Articleshttps://doi.org/10.1038/s41561-018-0249-7

1National Center for Atmospheric Research, Boulder, CO, USA. 2Pacific Northwest National Laboratory, Richland, WA, USA. 3Cornell University, Ithaca, NY, USA. *e-mail: fasullo@ucar.edu

Enhancement of the Earth’s albedo for the purpose of offsetting warming arising from anthropogenic greenhouse gas (GHG) emissions, or so-called solar radiation management (SRM), was formally proposed as a means to combat climate change over a decade ago1. Since that time, the consequences of various SRM approaches have been studied in some depth. In addition to the direct adverse effects of rising carbon dioxide levels under SRM, which include increases in ocean acidity2 and perturbed biogeochemical systems3, adverse climate impacts associated with stratospheric sulfur injec-tion arise from the shifting patterns and changing intensities of rainfall and persisting regional and seasonal shifts in temperature4–6. Recently, a strategic SRM approach to achieve constant global mean surface temperature with minimal changes to large-scale tempera-ture gradients (between hemispheres and between the tropics and polar regions) was developed7. This strategy was implemented in projections of the twenty-first century under an otherwise business-as-usual (Representation Concentration Pathway 8.5 (RCP8.5)) climate scenario8 in the Community Earth System Model (CESM) incorporating the Whole Atmosphere Community Climate Model (WACCM) as its atmospheric component (CESM1(WACCM))9 with the goal to achieve a stable, strategically geoengineered climate. The model is well-suited for this analysis as it includes a well-resolved stratosphere and realistic representation of stratospheric dynamics and chemistry, as well as a prognostic treatment of stratospheric sul-fate aerosols9, the adequate representation of which is essential for a realistic estimation of the climate response to stratospheric sulfate geoengineering10. In this experiment, an ensemble of 20 simula-tions of the geoengineered future climate that extend from 2020 to 2099 is produced, referred to hereafter as the Geoengineering Large Ensemble (GLENS), with the ensemble members differing only in exceedingly small (of the order 10−14 K) perturbations to their initial atmospheric states8. GLENS employs a strategy to inject aerosols at various latitudes to minimize changes in the temperature gradients

between hemispheres and between the equator and poles while also stabilizing the global mean surface temperature. To estimate the climate state in the absence of geoengineering, 20 simulations with RCP8.5 forcing were also carried out with this model from 2010 to 2030, with three of these simulations extended through the twenty-first century. Only continuously increasing stratospheric sulfur injections that start in 2020 differentiate the GLENS and RCP8.5 ensemble members. Reduced sulfur emissions are required to stabilize surface temperatures for a curtailed emissions pathway. GLENS is unique in that it allows for the separation of the forced cli-mate response (which arises from both stratospheric injections and elevated GHG concentrations) from the internal climate variability and is particularly valuable as it provides insight into the geoen-gineered response of low-frequency coupled modes of the internal climate variability.

Global mean climate responses in GLENSThe evolving large-scale climate state in the GLENS simulations demonstrates their success in achieving stability in planetary-scale temperature patterns8 (Fig. 1). The planetary radiative energy imbalance (RT) (Fig. 1a) increases steadily under an RCP8.5 cli-mate scenario in which atmospheric GHG concentrations rise throughout the twenty-first century due to the reduced long-wave radiation emitted to space and associated feedbacks and responses. Under GLENS, increasing amounts of stratospheric sulfate aerosol are used to reflect additional shortwave radiation to space to counter this imbalance. The net effect of these injec-tions is to reduce, but not eliminate, RT, which persists at above 0.3 W m−2 through 2100—a persistence that is explored further below. By design, global mean surface temperature (TG) (Fig. 1b) stabilizes in the GLENS simulations and global rainfall (PG) drops steadily in response to reductions in sunlight that reaches the sur-face and decreases the energy available to support evaporation

Persistent polar ocean warming in a strategically geoengineered climateJohn T. Fasullo   1*, Simone Tilmes1, Jadwiga H. Richter1, Ben Kravitz   2, Douglas G. MacMartin3, Michael J. Mills   1 and Isla R. Simpson1

