Carbon Cycle Response to Stratospheric Aerosol Injection With Multiple Temperature Stabilization Targets and Strategies

We analyze the global carbon cycle response to a set of stratospheric aerosol injection (SAI) simulations performed by the CESM2(WACCM6‐MA) model. The simulations are performed under the specified SSP2‐4.5 CO2 concentration pathway. It is found that both the temperature stabilization target and the SO2 injection strategy have important effects on the global carbon sink. Relative to the SSP2‐4.5 scenario, averaged over the last 20 years of our simulations (year 2050–2069), simultaneous multi‐location SO2 injection causes an increase in cumulative land carbon uptake of 45 and 23 PgC, and an increase in cumulative ocean carbon uptake of 6 and 2 PgC for temperature stabilization targets of 0.5°C and 1.5°C respectively. For a temperature stabilization target of 1.0°C, SO2 injections increase land and ocean carbon sinks by 22–42 PgC and 4–7 PgC, respectively, depending on the strategies of SO2 injections (low latitude, mid‐to‐high latitude, and multi‐objective injection). Relative to SSP2‐4.5, by year 2069, SAI increases diagnosed cumulative CO2 emissions by 25–53 PgC (3%–6%), implying a decrease in atmospheric CO2 if SO2 injections were performed under a prescribed CO2 emission pathway. Stratospheric SO2 injections slow permafrost thaw, but do not restore permafrost to the previous extent at the same warming level for all injection strategies. An abrupt termination of SO2 injection weakens both the ocean and land carbon sink, and causes a rapid decline of permafrost extent. A gradual phaseout of SO2 injection slows sharp decline of permafrost and delays the rebound of carbon sink.


Introduction
Since the pre-industrial period, emissions of greenhouse gases from human activities have caused a rapid rise in global temperature (IPCC, 2021).Current efforts of carbon emissions reduction and carbon dioxide removal may not be sufficient to avoid severe climate impacts (Bamber et al., 2019;MacMartin et al., 2022;Sherwood et al., 2020).Moreover, emission scenarios that aim to limit future warming to the 1.5°C threshold set by the Paris Agreement typically require net negative CO 2 emissions at an ambitious scale that may not be achieved in time (Matthews & Wynes, 2022;Rogelj et al., 2018).Solar radiation modification is proposed as an additional option to reduce some effects of anthropogenic warming (NAS, 2021).One of the most discussed SRM methods is stratospheric aerosol injection (SAI), which proposes to inject scattering aerosols or their precursors into the stratosphere to reflect sunlight to space and cool the planet.
• Both temperature stabilization goals and strategy for stratospheric aerosol injection are important in affecting land and ocean CO 2 uptake • Aerosol injections reduce CO 2 concentrations, diagnosed here as a 3%-6% increase in CO₂ emissions leading to specific CO 2 concentrations • A gradual phaseout compared to an abrupt termination is shown to result in a slower rebound of the carbon sink Supporting Information: Supporting Information may be found in the online version of this article.
A number of modeling studies have investigated the climate response to SAI, such as temperature, the hydrological cycle, ocean circulation, permafrost, and climate extremes (Irvine et al., 2016;NAS, 2015NAS, , 2021)).The responses of the carbon cycle to SAI have received limited attention.The carbon cycle regulates the amount of CO 2 in the atmosphere and plays a key role in the climate system.SAI, by affecting climate variables such as temperature, precipitation, soil moisture, and ocean circulation, would perturb the terrestrial and oceanic carbon fluxes (Cao, 2018).For instance, Tjiputra et al. (2016) showed that compared to the CO 2 emission scenario without SAI, SAI would strengthen both the land and ocean carbon uptake.Plazzotta et al. (2019) found that SAI would increase cumulative CO 2 emissions by enhancing CO 2 uptake by both the land and ocean.Yang et al. (2020) found that SAI would lead to increased land carbon sink due to reduced ecosystem respiration and disturbances.H. Lee et al. (2021) assessed changes in the terrestrial carbon and found that the effect of SAI on vegetation carbon storage varies by regional biome.
The permafrost contains a large amount of carbon.With continued global warming, the thawing of permafrost would release stored carbon to the atmosphere, posing great risks to the climate system (Schuur et al., 2015).Modeling studies showed that SAI could slow the degradation of permafrost and associated carbon loss (Chen et al., 2023;Chen et al., 2020;H. Lee et al., 2019;Liu et al., 2023).The effect of SAI on slowing permafrost thaw depends on its deployment strategy.For instance, W. Lee et al. (2023) showed that Arctic-focused SAI could more effectively reduce permafrost thaw and carbon release than a SAI strategy that injects similar quantities of SO 2 at lower latitudes.Morrison et al. (2023) showed that SAI limiting warming below 1.5°C above preindustrial is unlikely to stabilize permafrost extent, and a more aggressive SAI strategy may be required to fully prevent permafrost thaw.
Recently, two sets of SAI simulations were conducted in CESM2(WACCM6-MA).In the first, SO 2 is injected into the lower stratosphere at multiple locations to stabilize global mean temperature, the interhemispheric temperature gradients, and equator-to-pole temperature gradients at three different temperature targets of 0.5°C, 1.0°C, and 1.5°C warming (MacMartin et al., 2022).The second set of simulations uses the 1.0°C temperature target but considers a range of different SO 2 injection strategies (Bednarz et al., 2023;Zhang et al., 2023).In this study, we examine the carbon cycle response to both of these sets of SAI simulations, to explore how the carbon cycle response depends both on how much cooling is being done by SAI, and on how that cooling is being done.It is expected that this study would contribute to our understanding of the Earth system response to SAI.

