Title: The Climate Response to Stratospheric Aerosol Geoengineering Can Be Tailored Using Multiple Injection Locations
Abstract: Journal of Geophysical Research: AtmospheresVolume 122, Issue 23 p. 12,574-12,590 Research ArticleFree Access The Climate Response to Stratospheric Aerosol Geoengineering Can Be Tailored Using Multiple Injection Locations Douglas G. MacMartin, Corresponding Author Douglas G. MacMartin [email protected] orcid.org/0000-0003-1987-9417 Mechanical and Aerospace Engineering, Cornell University, Ithaca, NY, USA Department of Computing and Mathematical Sciences, California Institute of Technology, Pasadena, CA, USA Correspondence to: D. G. MacMartin, [email protected] for more papers by this authorBen Kravitz, Ben Kravitz orcid.org/0000-0001-6318-1150 Atmospheric Sciences and Global Change Division, Pacific Northwest National Laboratory, Richland, WA, USASearch for more papers by this authorSimone Tilmes, Simone Tilmes orcid.org/0000-0002-6557-3569 Atmospheric Chemistry, Observations, and Modeling Laboratory, National Center for Atmospheric Research, Boulder, CO, USA Climate and Global Dynamics Laboratory, National Center for Atmospheric Research, Boulder, CO, USASearch for more papers by this authorJadwiga H. Richter, Jadwiga H. Richter orcid.org/0000-0001-7048-0781 Climate and Global Dynamics Laboratory, National Center for Atmospheric Research, Boulder, CO, USASearch for more papers by this authorMichael J. Mills, Michael J. Mills orcid.org/0000-0002-8054-1346 Atmospheric Chemistry, Observations, and Modeling Laboratory, National Center for Atmospheric Research, Boulder, CO, USASearch for more papers by this authorJean-Francois Lamarque, Jean-Francois Lamarque orcid.org/0000-0002-4225-5074 Atmospheric Chemistry, Observations, and Modeling Laboratory, National Center for Atmospheric Research, Boulder, CO, USA Climate and Global Dynamics Laboratory, National Center for Atmospheric Research, Boulder, CO, USASearch for more papers by this authorJoseph J. Tribbia, Joseph J. Tribbia orcid.org/0000-0003-1639-9688 Climate and Global Dynamics Laboratory, National Center for Atmospheric Research, Boulder, CO, USASearch for more papers by this authorFrancis Vitt, Francis Vitt orcid.org/0000-0002-8684-214X Atmospheric Chemistry, Observations, and Modeling Laboratory, National Center for Atmospheric Research, Boulder, CO, USASearch for more papers by this author Douglas G. MacMartin, Corresponding Author Douglas G. MacMartin [email protected] orcid.org/0000-0003-1987-9417 Mechanical and Aerospace Engineering, Cornell University, Ithaca, NY, USA Department of Computing and Mathematical Sciences, California Institute of Technology, Pasadena, CA, USA Correspondence to: D. G. MacMartin, [email protected] for more papers by this authorBen Kravitz, Ben Kravitz orcid.org/0000-0001-6318-1150 Atmospheric Sciences and Global Change Division, Pacific Northwest National Laboratory, Richland, WA, USASearch for more papers by this authorSimone Tilmes, Simone Tilmes orcid.org/0000-0002-6557-3569 Atmospheric Chemistry, Observations, and Modeling Laboratory, National Center for Atmospheric Research, Boulder, CO, USA Climate and Global Dynamics Laboratory, National Center for Atmospheric Research, Boulder, CO, USASearch for more papers by this authorJadwiga H. Richter, Jadwiga H. Richter orcid.org/0000-0001-7048-0781 Climate and Global Dynamics Laboratory, National Center for Atmospheric Research, Boulder, CO, USASearch for more papers by this authorMichael J. Mills, Michael J. Mills orcid.org/0000-0002-8054-1346 Atmospheric Chemistry, Observations, and Modeling Laboratory, National Center for Atmospheric Research, Boulder, CO, USASearch for more papers by this authorJean-Francois Lamarque, Jean-Francois Lamarque orcid.org/0000-0002-4225-5074 Atmospheric Chemistry, Observations, and Modeling Laboratory, National Center for Atmospheric Research, Boulder, CO, USA Climate and Global Dynamics Laboratory, National Center for Atmospheric Research, Boulder, CO, USASearch for more papers by this authorJoseph J. Tribbia, Joseph J. Tribbia orcid.org/0000-0003-1639-9688 Climate and Global Dynamics Laboratory, National Center for Atmospheric Research, Boulder, CO, USASearch for more papers by this authorFrancis Vitt, Francis Vitt orcid.org/0000-0002-8684-214X Atmospheric Chemistry, Observations, and Modeling Laboratory, National Center for Atmospheric Research, Boulder, CO, USASearch for more papers by this author First published: 06 November 2017 https://doi.org/10.1002/2017JD026868Citations: 69 This article is a companion to Mills et al. (2017), https://doi.org/10.1002/2017JD027006, Richter et al., (2017), https://doi.org/10.1002/2017JD026912, Kravitz et al. (2017), https://doi.org/10.1002/2017JD026874, and Tilmes et al. (2017), https://doi.org/10.1002/2017JD026888. AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract By injecting different