Sulfate geoengineering impact on methane transport and lifetime : results from the Geoengineering Model Intercomparison Project ( GeoMIP )

Sulfate geoengineering (SG), made by sustained injection of SO2 in the tropical lower stratosphere, may impact the CH4 abundance through several photochemical mechanisms affecting tropospheric OH and hence the methane lifetime. (a) The reflection of incoming solar radiation increases the planetary albedo and cools the surface, with a tropospheric H2O decrease. (b) The tropospheric UV budget is upset by the additional aerosol scattering and stratospheric ozone changes: the net effect is meridionally not uniform, with a net decrease in the tropics, thus producing less tropospheric O(1D). (c) The extratropical downwelling motion from the lower stratosphere tends to increase the sulfate aerosol surface area density available for heterogeneous chemical reactions in the mid-to-upper troposphere, thus reducing the amount of NOx and O3 production. (d) The tropical lower stratosphere is warmed by solar and planetary radiation absorption by the aerosols. The heating rate perturbation is highly latitude dependent, producing a stronger meridional component of the Brewer–Dobson circulation. The net effect on tropospheric OH due to the enhanced stratosphere–troposphere exchange may be positive or negative depending on the net result of different superimposed species perturbations (CH4, NOy , O3, SO4) in the extratropical upper troposphere and lower stratosphere (UTLS). In addition, the atmospheric stabilization resulting from the tropospheric cooling and lower stratospheric warming favors an additional decrease of the UTLS extratropical CH4 by lowering the horizontal eddy mixing. Two climate–chemistry coupled models are used to explore the above radiative, chemical and dynamical mechanisms affecting CH4 transport and lifetime (ULAQ-CCM and GEOSCCM). The CH4 lifetime may become significantly longer (by approximately 16 %) with a sustained injection of 8 Tg-SO2 yr−1 starting in the year 2020, which implies an increase of tropospheric CH4 (200 ppbv) and a positive indirect radiative forcing of sulfate geoengineering due to CH4 changes (+0.10 W m−2 in the 2040–2049 decade and +0.15 W m−2 in the 2060–2069 decade).


Introduction
Many geoengineering methods have been proposed in order to temporarily balance out the direct effect of the increase of anthropogenic greenhouse gas emissions (Kravitz et al., 2011).Amongst those, stemming from the observations of the effects of large volcanic eruptions, is the injection of sulfate aerosol precursors (e.g., SO 2 ) into the stratosphere (Crutzen, 2006;Robock et al., 2011;Kravitz et al., 2012).The injection above the tropopause of very large amounts of particles and sulfur gases due to explosive volcanic eruptions Published by Copernicus Publications on behalf of the European Geosciences Union.D. Visioni et al.: Sulfate geoengineering and methane lifetime is able to increase the stratospheric aerosol optical depth by more than 1 order of magnitude.The initial volcanic SO 2 plume quickly nucleates into H 2 SO 4 vapor (Bluth et al., 1992), producing an optically thick cloud of sulfate aerosols (McCormick and Veiga, 1992;Lambert et al., 1993;Long and Stowe, 1994).The high reflectivity of these aerosols effectively decreases the amount of solar radiation reaching the Earth's surface, thus producing a net global cooling.In 1991, for example, the Pinatubo eruption produced a reduction of the global surface air temperature from 0.5 K (Soden et al., 2002) to 0.14 K using detrended analyses (Canty et al., 2013).
Besides the direct effect on surface temperatures, however, there is the need for a thorough examination of other effects on atmospheric circulation and chemical composition of the troposphere and stratosphere brought about by the increase in lower stratosphere optical thickness (Visioni et al., 2017).The interaction of the H 2 SO 4 particles with solar radiation is twofold: the aerosols increase the amount of radiation that is reflected and scattered but they also absorb part of it in the near-infrared wavelengths, increasing the lower stratospheric diabatic heating rates.This causes a local positive temperature change (Labitzke and McCormick, 1992) which induces a significant increase of westerly winds from the thermal wind equation with peaks at midlatitudes in the midstratosphere (Pitari et al., 2016c).These dynamical changes tend to increase the amplitude of planetary waves in the stratosphere and to enhance the tropical upwelling in the rising branch of the Brewer-Dobson circulation (Pitari et al., 2014;Aquila et al., 2014).For continuity, a stronger downward component is found in the lower branch of the Brewer-Dobson circulation (Aquila et al., 2013;Pitari et al., 2016b).
These dynamical changes can bring about modification in the concentration and growth rate of long-lived species that act as greenhouse gases, such as N 2 O and CH 4 , as observed in the case of the Pinatubo eruption (Schauffler and Daniel, 1994;Dlugokencky et al., 1994).An increase in the downward mid-and high-latitude fluxes in the lower stratosphere ends up advecting more stratospheric air below the tropopause, thus decreasing the tropospheric concentration of these gases.In addition, the horizontal eddy mixing in the upper troposphere and lower stratosphere (UTLS) is lowered as a consequence of the atmospheric stabilization resulting from the tropospheric cooling and lower stratospheric warming, thus decreasing the isentropic transport of CH 4 and N 2 O from the tropical pipe towards the midlatitudes.This favors an additional decrease of the UTLS extratropical downward fluxes of CH 4 and other long-lived species (Pitari et al., 2016b).The overall effect on tropospheric OH due this enhanced stratosphere-troposphere exchange and perturbed UTLS horizontal mixing may be positive or negative depending on the net result of different superimposed species perturbations in the UTLS (CH 4 , NO y , O 3 ).
Coupled with this perturbation of the stratospheretroposphere exchange, the lifetime of long-lived species with tropospheric OH sink can also be modified by other changes brought about by an injection of tropical stratospheric aerosols: (a) the surface cooling would directly lessen the amount of water vapor, thus lowering the tropospheric OH concentration; (b) the tropical tropospheric UV decrease due to enhanced radiation scattering would reduce the production of O( 1 D), thus decreasing OH production from O( 1 D) + H 2 O; (c) the increasing aerosol surface area density (SAD) would enhance heterogeneous chemistry in the mid-to-upper troposphere, which reduces the amount of NO x and the rate of O 3 production, both negatively affecting the amount of tropospheric OH.Since CH 4 is depleted by the OH radical, all these changes would mean an increase in methane lifetime (Bândǎ et al., 2013(Bândǎ et al., , 2015)).The aim of this study is to evaluate the chemical, radiative and dynamical effects of a sustained injection of SO 2 in the stratosphere on the lifetime and abundance of CH 4 .
The paper is organized in seven subsequent parts.Section 2 includes a description of participating models.In Sect.3, a model evaluation for long-lived species stratospheric abundance and transport is presented using available satellite observations.Section 4 analyzes the sulfate geoengineering (SG) induced perturbations on stratospheric species transport, while Sect. 5 discusses the effects on tropospheric chemistry and CH 4 direct and indirect radiative forcing components, with the overall main conclusions discussed in Sect.6.