Enhancement of the Earth’s albedo through the injection of sulfate aerosols into the stratosphere has been proposed as an approach to offset some of the adverse effects of climate change. Here we analyse an ensemble of simulations of the twenty-first century climate designed to explore a strategic geoengineering approach. Specifically, stratospheric sulfur injections are imposed at 15° and 30° in both hemispheres with the aim to minimize the changes in surface temperature, both in the global mean and in its gradients between hemispheres and from equator to pole. The approach accomplishes these goals and reduces previously noted adverse impacts of solar radiation management, such as excessive cooling in the tropics and weakening rain-fall over land. Nonetheless, hydrological responses over the North Atlantic Ocean lead to an acceleration of the Atlantic meridi-onal overturning circulation and to continued warming of the deep and polar oceans, particularly in the vicinity of southern Greenland. These changes could cause continued, albeit slower, cryospheric melt and global sea level rise. Our simulations demonstrate the complexity of the coupled climate response to geoengineering and highlight the need for significant advances in our ability to simulate the coupled climate system and the continued refinement of geoengineering strategies as a prerequi-site to their successful implementation.

NaTuRE GEoScIENcE | www.nature.com/naturegeoscience

mailto:fasullo@ucar.eduhttp://orcid.org/0000-0003-1216-892Xhttp://orcid.org/0000-0001-6318-1150http://orcid.org/0000-0002-8054-1346http://www.nature.com/naturegeoscience

Articles NATurE GEOSciENcE

and thereby decreases precipitation. Changes in both TG and PG contrast with the increases that occur under an RCP8.5 climate, in which enhancements in the surface downwelling net radiation occur, which warms the surface and enhances evaporation. Over land, the mean temperature (TL) is relatively constant from 2020 through 2100 in the GLENS projections (Fig. 1c), a direct con-sequence of TG stabilization, and rainfall (PL) experiences only a slight reduction and decreases less than under more crude SRM approaches, in which PL reductions are otherwise comparable in magnitude to those in PG (refs 4–6). The relatively small reduc-tion in PL implies that larger decreases are displaced to ocean regions. Ocean heat content (OHC) integrated across various depths increases considerably under RCP8.5 conditions as a con-sequence of the positive net surface radiative heating associated with increases in RT, whereas under GLENS, the increases are gen-erally small. A notable exception is below 2,000 m, where global OHC increases by 2100 are larger in GLENS (24 × 1022 J) than in RCP8.5 (20 × 1022 J) (Fig. 1d, the differences between the solid and dashed lines).

The regional coupled climate response in GLENSUnderstanding the perturbations in ocean–atmosphere exchanges is key to understanding the coupled climate response in GLENS, par-ticularly as it relates to the persistence in RT and deep-ocean warming identified above. The pattern of the forced response of surface fluxes to geoengineering in the twenty-first century can be estimated from the change in the GLENS ensemble mean fluxes of heat (FS) (Fig. 2a) and moisture (precipitation minus evaporation (PE)) (Fig. 2b). The ongoing anomalous flux of heat into the ocean, suggested by RT (Fig. 1a), occurs with a strong regional structure in which areas of net uptake span much of the subtropical oceans, driven largely by a reduced evaporation, and the northern fringes of the oceans’ equa-torward-displaced western boundary currents (along the northeast coasts of North America and Eurasia). Various regions of anoma-lous heat flux out of the ocean (FS > 0) also exist, associated with the boundary currents’ equatorward displacement and in various

2020 2040 2060 2080 2100

1

2

4

a

RT (

W m

–2)

RCP8.5

GLENS

2020 2040 2060 2080 2100

288

289

290

291

292

293

294

Tem

pera

ture

(K

)

b

RCP8.5

GLENS

230

235

240

AS

R (W

m–2)

245

2.3

2.5

2.6

2.7

2.8

Pre

cipi

tatio

n (m

m d

–1)

Pre

cipi

tatio

n (m

m d

–1)

c

RCP8.5

GLENS

2020 2040 2060 2080 2100

0

1

2

3

Glo

bal O

HC

(10

24J)

dFull0–2,000 m0–700 m0–300 m

RCP8.5

GLENS

2020 2040 2060 2080 2100

284

286

288

290

Tem

pera

ture

(K

)

