Multi‐Model Simulation of Solar Geoengineering Indicates Avoidable Destabilization of the West Antarctic Ice Sheet

Heat transported in Circumpolar Deep Water is driving the break‐up of ice shelves in the Amundsen Sea sector of Antarctica, that has been simulated to be unavoidable under all plausible greenhouse gas scenarios. However, Solar geoengineering scenarios remain largely unexplored. Solar geoengineering changes global thermal radiative balance, and atmospheric and oceanic transportation pathways. We simulate stratospheric aerosol injection (SAI) designed to reduce global mean temperatures from those under the unmitigated SSP5‐8.5 scenario to those under the SSP2‐4.5 scenario with six CMIP6‐class Earth System Models. These consistently show intensified Antarctic polar vortex and sub‐polar westerlies, which mitigates changes to easterly winds along the Amundsen Sea continental shelf compared with greenhouse gas scenarios. The models show significantly cooler Amundsen Sea waters and lower heat content at 300–600 m under SAI than with either solar dimming or the SSP5‐8.5 unmitigated greenhouse gas scenarios. However, the heat content increases under all scenarios compared with present day suggesting that although vulnerable ice shelves would continue to thin, the rate would be lower for SAI even with SSP5‐8.5 specified greenhouse gases, than for the moderate (SSP2‐4.5) scenario. The simulations here use solar geoengineering designed to meet global temperature targets; interventions targeted at preserving the frozen high latitudes have also been proposed that might be expected to produce bigger local effects, but potentially deleterious impacts elsewhere. Considering the huge disruptions to society of ice sheet collapse, more research on avoiding them by intervention technology is a moral imperative.


Introduction
Ice shelves tend to breakup and retreat when they are exposed to increased basal melt rates via ingress of relative warm, dense deep water derived from Circumpolar Deep Water (CDW) to their grounding lines (Liu et al., 2015;Turner et al., 2017).These deep warm currents circulate around Antarctica offshore of the continental shelf break.The CDW flows closest to the ice sheet in the Amundsen Sea sector (Jacobs et al., 2011), where the CDW waters overflow the 500-600 m deep sedimentary shelf outer margin (Heywood et al., 2016) and can then penetrate far onto the continental shelf via deeply incised bathymetric channels and directly access sub-ice shelf cavities (Wåhlin et al., 2013(Wåhlin et al., , 2020)).In the Amundsen Sea sector large glaciers are resting on bedrock that deepens inland away from their grounding lines, and which are thought to be already vulnerable to Marine Ice Shelf Instability (Favier et al., 2014;Joughin et al., 2014).As the buttressing force stabilizing the glaciers is reduced by ice shelf basal melting, the inland ice accelerates across the grounding line potentially leading to rapid grounding line retreat.Global sea level rises of several meters are possible if Thwaites and neighboring glaciers collapse because this would likely destabilize much of West Antarctica that presently rests on beds far below sea level (DeConto & Pollard, 2016;Turner et al., 2017).
The oceans have absorbed 93% of the increased radiative heating from greenhouse gases, and hence are a huge reservoir of energy (Trenberth and Fasullo, 2016).Any changes in ocean circulation are a concern if this supply of energy is directed toward sensitive environments-such as the West Antarctic ice sheet (WAIS).Solar geoengineering is designed to act on incoming solar radiation whereas as greenhouse gases absorb long wave outgoing radiation and the net result is spatial and temporal differences in radiative forcing that will impact, for example, equatorial-polar temperature gradients, and atmosphere-ocean circulation patterns (Gertler et al., 2020;Xie et al., 2022).Despite this, simulations of impacts of solar geoengineering show global benefits in moderating temperature rise compared with the impacts expected under modest greenhouse gas forcing scenarios such as SSP245, and unmitigated scenarios such as SSP585 (Harding et al., 2020;Irvine & Keith, 2020).