Climate Model
We use the Community Earth System Model version 2 with the "middle atmosphere" (MA) chemistry version of the Whole Atmosphere Community Climate Model version 6 (WACCM6) as its atmospheric component, denoted as CESM2(WACCM6-MA) (Danabasoglu et al., 2020;Davis et al., 2023;Gettelman et al., 2019).CESM2 (WACCM6-MA) is a fully coupled Earth system model with the WACCM6-MA as the atmospheric component (Gettelman et al., 2019), the Community Land Model version 5 (CLM5) as the land component (Lawrence et al., 2019), the Parallel Ocean Program version 2 (POP2) as the ocean component (Danabasoglu et al., 2012), the Los Alamos Sea Ice Model version 5 (CICE5) as the sea ice component (Hunke et al., 2015) and the Community Ice Sheet Model (CISM2) as the land ice component (Lipscomb et al., 2019).For computational efficiency, the atmospheric component used here is the simplified middle atmosphere chemistry scheme, which includes a comprehensive stratospheric chemistry but simplified tropospheric chemistry.It is similar to the earlier CESM1 (WACCM) model (Mills et al., 2017) that has been extensively used for SAI simulations.The horizontal resolution is 1.25°by longitude and 0.95°by latitude for atmosphere and land grids.The atmosphere has 70 vertical layers with a model top at roughly 140 km.The ocean component has a displaced North Pole grid with a nominal 1°horizontal resolution by both latitude and longitude.The model ocean has 60 vertical layers with thicknesses ranging from 10 m for the top 160 m ocean, monotonically increasing from 10 to 250 m for the ocean between 160 and 3,500, and 250 m for the deep ocean between 3,500 and 5,500 m.