Model experiments
The characteristics of the experiment follow the description of experiment G4 in the Geoengineering Model Intercomparison Project (GeoMIP) (Kravitz et al., 2011).The G4 experiment consists of a constant yearly injection of SO 2 in the tropical lower stratosphere.The SO 2 injection is handled by the single models in the same way they simulate the Pinatubo eruption in terms of injection height.The background anthropogenic forcing corresponds to the one from the Representative Concentration Pathway 4.5 (RCP4.5)(Taylor et al., 2012).Starting from 2020, 8 (or 5) Tg-SO 2 yr −1 are injected in the stratosphere with a sudden stop after 50 years.An additional 20 years of model simulations are performed (up to 2090) in order to assess the termination effects of the sulfur injection.The choice of the different amounts of injected SO 2 follows two reasons: for some of the analyses, we have decided to use the same simulations used in Pitari et al. (2014) with 5 Tg-SO 2 yr −1 .However, two experiments with varying sea surface temperatures (SSTs) have also been carried out with ULAQ-CCM to identify possible changes due to these dynamics-driving mechanisms; for this reason, an injection of 8 Tg-SO 2 yr −1 was performed with ULAQ-CCM in order to use the CCSM-CAM4 surface temperatures that resulted from a 8 Tg-SO 2 yr −1 injection.The main features of the participating models are summarized in Table 1.

Model
Resolution The model is the same as described in Tilmes et al. (2016), but in this case it was run with no interactive chemistry.d Quasi-biennial oscillation (QBO) internally generated using a gravity wave drag parameterization and resolved wave forcing.e ULAQ-CCM includes aerosol microphysics (RCP4.5 r eff = 0.19 µm).
One of these models (CCSM-CAM4) is an atmosphereocean coupled model and it has been used (without interactive chemistry) to calculate the surface temperature evolution from 2010 to 2090 for a reference RCP4.5 case and a geoengineering G4 perturbed case with 8 Tg-SO 2 yr −1 injected continuously from 2020 to 2070 (Kravitz et al., 2011).One of the other two models (ULAQ-CCM) has assimilated surface temperatures calculated in the CCSM-CAM4 atmosphereocean coupled model for the reference RCP4.5 and the perturbed G4 cases (i.e., two different datasets for surface temperatures), whereas the third model (GEOSCCM) has run the G4 case with RCP4.5 SSTs assimilated from the CESM atmosphere-ocean coupled model.Both models prescribe CH 4 mixing ratios at the surface (except in one numerical experiment of ULAQ-CCM where emission fluxes are used, as discussed below) and do not include changes in emission fluxes due to surface temperature modifications.A more detailed description of these numerical models can be found in Tilmes et al. (2016) and Pitari et al. (2014).
In order to properly assess the different contributions to CH 4 changes discussed before, three different experiments have been carried out with the ULAQ-CCM model: experiments (a) and (b) use appropriate surface temperatures for RCP4.5 and G4 cases (as previously explained), with surface CH 4 treated under mixing ratio boundary condition (MBC) and flux boundary condition (FBC) approaches for (a) and (b), respectively.Experiment (c), on the other hand, uses the same SST for both RCP4.5 and G4 cases (as in GEOSCCM), with the purpose of highlighting the impact of SST changes on the G4-RCP4.5 large-scale transport perturbations.The full list of numerical experiments completed with the three models is presented in Table 2.
The ULAQ-CCM sensitivity cases run with the FBC approach will help in assessing the role of temperature and wind changes in the CH 4 lifetime perturbation under geoengineering conditions.