3.0

3.1

3.2

3.3

3.4

2.9

2.8

2.4

Fig. 1 | Global mean climate responses. a–d, Twenty-first century evolution of the ensemble annual mean global net top-of-atmosphere RT (full lines) and net incoming shortwave flux (dashed lines) (a), TG (full lines) and PG (dashed lines) (b), TL (full lines) and PL (dashed lines) (c) and global OHC integrated to varying depths (d). The shading corresponds to twice the ensemble standard error. All the available ensemble members are included in the RCP8.5 lines, which include 20 members from 2010 through to 2030 and 3 through to 2100.

.1

.5

–20

–16

–12 –8 –4 –3 –2 –1 0 1 2 3 4 8 12 16 20

–1–0.8

–0.6

–0.4

–0.2

–0.15–0.1

–0.05 0

0.05 0.

10.15 0.

20.4

0.6

0.8 1

a

b

Fig. 2 | Responses of air–sea fluxes of heat, freshwater and momentum. a,b, GLENS annual mean changes from 2010–2030 to 2075–2095 in the net upward surface heat flux FS (W m−2) and near-surface winds (vectors (m s–1)) (a) and freshwater PE flux (mm day–1) and surface stress (vectors (N m−2)) (b). All the vectors and coloured regions are statistically significant (> 2 standard error).

NaTuRE GEoScIENcE | www.nature.com/naturegeoscience

http://www.nature.com/naturegeoscience

ArticlesNATurE GEOSciENcE

upwelling zones, such as the eastern Pacific Ocean, Arabian Sea and Southern Oceans. Although these positive regional anomalies can be considerable, the north Atlantic Ocean stands out as unique, with anomalous heat fluxes out of the ocean that exceed 20 W m−2 and extend from the Labrador Sea into the Denmark Strait, a region that is noteworthy for its role in driving the global ocean’s thermohaline cir-culation11–13. Positive heat flux anomalies also span much of the Arctic Ocean, which, together with anomalous fluxes in the North Atlantic, imply a substantial increase in ocean heat transport into the region. As suggested by the evolution of global means of other fields (Fig. 1), these changes occur progressively through the course of the twenty-first century, with the mid-century mean anomalies in GLENS being about half as large (not shown) as those at the end of the century.

Anomalies in the surface water balance PE (Fig. 2b) are crucial to understanding the evolution of both FS in GLENS and its broader climate response. Much of the subtropics experience an increased net water flux (PE > 0), largely due to the aforementioned reduc-tions in evaporation, whereas higher latitudes (45–60° in both hemispheres) exhibit negative anomalies in PE as a consequence of both increases in evaporation and reductions in rainfall. In the North Atlantic, where rainfall increases are modest, increases in evaporation are large and drive pronounced PE deficits. On the southern fringe of these negative PE anomalies, easterly wind and surface stress anomalies arise from the forced climate response and imply a northward ocean mass Ekman transport, enhanced north-ward currents and associated advection of warm and saline water.

Inferences based on surface fluxes are borne out in the ocean’s forced response (Fig. 3), which is characterized by an increase in the upper layer salinity across much of the Northern Hemisphere (Fig. 3a), strongly regional surface temperature anomalies at mid and high

latitudes (Fig. 3b), and increases in OHC, both in the upper 300 m at high latitudes (Fig. 3c) and integrated to depth across most of the globe (Fig. 3d). Although the relationship between PE and salinity is complicated by the influence of ocean advection and mixing, surface salinity in the Southern Hemisphere decreases across much of the subtropics and varies in a manner consistent with the PE anomalies generally. In contrast, salinity increases in the Northern Hemisphere are pervasive, even in regions where the PE changes are positive. Salinity increases in the North Atlantic’s deep-water formation region are particularly pronounced and coincide with the area of strong sur-face warming, currents and freshwater flux deficit already discussed (Fig. 2). Anomalous northward surface ocean currents that flow into the North Atlantic Ocean are evident, consistent with the implica-tions of surface winds and related Ekman drift, and a narrow return flow along the North American coast is also evident. Positive OHC anomalies in the upper 300 m span much of the Northern Hemisphere and are strongest in regions of strong surface warming. Integrated to depth, OHC anomalies are positive in almost all regions of the globe, with a particularly pronounced warming that occurs at high latitudes.