Simulation Experiments
The background CO 2 concentration scenario used here is the Shared Socioeconomic Pathway (SSP) 2-4.5 (Meinshausen et al., 2020), which represents a moderate-mitigation scenario.On top of the SSP2-4.5 scenario, stratospheric aerosol injection simulations begin in 2035 and end in 2069.We always analyze the last 20 years of Earth's Future 10.1029/2024EF004474 ZHAO ET AL. each simulation, after the specific cooling targets have been reached (MacMartin et al., 2022) to focus on the quasi steady-state response of the carbon cycle at different levels of cooling, ignoring the initial transient periods.
The first series of SAI simulations analyzed herein are performed aiming for three temperature stabilization targets: 0.5°C, 1.0°C, and 1.5°C global mean surface warming above the pre-industrial level (described in MacMartin et al., 2022).These three levels of warming correspond to the simulated warming level without SAI experienced in three periods: Historical (mean of 1993-2012), mean of 2008-2027, and mean of 2020-2039(Visioni, Bednarz, et al., 2023).SAI is implemented throughout the period of 2035-2069 to meet global mean temperature targets of 1.5°C (for the period of 2035-2069), 1.0°and 0.5°C (for the period of 2050-2069).For each of these three targets, SO 2 is injected into the lower stratosphere at the height of 21.5 km over four locations simultaneously (180°E at 30°N, 15°N, 15°S, and 30°S).The amount of SO 2 injected is adjusted every year to maintain the global mean temperature (T0), the interhemispheric temperature gradients (T1) and equator-to-pole temperature gradients (T2) at the same level as they were under 0.5°C, 1.0°C, and 1.5°C of warming.These simulations are named as SAI 0.5 Target Multi-objective, SAI 1.0 Target Multi-objective, and SAI 1.5 Target Multi-objective.Each set of SAI and SSP2-4.5 simulations has three ensemble members and the average of the ensemble simulations are used for analysis.
To investigate the effect of different SO 2 injection strategies, four additional SAI simulations are performed for the "middle" stabilization target of 1.0°C warming (described in Zhang et al., 2023).In these simulations, SO 2 is injected at the height of 21.5 km over the equator (SAI 1.0 Target EQ), over both 15°N and 15°S (SAI 1.0 Target 15N + 15S), over both 30°N and 30°S (SAI 1.0 Target 30N + 30S), and at the height of 15 km over both 60°N and 60°S (SAI 1.0 Target 60N + 60S), respectively.Unlike SAI 1.0 Target Multi-objective, these simulations maintain only T0 (Zhang et al., 2023).For the above simulations, SO 2 is continuously injected throughout the year except for the simulation of SAI 1.0 Target 60N + 60S in which SO 2 is only injected in spring (W. Lee et al., 2021).
For SAI simulations, the injection rates are determined by a controller consisting of a feedforward component and a feedback component (Kravitz et al., 2017;MacMartin et al., 2017;Zhang et al., 2023).The feedforward component calculates the best guess of the injection required to maintain temperature objectives, and the feedback component corrects for uncertainty by a Proportional Integral controller.
To investigate the effects of SAI termination, simulations have also been performed in which SAI is terminated abruptly or gradually (MacMartin et al., 2022).In particular, branching from the SAI 1.5 Target Multi-objective, two simulations are performed: (a) SAI terminates abruptly in 2055; (b) SAI gradually phases out to zero during the 10-year period of 2055-2064.A list of simulations is provided in Table S1 in Supporting Information S1.

Results
Model-simulated changes in climate including temperature, precipitation, and ocean circulation in response to the above SAI simulations are detailed in MacMartin et al. (2022) and Zhang et al. (2023).As we expect SAI induced changes in global climate to further perturb the terrestrial and ocean carbon cycles, here we examine simulated carbon cycle response to the different SAI scenarios.

Response in Carbon Uptake
Figure 1 and Figure S1 in Supporting Information S1 present model-simulated terrestrial carbon fluxes and stocks at different levels of global mean temperature change under SSP2-4.5 and different SAI scenarios.Under SSP2-4.5, gross primary productivity (GPP) increases by 10.0% from year 2020-2039 (143.3 ± 1.5 PgC yr 1 , ±denotes one standard error) to year 2050-2069 (157.6 ± 0.6 PgC yr 1 ) mainly as a result of CO 2 -fertilization effect that refers to the general increase in plant photosynthesis with increasing atmospheric CO 2 .Globally integrated, GPP changes little under all SAI simulations scenarios compared to that of SSP2-4.5 (Figure S1a in Supporting Information S1).Compared to the corresponding baseline period with the same global mean warming without SAI (0.5°C for the mean of 1993-2012, 1.0°C for the mean of 2008-2027, and 1.5°C for the mean of 2020-2039), GPP under all SAI scenarios is substantially larger, indicating the dominant role of CO 2 -fertilization.Plants consume organic carbon through autotrophic respiration (AR).Under SSP2-4.5, AR increases by 8.4% from year 2020-2039 (84.2 ± 0.9 PgC yr 1 ) to year 2050-2069 (91.3 ± 0.3 PgC yr 1 ) mainly as a result of increased CO 2 .Compared to SSP2-4.5, global AR changes little in response to SAI (Figure S1d in Supporting Information S1).
Thus, net primary productivity (NPP), which is the difference between GPP and AR, also show little change in response to SAI (Figure 1a).SAI-induced changes affect the terrestrial carbon cycle through a variety of factors including temperature, precipitation, soil moisture, and sunlight.The spatial pattern of changes in these variables are shown in Figure S2 in Supporting Information S1 and their correlation with NPP are shown in Figure S3 in Supporting Information S1.The dominant factors controlling NPP differ for different regions.At northern high latitude regions, relative to the simulation of SSP2-4.5, SAI-induced cooling decreases NPP, as shown by the positive correlation between temperature and NPP at this region (Figure S3a in Supporting Information S1).At low-to-mid latitudes, relative to the simulation of SSP2-4.5, SAI-induced cooling increases NPP.The effects of different injection strategies on NPP can be explained by these different temperature influences.For example, an equatorial injection strategy "overcools" the tropics relative to high latitudes and thus causes more increase in NPP in the tropics, while a more polar focused strategy causes less increase in NPP.As shown in Figure S3d in Supporting Information S1, SAIinduced cooling reduces vapor pressure deficit (VPD), which negatively correlates with NPP.The negative correlation between VPD and NPP is also found in previous studies (Fu et al., 2022;Mirabel et al., 2023) which argued that reduced VPD enhances stomatal conductance and promotes plant growth and NPP.However, it is likely that not all relevant VPD-related effects are included as feedbacks in climate models (Zhong et al., 2023) and that further focus should be given to analyze the range of potential interactions (Novick et al., 2016).