amounts of SO2 at multiple different latitudes, the spatial pattern of aerosol optical depth (AOD) can be partially controlled. This leads to the ability to influence the climate response to geoengineering with stratospheric aerosols, providing the potential for design. We use simulations from the fully coupled whole-atmosphere chemistry climate model CESM1(WACCM) to demonstrate that by appropriately combining injection at just four different locations, 30°S, 15°S, 15°N, and 30°N, then three spatial degrees of freedom of AOD can be achieved: an approximately spatially uniform AOD distribution, the relative difference in AOD between Northern and Southern Hemispheres, and the relative AOD in high versus low latitudes. For forcing levels that yield 1–2°C cooling, the AOD and surface temperature response are sufficiently linear in this model so that the response to different combinations of injection at different latitudes can be estimated from single-latitude injection simulations; nonlinearities associated with both aerosol growth and changes to stratospheric circulation will be increasingly important at higher forcing levels. Optimized injection at multiple locations is predicted to improve compensation of CO2-forced climate change relative to a case using only equatorial aerosol injection (which overcools the tropics relative to high latitudes). The additional degrees of freedom can be used, for example, to balance the interhemispheric temperature gradient and the equator to pole temperature gradient in addition to the global mean temperature. Further research is needed to better quantify the impacts of these strategies on changes to long-term temperature, precipitation, and other climate parameters. Key Points Injecting SO2 at multiple latitudes allows some tailoring of the resulting climate response Three independent degrees of freedom of aerosol optical depth can be achieved Multiple injection locations may improve climate change compensation relative to equatorial injection Plain Language Summary Solar geoengineering, by adding aerosols to the stratosphere that reflect some sunlight, could be used to partially offset the climate change from increased carbon dioxide (CO2). However, one of the concerns is that this does not affect the climate the same way that increased CO2 does, leading to some regions cooling more than others. Previous simulations only injected aerosols at a single latitude. We show that if you were to inject aerosols at a combination of multiple different latitudes, you could better tailor the resulting climate response, providing a way of designing solar geoengineering to better meet climate goals. One could, for example, adjust not only the injection rate to maintain the global mean temperature at some desired value but also the temperature difference between Northern and Southern Hemispheres (which influences tropical precipitation) and the temperature difference between tropics and high-latitude regions. 1 Introduction Solar geoengineering using stratospheric aerosols has been suggested as a possible approach to reduce future impacts of climate change (e.g., Crutzen, 2006; National Academy of Sciences, 2015). One of the key questions is what the climate effects would be, including concern over differential regional effects (Kravitz et al., 2014; Ricke et al., 2010). However, one can choose not only how much aerosol to add but also where (e.g., latitude and altitude), and the choice of injection location influences the resulting climate (Tilmes et al., 2017). This motivates the need to explore the design space of stratospheric-aerosol geoengineering, and more fundamentally, to explicitly treat geoengineering as a design problem, bringing an engineering perspective into research on solar geoengineering. Simulations have typically tested specific injection scenarios, with the majority considering only equatorial injection as in the Geoengineering Model Intercomparison Project (GeoMIP) scenarios G3 and G4 (Kravitz et al., 2011; Pitari et al., 2014), and others (e.g., Aquila et al., 2014; Crook et al., 2015; Kalidindi et al., 2015; Niemeier et al., 2013; Niemeier & Timmreck, 2015). These simulations may lead to regional differences in climate outcomes, but it is unclear whether such differences are an inevitable consequence of any strategy for deploying stratospheric aerosol geoengineering or only a consequence of the specific scenario that was simulated. The ability to optimize climate response has previously been explored only with idealized latitude-dependent solar reductions (Ban-Weiss & Caldeira, 2010; Kravitz et al., 2016; MacMartin et al., 2013). These studies all considered three independent spatial degrees of freedom: a uniform solar reduction, a reduction that varied linearly with the sine of latitude, and a quadratic variation. These correspond to the ability to separately influence the global mean