The evolution of the Atlantic meridional overturning circulation (AMOC), the key downwelling branch of the thermohaline circu-lation and the vertical structure of the ocean forced response pro-vide a physical context for the patterns of OHC change and their connection to surface forcing (Fig. 4). The standardized AMOC index14 strengthens considerably through the course of the twenty-first century in GLENS simulations, in marked contrast to an initial strengthening, perhaps influenced by ocean initialization8, and then persistent weakening under RCP8.5 forcing (Fig. 4a), changes that are both well outside the range of present-day variability (shading) and correspond to an approximately 20% increase and 30% decrease

–1 –2

–100 –8

0–60

–40

–20

–15

–10 –5 0 10 15 20 40 60 80 10

05

–1.6

–1.2

–0.8

–0.4

–0.3

–0.2

–0.1 0 0.

10.2

0.3

0.4

0.8

1.2

1.6 2

–0.8

–0.6

–0.4

–0.2–0.15

–0.1 0 0.

1–0.05

0.05

0.15 0.

20.4

0.6

0.8 1

2a b

c d2 2

2

Fig. 3 | Responses of the upper ocean and oHc. a–d, GLENS ensemble annual mean changes from 2010–2030 to 2075–2095 in the surface ocean currents (vectors (cm s–1)) and ocean surface salinity (g kg–1) (a), surface temperature (K) (b), upper ocean (0–300 m) heat content (108 J m−2) (c) and full-depth OHC (d). All the vectors and coloured regions are statistically significant (> 2 standard error).

NaTuRE GEoScIENcE | www.nature.com/naturegeoscience

http://www.nature.com/naturegeoscience

Articles NATurE GEOSciENcE

in the mass overturning circulation during the twenty-first century. Meridional currents exhibit northward anomalies in the ocean’s upper 1,000 m from 30° N to 60° N, with compensating south-ward anomalies at depth, and an opposing overturning in the high Arctic, which suggests an anomalous surface convergence from 60° N to 80° N in the ocean’s upper 1,000 m (Fig. 4b). Co-located with the convergence are positive anomalies in potential density (contour lines), which suggests a direct role for salinity in driving the enhanced circulation. Increases in full-depth OHC at high lati-tudes are consistent with deep-ocean warming poleward of 40° in both hemispheres, with a particularly pronounced warming in the Arctic (Fig. 4c). As seen at the surface (Fig. 3a), salinity changes exhibit a strong hemispheric dependence with considerable fresh-ening of the Southern Hemisphere’s upper 1,000 m and enhanced salinity in much of the Northern Hemisphere. Salinity increases are intense and pervasive in the ocean’s upper 2,000 m north of 40° N. Together these fields provide a physically consistent explanation of the ocean’s forced response in the GLENS simulations. In the North Atlantic Ocean, northward salinity advection anomalies and PE deficits drive an increased density and deep-water formation, which induces a global ocean circulation response that sequesters heat in the high latitude and deep oceans, but appears to be otherwise innocuous in TG and TL (Fig. 1).

Broader consequencesThe acceleration of the AMOC in the GLENS simulations has poten-tially far-reaching implications. First, it demonstrates the potentially

distinct evolution of TG, OHC and global mean sea level (GMSL) that can exist in some climate-forcing conditions. These fields are often assumed to be tightly coupled, such as in recent efforts to quantify sea level responses to geoengineering15, and their divergence further motivates the use of comprehensive physically coupled models that incorporate a variable volume ocean and a fully resolved cryosphere in estimating the geoengineered climate response16,17. Increases in OHC at high latitudes driven by an increased ocean heat transport are particularly relevant to GMSL given the role that subsurface ocean temperatures play in driving ice sheet and ice cliff instabilities in the Arctic and Antarctic18–21. In GLENS, warming of the Southern Ocean persists after 2020 and full-depth OHC near Greenland is comparable to that under RCP8.5 (that is, the 2𝛔 ranges overlap) until about 2050. This persistent rise in high latitude OHC suggests a potential limita-tion in its ability to stem the GMSL rise, one of the main drivers of impacts in a changing climate. Revisions to the implementation are therefore likely to be desirable to reduce such changes. In addition, the surface expression of the AMOC is itself associated with Atlantic mul-tidecadal variability and related teleconnections are known to bring with them a range of potential impacts, which include changes in the intensities of the South American, African and Indian monsoons22,23 and linkages to Atlantic hurricane activity22. Our knowledge of the potential intensity of such changes is limited by our coarse under-standing of the historical Atlantic multidecadal variability and the degree to which past changes were driven by external versus internal influences24. Considerable uncertainty therefore surrounds the poten-tial impacts of such shifts, and the relative magnitude of such impacts