Response of Carbon Storage
Under SSP2-4.5, the terrestrial ecosystem as a whole continues to take up carbon and accumulate carbon in both soil and vegetation (Figure 1g).Relative to SSP2-4.5, SAI increases terrestrial carbon storage primarily due to reduced soil respiration and enhanced accumulation of carbon in soil (Figure S1j in Supporting Information S1).
As shown in Figure 1g, compared with SSP2-4.5, a lower SAI temperature stabilization target causes more carbon to be stored in the terrestrial ecosystem.Relative to SSP2-4.5, averaged over year 2050-2069, terrestrial carbon storage increases by 17.7 ± 4.5, 15.0 ± 5.4, and 8.8 ± 5.0 PgC for the multiple target SO 2 injection strategy at temperature stabilization levels of 0.5°C, 1.0°C, and 1.5°C.The effect of different SAI temperature stabilization targets on terrestrial carbon storage differ for regions.In the low latitudes, lower temperature stabilization targets cause more carbon to be stored in vegetation (Figures S1i and S4 in Supporting Information S1), causing more soil carbon input from litterfall (Figures S5a and S5b in Supporting Information S1).Meanwhile, lower temperature suppresses the rate of soil respiration (Figure S5c in Supporting Information S1).As a result, more carbon is stored in soil for lower temperature stabilization targets (Figures S1 and S6 in Supporting Information S1).At northern mid-to-high latitude regions, on one hand, lower temperature stabilization targets cause less carbon to be stored in vegetation (Figures S1h and S5d in Supporting Information S1).On the other hand, lower stabilization temperature targets are more effective in preventing the decomposition of permafrost carbon, causing more carbon to be stored in soil (Figure S1k in Supporting Information S1).Thus, at mid-to-high latitude regions, different temperature stabilization targets have a small effect on total terrestrial ecosystem carbon storage (Figure 1i).
For the same 1°C global mean temperature stabilization target, SAI-induced increase in global terrestrial carbon storage ranges from 6.1 ± 0.5 to 18.1 ± 0.8 PgC (averaged over 2050-2069) for different SO 2 injection strategies (Figure 1g).Equatorial injection cools the tropics more, causing more carbon to be stored in vegetation and soil, while polar injection cools high-latitudes more and tropical regions less, causing less carbon to be stored.