temperature, the Northern versus Southern Hemisphere cooling (which influences the location of the Intertropical Convergence Zone (ITCZ) and hence tropical precipitation changes, e.g., Haywood et al., 2013), and the relative cooling between high and low latitudes. As an example, a robust result from simulations is that uniformly reducing insolation leads to less polar amplification than in the climate response to greenhouse gases, and hence undercools high latitudes relative to low (Govindasamy & Caldeira, 2000; Kravitz et al., 2013). Thus, the ability to increase how much sunlight is reflected at high latitudes relative to low, for example, can, in principle, improve the compensation for greenhouse gas climate changes, potentially reducing regional disparities in outcomes (Kravitz et al., 2016). The three spatial degrees of freedom considered in these earlier optimization studies were motivated by what was believed to be achievable using stratospheric aerosols, based on the primarily poleward air transport of the stratospheric Brewer-Dobson circulation. Hence, injecting sufficiently far from the equator in one hemisphere is expected to predominantly increase the aerosol optical depth (AOD) in that hemisphere, while injecting further from the equator will predominantly increase AOD at higher rather than low latitudes. Indeed, Robock et al. (2008) showed, with a simpler model than used here, that Arctic and tropical SO2 injections lead to different outcomes, with the former yielding less tropical cooling relative to Arctic cooling than the latter, although it is clear from this earlier work that energy transport within the climate system will limit what climate states can be achieved; see also MacCracken et al. (2013) and Tilmes et al. (2014). Here we consider, for the first time, whether simultaneous injection of stratospheric SO2 at multiple different latitudes can be used to obtain multiple independent degrees of freedom of the spatial pattern of AOD, and in turn if these lead to multiple independent degrees of freedom of the spatial pattern of surface air temperature. The amplitude and spatial pattern of AOD that results from injection at any given latitude depends on a complex interplay of factors (MacMartin et al., 2016; Pitari et al., 2016), and thus, it is not immediately clear what spatial patterns of AOD are actually achievable given the constraints imposed by stratospheric circulation, nor whether the (nonlinear) interactions between injections at different latitudes preclude useful design. Injecting SO2 leads to sulfate aerosols through oxidation, followed by nucleation, condensation, and coagulation; these microphysical processes affect the resulting size distribution (English et al., 2012; Heckendorn et al., 2009; Pierce et al., 2010), which affects both lifetime (through sedimentation) and AOD for a given sulfate mass. The aerosols affect ozone chemistry (Solomon et al., 2016; Tilmes et al., 2008). Aerosol absorption of radiation leads to heating, affecting both the mean circulation pattern (Pitari, 1993) and variability about the mean (Aquila et al., 2014; Richter et al., 2017); these are also influenced by radiative effects from changes in ozone concentration. Heating of the tropical tropopause layer leads to increased water vapor transport into the stratosphere (Heckendorn et al., 2009; Pitari et al., 2014) that in turn influences radiative forcing (Solomon et al., 2010). All of these factors interact, and thus, to properly assess how the AOD depends on injection latitude requires a climate model that includes all of these processes: aerosol microphysics, stratospheric ozone chemistry, stratosphere-troposphere exchange, and a sufficiently well-resolved stratosphere to reasonably reproduce both the Brewer Dobson Circulation and the dominant mode of variability, the quasi-biennial oscillation (QBO). Simulations are conducted using a fully coupled chemistry climate model that includes all of these effects; the model is described briefly in section 2 and in more detail by Mills et al. (2016, 2017). Simulations were first conducted using injection at several different latitudes and altitudes, as described by Tilmes et al. (2017), to determine how the spatial pattern of AOD and climate response depend on injection location. Results herein rely on linear superposition of the AOD and temperature response obtained from these simulations in order to predict what would occur with some linear combination of different amounts of injection at different locations. Linearity has previously been shown to be a reasonable approximation for the climate response to different spatial patterns of solar reduction (e.g., MacMartin et al., 2013). However, there are additional physical processes involved in the response to SO2 injection, which lead to increased nonlinearity. As the injection rate increases, some of the additional mass adds to the