2020 2040 2060 2080 2100

–2

–1

0

1

2

3a

c

3,000

60° S 30° S 0°

–0.25 –0.05 0 0.05 0.25

30° N 60° N

60° S 30° S 0°

–1.5

–0.9 0.

9–0

.3 0.3

–0.1

51.

50.

150 –0.5

–0.3 0.

3–0

.1 0.1

–0.0

50.

50.

050

30° N 60° N 60° S 30° S 0° 30° N 60° N

2,000

1,000800600

400

200

800600

400

200

0

b

Dep

th (

m)

3,000

2,000

1,000

0d

Dep

th (

m)

–3

RCP8.5

GLENS

Fig. 4 | ocean responses with depth. a–d Evolution of the ensemble annual mean AMOC index20 (a) and latitude–depth view of changes between 2075–2095 and 2010–2030 in ocean meridional currents (b), temperature (c) and salinity (d). The shading in a depicts the s.d. of the annual mean AMOC values and the contour lines in b–d depict changes in potential density (at 0.025 g kg–1 increments; negative changes are shown by the dashed contour lines). All the coloured regions and contour lines are statistically significant (> 2 standard error). Hatched regions correspond to bathymetry.

NaTuRE GEoScIENcE | www.nature.com/naturegeoscience

http://www.nature.com/naturegeoscience

ArticlesNATurE GEOSciENcEto those where geoengineering is not implemented. Finally, given the strong dependence of CESM1(WACCM)’s forced response on regional projections of the water cycle and surface winds, its estima-tions of AMOC variability, trends and impacts are themselves uncer-tain, with potentially important model dependencies. Investigation is currently underway to isolate the relative importance of the dynami-cal response to the stratospheric heating anomalies from the impact of the altered radiative fluxes below the tropopause. This will be pre-sented in a forthcoming study, but preliminary results indicate that, although the reductions in evaporation in the subtropics are primarily due to the reduced downward radiative fluxes, there is an important role for the dynamical response to the stratospheric heating anomalies in producing the extratropical surface wind stress anomalies. Which of these factors is key to producing the ocean circulation changes dis-cussed here remains to be fully understood, but the suggestion is that solar dimming experiments may be inadequate to simulate accurately the coupled response25–29. Together, these findings demonstrate the risks involved in geoengineering associated with the potential com-plexity of the coupled climate system response and the need to better develop our understanding of the climate system before the character of a geoengineered climate can be estimated with confidence.

online contentAny methods, additional references, Nature Research reporting summaries, source data, statements of data availability and asso-ciated accession codes are available at https://doi.org/10.1038/s41561-018-0249-7.

Received: 20 March 2018; Accepted: 26 September 2018; Published: xx xx xxxx

References 1. Crutzen, P. J. Albedo enhancement by stratospheric sulfur injections: a

contribution to resolve a policy dilemma? Clim. Change 77, 211–219 (2006). 2. Kleypas, J. et al. Geochemical consequences of increased atmospheric carbon

dioxide on coral reefs. Science 284, 118 (1999). 3. Cao, L. The effects of solar radiation management on the carbon cycle. Curr.

Clim. Change Rep. 4, 41–50 (2018). 4. Haywood, J. M. A., Jones, N., Bellouin & Stephenson, D. Asymmetric forcing

from stratospheric aerosols impacts Sahelian rainfall. Nat. Clim. Change 3, 660–665 (2013).