Net Land Carbon Uptake
We utilize net ecosystem exchange of carbon (NEE) to examine the net carbon uptake fluxes in terrestrial ecosystems.Positive values denote carbon uptake in land while negative values denote carbon losses (Figure 2a).NEE excludes net carbon release from land-use change and crop harvest disturbance.
Under SSP2-4.5, NEE increases from 6.4 ± 0.2 PgC yr 1 (averaged over 2020-2039) to 8.3 ± 0.1 PgC yr 1 (averaged over 2050-2069).Increasing atmospheric CO 2 enhances the carbon uptake by plants, which is larger than the losses of carbon due to increased heterotrophic respiration (HR) (Figure 1d, Table S2 in Supporting Information S1), resulting in increased net carbon uptake by the terrestrial biosphere.Relative to SSP2-4.5, lower temperature stabilization targets for SAI leads to more positive NEE mainly as a result of decreased rate of HR (Figure 1d).For the same temperature target, SO 2 injected near the equator causes more positive NEE.During the period of 2016-2069, compared with the simulation of SSP2-4.5, for the multiple target SO 2 injection strategy at temperature stabilization levels at 0.5°C, 1.0°C, and 1.5°C, SAI-induced increase in cumulative land carbon sink is 44.9 ± 2.4, 35.5 ± 2.8, and 23.1 ± 2.0 PgC (Figure S7a in Supporting Information S1).For different SAI strategies at a temperature stabilization target of 1.0°C, SAI-induced increase in cumulative land carbon sink ranges from 21.5 to 42.4 PgC (42.4 ± 3.4 PgC for SAI 1.0 Target EQ, 37.4 ± 3.0 PgC for SAI 1.0 Target 15N + 15S, 28.4 ± 2.6 PgC for SAI 1.0 Target 30N + 30S and 21.5 ± 1.7 PgC for SAI 1.0 Target 60N + 60S, respectively).The equatorial injection causes more carbon uptake than that of mid-to-high latitude injection (Figure S7b in Supporting Information S1), indicating the important role of SAI strategies on the terrestrial carbon sink.

Response of the Ocean Carbon Cycle
Under SSP 2-4.5, the modeled ocean carbon sink continues to grow and peak around mid-century at ∼3.2 PgC yr 1 .After that, the oceanic CO 2 sink decreases mainly as result of decreased buffering capacity of the carbonate system.Relative to the SSP2-4.5 scenario, SAI increases global ocean CO 2 uptake (Figure 2b, Table S2 in Supporting Information S1).SAI-induced cooling increases CO 2 solubility in seawater, causing more CO 2 uptake by the ocean.Also, SAI-induced cooling suppresses ocean stratification and reduction in Atlantic meridional overturning circulation (AMOC) (Zhang et al., 2023), enhancing oceanic CO 2 uptake.Compared with SSP2-4.5, SAI increases CO 2 uptake over large parts of the global ocean (Figures S8 and S9 in Supporting Information S1).
In the Arctic and high-latitude Southern Ocean, SAI decreases oceanic CO 2 uptake as a result of larger sea ice area that blocks air-sea CO 2 exchange (Figure S10 in Supporting Information S1).In the equatorial Pacific Ocean and around 60°S of the Southern Ocean, SAI decreases CO 2 uptake.
Compared with the simulation of SSP2-4.5, for the multiple target SO 2 injection strategy at temperature stabilization levels at 0.5°C, 1.0°C, and 1.5°C, SAI-induced increase in cumulative ocean carbon sink is 6.2 ± 0.4, 4.9 ± 0.2, and 2.2 ± 0.2 PgC (Figure S7c in Supporting Information S1).For different SAI strategies at a temperature stabilization target of 1.0°C, SAI-induced increase in cumulative ocean carbon sink ranges from 3.8 to 6.6 PgC (3.8 ± 0.3 PgC for SAI 1.0 Target EQ, 4.4 ± 0.3 PgC for SAI 1.0 Target 15N + 15S, 6.6 ± 0.4 PgC for SAI 1.0 Target 30N + 30S and 4.4 ± 0.3 PgC for SAI 1.0 Target 60N + 60S, respectively).Different injection strategies have different effects on oceanic carbon uptake in different regions (Figure S9 in Supporting Information S1).For example, high-latitude injection cools the Southern Ocean more, hence increasing CO 2 solubility and causing more ocean CO 2 uptake in the Southern Ocean compared to low-latitude injection.However, the lowlatitude injection is more efficient in recovering AMOC compared to high-latitude injection (Zhang et al., 2023), thus causing more carbon uptake in the North Atlantic.