size of existing aerosols rather than forming new aerosols (English et al., 2012; Heckendorn et al., 2009; Niemeier & Timmreck, 2015; Pierce et al., 2010). As a result, the AOD per unit mass will decrease since larger particles are less efficient scatterers, and the total stratospheric sulfate mass may decrease due to increased gravitational sedimentation. Nonlinearities can also arise because the impact of aerosols on stratospheric circulation increases with concentration, making the spatial distribution of aerosols dependent on injection rate. Nonlinearity is evaluated for simulations of SO2 injection at a single latitude in Tilmes et al. (2017). Before considering what can be achieved by combining injections at different latitudes, additional simulations were conducted to evaluate the extent to which simultaneous injection at different locations yields similar results to what would be predicted from simulations of injection at each location separately; these are described in section 3 and revisited in sections 4 and 5. Section 4 demonstrates that multiple different spatial distributions of AOD are indeed achievable through choice of injection location and that these are broadly consistent with the idealized patterns of solar reduction considered in previous studies; this section also considers the temporal evolution and the seasonal dependence of the AOD. The ability to achieve multiple independent patterns of AOD enables the ability to independently influence multiple degrees of freedom of surface climate response; this is explored in section 5 to evaluate using these degrees of freedom to improve the compensation of the spatial pattern of surface air temperature change that results from increased atmospheric CO2 concentrations. 2 Simulations Simulations were conducted with the Community Earth System Model CESM1, with the Whole Atmosphere Community Climate Model (WACCM) as the atmospheric component, described by Mills et al. (2017). The model horizontal resolution is 0.95∘ in latitude by 1.25∘ in longitude, with 70 vertical layers extending up to 145 km altitude. The model includes prognostic aerosols using a trimodal treatment (MAM3; Liu et al., 2012), fully interactive chemistry in order to capture changes in ozone concentrations, and is coupled to ocean and land models. The model has been verified against observations for a present-day simulation (1975–2010) and has been shown to have a good representation of the observed AOD after explosive volcanic eruptions (see Mills et al., 2016, 2017 for details). This model yields an internally generated QBO without any need for nudging; this is important for capturing amplitude-dependent nonlinearities that can occur if QBO is itself influenced by the addition of aerosols as in Aquila et al. (2014). For this study, we inject SO2, as in a variety of previous studies (e.g., Aquila et al., 2014; English et al., 2012; Heckendorn et al., 2009; Niemeier & Timmreck, 2015; Pitari et al., 2014). The detailed processes involved in the conversion of SO2 into sulfate aerosols are represented, including oxidation to sulfuric acid, nucleation and condensation, particle growth through coagulation, and sedimentation. An important caveat is that while the model was validated by comparison with observations of volcanic aerosol properties, there may be greater aerosol coagulation and condensation with continuous SO2 injection than there is with impulsive volcanic eruptions, and hence the aerosol size distribution may not be as well captured by the same set of parameters of the modal aerosol model (see Mills et al., 2016 for radius and standard deviation of Aitken, accumulation, and coarse modes). While the size distribution in volcanic comparisons are largely consistent with the results of sectional models (as in English et al., 2013, Kokkola et al., 2009; Weisenstein et al., 2007), a sectional model would allow more freedom in the evolution of the size distribution for geoengineered aerosols. At higher injection rates, it may be more effective to inject as H2SO4 directly (Benduhn et al., 2016; Pierce et al., 2010), as new particle formation then occurs within the high concentrations of the initial plume, leading to less condensation onto and coagulation with existing aerosols, reducing the nonlinearities arising from aerosol microphysical growth. Different aerosols will also have different scattering properties and different chemical properties (Keith et al., 2016; Weisenstein et al., 2015), leading to different effects on stratospheric dynamics and ozone for a given radiative effect. However, while the details will differ, different aerosol choices are not expected to affect the broader conclusions in this paper. We characterize the climate response to increased greenhouse gas concentrations without geoengineering with a simulation, from 1975 through to 2100, using anthropogenic forcing from the