There have been three previous studies of SRM impact on WAIS (Table 1).Two recent studies on the simulation of the impacts of geoengineering on the stability of the WAIS contrast with early work (McCusker et al., 2015) that concluded that SAI would promote warm water flows to the vulnerable ice shelves, destabilizing the ice sheet.Sutter et al. (2023) show that on a high-GHG pathway, solar geoengineering to designed to limit global mean temperatures rises to 1.5°C would only be effective at stabilizing WAIS if started immediately (2020), whereas on an intermediate GHG pathway, SAI from mid-century may be effective.While, Goddard et al. (2023) use a suite of stratospheric sulfate aerosol injection (SAI) simulations at different latitudes to show that collapse depends on aerosol injection strategy, with greater cooling of key waters adjacent to the vulnerable ice shelves if the southern hemisphere were preferentially targeted.Recent greenhouse gas only simulations driving a high resolution ocean model of Amundsen Sea found accelerated ice shelf melt rates under every future climate pathway (Naughten et al., 2023), implying that mitigation was no longer a viable option to prevent ice sheet collapse.In contrast with previous studies on interventions that relied on climates from single models (Table 1), in this paper we simulate SAI designed to reduce global mean temperatures from those under an unmitigated emissions scenario to those under a more moderate scenario that more closely tracks the Paris 2015 multi-country emissions pledges, with six CMIP6-class Earth System Models.

Data and Methods
In this article we consider the impacts on the Amundsen Sea with six state of the art climate models in contrast with single model studies that have been used earlier (Table 2) and we consider here both the largely theoretical concept of solar dimming geoengineering by which the solar constant is lowered, mimicking the idea of a space parasol, and the more practical SAI method, mimicking volcanism.
SSP245 is a mid-range greenhouse gas scenario where radiative forcing peaks in mid-21st century and global mean temperatures rise throughout the century before cooling in the 22nd and SSP585 is a no mitigation scenario where radiative forcing and temperatures continue to rise throughout the 21st century (O'Neill et al., 2016).The geoengineering simulations we use here are the GeoMIP6 (Geoengineering Model Intercomparison Project; Visioni et al., 2023), G6solar and G6sulfur.G6solar (Kravitz et al., 2015) is a solar dimming experiment where greenhouse gas forcing follows the SSP585 scenario, but incoming shortwave radiation is lowered from 2020 so that global mean top of atmosphere radiative forcing matches closely to the SSP245 scenario which is close to the mitigation commitments agreed towards the Paris 2015 accord (Kitous & Keramidas, 2015).G6sulfur (Kravitz et al., 2015) also reduces SSP585 radiative forcing to SSP245, but using SAI into the lower tropical stratosphere, again starting in 2020 and running until 2099.In fact, several models only have data until 2089.In our analysis we compare the available final 20 years of geoengineering (2070)(2071)(2072)(2073)(2074)(2075)(2076)(2077)(2078)(2079)(2080)(2081)(2082)(2083)(2084)(2085)(2086)(2087)(2088)(2089) in the different scenarios, with SSP585 as the reference since we are concerned how the scenarios reverse or avoid the large greenhouse gas forced changes in SSP585.All four scenarios used produce higher radiative response than at present until mid-century, and consequently show increases in surface temperature (Visioni et al., 2021) and sea level rise (Yue et al., 2023).The simulations here do not reach any steady state, and Sutter et al. (2023) show that the ice sheet response continues for many centuries, but air and continental shelf ocean temperatures respond far faster, so that useful inferences may be made with simulations just for the 21st century (e.g., Goddard et al., 2023;McCusker et al., 2015).
These simulations have been done by six CMIP6 generation earth system models (ESM; Table 2; Visioni et al., 2021).CMIP6 models significantly improve Southern Ocean simulation quality, especially in realism of wind driven transports compared with earlier generation models (Beadling et al., 2020).The CCSMv4 model used by McCusker et al. (2015) was of similar resolution to the six ESMs we use (Table 2), but it was two model generations earlier than the CESM2 we use; similarly the HadGEM2-ES used by Sutter et al. (2023) is a CMIP5 generation model of reduced complexity compared with its CMIP6 successor, UKESM-1, that we use here (Table 2).Southern Ocean simulations have large model spread, but significant changes in wind speed occur under both greenhouse gas scenarios and solar geoengineering (Gertler et al., 2020;Wei et al., 2018).Thus, the most robust estimates of stratospheric to surface winds and ocean temperatures will come from a multi-model ensemble of CMIP6 generation models, and here we use all available realizations of the G6 experiments (Table 2) with matching realizations for SSP scenarios.