Response in Permafrost
Large amounts of organic carbon are stored in permafrost around the Arctic.Here, we use annual maximum active layer thickness (ALTMAX), which is the depth of soil that seasonally thaws and freezes, to define permafrost.In this study, a region where ALTMAX is less than 3 m is considered to be permafrost (W. Lee et al., 2023).This definition uses the model output ALTMAX and 3 m is chosen to indicate that only the surface soil thaws and there is soil remaining frozen underneath.Under SSP2-4.5, with continued warming, the permafrost extent continues decreasing (Figure 3 and Figure S11 in Supporting Information S1).Permafrost extent averaged over 2060-2069 is 32.9 ± 0.6% smaller than that over the historical period of 1993-2012 (12.8 million km 2 ).The losses of permafrost under SSP2-45 mainly appear at the southern edge of the permafrost zones (Figure 3b).SAI effectively slows down the degradation of permafrost.Averaged over 2050-2069, for the multiple target SAI strategy aiming at temperature stabilization levels of 0.5°C, 1.0°C, and 1.5°C, permafrost extents are 11.5 ± 1.0%, 13.9 ± 0.8%, and 18.9 ± 0.8% smaller than that of historical period (mean of 1993-2012).For different injection strategies aiming for the 1°C stabilization target, permafrost extents are 16.4 ± 1.1%, 14.4 ± 1.0%, 13.8 ± 0.7%, 13.1 ± 0.9% smaller than that of historical period for SAI 1.0 Target EQ, SAI 1.0 Target 15N + 15S, SAI 1.0 Target 30N + 30S and SAI 1.0 Target 60N + 60S, respectively.However, none of the temperature stabilization targets or strategies are able to restore permafrost extent to the previous level with the same temperature (Figure 3).For the same 1.0°C temperature stabilization target, compared with low latitude injections (SAI 1.0 Target EQ and SAI 1.0 Target 15N + 15S), SO 2 injected at higher latitudes (SAI 1.0 Target 30N + 30S and SAI 1.0 Target 60N + 60S)is more effective in preventing permafrost thaw because of larger soil cooling over permafrost regions (Figure S12 in Supporting Information S1).
The effect of SAI on permafrost differs across regions (Figure 3b and Figure S13 in Supporting Information S1).For example, compared to the period of 2008-2027 with 1°C warming, SAI 1.0 Target Multi-objective produces a deeper active layer (i.e., ALTMAX increases) over northeastern Asia, northern Europe, and northern U.S., and a shallower active layer over northwestern Canada and some parts of Eurasia.For different strategies with 1°C Earth's Future 10.1029/2024EF004474 ZHAO ET AL. temperature stabilization target, high-latitude injection is more effective in maintaining ALTMAX in the majority of permafrost regions (Figure 3b).The regional response in active layer thickness is broadly consistent with the response in surface soil temperature (Figures S12 and S13 in Supporting Information S1), with more residual warming causing deeper active layer and vice versa.

Consequences of SAI Termination
A sudden termination of SAI at year 2055 causes a rapid increase in surface temperature, which approaches the level of SSP2-4.5 within approximately 15 years (Figure S14a in Supporting Information S1).Following SAI termination, both land and ocean carbon uptake decrease and rapidly return to the level simulated in SSP2-4.5 (Figure 4).The decrease in oceanic CO 2 uptake is mainly associated with the increase in sea surface temperature, and the decrease in land CO 2 uptake is mainly attributed to the increased HR associated with increased temperature (Figure S14b in Supporting Information S1).In response to SAI termination, soil carbon storage rebounds rather slowly and would not return to the level of SSP2-4.5 during the simulated period.The permafrost extent decreases rapidly in response to SAI termination and almost returns to the level of SSP2-4.5 at the end of the simulation (Figure 4d).
For a gradual phaseout of SAI over 10 years, the rebound of terrestrial and ocean carbon uptake is delayed compared with abrupt SAI termination.Substantial interannual fluctuations are observed in both terrestrial and oceanic CO 2 uptake, and the delayed rebound of CO 2 flux in response to gradual phaseout of SAI is more clearly seen in the 20-year (2050-2069) mean CO 2 flux.On the other hand, the response of soil carbon storage is similar between the scenario of abrupt SAI termination and gradual phaseout of SAI.The gradual phaseout of SAI delays permafrost thaw as compared to sudden SAI termination (Figure 4d).
A rapid phase-in to a lower temperature target might also have implications for the global carbon cycle.For our simulations with a 1.0°C or 0.5°C target, global mean temperatures are reduced from roughly 1.5°C over a decade (Figure S14a in Supporting Information S1).The carbon cycle response in Figure 2 shows a smooth transition toward the longer-term behavior without any unexpected transient.Nonetheless, this rate of cooling (0.5°C or 1°C per decade) is rapid compared to current rates of warming and could have ecological consequences (Hueholt et al., 2024); these may not be accurately represented in this climate model.