5. Tilmes, S. et al. The hydrological impact of geoengineering in the Geoengineering Model Intercomparison Project (GeoMIP). Geophys. Res. Lett. 118, 11–036 (2013).

6. Jones, A. C. et al. Impacts of hemispheric solar geoengineering on tropical cyclone frequency. Nat. Commun. 8, 1382 (2017).

7. Kravitz, B. et al. First simulations of designing stratospheric sulfate aerosol geoengineering to meet multiple simultaneous climate objectives. J. Geophys. Res. Atmos. 122, 12616–12634 (2017).

8. Tilmes, S. et al. CESM1(WACCM) Stratospheric Aerosol Geoengineering Large Ensemble (GLENS) project. Bull. Am. Meteorol. Soc. https://doi.org/10.1175/BAMS-D-17-0267.1 (2018).

9. Mills, M. J. et al. Radiative and chemical response to interactive stratospheric sulfate aerosols in fully coupled CESM1(WACCM). J. Geophys. Res. Atmos. 122, 13061–13078 (2017).

10. Richter, J. H. et al. Stratospheric dynamical response and ozone feedbacks in the presence of SO2 injections. J. Geophys. Res. Atmos. 122, 12557–12573 (2017).

11. Kushnir, Y. Interdecadal variations in the North Atlantic sea surface temperature and associated atmospheric conditions. J. Clim. 7, 141–157 (1994).

12. Delworth, T. L. & Mann, M. E. Observed and simulated multidecadal variability in the Northern Hemisphere. Clim. Dyn. 16, 661–676 (2000).

13. Zhang, R. On the persistence and coherence of subpolar sea surface temperature and salinity anomalies associated with the Atlantic multidecadal variability. Geophys. Res. Lett. 44, 7865–7875 (2017).

14. Phillips, A., Deser, C. & Fasullo, J. Evaluating modes of variability in climate models. Eos Trans. AGU 95, 453–455 (2015).

15. Moore, J. C., Jevrejeva, S. & Grinsted, A. Efficacy of geoengineering to limit 21st century sea-level rise. Proc. Natl Acad. Sci. USA 107, 15699–15703 (2010).

16. Irvine, P. J. et al. The fate of the Greenland Ice Sheet in a geoengineered, high CO2 world. Environ. Res. Lett. 4, 045109 (2009).

17. Irvine, P. J., Keith, D. W. & Moore, J. Brief communication: understanding solar geoengineering’s potential to limit sea level rise requires attention from cryosphere experts. Cryosphere 12, 2501–2513 (2018).

18. Overpeck, J. T. et al. Paleoclimatic evidence for future ice-sheet instability and rapid sea-level rise. Science 311, 1747–1750 (2006).

19. Favier, L. et al. Retreat of Pine Island Glacier controlled by marine ice-sheet instability. Nat. Clim. Change 4, 117–121 (2014).

20. Pollard, D., DeConto, R. & Alley, R. Potential Antarctic ice sheet retreat driven by hydrofracturing and ice cliff failure. Earth Planet. Sci. Lett. 412, 112–121 (2015).

21. Chylek, P., Dubey, M. K. & Lesins, G. Greenland warming of 1920–1930 and 1995–2005. Geophys. Res. Lett. 33, L11707 (2006).

22. Zhang, R. & Delworth, T. L. Impact of Atlantic multidecadal oscillations on India/Sahel rainfall and Atlantic hurricanes. Geo. Res. Lett. 33, L17712 (2006).

23. Krishnamurthy, L. & Krishnamurthy, V. Teleconnections of Indian monsoon rainfall with AMO and Atlantic tripole. Clim. Dyn. 46, 2269–2285 (2016).

24. Frankignoul, C., Gastineau, G. & Kwon, Y. Estimation of the SST response to anthropogenic and external forcing and its impact on the Atlantic multidecadal oscillation and the Pacific decadal oscillation. J. Clim. 30, 9871–9895 (2017).

25. Ineson, S. et al. Solar forcing of winter climate variability in the Northern Hemisphere. Nat. Geosci. 4, 753–757 (2011).

26. Muthers, S., Raible, C. C., Rozanov, E. & Stocker, T. F. Response of the AMOC to reduced solar radiation—the modulating role of atmospheric chemistry. Earth Syst. Dynam. 7, 877–892 (2016).