Model evaluation
Both ULAQ-CCM and GEOSCCM have already been extensively reviewed in the past, both on their general features (Morgenstern et al., 2017) or for issues related to this study, such as the extratropical UTLS (Hegglin et al., 2010), or surface UV (Bais et al., 2011).The shortwave radiative transfer module of the ULAQ-CCM was carefully evaluated in the AeroCom intercomparison exercise of Randles et al. (2013).
In order to properly evaluate the models regarding the specific points of this paper, however, further evaluations have been done with different sets of observations.A list of these is available in Table 3. CH 4 measurements are taken by the Halogen Occultation Experiment (HALOE),   TES 2004-20102004-2010HALOE 1991-2005SMR 2001-2005 which is onboard the Upper Atmosphere Research Satellite (UARS), launched in 1991 (Russell et al., 1993).Climatologies are formed for the period 1991-2005 based on extended data from Grooß and Russell III (2005).HALOE measurements range from 15 to 60-130 km altitude (depending on the species) and cover 80 • S to 80 • N in latitude within 1 year.In all intercomparisons, the HALOE climatological mean and the interannual standard deviation (1σ ) are shown.CH 4 and N 2 O profiles are estimated by Aura Tropospheric Emission Spectrometer (TES) thermal infrared radiances at λ = 8 µm with the version 5 retrieval algorithm, where CH 4 is corrected using co-retrieved N 2 O estimates (Worden et al., 2012).Climatological mean and interannual standard deviations for both species are calculated for the period 2004-2010.Climatological mean and interannual standard deviation of N 2 O between 2001-2005 are based on the Odin/SMR product (Urban et al., 2009).A further discussion regarding TES and HALOE differences can be found in Pitari et al. (2016a), together with a more in-depth evaluation of ULAQ-CCM CH 4 predictions.CH 4 diagnostics largely reflect the skill of the transport representation in the models.We examined climatological zonal profiles at selected latitudes, months and pressure levels for both model outputs and observations (Fig. 1).The climatologies refer to the years 1990-2010, in order to include the range of HALOE and TES observations.Both ULAQ-CCM and GEOSCCM compare well with observations and are normally in the ±1σ deviation interval, relative to the climatological zonal mean.Some spread between models appears, more evidently in the polar regions at 100 hPa.This might be due to a combination of insufficient advective high-latitude downwelling and too-strong eddy mixing in the Southern Hemisphere during the autumn season in ULAQ-CCM.GEOSCCM values are generally closer to observations than those of ULAQ-CCM.Otherwise, models perform quite similarly, and overall these diagnostics do not reveal major weaknesses in the simulations.
A more in-depth evaluation of transport properties in the models can be found in the Supplement regarding the correlation between CH 4 and N 2 O and the mean age of air.The correlation between CH 4 and N 2 O can be used to investigate transport properties relative to model and observations (SPARC-CCMVal, 2010).Figures S1 and S2 in the Supplement show CH 4 vs.N 2 O correlations between 100 and 1 hPa.In Table S1 in the Supplement, we present Pearson correlation coefficients relative to the different latitude bands.All these panels show a compact correlation and a good agreement with the observations; the relative Pearson coefficients in Table S1a and b are always significant.Panels regarding polar regions (Figs.S1a, d and S2a, d) present a larger spread with a slightly lower (but still significant) Pearson coefficient between 90 and 60 • S. In the lower stratosphere at tropical and midlatitudes, there is a strong compact relationship between CH 4 and N 2 O related to the slope equilibrium (Sankey and Shepherd, 2003): the mixing happens on a faster timescale than the chemical loss and transport to the surface.At polar latitudes, the correlation is affected by vortex edge, which represents a mixing barrier during the winter-spring season (Fig. S3).
Another important diagnostic for the evaluation of the model transport is based on the mean age of air (AoA).In particular, the latitudinal gradient between tropics and midlatitudes can be used to assess tropical ascent independently of quasi-horizontal mixing (SPARC-CCMVal, 2010).Following Strahan et al. (2011), tropical mean AoA profiles combine the effect of ascent rate and horizontal mixing.The agreement of model and observations only shows that the combined effects of ascent and mixing produce a realistic mean AoA in the models.Figure S3c identifies how ascent contributes to the overall tropical transport.
The horizontal gradient of mean age is able to reveal some characteristics of the Brewer-Dobson circulation (BDC) ( Neu and Plumb, 1999), namely the ascent rate.In fact, differences between midlatitude and tropical values exclude horizontal mixing, since that equally affects both the tropics and midlatitudes.In GEOSCCM and ULAQ-CCM, the horizontal gradient is smaller than observations by up to 21 km, indicating a fast ascent, but still included in the range of observed variability.The analysis of the relationship between mean AoA and N 2 O (Fig. S3d) evaluates the lower stratospheric transport and our use of the well-measured N 2 O in In order to properly asses the temperature of the polar stratosphere and its interannual variability, the models must correctly simulate the vertical propagation of planetary waves from the troposphere to the stratosphere.Since it is possible to use the correlation between winter polar temperatures and eddy heat fluxes in the lower stratosphere as a proxy for planetary wave propagation, we looked at the correlation between the meridional heat flux at 100 hPa (40 to 80 • for the two hemispheres) and the 50 hPa polar temperatures (60 to 90 • for the two hemispheres), following Eyring et al. (2006).Table S2 in the Supplement compares the coefficients of the linear fit between the two quantities for ULAQ-CCM, GEOSCCM and the ERA40 reanalysis.The positive slope is found in both models and reanalysis, with a greater similarity in the Northern Hemisphere with respect to the Southern Hemisphere; this difference was already shown in Eyring et al. (2006).
In Fig. 2, the vertical mass fluxes are evaluated by looking at the CH 4 and N 2 O measurements combined with the vertical velocities measured by MERRA, defining the flux as [ρw * ].A good agreement between measurements and models is found in the 5 to 100 hPa profile, with GEOSCCM underestimating the vertical flux between 50 and 30 hPa. heat fluxes in order to further evaluate the transport skill of the two models.

