climate change simulations with EMAC with comparable magnitude of RI (Rieger et al., 2017), suggesting an 470 efficacy of CH 4 ERF close to one. The estimate of λ corresponding to 2×CH 4 is smaller than the respective value for 5×CH 4 , but has a large uncertainty. Considering the large uncertainty and intermodel spread (Richardson et al., 2019) of this parameter, we conclude that a more targeted experimental design is necessary to exactly quantify the effect of chemical feedbacks on the climate sensitivity in CH 4 -driven scenarios and its efficacy with respect to CO 2 forcing.
The RIs from the purely SST-driven response of CH 4 and O 3 are small. The RIs resulting from changes of tropospheric and 475 stratospheric H 2 O are enlarged by SST-driven climate feedbacks. Increased tropospheric humidity in a warming troposphere enhances the RI. The reason for the enlarged RI from SWV is its more pronounced increase in the lower stratosphere, where its changes dominate the induced RI (Solomon et al., 2010). As the increase of SWV in this region is likely induced by transport from the warmer tropical troposphere, this part of the RI increase cannot be regarded to be a chemically induced rapid adjustment. The associated responses of stratospheric adjusted temperatures from the purely SST-driven response are 480 dominated by the just explained changes of SWV and by decreases of stratospheric O 3 in the lowermost tropical stratosphere.
It is worth noting, that tropospheric CH 4 mixing ratios do not respond to changes in tropospheric sinks (e.g. OH) in the used simulation set-up, as its mixing ratio is prescribed at the lower boundary. The prolongation of the tropospheric CH 4 lifetime indicates a positive feedback on the CH 4 mixing ratio, and thus on the induced RI. In a future study, climate change scenario simulations conducted with a CCM with realistic CH 4 emission fluxes are planned to quantify this chemical feedback of CH 4 .

485
In the present study we are able for the first time to quantify the effects of slow climate feedbacks on the chemical composition and circulation in CH 4 -forced climate change scenarios and further evaluate them in comparison to the quasi-instantaneous atmospheric response. Competing interests. The authors declare that they have no conflict of interest.
internal review of the manuscript and Hella Garny for her helpful comments on the interpretation of the dynamically induced temperature response.