The 6 ESM (Table 2) tend to the "hot" in terms of their equilibrium climate sensitivity.Hausfather et al. (2022) suggest only using models which fall in the likely range of ECS 2.5-4°C (thereby excluding all but 2 ESM here), or within the very likely range of 2.3-4.7°C,excluding 2 out of 6 models.These high ECS may be erroneous, but we do not observe especially different behavior of these "hot" models in the particular anomaly fields we examine, and so we decide to keep all ESM in the ensemble results.Some details of the aerosol injection strategy for G6sulfur are listed in Table 2, and a discussion of the various methods for simulating the SAI can be found in Visioni et al. (2021) along with the comparison of ESM with the 1991 Mt. Pinatubo eruption.The G6sulfur specifies tropical injection and all 6 ESM have higher concentrations in the tropical stratospheric pipe than at higher latitudes as observed with Pinatubo, although with wide model spread.The quantity of aerosol needed to reduce SSP585 to SSP245 radiative forcing in the last decades of simulation is 29 ± 9 Tg SO 2 /yr (Table 2), which indicates a measure of the cross-model differences in aerosol model physics.
The sub-100 km scale details of Antarctic continental shelf bathymetry are not resolved in ESMs, but what we can examine is the thermal structure of waters near the 500-600 m deep shelf break (Heywood et al., 2016;Hogan et al., 2020).The draft of Thwaites ice shelf and Pine Island Glaciers are about 300-500 m (Morlighem et al., 2020), while the depth of the thermocline is at 200-700 m (Heywood et al., 2016), hence temperatures of waters in the 300-600 m depth range are the most critical in assessing changing thermal impacts on glacier grounding lines in the Amundsen Sea sector.

Results
SAI by design involves introducing radiatively active components to the atmosphere, and sulfate has been widely simulated since it produces global cooling after large volcanic eruptions (Robock, 2000).However, the radiative effects of sulfate aerosol also produces stratospheric heating (Visioni et al., 2020), and various chemical reactions that are expected to deplete stratospheric ozone (Tilmes et al., 2022).SAI in models leads to an intensification of the spring time polar vortex (Visioni et al., 2020), especially over Antarctica (Figure 1), whereas projections of changes to the vortex under future global warming scenarios are more uncertain since ozone recovery rates will play a key role (Eyring et al., 2013).
Changes in ocean circulation are driven by surface winds, which can in turn be driven by downward propagation of stratospheric anomalies (Baldwin & Dunkerton, 2001).Intensification of the polar stratospheric vortex occurs most clearly under the G6sulfur experiments relative to G6solar (Figure 1).This is especially noticeable in the polar spring.There is clear similarity between G6solar and SSP2-4.5 in both northern and southern spring, which may be surprising considering the differences in radiative forcing between greenhouse gas and solar dimming.Part of the change in vortex with SAI is due to stratospheric ozone depletion, but it is mainly a response to the changing thermal balance in the global stratosphere (Visioni et al., 2020).The induced strengthening and expansion equatorward of the polar vortex leads to increases in tropospheric winds, in particular the surface westerly winds in the Southern Ocean (Figure 2h), although model spread means localized impacts can vary.Impacts in the Arctic are much smaller because of the geographic placement of the Arctic Ocean and surrounding land masses compared with the high-elevation Antarctic continent and surrounding Southern Ocean.