Perturbation of stratospheric species transport
Absorption of solar near-infrared (NIR) and planetary radiation by the geoengineering aerosols produces an increase of diabatic heating rates in the tropical lower stratosphere, resulting in local warming, changes in the latitudinal distribution of zonal winds, changes of the equatorial quasi-biennial oscillation (QBO) (Aquila et al., 2014) and a strengthening of the stratospheric Brewer-Dobson circulation (BDC) (Pitari et al., 2014).Enhanced tropical upwelling (about 5-10 % increase in vertical velocities in the lower stratosphere) and extratropical descent tend to move CH 4 poor air more efficiently towards the extratropical UTLS, as well as for other stratospheric long-lived species.The net impact on tropospheric OH and CH 4 lifetime depends on the net result of superimposed species perturbations in the UTLS (CH 4 , NO y , O 3 , SO 4 ), in addition to tropospheric chemistry perturbations due to changes in water vapor content, UV radiation and heterogeneous reactions on sulfate aerosols that affect the NO x balance.
The 5-10 % increase of stratospheric tropical upward mass fluxes of both CH 4 and N 2 O, as shown in Fig. 3a, b, is predicted by the models in geoengineering conditions as a consequence of the increasing tropical midstratospheric upwelling, with a larger anomaly in GEOSCCM with respect to both MBC experiments run with the ULAQ-CCM (cases a and c in Table 1, with 8 and 5 Tg-SO 2 injected, respectively).The choice to only include MBC experiments when discussing vertical mass flux anomalies is made in order to bet-ter highlight transport anomalies, because in the FBC experiment the anomaly would be largely masked by the increasing amount of tropospheric CH 4 .The larger GEOSCCM anomaly could be explained by the QBO modification produced by geoengineering aerosols, since the prolonged lower stratospheric westerly phase produces a better tropical confinement (Trepte and Hitchman, 1992;Aquila et al., 2014;Visioni et al., 2017).This effect is absent in the ULAQ-CCM model, which does not have an internally generated QBO, but specifies the QBO with observed equatorial zonal wind data using a nudging procedure (Morgenstern et al., 2010).
The UTLS horizontal mixing anomalies (Fig. 3c, d) are larger in case (a) of ULAQ-CCM with respect to ULAQ-CCM (c) and GEOSCCM.In the latter, two model simulations, RCP4.5 SSTs are used for both the baseline and the geoengineering perturbed experiments, whereas ULAQ-CCM (a) is driven in the latter experiment by G4 surface temperatures (from CCSM-CAM4).In this case, the larger decrease of the UTLS horizontal mixing can be explained by the increased atmospheric stabilization caused by the sea surface cooling, which is not present in GEOSCCM and ULAQ-CCM (c).The ULAQ-CCM (c) results do not change significantly in a sensitivity simulation made by increasing the stratospheric sulfur injection from 5 to 8 Tg-SO 2 yr −1 (see Table 2), pointing out the important role of the decreasing horizontal mixing resulting from sea surface cooling as in ULAQ-CCM (a).
The time series of model calculated CH 4 and N 2 O changes in the UTLS is presented in Fig. 4 for ULAQ-CCM and GEOSCCM.If we compare the ULAQ-CCM case (c) with GEOSCCM, the results of the two models are similar for N 2 O and are consistent with changes of lower stratospheric heating rates and BDC (due to aerosols and O 3 ).The N 2 O anomalies are of the order of −1 ppbv in both models (that is about −0.3 %), while those of CH 4 are of the order of −5 ppbv in the ULAQ model and about a factor of 2 smaller in GEOSCCM.This is due to missing chemical processes in the upper troposphere in GEOSCCM, where tropospheric OH is kept fixed at RCP4.5 values.As already discussed in Fig. 3, the UTLS anomalies G4-RCP4.5 are rather different for ULAQ-CCM (a), mostly as a consequence of the changing SSTs in G4, with decreased horizontal mixing in the UTLS and enhanced isolation of the tropical pipe.The negative anomaly of N 2 O (a quasi-passive tracer) increases up to 2-4 ppbv after 2030, whereas the negative CH 4 anomaly increases up to approximately 10 ppbv between 2030 and 2050.A clear sign inversion is predicted after 2050 for the CH 4 anomaly in geoengineering conditions as a consequence of a negative OH trend resulting from superimposed effects of NO x and O 3 .A positive trend of stratospheric O 3 is, in fact, predicted in G4 with respect to RCP4.5 due to the lowering chlorine-bromine loading in the atmosphere in the 21st century (Pitari et al., 2014).
The zonally averaged changes of N 2 O and CH 4 are presented in Fig. 5, with a comparison of model results from GEOSCCM and ULAQ-CCM (a).The midstratospheric changes are quite comparable between the two models, whereas the UTLS negative anomalies in ULAQ-CCM  1, respectively) (decadal averages).Units are ppbv.
(a) are significantly larger for the reasons discussed above in Figs.3-4, they are fully comparable when considering GEOSCCM and ULAQ-CCM (c) results, as shown in Fig. S5.Again, this points out the sea surface cooling role on the UTLS horizontal mixing in sulfate geoengineering conditions.Further remaining differences between GEOSCCM and ULAQ-CCM (c) regarding horizontal mixing can be explained by a different treatment of QBO, which is modified in GEOSCCM with a prolonged e-shear in the G4 simulation.Interhemispheric asymmetries in the lower stratospheric mixing ratio anomalies of ULAQ-CCM (a) and their differences with respect to GEOSCCM can be explained by a combination of vertical and horizontal mass flux changes and will be addressed later on in the discussion.
To better understand the differences between the cases with fixed SSTs and the one with changing SSTs, in Fig. 6 we show the anomalies in sea surface temperatures used in ULAQ-CCM (a) and (b) against the ones used in ULAQ-CCM (c); surface temperatures are taken from the CCSM-CAM4 atmosphere-ocean coupled model for RCP4.5 and G4 simulations (with an injection of 8 Tg-SO 2 ), as described in Table 2.The zonally averaged surface temperature anomalies G4-RCP4.5 are presented in Fig. 6a for the various decades from 2020 to 2090.A strong interhemispheric asymmetry is evident, with a negative anomaly more pronounced in the Arctic region by approximately 1.5 K with respect to the latitude range 50-70 • S. The geoengineering cooling impact on Arctic sea ice is the main driver for the larger negative temperature anomaly in the Northern Hemisphere high latitudes, which favors a more pronounced atmospheric stabilization in the Northern Hemisphere winter-spring months with respect to the Southern Hemisphere.The time series of the globally averaged surface temperature anomalies is shown in Fig. 6b for the RCP4.5 and G4 cases: the slow oceanic response coupled to the atmospheric perturbation of D. Visioni et al.: Sulfate geoengineering and methane lifetime long-lived species delays the surface temperature return in G4 to RCP4.5 values by more than one decade.
The decreased horizontal fluxes of long-lived species discussed in Fig. 3 for the ULAQ-CCM simulations with changing SSTs are a direct consequence of the atmospheric stabilization.As shown in Fig. 6c, the increased atmospheric stability in sulfate geoengineering conditions may be partially counterbalanced by the increased longitudinal variability of the induced cooling, in particular in the Northern Hemisphere, which may enhance the amplitude of planetary waves.Regions of oceanic warming in the sub-Arctic are a consequence of the increasing amount of sea ice in G4 and related enhanced transport of colder and saltier waters towards the subpolar regions (Tilmes et al., 2009).This favors cold sea water downwelling and thus positive anomalies of SSTs with respect to reference RCP4.5 conditions, mainly in the North Atlantic region (where the decrease of sea ice would produce less saltier waters, followed by less downwelling, leading to cooler SSTs).
Lastly, we show the anomalies of vertical and horizontal fluxes in Figs.7 and 8, respectively, for ULAQ-CCM (a) and for GEOSCCM.For ULAQ-CCM (a), a 5 % increase of the midstratospheric tropical upward fluxes is predicted in G4 with respect to the reference RCP4.5 case, with a pronounced interhemispheric asymmetry.The Southern Hemisphere increase of downward mass fluxes is much larger than in the Northern Hemisphere, both in absolute and relative units.The stratospheric mean meridional circulation is more efficiently perturbed in the Southern Hemisphere due to the more effective atmospheric stabilization in the Northern Hemisphere (see above; Fig. 6).A 5-10 % decrease of the extratropical horizontal mass fluxes is also predicted, as expected from the discussion above for Fig. 6.The isolation of the tropical pipe is increased in a dynamical regime with increased tropical upwelling and enhanced atmospheric stabilization.The importance of SST changes due to geoengineering is highlighted by the much smaller interhemispheric difference shown by GEOSCCM for the downward fluxes, as well as in ULAQ-CCM (c) (not shown), while the increase in the tropical upward fluxes in Fig. 7 is comparable to the ULAQ-CCM results.Furthermore, due to less atmospheric stabilization, GEOSCCM shows much smaller changes in extratropical horizontal fluxes (Fig. 8).This is further highlighted in Fig. S6, where the horizontal mass flux anomalies are also shown for ULAQ-CCM (c).In this figure, the difference between the two ULAQ-CCM simulations regarding the horizontal mass flux anomalies is clearly visible, with ULAQ-CCM (c) having latitudinal means 1 order of magnitude smaller compared to ULAQ-CCM (a) and much more comparable to GEOSCCM in the extratropics.
Another highlight of the different effects of transport and chemical effects on lifetimes is shown in Table 4, where atmospheric lifetime anomalies are shown for five species with stratospheric photolysis and O( 1 D) reaction, as calculated in ULAQ-CCM (b).The net lifetime changes G4-RCP4.5 result from the superposition of two effects: perturbation of species transport and sulfate-aerosol-induced changes in O 3 via NO x depletion from heterogeneous chemical reactions.The increased tropical upwelling moves these long-lived species more efficiently at higher altitudes in the midstratosphere where the photolysis sink is enhanced, thus decreasing the lifetimes.On the other hand, the chemically induced ozone increase (due to the NO x sink by sulfate aerosols) tends to increase the overhead column, with a decreased midstratospheric UV flux.As a consequence, the photolysis rates decrease, thus prolonging the lifetimes.As shown in Pitari et al. (2014), however, the net effect on ozone of the aerosolinduced NO x depletion is not constant in time due to the decreasing amount of Cl-Br species during the 21st century.