Discussions and Conclusions
We analyze the CESM2(WACCM6-MA) model ensemble results of SAIs using a set of simulations with different temperature stabilization targets and SO 2 injection strategies on top of the SSP2-4.5 scenario.
In the SAI simulations, SO 2 is injected into the lower stratosphere during 2035-2069 to stabilize global mean warming at 0.5°C, 1.0°C, and 1.5°C, respectively.The amount of SO 2 injected increases from 0 to, approximately, up to 28 Tg yr 1 , 17-25 Tg yr 1 (the range represents different SO 2 injection strategies), and 10 Tg yr 1 between the period of 2035-2069 for the temperature stabilization targets of 0.5°C, 1.0°C, and 1.5°C, respectively (MacMartin et al., 2022;Zhang et al., 2023).Our results show that relative to SSP2-4.5 without SAI, SAI-induced cooling enhances CO 2 uptake in both the terrestrial biosphere and the ocean.Lower temperature stabilization target leads to larger terrestrial and oceanic carbon sink.
Averaged over the period of 2050-2069, relative to SSP2-4.5, for the temperature stabilization targets of 0.5°C, 1.0°C, and 1.5°C, the SAI-induced increases in cumulative terrestrial carbon sink are 44.9, 21.5-42.4(the range represents different SO 2 injection strategies for 1.0°C temperature stabilization target), and 23.1 PgC, respectively.SAI 1.0 Target EQ causes more terrestrial carbon uptake (42.4 PgC) compared to that of SAI 1.0 Target , and SAI 1.0 Target 60N + 60S causes less carbon uptake (21.5 PgC).SAI-induced increases in cumulative oceanic carbon sink are 6.2, 3.8-6.6(the range represents different SO 2 injection strategies for 1.0°C temperature stabilization target), and 2.2 PgC, respectively.SAI 1.0 Target 30N + 30S causes the most ocean carbon uptake (6.6 PgC), and SAI 1.0 Target EQ causes the least ocean carbon uptake (3.8 PgC).Our results show that the different SAI strategies aiming for the same global mean temperature target could have a comparable effect on the CO 2 sink as those aiming for different global mean temperature targets by SAI.These results demonstrate the importance of both temperature stabilization targets and SO 2 injection strategies on terrestrial and oceanic carbon sink.Compared to SSP2-4.5, by year 2069, SO 2 injections increase diagnosed cumulative CO 2 emissions by 25-53 PgC (3%-6%), depending on temperature stabilization targets and injection strategies.Following an abrupt termination of SAI, both the ocean and terrestrial carbon sink rapidly rebound to the level of SSP2-4.5.Gradual phaseout of SAI somewhat slows the rebound of the carbon sink.Our results show that SAI could prevent permafrost thaw to some extent, but none of the SAI temperature stabilization targets (0.5°C, 1.0°C, and 1.5°C) could restore the permafrost extent to the corresponding baseline level at the same temperature.Compared to other SO 2 injection strategies, high-latitude injection is more effective in preventing permafrost thaw.This is consistent with the finding of W. Lee et al. (2023) which showed that compared to low-latitude injection, SO 2 injected in the Arctic could more effectively reduce the thawing of permafrost and the associated carbon release.In response to a termination of SAI, the evolution of permafrost extent rapidly follows that under SSP2-4.5.
A number of previous studies also found that the SAI would enhance the carbon uptake by the terrestrial biosphere and ocean compared to the high CO 2 scenario without SAI.For example, Tjiputra et al. (2016) linearly increases the aerosol concentration in the stratosphere to bring global warming under RCP8.5 down to the level of RCP4.5.They find that relative to the RCP8.5 scenario without SAI, during the period of 2020-2100, SAI increases the terrestrial carbon sink by 2 PgC and oceanic carbon sink by 17 PgC.In the study of Muri et al. (2018), SO 2 is injected to reduce radiative forcing from RCP8.5 to that of RCP4.5.They find that relative to the simulation of RCP8.5, during the period of 2020-2100, SAI increases the land carbon uptake by 6 PgC and the ocean carbon uptake by 10 PgC.Plazzotta et al. (2019) analyzed the results of six Earth system models simulating a continuous SO 2 injection of 5 Tg yr 1 under the scenario of RCP4.5.They find that relative to the simulation without SAI, a 50-year SAI increases the cumulative CO 2 emissions by 40 PgC: the land carbon sink increases by 34 PgC and ocean carbon sink increases by 6 PgC.Because of different background CO 2 concentrations, different SAI intensities and strategies used, it is difficult to directly compare these results with ours.Also, different Earth system models would have different response in physical climate and the carbon cycle.A recent study (Cox et al., 2024) showed that CESM2-WACCM simulated cumulative CO 2 emissions under the SSP2-4.5 scenario is at the lower end of the simulation results from nine Earth system models, indicating that our simulated carbon sink in response to SAI might be a conservative estimate.Nevertheless, these previous studies have not considered different amounts of SAI or different strategies of SAI on the carbon cycle.Our study emphasizes the important effects of both SAI temperature stabilization targets and SO 2 injection strategies on the terrestrial and oceanic carbon uptake.
Our results are subject to limitations and uncertainties.The simulation of the terrestrial carbon cycle here does not include the effect of nitrogen limitation, which would have important effect on the terrestrial carbon cycle response to increasing atmospheric CO 2 and SAI (Duan et al., 2020;Glienke et al., 2015;Tjiputra et al., 2016;Yang et al., 2020).In terms of the oceanic carbon cycle, the responses of marine ecosystem processes and effects on the ocean CO 2 uptake are of substantial uncertainty and merit further studies (Hauck et al., 2015;Moore et al., 2013;Thomalla et al., 2023).In this study, we define permafrost solely by near-surface active layer depth, and do not account for the changes in deep permafrost, which is resistant to surface temperature changes (Morrison et al., 2023).Furthermore, the response of the slow components of the Earth climate system such as deep ocean dynamics and ice sheet would further affect the carbon cycle.Therefore, longer simulations would be useful in obtaining a more comprehensive assessment of the effects of SAI on the carbon cycle.Here, atmospheric CO 2 concentrations are prescribed.In fully coupled simulations with prescribed CO 2 emissions, SAI would change atmospheric CO 2 by perturbing the terrestrial and oceanic CO 2 uptake, which could in turn affect global climate and the effectiveness of SAI (Cao, 2018;Cao & Jiang, 2017;Keith et al., 2017;Keller et al., 2014;Matthews & Caldeira, 2007).Multi-model simulations are needed to better understand the coupled climatecarbon cycle responses to different SAI strategies (Visioni, Robock, et al., 2023).