Sea surface temperatures will rise in all climate scenarios relative to the recent past (Figure 2).Westerly wind stresses also exhibit increases in the Southern Ocean relative to historical times in all scenarios, while mixed layer depths decrease in consequence (Figure 2, second and third rows).Potential temperatures at 66°S show strongest warming in the Amundsen and Bellingshausen Seas (around 100°W) and cooling in the Prydz Bay, Amery ice shelf sector (around 80°E; Figure 2, bottom row).Differences between the SSP585 and the other three scenarios in surface temperature are obvious (Figure 2).By design the G6 scenarios match the global mean temperatures in SSP245, and in general the spatial patterns are similar.There are more obvious differences in surface winds between the G6 scenarios, with G6solar similar to SSP245, and G6sulfate closer to SSP585 in the Southern Ocean.
Figure 3 compares the scenario anomalies relative to SSP585.While there are few obvious difference between the G6 anomalies in sea surface temperatures, the westerly winds under G6sulfate in the Southern Ocean are clearly different from the G6solar anomalies (Figures 3g and 3h).Potential temperatures deeper than 300 m are warmer in G6solar than SSP585 (Figure 3j), while they are cooler under G6sulfate.The potential temperatures in the Amundsen Sea sector deeper than 300 m are even cooler than under SSP245 all around Antarctica (cf Figures 2n  and 2p).
In the Southern Hemisphere, the accelerated westerlies under all scenarios, but most visibly in all ESM (Figure S2 in Supporting Information S1) under SSP585 (Figure 2e) and G6sulfur (Figure 2h), in combination with the Coriolis force increases the northward Ekman transport in the upper ocean.Changes in the Ekman transport result in geostrophic adjustments that lead to changes in regional sea level, for instance due to the poleward migration of the Antarctic convergence (Rintoul & Naveira Garabato, 2013).Hence, the G6solar and SSP245 sea level anomalies relative to SSP585 are larger (Figures 4a and 4c) than the G6sulfur anomaly (Figure 4b), which produces a fall in sea level around the Antarctic coast in 5 of 6 ESM (Figure S4 in Supporting Information S1), with the UKESM being slightly exceptional, but still with lower rise than G6solar or SSP245 relative to SSP585.Importantly, the continuity of fluid volume dictates that deep waters, such as CDW, upwell under G6solar and SSP245 (Figures 4d and 4f) to compensate for the increased northward upper ocean Ekman transport (Figures 4g and 4i).
This deep-water upwelling might be expected to increase temperatures offshore of the Antarctic continental break, but we see significant cooling in water temperatures in the upper 300 m of waters at 66°S (Figures 3j and  3l).These waters are shallower than CDW, and called Antarctic Surface Water (ASW), which under some conditions are entrained in the Antarctic Circum-Polar Current and gain access to the sub-ice shelf cavities (Jenkins et al., 2016), furthermore these waters are expected to be most sensitive to surface wind conditions.Indeed, in the Amundsen Sea there is greater than average cooling under both G6solar (Figure 3j) and G6sulfur (Figure 3k) compared with the rest of the 66°S meridian.Only in the Weddell Sea is there any degree of warming under G6solar relative to SSP585, where there are also corresponding increases in mixed layer thickness and surface temperatures (Figures 3a-3g).The relative cooling of the tropical oceans is important because conditions there are statistically better correlated with Amundsen Sea ice shelf episodes of retreat and advance than purely regional drivers (Holland et al., 2019).
Changes in ice shelf melt rate due to water import onto the continental shelf are correlated with the extra-tropical response of ENSO that brings westerly air flow over the shelf edge (Holland et al., 2019).The G6 experiments by design reduce the energy absorbed by the oceans with both G6 experiments reducing SSP585 surface temperature anomaly fields close to those under SSP245 (Figures 2 and 3).Thus, on long timescales oceans heat up at a much slower rate than under SSP585, and hence the oceans around Antarctica would never reach temperatures that they would under SSP585 as long as the interventions were to continue, or eventually atmospheric greenhouse gases had been reduced.