Perturbation of tropospheric chemistry
Stratosphere-troposphere exchange of geoengineering sulfate enhances the aerosol SAD in the upper troposphere, thus favoring NO x depletion through heterogeneous chemical reactions (i.e., hydrolysis of N 2 O 5 and BrONO 2 ) (Tilmes et al., 2009).Again, this implies less OH production and a longer CH 4 lifetime (mostly via NO + HO 2 → NO 2 + OH). Figure 9 compares the G4-RCP4.5 anomalies of sulfate aerosol mass and surface area density in the UTLS, as calculated in ULAQ-CCM (c) and GEOSCCM.The ULAQ-CCM model results are taken from numerical experiments (c) in Table 1 in order to make a more meaningful comparison with GEOSCCM (same injection of 5 Tg-SO 2 yr −1 ; SSTs in G4 with respect to RCP4.5).
A combination of isentropic SO 4 transport above the tropopause and tropical upwelling/extratropical descent produces aerosol accumulation in the extratropical lower stratosphere with a clear maximum of mass density in the Northern Hemisphere (>2 µg m −3 at ∼ 12-14 km altitude).Larger values in the ULAQ-CCM of both SAD and mass density in the tropical upper troposphere are due to a more efficient gravitational settling of the particles.An important difference between the two models is that ULAQ-CCM includes an aerosol microphysics code for predicting the particle size distribution, which, on the other hand, is assigned in GEOSCCM.A comparison of the simulated stratospheric distribution of the SO 4 SAD is shown in Fig. S7 in order to highlight the ability of both models to correctly simulate the tropical aerosol confinement, and a further discussion of the differences between the two models in this aspect, together with profile evaluation using SAGE II data, is presented in Pitari et al. (2014).
The two models predict an increase of SAD ranging between 2 and 10 µm 2 cm −3 in the extratropical upper troposphere, and this increase is the major diver for tropospheric NO x changes in geoengineering conditions.Enhanced heterogeneous NO x conversion to HNO 3 on the aerosol surface ends up limiting the efficiency of reac-  1. Units are ppbv.In panels (a, c), the contour line increment is 10; in panels (b, d), the contour line increment is 2. Table 4. Atmospheric lifetimes (years) calculated in the ULAQ-CCM (case b in Table 1), relative to five species with stratospheric photolysis and O( 1 D) reaction sink (i.e., N 2 O, CFC-11, H1301, CFC-12, CFC-114).The first column shows year 2000 values (as an average over the 1996-2005 decade); the second column shows a model mean from the SPARC (2013) report on lifetimes.Subsequent columns show the calculated lifetime anomalies due to sulfate geoengineering (average 2030-2069).Inside the square brackets we highlight the physical and chemical effects driving the lifetime changes: changing stratospheric transport in the fourth column and changing stratospheric O 3 in the fifth column (due to the aerosol-induced NO x loss).Results in the rightmost two columns are obtained through G4 sensitivity experiments (sn1, sn3) explained in Table 2  depletion in geoengineering conditions (Pitari et al., 2014;Xia et al., 2017).High-latitude stratospheric ozone depletion produces significant UVB increase at the surface (Tilmes et al., 2012).On the other hand, the enhanced radiation scattering in the tropical lower stratosphere overbalances the UVB increase due to tropical stratospheric ozone losses, ending up in a net decrease of tropical tropospheric UVB, which means again less OH production and longer CH 4 lifetime (regulated essentially by tropical OH). Figure 11 shows the percent anomalies of UVB as calculated in GEOSCCM and ULAQ-CCM (c) for the two components that are explicitly online in the models (O 3 and sulfate aerosols).A 1.5 to 2.0 % UVB decrease is predicted by the models equatorward of 40 • latitude in both hemispheres (−1.60 % for GEOSCCM and −1.94 % for ULAQ-CCM).The sulfate geoengineering impact on tropospheric UV penetration and heterogeneous chemistry changes has been widely discussed in Xia et al. (2017), along with their effects on surface ozone concentration.
Solar radiation reflection by geoengineering aerosols increases the planetary albedo and cools the surface, with a tropospheric water vapor decrease as a response to this cooling: less OH is produced by reaction H 2 O + O( 1 D), thus prolonging the CH 4 lifetime.The combination of this climatechemistry effect with the others discussed above (NO x , UV, strat-trop O 3 transport) produces the net OH perturbation in G4 with respect to RCP4.5 (Fig. 12a) and the resulting CH 4 change (Fig. 12b).The calculated average tropospheric anomaly of CH 4 is +190 ppbv, i.e., 10.6 % with respect to the RCP4.5 base case average mixing ratio in the years 2040-2049.The stratospheric anomalies are consistent with those discussed in Fig. 5c, obtained with the same G4 perturbation, but using the MBC approach (ULAQ-CCM a).
Any attempt to assess the long-term atmospheric response of CH 4 to OH changes needs the surface mixing ratio to be allowed to respond freely to tropospheric perturbations of its main sink process (i.e., oxidation by OH), which determines the CH 4 lifetime.The usual modeling approach of adopting an assigned time-dependent mixing ratio as a surface boundary condition (MBC) can still be used to calculate climatechemistry-induced changes in CH 4 lifetime, but this cannot provide information on the tropospheric mass changes of CH 4 induced by the OH perturbations.In addition, to ob-  1. Units are µm 2 cm −3 for the surface area and µg m −3 for the mass density.In panels (a, b), the contour line increment is 0.5 for values less than 12 and 2.0 from 14 to larger values.In panels (c, d), the contour line increment is 0.1 for values less than 2.5 and 1.0 from 3.0 to larger values.to the missing feedback of lower tropospheric CH 4 changes on HO x chemistry (Myhre et al., 2011).
The alternative approach of using a surface flux boundary condition (FBC) would, in principle, resolve these issues.Table 5 summarizes CH 4 surface emissions, sinks, global mass  (2040)(2041)(2042)(2043)(2044)(2045)(2046)(2047)(2048)(2049).UVB changes are shown for the two components that are explicitly online in the models (i.e., O 3 and aerosols) and for their net.ULAQ-CCM results are taken from numerical experiment (c) in Table 1 in order to make a more meaningful comparison with GEOSCCM, as in Fig. 9.  1. Units are 10 6 molec cm −3 for OH and ppbv for CH 4 .The contour line increment is 0.1 × 10 6 molec cm −3 for OH and 25 ppbv for CH 4 .