Figure 1 .
Figure 1.Model-simulated terrestrial carbon fluxes (net primary production, NPP; heterotrophic respiration and stocks (terrestrial ecosystem carbon) at different levels of global mean warming.For simulations under SSP2-45 without stratospheric aerosol injection (SAI), what is shown are values averaged over years 1993-2012, 2008-2027, 2020-2039, and 2050-2069 when global warming reaches 0.5°C, 1.0°C, 1.5°C, and 2.3°C relative to pre-industrial level.For simulations with different SAI strategies, what is shown are values averaged over years 2050-2069 for temperature stabilization targets of 0.5°C, 1.0°C, and 1.5°C.For each variable, the left column shows global values, the middle column shows values for Northern Hemisphere mid-to-high latitude (45°N-90°N), and the right column shows values for low latitude (30°S-30°N).

Figure 2 .
Figure 2. Model-simulated time series of annual mean (a) land carbon uptake, (b) ocean carbon uptake, and (c) diagnosed cumulative CO 2 emission.The dashed lines represent the 20-year average values for the baseline period of 1993-2012, 2008-2027 and 2020-2039, corresponding to global mean warming of 0.5°C, 1.0°C, and 1.5°C.The dots on the right panels represent the 20-year average over 2050-2069 and uncertainties are estimated by one standard error.

Figure 4 .
Figure 4. Model-simulated time series of annual mean (a) land carbon uptake, (b) ocean carbon uptake, (c) total soil carbon stocks, and (d) the permafrost extent of the area north of 50°N.The green dashed lines represent the 20-year average values for the baseline period of 2020-2039, corresponding to global mean warming of 1.5°C.The black dashed lines represent the year (2055) that stratospheric aerosol injection is terminated abruptly or gradually.The dots on the right panels represent the 20year average over 2050-2069 and uncertainties are estimated by one standard error.