Discussion and Conclusion
The temperatures in the 300-600 m depth range at 66°S cool by a few tenths of degree by 2090 in all ESM under G6sulfur, and with more consistency (Figure S3 in Supporting Information S1) than either G6solar (3 from 6) or SSP245 (4 from relative to SSP585.A cooling of 0.1°C can thicken an Antarctic ice shelf by 1 m/year (Rignot & Jacobs, 2002), although a sophisticated ice/ocean model would be required to adequately quantify melt for any particular glacier.The only study to date with a relatively high resolution ice dynamics model driven by from HadGEM2 when forced by the RCP8.5 (comparable with SSP585) emissions scenario leads to eventual collapse of the vulnerable parts of the WAIS after the year 2300 (Sutter et al., 2023).Only low emissions scenarios and early SRM interventions (2020 for RCP 8.5; 2040 for RCP 4.5) stabilize the ice sheet (Sutter et al., 2023).Tipping point behavior, that is irreversible hysteresis in the response of the WAIS to temperature rises is thought to occur for global mean near surface air temperature (GMT) rises of 1.5°C above pre-industrial (Armstrong Mackay et al., 2022 and references therein).This is the closest tipping point to present day conditions, and some parts of the WAIS may have already entered the unstable ice dynamical regime known as marine ice sheet instability (MISI; Schoof, 2007) whereby loss of mechanical buttressing from ice shelf thinning leads to unstable retreat when the ice sheet grounding line rests on a retrograde sloping bed (Favier et al., 2014;Joughin et al., 2014;Rignot. et al., 2014).This critical thinning of ice shelves is exclusively driven by oceanic thermal forcing in the Amundsen Sea since temperatures are not yet warm enough for any significant surface melt.Hence the oceanic thermal forcing can be viewed as a proxy for WAIS stability if a threshold temperature can be established (Armstrong Mackay et al., 2022).Sutter et al. (2023) demonstrate in their Figure 1 that in their ice model, and driven by a single ESM, the WAIS collapse likelihood dramatically increases when ocean temperatures between 400 and 700 m depths around the Antarctic ice fronts rise by 1-1.5°C above those at the end of the 20th century.
Note that this water temperature rise is coincidentally the same value as the tipping point GMT threshold from Armstrong Mackay et al. (2022).Figure 2 shows that ocean temperatures rise under all 4 scenarios by more than this offshore of the Amundsen Sea continental shelf.In contrast Figure 5 shows much smaller temperature rises around all Antarctica at the 300-600 m ocean depths (Figure 5c), and also in the specific region relevant for ice sheet collapse in the Amundsen Sea (Figure 5d), at least as far as the simulations extend, to 2089.This threshold is likely model dependent but it is clearly different for the 6 models analyzed here than for the HadGEM2 model in Sutter et al. (2023).
The cooling in Figure 3 contrasts with previous experiments using a single ESM with much higher levels of greenhouse gas and SAI (McCusker et al., 2015) which found SAI produced a 0.5°C rise in the Amundsen Sea sector at depths of 200-500 m.McCusker et al. (2015) also found that solar dimming reduced temperatures at the same locations by 0.5°C relative to those in 2035 when their SAI began.The change in deep water temperatures we see may be driven by 21st century differences in radiative forcing directly since winds do affect the upper 1 km of the ocean (Vallis, 2017) and all 6 ESM simulate cooler deeper waters (300-600 m) under G6sulfur than the other scenarios.2023) simulations by consideration of the global mean near surface temperatures targets they aim for, and the injection strategy used (Table 1).Injections in both articles start when surface temperatures are already above 1.5°C above pre-industrial.Sutter et al. (2023) explicitly target 1.5C above per-industrial and McCusker et al. (2015) aims to stabilize at 2035 temperatures.Injection is tropical in these simulations, as it is in the G6sulfur scenario explored here.Goddard et al. (2023) show that reduced warming around the Antarctic coastal waters can be achieved with southern hemisphere preferential injection strategies.Sutter et al. (2023) show with a very long simulation (to the year 3000) that the only scenarios that may (with about 2:1 odds) prevent WAIS collapse are where SAI is started early (2040) and GMT is kept to about 1.5°C above pre-industrial.Goddard et al. (2023) show that the odds might be greatly increased with lower temperature targets, for example, below the temperatures at the start of the deployment, and with an early start to minimize temperature rises in the key CDW waters that access the ice shelf cavities fed by the Amundsen Sea.