burden and lifetime in ULAQ-CCM (b) for the year 2000.The major atmospheric sink of CH 4 is the reaction with OH and this determines the CH 4 lifetime, except for an additional smaller contribution from soil deposition and an additional stratospheric sink due to CH 4 reactions with O( 1 D) and Cl.The calculated OH abundance is then critical in the determination of a realistic global burden and lifetime of CH 4 .Tropospheric OH concentrations have been evaluated in Pitari et al. (2016a) using climatological values from Spivakovsky et al. (2000).In the same published work, a comparison of calculated tropospheric CH 4 mixing ratios is made with observations from TES/Aura radiances.The ULAQ-CCM calculated time series of CH 4 lifetime and surface mixing ratio is presented in Fig. 13a, for both reference RCP4.5 and perturbed G4 cases, using the FBC approach (experiment b in Table 1).A simple approach was used for the time evolution of CH 4 emission fluxes: the geographical distribution was fixed at year 2000 values, but the net global value was linearly scaled to the ratio of RCP4.5 recommended surface mixing ratios in future years (dotted line in Fig. 13a) with the year 2000 recommended value (1754 ppbv).An in-depth study of future climate change effects on CH 4 natural emissions or future changes on the geographical distribution of anthropogenic emissions is beyond the purposes of the present study.The lifetime change G4-RCP4.5 shown in Fig. 13a increases up to 1.7 years in 2070 during the time period of geoengineering implementation, then slowly decreases in the so-called termination period (2070-2090) down to 1.2 years in 2090.Similarly, the surface mixing ratio change increases up to 250 ppbv in 2070 and then slowly decreases in the termination period down to 150 ppbv in 2090.These slow decreases are due to the long time needed for atmospheric CH 4 to return to baseline RCP4.5 values.In addition, sea surface temperatures need a few decades to recover to RCP4.5 values (Fig. 6a, b), thus triggering a prolonged perturbation of the stratospheric circulation.
A summary of gas-phase radiative forcing (RF) components related to the CH 4 perturbation is presented in Fig. 13b.Direct stratospheric aerosol RF obviously dominates in sulfate geoengineering (∼ −1.2 W m −2 ), as discussed in Visioni  1), with bars referring to decadal averages (gray for RCP4.5 and white for G4).Superimposed are globally averaged CH 4 surface mixing ratios (ppmv, right scale) for the corresponding RCP4.5 and G4 simulations (black solid and red curves, respectively).The dotted curve shows globally averaged CH 4 surface mixing ratios for the RCP4.5 MBC case (experiment a in Table 1), i.e., using prescribed fixed mixing ratios at the surface (Eyring et al., 2013).(b) Time series of G4-RCP4.5 radiative forcing of CH 4 (mW m −2 ).Black, purple and blue curves show the direct and indirect effects (purple and blue curves are for CO 2 and stratospheric H 2 O from CH 4 oxidation, respectively).Dashed blue curve is for stratospheric H 2 O changes resulting from G4-RCP4.5 temperature anomalies at the tropopause tropical layer (TTL).
et al. (2017), using independent estimates available in the literature.Among gas species, CH 4 produces the largest indirect RF (∼ +0.1 W m −2 ), in addition to contributions from O 3 (negative) and stratospheric H 2 O (positive), with the latter due to slight warming of the tropopause tropical layer (TTL) (see Pitari et al., 2014).Small indirect CH 4 contributions come from increasing amounts of CO 2 and H 2 O in the CH 4 oxidation chain.This chemical increase of stratospheric H 2 O, however, is normally smaller than the one driven by the geoengineering aerosol warming at the TTL cold point (as shown in Fig. 13b).Table 6 summarizes our calculations for OH-dependent species lifetimes under geoengineering conditions.The ULAQ-CCM calculated lifetimes under year 2000 conditions are fully comparable with the values in the SPARC (2013) report on lifetimes.G4-RCP4.5 anomalies averaged between 2030-2069 range between +1.33 years for CH 4 and +0.5 years for CH 3 CCl 3 .The FBC approach was used for CH 4 in order to properly evaluate its feedback on HO x chemistry.The rightmost three columns in Table 6 show the different contributions to the lifetime changes, through G4 sensitivity experiments (sn1, sn2, sn3) explained in Table 2.The major contribution to the CH 4 lifetime change (but also for HCFC-22 and CH 3 CCl 3 ) comes from the presence of aerosols with their feedback on NO x -HO x -O 3 photochemistry, as discussed before in Figs. 9, 10 and 11 (temperature and winds are kept unchanged with respect to RCP4.5 in the G4-sn1 sensitivity case, in the chemistry module and continuity equations of chemical tracers).
The effects of tropospheric cooling with decreased water vapor (due to solar radiation scattering by the stratospheric aerosols) and strengthening of the BDC with enhanced strattrop downward flux (due to heating rates by the stratospheric aerosols) tend to partially or completely cancel each other.The impact of tropospheric cooling on OH-driven lifetimes is limited by the fact that the lowered H 2 O and OH production is partially counterbalanced by a less efficient reaction of NO + O 3 → NO 2 + O 2 in a colder troposphere (see Fig. S8).This decreases NO 2 and the NO x sink to HNO 3 , which implies an OH increase, mostly in the upper troposphere.In addition, OH formation from NO + HO 2 reaction is enhanced if the NO loss on O 3 is less efficient.
The strengthening of the Brewer-Dobson circulation affects essentially the upper tropospheric amount of SO 4 , CH 4 , NO y and O 3 .This results in a negative anomaly for geoengineering SO 4 and for CH 4 (due to the enhanced lower stratospheric tropical confinement; see Figs. 8 and 5c) and a positive anomaly for NO y and O 3 (due to the enhanced strat-trop downward flux).The induced OH anomaly is negative from CH 4 (a net HO x source) and O 3 (which is an OH sink in the upper troposphere).On the other hand, it is positive from SO 4 and NO y (due to the increasing NO x amount, their negative or positive anomaly will produce).This positive NO x anomaly induced in the upper troposphere by the enhanced stratospheric circulation mostly regulates the net positive OH change in the ULAQ-CCM with decreasing lifetimes (fifth column in Table 6).