historical period and Representative Concentration Pathway RCP8.5 (Meinshausen et al., 2011). SO2 injection cases are branched from this RCP8.5 simulation in 2040 and simulated for 10 years. Simulations were conducted for SO2 injection at 0° (equatorial), and at 15°, 30°, and 50° latitude in each hemisphere, at 180°E in a single grid point, and at a fixed altitude roughly 1 km and 5 km above the annual-mean tropopause height at each latitude (the actual height above the tropopause will vary seasonally). For each location, injection rates of 6, 8, and 12 Tg per year of SO2 were simulated. These single-point injection simulations are described in detail by Tilmes et al. (2017), including the stratospheric aerosol properties and aerosol spatial distribution, the impact on radiative forcing, and some estimates of the climate response. The resulting changes in stratospheric dynamics are described by Richter et al. (2017). This paper builds on these studies to assess what can be achieved by combining injection at multiple latitudes. Here we only consider the high-altitude simulations (5 km above the annual-mean tropopause), as the spatial AOD patterns are similar for high- and low-altitude injection, but with greater AOD per unit injected mass for the high-altitude cases and also greater surface air temperature response per unit AOD. Future work could more thoroughly explore the trade-off with injection altitude and also with the time of year. Because aerosols are well mixed longitudinally, we focus on zonal-mean behavior. We also focus here on annual-mean behavior, although even with a constant rate of SO2 injection, the resulting AOD is seasonally dependent (see section 4), which may be important in influencing changes in surface climate. Ten year simulations are long enough to evaluate the response of stratospheric aerosol concentrations and AOD to the injection, with one caveat noted below. The aerosol concentrations increase during the first 2 years after injection starts, but the year 3 response is statistically no different from that in year 10; see Figure 1. Interannual variability in AOD is relatively small, and the stratospheric aerosol concentrations in the absence of injection are small, leading to high signal-to-noise ratio. For evaluating AOD, aerosol effective radius, and total sulfate mass, averaging over the last 8 years of each simulation gives an accurate assessment of the model response for each injection scenario with some small uncertainty due to interannual variability. Because the surface air temperature does not adjust instantaneously to changes in AOD, and the AOD does not adjust instantaneously to changes in injection, we average over the last 7 years to evaluate the expected temperature response to each injection case. The surface temperature response continues to evolve beyond 10 years, as shown in Figure 1, but the first-decade response does give useful information about how the temperature responds to forcing; section 5 discusses this limitation in more detail. Long-term changes in the tropospheric climate response will also have some influence on stratospheric circulation and hence on the spatial distribution of AOD that will not be captured in these simulations, although Figure 1 suggests that the impact of this is likely to be small. Averaging over time, each simulation thus gives us an estimate of the response pattern corresponding to a particular choice of SO2 injection. Figure 1Open in figure viewerPowerPoint (top) Annual-mean aerosol optical depth and (bottom) surface air temperature resulting from equatorial SO2 injection starting in 2040. The AOD is in steady state by year 3, but the temperature continues to evolve throughout the simulation, with the relative error most notable over the Southern Ocean. The right-hand panels show 8 year averages (for AOD) for years 3–10, used throughout this study, and years 13–20 for comparison and 7 year averages (for temperature) for years 4–10 used herein and 14–20 for comparison. All remaining simulations were conducted for only 10 years, introducing uncertainty in the estimated temperature response to SO2 injection. In addition to these single-latitude aerosol injection cases, we also consider several multiple-latitude injection cases in order to evaluate linearity; these are described next. 3 Discussion of Linearity Assumption A first step in determining whether one can combine injections at different latitudes to achieve different outcomes is whether the response to some combination is predictable from the response to individual injection cases. If the response is linear, so that there are no significant interactions, then one can simply scale and add the responses estimated from the single-latitude simulations. This clearly will not hold exactly, but nonetheless may be a sufficiently good approximation at relevant forcing levels to project what patterns of AOD