In addition to water temperatures, CDW (or modified CDW) salinity changes impact density more than temperature in cold waters.Cooler surface temperatures also increase sea ice production and salt rejected which could trigger convection and deeper mixed layers.Moreover, any changes in CDW flux onto the continental shelf will impact ice shelf melt rates.Processes involved are far from understood and many are unresolvable with the ESM analyzed here, such as onshore transport in bottom Ekman layers and the topographically steered undercurrent Earth's Future 10.1029/2024EF004424 MOORE ET AL. (Heywood et al., 2016).However, observations and high-resolution model analyses highlight the importance of easterly winds at, and north of, the shelf break, especially in Antarctic spring (Assmann et al., 2013;Dotto et al., 2019;Holland et al., 2019;Spence et al., 2014;Thoma et al., 2008;Walker et al., 2013), when the impact of the polar vortex circulation imposed by is most intense (Figure 1).
Easterlies near the Amundsen Sea shelf break have weakened over the twentieth century (Holland et al., 2019), which has favored the driving of deep waters onto the shelf via Rossby wave teleconnections to the extra tropical Pacific on decadal scales (Holland et al., 2019).All models here, except CNRM-ESM2.0,simulate under all four scenarios strengthened westerlies, associated with the weakened easterlies.Specifically, the weakened easterlies reduce the speed of the westward geostrophic coastal flow at the surface, decrease the Ekman transport of cold Antarctic Surface Water (ASW) and its downwelling against the coast, weaken the Antarctic Slope Front, and as a result the warm CDW advances further toward the ice shelf (Jenkins et al., 2016; Figure 6).Moreover, Azaneu et al. (2023) suggest that on-shore heat flux drives melt whereby an eastward wind anomaly drives a barotropic acceleration of the eastward geostrophic undercurrent by changing the meridional sea surface slope between the shelf and offshore.In support to these arguments, westerly wind anomalies in the region are well correlated with surface temperatures in the tropical and extra-tropical Pacific (Holland et al., 2019), and historically these warm incursions have led to increases in ice shelf melt rates even without an increase in CDW temperatures (Holland et al., 2019).
Ensemble annual mean monthly wind stresses under the scenarios in the region adjacent to the shelf break are statistically different for each scenario except between SSP245 and G6solar (Figure 5), with largest westerlies under SSP585, while the other three scenarios show much smaller trends toward increased westerlies.The time series of 300-600 m ocean temperatures in the region just off the continental shelf break (Figure 5d) shows clear separation between scenarios, especially after 2060, despite CESM2(WACCM) displaying only small heat content changes, all models indicate significant differences between G6sulfur and SSP585.G6sulfur has the lowest simulated ocean temperatures, while SSP585 has the largest.All scenarios show higher sea surface and 300-600 m ocean temperatures by 2050 than at present, but G6sulfur is the only scenario where the 300-600 m temperatures slow their rate of increase, potentially plateauing at about 0.3°C warmer than 2020 levels, with both SSP245 and G6solar continually rising until end of century.So, the combination of reduced temperatures in the waters offshore of the continental shelf together with lower ocean temperatures at the shelf break would likely lead to reduced melting under the G6sulfur scenario compared with SSP585, and probably SSP245.
The driver for the changing surface winds is the globally applied SAI which overcools the tropical regions relative to the poles and strengthens polar vortex (Figure 6).The resulting reduction in the pole-equator temperature gradient should preferentially reduce the wind variability near the continental shelf break in the Amundsen Sea, but may also affect for example, teleconnection patterns (Rezaei et al., 2023).Low-latitude injections of sulfate in the CESM model (Visioni et al., 2020) change the thermal wind balance due to localized heating, increasing the equator-to-pole horizontal temperature gradient in the stratosphere and strengthening the polar vortex in both hemispheres.Such strategies (like G6sulfur) would therefore make it difficult to specifically cool the polar regions by SAI.However, injections at 60°N and S may cool the high latitudes and avoid heating the tropical stratosphere to the same extent (Lee et al., 2021).Hence, surface wind impacts will be SAI strategy-dependent and CDW flux depends on local conditions and processes acting on various time scales; topography, winds, oceanic fronts and water masses.