Conclusions
In the present work, we have described how an injection of 5-8 Tg of SO 2 per year would modify the large-scale transport and lifetime of CH 4 , using two climate-chemistry coupled models, ULAQ-CCM and GEOSCCM.Both models use prescribed SST coming from two atmosphere-ocean coupled models: CCSM-CAM4 for ULAQ-CCM and CESM for GEOSCCM.The model evaluation has shown that both models correctly simulate the vertical profiles for the chemical species under analysis (N 2 O as a quasi-passive tracer and CH 4 ), the mean age of air and the vertical velocity w * .Furthermore, the latitudinal heat fluxes have been compared with ERA40 reanalysis in order to evaluate the skill of the models in correctly simulating the meridional transport.
We have shown that changes in the BDC due to lower stratospheric aerosol heating reduce the amount of CH 4 in the extratropical UTLS.This is both because of the strengthening of the downward branches of the BDC which brings more stratospheric air (poorer in CH 4 ) down in the upper Table 6.Atmospheric lifetimes (years) calculated in the ULAQ-CCM (experiment b in Table 1), relative to three species that include an OH reaction sink (i.e., CH 4 , HCFC-22, CH 3 CCl 3 ).CH 4 is predicted with the FBC approach; the other two species with specified surface mixing ratios (unchanged between G4 and RCP4.5).The first column shows year 2000 values (as an average over the 1996-2005 decade); the second column shows a model mean value from the SPARC (2013) report on lifetimes.Subsequent columns show the calculated lifetime anomalies due to sulfate geoengineering (average 2030-2069).Inside the brackets we highlight the physical and chemical effects driving the lifetime changes (see text   troposphere and because of a greater isolation of the tropical pipe that reduces the amount of horizontal mixing.However, in order to properly assess the magnitude of the transport perturbation (whether it is horizontal mixing or vertical fluxes), the addition of the feedback of the ocean has proven crucial.Cooler oceans allow for a further atmospheric stabilization of the atmosphere, and the cooling of the sub-Arctic regions produces important hemispheric asymmetries that are not found in fixed SSTs simulations.This points to a important limitation of pure CCM studies, with prescribed time-dependent SSTs consistent with a given RCP scenario.The large-scale transport effects of sulfate geoengineering on trace species can only be captured on all their nonlinear aspects using coupled atmosphere-ocean global circulation model (AOGCM) simulations, which may quantify the SGinduced changes on SSTs.These can, in turn, be used as input for the aerosol-chemistry-radiation-dynamics fully interactive CCM experiments.
Furthermore, we have shown that the changes in CH 4 lifetime and concentration take place because of a reduction of atmospheric OH, mostly due to three overlapping factors: (1) reduction in tropospheric water vapor caused by the surface cooling; (2) decrease in O( 1 D) caused by a decrease in tropical tropospheric UV (because part of the incoming so-lar radiation is scattered by the stratospheric aerosols, which also deplete stratospheric ozone); (3) decrease in NO x production caused by the enhancing of heterogeneous chemistry (see visual summary in Fig. 14).Changes in stratospheric large-scale transport and strat-trop exchange may also contribute to perturb the tropospheric amount of OH, with a net effect whose sign results from simultaneous changes of CH 4 , NO y , O 3 and SO 4 .All of these effects may cause a CH 4 lifetime increase of more than 1 year in the central decades of the experiment, leading, in turn, to an increase in methane mixing ratio of over 200 ppbv.
Overall, these changes produce a positive radiative forcing of more than +0.1 W m −2 in our radiative transfer model calculations, a result that it is still 1 order of magnitude smaller than the direct negative radiative forcing of the aerosols, which has been estimated to be −1.2 ± 0.5 W m −2 for a 5 Tg SO 2 yr −1 injection, considering simulations from a vast array of models (Visioni et al., 2017).In addition, gas species concentration changes (especially ozone) would also affect air quality and surface UV concentrations, which might have implications on human health, as already noted in Xia et al. (2017) and Nowack et al. (2016).As discussed in the present study, as well as in Nowack et al. (2016), Tilmes et al. (2012) and Pitari et al. (2014), the stratospheric ozone depletion induced by geoengineering solar radiation management techniques directly impacts the tropospheric UV budget.The health impact of a surface UV enhancement (located only at mid-to-high latitudes in the case of sulfate geoengineering) may be partly counterbalanced by the decreased tropospheric OH concentration and O 3 production.
Our analysis is limited to an atmospheric perturbation produced by sulfate geoengineering on photochemistry and large-scale transport; other important changes that would happen under this hypothetic scenario are the ones in natural surface emissions of CH 4 that would occur following changes in surface temperatures.Natural emissions would be reduced under sulfate geoengineering for three main reasons: (1) a reduction in surface temperatures that would, in turn, be connected with a highly probable reduction in rainfall, compared with the predicted increase under most future warming scenarios (Trenberth, 1998;Pandey et al., 2017); this would reduce the amount of CH 4 produced by wetland areas, thus affecting the atmospheric methane concentration; (2) the increased surface deposition of sulfate under SG conditions would itself produce changes in emissions from wetlands (Gauci et al., 2008); (3) SG could help avert one of the possible risks of global warming, i.e., the emission of methane from permafrost thawing (Kohnert et al., 2017).It remains to be investigated how much these effects, together, could offset the photochemical CH 4 increase resulting from our study.
Figs. S1 and S2.The model values of mean AoA and N 2 O shown represent the climatological mean (1980-2005) in the range 10-100 hPa and 10 • S-10 • N, while observed values of mean age of air are the same as in Fig. S3a and observed values of N 2 O are the SMR/Odin climatological mean (2001-2005).The correlation for N 2 O > 150 ppbv looks compact, ad the slope of the model curves is similar to the observed curve; model values of N 2 O and mean AoA are in the same range as the observations.Figure S3e presents the evaluation of latitudinal sections of N 2 O at 50 hPa against SMR/Odin data.For tropical values, GEOSCCM and ULAQ-CCM agree very well with the observations; overall model values fall inside the 2σ interannual variability.At northern midlatitudes, ULAQ-CCM overestimates SMR; in the Southern Hemisphere, GEOSCCM values are larger than SMR and ULAQ-CCM values lower.