27. Otterå, O. H., Bentsen, M., Drange, H. & Suo, L. External forcing as a metronome for Atlantic multidecadal variability. Nat. Geosci. 3, 688–694 (2010).

28. Scaife, A. A. et al. A mechanism for lagged North Atlantic climate response to solar variability. Geophys. Res. Lett. 40, 434–439 (2013).

29. Swingedouw, D. et al. Natural forcing of climate during the last millennium: fingerprint of solar variability. Clim. Dynam. 36, 1349–1364 (2011).

acknowledgementsThe authors acknowledge support from NASA award no. 80NSSC17K0565 and NSF award no. 1243107. The Pacific Northwest National Laboratory is operated for the US Department of Energy by Battelle Memorial Institute under contract DE-AC05-76RL01830. This research was developed with funding from the Defense Advanced Research Projects Agency. The views, opinions and/or findings expressed are those of the authors and should not be interpreted as representing the official views or policies of the Department of Defense or the US Government.

author contributionsJ.T.F. performed the core analysis, interpretation of results and writing. S.T. and J.H.R. performed the experiments and contributed to manuscript revisions. B.K. envisioned and designed the experimental set-up. D.G.M. contributed to the experimental set-up and manuscript revisions. M.J.M. and I.R.S. contributed to group discussions and manuscript revisions.

competing interestsThe authors declare no competing interests.

additional informationReprints and permissions information is available at www.nature.com/reprints.

Correspondence and requests for materials should be addressed to J.T.F.

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© The Author(s), under exclusive licence to Springer Nature Limited 2018

NaTuRE GEoScIENcE | www.nature.com/naturegeoscience

https://doi.org/10.1038/s41561-018-0249-7https://doi.org/10.1038/s41561-018-0249-7https://doi.org/10.1175/BAMS-D-17-0267.1https://doi.org/10.1175/BAMS-D-17-0267.1http://www.nature.com/reprintshttp://www.nature.com/naturegeoscience

Articles NATurE GEOSciENcEMethodsThe simulations in GLENS were designed to achieve multiple climate goals by injecting sulfur in a strategic manner at four different locations in the stratosphere in an iterative self-adjusting fashion based on the preceding simulated temperature anomalies. This contrasts with many previous studies that stabilize only the global mean surface temperature (through equatorial injections). Our novel approach reduces some adverse impacts identified in previous studies, such as an overcooling of the tropics and shifts in tropical precipitation.

The CESM1(WACCM) model provides a comprehensive representation of both the troposphere and stratosphere9, and is coupled to terrestrial, sea-ice and ocean component models. SO2 injections occur at four predefined locations (on the dateline at 30° N, 30° S, 15° N and 15° S) roughly 5 km above the tropopause. The efficacies of the injection altitudes of 1 km and 5 km above the tropopause were tested and the higher altitude was found to be more efficient in reducing surface warming and was therefore used5. Aerosol radiative effects are designed to achieve multiple stability objectives, which include in the global mean, interhemispheric gradient and equator-to-pole difference of surface temperature using a feedback-control algorithm that governs the annual injection amounts at each latitude7.

The ensemble consists of 20 members, which allows for a robust assessment of the regional climate response within the variability of the climate system. The atmospheric model, WACCM, uses a 0.9° latitude × 1.25° longitude grid with 70 vertical layers that reach up to 140 km (~10− 6 hPa). The model includes fully interactive middle atmosphere chemistry. The chemical scheme includes gas phase and heterogeneous reactions relevant to stratospheric ozone chemistry9; however, a simplified chemistry scheme is used in the troposphere that renders the simulations unsuitable for investigating changes in the tropospheric ozone or other tropospheric trace gases. The model has the ability to simulate the formation of stratospheric sulfate aerosols through oxidation (after the injection of SO2) using a modal aerosol model (MAM3), which is interactively coupled to chemistry and radiation30. MAM3 simulates aerosol microphysical processes relevant to sulfate aerosols and includes nucleation, coagulation, condensational growth, evaporation and sedimentation. The coupling between tropospheric aerosols, clouds and radiation is explicitly resolved31.