This study is but a small step to researching alternative ideas that might contribute to the accepted paradigm of greenhouse gas mitigation.Indeed, if MISI is already underway then greenhouse gas concentrations are not relevant at any concentration because the ice sheet is then just undergoing a geometrically driven collapse that can only be arrested by additional buttressing (Schoof, 2007).Our study has several weaknesses: (a) simulations are only 70 years long; (b) with relatively low resolution; (c) are not coupled to a dedicated ice sheet model.Coupled ice-ocean simulations have recently been designed (e.g., Gladstone et al., 2021) that can capture more of the physical processes than the relatively simplified model employed by Sutter et al. (2023) which explored a wide space of ice sheet response on long timescales.Fully coupled ice-ocean models that include the boundary layer physics needed at the grounding line where the ice sheet goes afloat may need vast computing resources given the ice sheet requires resolutions of order 1 km and the ocean must be simulated (at least theoretically) on time steps of only several minutes and horizontal resolutions of tens of meters, across regions as large as many countries (Durand et al., 2011;Gladstone et al., 2021).Additionally, only very sparse data exists on ocean currents and thermal forcing in the Amundsen Sea that drives ice shelf thinning (Wåhlin et al., 2020), and collecting such data is both expensive and hampered by very bad weather even for Antarctica (Turner et al., 2017).The calving process is critical for understanding rates of ice shelf collapse but remains one the huge challenges to ice dynamics models since cracking happens at the speed of sound and needs to be incorporated in models along with viscous deformation and sliding that is about 7 orders of magnitude slower (Benn et al., 2022).
We are not advocating for any interventions as simulated to be started; immediate and massive greenhouse gas emissions cuts must be done to preserve vital elements of the global system (Wieners et al., 2023).Instead, we want to draw attention to the possibility that there may be other options for prevention of ice sheet collapse in addition to greenhouse gas scenarios.Those alone are no longer sufficient to maintain the stability of the vulnerable ice shelves in the Amundsen Sea (Armstrong McKay et al., 2022;Naughten et al., 2023).The economic costs of sea level rise are estimated at $40 billion per meter per year by 2100 (Brown et al., 2021).Furthermore, the social cost of mass migration away from the inundated coasts-around 200 million people per meter of sea level rise, and the ecological costs of destroyed wetlands, demands that research on all options be considered.These socio-economic costs are far larger than the costs associated with exploring many of the ideas (van Wijngaarden et al., 2024) proposed for stabilizing ice sheets.Ultimately, after urgent multi-disciplinary research ranging from ethics to engineering, one or more of these interventions may offer a better option than accepting the massive damages and consequences, especially in the Global South which will bear the brunt of sea level rise from ice sheet collapse (Brown et al., 2021).

Figure 1 .
Figure 1.6-ESM ensemble mean spring time (Sep-Oct and Feb-Mar) anomalies in Ertel's potential vorticity gradient in the atmosphere across southern (a-c) and northern (d-f) mid to high latitudes under SSP585, G6solar and G6 sulfur experiments relative to SSP585.Contour lines indicate the background potential vorticity gradient in SSP585.Stippling indicates places where anomalies are insignificant at the 95% level.Individual ESM are in Figure S1 in Supporting Information S1.

Figure 2 .
Figure 2. 6-ESM ensemble mean southern hemisphere climate anomalies relative to 1995-2014 for the four scenarios: SSP585, SSP245, G6solar and G6sulfur.Top row (a-d) shows sea surface temperatures, second row (e-h): westerly wind stress; third row (i-l): mixed layer depth and fourth (m-p) row: potential temperatures in the top 1500 m of the Southern Ocean as a function of longitude at 66°S.The Amundsen Sea sector is marked with black lines on the bottom row (102-115°W) Stippling indicates places where anomalies are insignificant at the 95% level.Individual ESM results are in Figure S2 in Supporting Information S1.