Figure 2 .
Figure 2. Tropical stratospheric vertical mass fluxes (20 • S-20 • N) of (a) CH 4 and (b) N 2 O for GEOSCCM (blue) and ULAQ-CCM (red) results; the vertical mass fluxes are defined as [ρw * ], where w * is the zonal mean residual vertical velocity and ρ is the zonally averaged mass concentration of CH 4 and N 2 O, respectively.A model evaluation is made with flux data obtained with w * from MERRA reanalysis and CH 4 , N 2 O mixing ratios from HALOE and SMR results (black) (kg km −2 yr −1 ).CH 4 and N 2 O fluxes are averaged over 1991-2005 and 2001-2005, respectively, to keep consistency with the adopted HALOE and SMR mixing ratio values.

Figure 3 .
Figure 3. G4-RCP4.5 anomalies of (a, b) vertical and (c, d) horizontal mass fluxes of (a, c) CH 4 and (b, d) N 2 O (years 2040-2049 time average).Vertical mass fluxes in panels (a, b) (defined as in Fig. 2) are averaged over the tropics (20 • S-20 • N) in the 5-50 hPa vertical layer, with GEOSCCM results in blue and ULAQ-CCM results in red and magenta for cases (a) and (c) as in Table 1, respectively (kg km −2 yr −1 ).Horizontal mass fluxes in panels (c, d) (defined as vρ, with v and ρ the 3-D meridional wind component and mass concentration of CH 4 and N 2 O, respectively) are averaged (in absolute values) over the extratropics (90-20 • S and 20-90 • N) in the 50-150 hPa vertical layer, with model results as in panels (a, b) (kg m −2 yr −1 ).

Figure 4 .
Figure 4. Time series of globally averaged changes of CH 4 (a) and N 2 O (b) in the 50-150 hPa vertical layer for GEOSCCM (blue) and ULAQ-CCM (red and magenta, for cases (a) and (c) as in Table1, respectively) (decadal averages).Units are ppbv.

Figure 6 .
Figure 6.(a) Zonally averaged surface temperature changes G4-RCP4.5 (K) in the ULAQ-CCM (cases a and b), using sea surface temperatures from the atmosphere-ocean coupled model CCSM-CAM4 (decadal time averages from 2020 to 2089; see legend for the different colors).(b) Time series of the globally averaged surface temperatures (K) from 2020 to 2090 (RCP4.5 in red and G4 in blue).(c) Annually averaged surface temperature anomalies G4-RCP4.5 (K) from the atmosphere-ocean coupled model CCSM-CAM4 (time average 2030-2069).Shaded areas are not statistically significant within ±1σ .

Figure 10 .
Figure 10.G4-RCP4.5 anomalies of NO + NO 2 mixing ratios in the upper troposphere and lowermost stratosphere from experiment (b) of the ULAQ-CCM (time average 2040-2049).Panels (a, b) are for absolute (ppbv) and percent NO x changes, respectively.The contour line increments are 0.025 ppbv and 5 % in panels (a, b), respectively.

Figure 13
Figure 13.(a) Time series of CH 4 global mean atmospheric lifetime (years, left scale, bars) calculated in the ULAQ-CCM FBC case (experiment b of Table1), with bars referring to decadal averages (gray for RCP4.5 and white for G4).Superimposed are globally averaged CH 4 surface mixing ratios (ppmv, right scale) for the corresponding RCP4.5 and G4 simulations (black solid and red curves, respectively).The dotted curve shows globally averaged CH 4 surface mixing ratios for the RCP4.5 MBC case (experiment a in Table1), i.e., using prescribed fixed mixing ratios at the surface(Eyring et al., 2013).(b) Time series of G4-RCP4.5 radiative forcing of CH 4 (mW m −2 ).Black, purple and blue curves show the direct and indirect effects (purple and blue curves are for CO 2 and stratospheric H 2 O from CH 4 oxidation, respectively).Dashed blue curve is for stratospheric H 2 O changes resulting from G4-RCP4.5 temperature anomalies at the tropopause tropical layer (TTL).

Figure 14 .
Figure 14.Visual representation of the photochemical and transport effects of sulfate geoengineering on CH 4 , as studied in this paper.Effects connected to perturbed CH 4 emissions due to surface cooling are not shown because they were not explicitly considered in this study.These effects are essentially a decrease in wetland areas connected to reductions in rainfall and halting of permafrost thawing.

Table 2 .
Summary of numerical experiments with ensemble size.The amount of injected SO 2 (per year) is specified between brackets in the G4 column.

Table 3 .
Summary of CH 4 and N 2 O satellite observations used in this study. .
4 lifetime.Figure10shows the ULAQ model calculated anomaly of UTLS NO x in experiment (b) of Table1, with values ranging between −0.02 and −0.2 ppbv in the upper troposphere (10 to 30 % reduction).

Table 5 .
CH 4 surface emissions, sinks, global mass burden and lifetime in the ULAQ-CCM for experiment (b) (year 2000).