The forcings of the GLENS simulations are based on RCP8.5, that is, a high anthropogenic emission scenario, which specifies anthropogenic aerosol and GHG concentrations along with other forcings. The simulations are initialized in 2010 using a free-running historical simulation of CESM1(WACCM) forced by Coupled Model Intercomparison Project Phase 5 historical forcings and RCP8.5 after 2005, as described previously9. The ocean and sea-ice initial conditions were taken from year 2010 of member 001 of the CESM Large Ensemble Project32. This approach provides a reasonable sampling of the atmospheric internal variability but not the full diversity of the ocean initial states. Other details about the set-up of the simulations are already documented8. The 20-member ensemble starts in 2010 and continues until 2030. Three ensemble members are extended at least until 2097, with only one (ensemble number three) completing year 2099 due to instabilities in

simulating RCP8.5 towards the end of the twenty-first century. For each ensemble member, the atmospheric state is initialized with 1 January conditions taken from different years 2008 to 2012 of the reference simulation and a round-off (order of 10−14 K) air temperature perturbation. The land, sea-ice and ocean start from the same initial conditions for each ensemble member. The geoengineering simulations start in 2020, branch from each of the 20 control simulations and continue until 2099. Slightly different injection amounts per location in each simulation are used at a given time due to the anticipated differences in the evolution of temperature targets that arise from internal variability. Note that, although the ocean conditions diverge somewhat between 2010 and 2020 in the control simulations, some memory of the ocean state persists, particularly in the Atlantic Ocean. For example, all the RCP8.5 and geoengineering ensemble members exhibit a negative phase of the Atlantic multidecadal oscillation33 at around 2020 and then slowly diverge over the next 15–20 years. The manifestation of initialization-related changes in AMOC are, however, small in relation to twenty-first century forced changes (Fig. 4a).

Model simulations were performed on the Cheyenne high-performance computing platform built for the National Center for Atmospheric Research by Silicon Graphics International. A comprehensive output was produced to enable the broader community to perform analysis of the ensemble. A general diagnostic assessment of the ensemble and information on how to download the output from the National Center for Atmospheric Research Earth System Grid are available at http://www.cesm.ucar.edu/projects/community-projects/GLENS

Code availability. The software code used to compute the AMOC time series is available at http://www.cesm.ucar.edu/working_groups/CVC/cvdp/.

Data availabilityThe data sets analysed during this study are available on the Earth System Grid (https://www.earthsystemgrid.org and http://www.cesm.ucar.edu/projects/community-projects/GLENS/).

References 30. Mills, M. J. et al. Global volcanic aerosol properties derived from emissions,

1990–2014, using CESM1(WACCM). J. Geophys. Res. Atmos. 121, 2332–2348 (2016).

31. Liu, X. et al. Toward a minimal representation of aerosols in climate models: description and evaluation in the Community Atmosphere Model CAM5. Geosci. Model Develop. 5, 709–739 (2012).

32. Kay, J. E. et al. The community earth system model (CESM) large ensemble project: a community resource for studying climate change in the presence of internal climate variability. Bull. Am. Meteorol. Soc. 96, 1333–1349 (2015).

33. Trenberth, K. E. & Shea, D. J. Atlantic hurricanes and natural variability in 2005. Geophys. Res. Lett. 33, 12 (2006).

NaTuRE GEoScIENcE | www.nature.com/naturegeoscience

http://www.cesm.ucar.edu/projects/community-projects/GLENShttp://www.cesm.ucar.edu/working_groups/CVC/cvdp/https://www.earthsystemgrid.orghttp://www.cesm.ucar.edu/projects/community-projects/GLENS/http://www.cesm.ucar.edu/projects/community-projects/GLENS/http://www.nature.com/naturegeoscience

Persistent polar ocean warming in a strategically geoengineered climateGlobal mean climate responses in GLENSThe regional coupled climate response in GLENSBroader consequencesOnline contentAcknowledgementsFig. 1 Global mean climate responses.Fig. 2 Responses of air–sea fluxes of heat, freshwater and momentum.Fig. 3 Responses of the upper ocean and OHC.Fig. 4 Ocean responses with depth.