Figure 3 .
Figure 3.The six model ensemble southern hemisphere mean anomalies relative to SSP585 during the years 2070-2089, of: (a-c) Surface temperature; (d-f) Westerly surface wind stress; (g-i) Mixed layer depth (m) changes; (j-l) Potential temperatures in the top 1,500 m of the Southern Ocean as a function of longitude at 66°S.The Amundsen Sea sector is marked on the bottom row (102-115°W) Stippling indicates places where anomalies are insignificant at the 95% level.Individual ESM are in Figure S3 in Supporting Information S1.

Figure 4 .
Figure 4.The six model ensemble zonal (30-75°S) mean anomalies relative to SSP5.85 in the Southern Ocean, of: Panels (a)-(c) Regional sea level due to thermosteric and dynamic effects (no data for CESM2-WACCM); (d-f) vertical velocity in the upper 0-1,500 m; (g-i) Meridional velocity in the upper 150 m.Currents almost everywhere in all scenarios in the same direction and positive (red) anomalies indicate more upwelling or northerly flow.Region used to calculate Figure 5 relations is marked by the red box in the Amundsen Sea in panels (a)-(c).Stippling indicates places where anomalies are insignificant at the 95% level.Individual ESM are in Figure S4 in Supporting Information S1.
Both the McCusker et al. (2015) and the Sutter et al. (2023) results were reconciled with the Goddard et al. (

Figure 5 .
Figure 5.The six model ensemble 10-year smoothed mean time series of ocean properties anomalies relative to 1995-2014.(a) westerly wind stress in the Amundsen Sea (defined as 70-72°S by 102-115°W); (b) Amundsen Sea surface temperature; (c) temperatures at 300-600 m depth in the ocean off the circum-Antarctic continental shelf; (d) Amundsen Sea 300-600 m depth temperatures, for each of the four scenarios for all months.The inset shows the Wilcoxon paired rank differences significant at 95% as solid arrows, with dotted arrows denoting insignificantly different pairings.Individual ESMs are in Figure S5 in Supporting Information S1.

Figure 6 .
Figure6.Schematic of processes under SRM that lead to cooler conditions (a) and under greenhouse gas forcing alone that lead to warmer conditions (b), modified afterJenkins et al. (2016).The key process allowing warm deep waters in the CDW into the ice shelf cavity is the easterlies that are promoted by the strengthened polar vortex under SAI which enhances downwelling, lowers the thermocline (the dashed line) and suppresses the Antarctic Slope Front undercurrent bringing CDW through Pine Island Trough.These processes are presently controlled from the tropical Pacific Ocean.SRM suppresses the GHG induced increases in thermal ocean forcing.All scenarios predict increases in westerlies over the Amundsen Sea, that further promote warming conditions in the sub-shelf cavities, but G6sulfur has smallest increases in westerlies due to its enhanced polar vortex from stratospheric heating.

Table 1
Previous Studies of Antarctic Response to Solar Radiation Management (SRM) Note.The climate models describe the ESM model analyzed, the SAI strategy used and the targeted global mean temperature relative to pre-industrial levels (GMT-PI), or SSP scenario, anomalies are the experiment period and control used in the analyses, and the SRM window is the duration of the SRM forcing.MOORE ET AL.

Table 2
The Earth System Models Used Here Sellar et al. (2019) Note.Time spans in G6solar and G6sulfur for MPI-ESM1.2-LRand MPI-ESM1.2-LRare available in 2020-2089.We bi-linearly interpolated each ESM to a common grid of 0.5°× 0.5°resolution to facilitate multi-model ensemble mean calculation.a We use all the ensemble members available in each ESM for both historical and future scenarios and weight all ESM equally in the ensemble mean.b Equilibrium climate sensitivity from Meehl et al. (2020), with mean of 4.2 ± 1°C.c Data from Visioni et al. (2021), na means not available.d Only 1 realization was available for G6solar.MOORE ET AL.