or temperature could be achieved by combining different amounts of forcing at different locations. The spatial distribution of aerosols that results from small-amplitude injection will depend primarily on the baseline stratospheric circulation. However, as the injection rate increases, the aerosol size, lifetime, and spatial distribution will shift due to nonlinearities (Tilmes et al., 2017). There are two main sources of nonlinearity: higher SO2 injection rates lead to larger aerosols and hence a reduction in AOD per unit injection, and aerosol heating and other radiative and chemical interactions will change stratospheric circulation and transport, altering the spatial distribution as the injection rate increases. Tropospheric changes due to the aerosol radiative forcing will also have some impact on stratospheric circulation. To evaluate whether it is a reasonable approximation to predict the response to combined injection from single-latitude cases, several additional 10 year simulations were conducted with simultaneous injection at multiple latitudes to evaluate the AOD and (short-term) temperature response. A comparison between the linear prediction and actual simulation is shown for several cases in Figure 2; this figure also illustrates the zonal-mean response patterns that result from some of the single-latitude injection cases. As described by Tilmes et al. (2017), the dominant mechanism of nonlinearity is due to increased aerosol size through condensation and coagulation close to the injection point in the region of new particle formation. Figure 2 (left column) corresponds to simultaneous injection of 12 TgSO2 per year each at 15°N and 15°S. When injecting only in one hemisphere, new particle formation occurs primarily in that hemisphere (Tilmes et al., 2017), and thus, there is very little interaction between the injections in each hemisphere in this case, and the total AOD is roughly the sum of the AODs from each injection separately. Figure 2 (middle column) corresponds to simultaneous injection of 12 Tg SO2 per year each at 15°N and 30°N. In this case, the region of aerosol formation for each of these overlaps. Injecting at both locations simultaneously leads to larger aerosols than from either case alone, as evident from the effective radius. This leads to both lower SO4 mass (due to increased sedimentation) and lower AOD per unit aerosol mass, as compared with what one would predict from the individual simulations. At these injection levels, the difference in AOD is at most roughly 20% at any latitude, with roughly a 10% difference in the global average. Nonlinear interactions may limit the applicability of the conclusions herein at sufficiently high injection rates (see also, e.g., Niemeier & Timmreck, 2015). The impact of nonlinearities on achievable patterns of AOD and surface air temperature response will be further discussed in sections 4 and 5. Figure 2 (right column) considers 6 Tg SO2 per year at each of the four locations; from the first case considered we can expect that the Northern and Southern Hemisphere injections do not strongly interact, and so the difference between the linear sum and the actual simulation should be primarily due to the interactions between the injections in the same hemisphere and are smaller than the corresponding interactions at the higher injection rate shown in Figure 2 (middle column). Figure 2Open in figure viewerPowerPoint Response to simultaneous aerosol injection at multiple locations, compared with the response predicted from linear superposition of individual injection scenarios. (first row) Zonal-mean AOD, (second row) the stratospheric SO4 mass burden, (third row) the maximum aerosol effective radius in the stratosphere, and (fourth row) the zonal-mean surface air temperature. Temperature responses throughout are averages over years 4–10; aerosol properties are all averages over years 3–10. Uncertainty due to natural variability is indicated with shaded bands. Cases considered are simultaneous injection at (left column) 15°S and 15°N, (middle column) 15°N and 30°N, and (right column) 30°S, 15°S, 15°N, and 30°N; the individual simulation results are shown with solid lines for Northern Hemisphere injection and dashed for Southern Hemisphere, blue for ±15° and red for ±30°. All cases are shown for a total of 24 Tg/yr injection; either 12 Tg each at two locations (left column and middle column) or 6 Tg at each of four locations (right column); this leads to roughly 2°C cooling in each case. A dominant mechanism of nonlinearity is the formation of larger aerosols, which both increases sedimentation (reducing concentrations) and results in lower AOD per unit mass. From Figure 2 we conclude that for AOD there is a reasonably linear relationship at forcing levels that result in several degrees centigrade of cooling, with the largest difference occurring at latitudes where th