Commentary | Published:

Solar geoengineering reduces atmospheric carbon burden

Nature Climate Change volume 7, pages 617619 (2017) | Download Citation

Solar geoengineering is no substitute for cutting emissions, but could nevertheless help reduce the atmospheric carbon burden. In the extreme, if solar geoengineering were used to hold radiative forcing constant under RCP8.5, the carbon burden may be reduced by 100 GTC, equivalent to 12–26% of twenty-first-century emissions at a cost of under US$0.5 per tCO2.

Failure to address the accumulation of atmospheric carbon is among the most frequently noted disadvantages of solar geoengineering1,2,3, an attempt to reflect a small fraction of radiation back into space to cool the planet. The latest US National Academy of Science solar geoengineering report1 states it “does nothing to reduce the build-up of atmospheric CO2”.

This is not so. Solar geoengineering reduces the carbon burden, and therefore ocean acidification, due to the three pathways explored here: carbon-cycle feedback4,5,6,7,8, reduced permafrost melting, and reduced energy-sector emissions.

While it is appropriate to treat solar geoengineering as distinct from carbon mitigation or geoengineering approaches that tackle carbon directly9, the impact of solar geoengineering on the carbon cycle calls for more integrated research. Solar geoengineering or solar radiation management (SRM) is, in this sense alone, arguably a form of carbon dioxide removal (CDR).

Carbon impacts of solar geoengineering

We calculate the total carbon burden in 2100 and carbon emissions impacts during the twenty-first century by estimating, based on diverse prior literature, the difference between a Representative Concentration Pathway (RCP) 8.5 scenario and one in which solar geoengineering is used to hold radiative forcing at current levels. This is not a complete analysis, but rather a call for further research. It is also a call for assessing solar geoengineering scenarios that go well beyond oft-modelled extreme scenarios that offset total anthropogenic radiative forcing10. These rough estimates alone, however, provide suggestive evidence of the potentially large impact of solar geoengineering on the carbon burden and emissions.

Warming can increase the atmospheric carbon burden by increasing ecosystem respiration, decreasing primary productivity, and decreasing oceanic carbon uptake. These carbon cycle feedbacks amplify climate responses to anthropogenic emissions. Point estimates differ widely (see Supplementary Table 1).

We derive an overall range in two steps. First we take estimates of 31 GtC (ref. 5) and 251 GtC (ref. 4) from the only two models that directly simulate the carbon cycle response to RCP8.5 and solar geoengineering, and combine them with the full range of results from a study11 estimating the carbon response with and without CO2 impact on climate: 42–420 GtC. The latter provides systematic sampling of the uncertainty in the carbon cycle feedback under assumptions that are similar — though not equal — to those that would be used to simulate solar geoengineering to stabilize radiative forcing under an RCP8.5 scenario. We then combine the two ranges using equal weights and uncorrelated error propagation to yield an overall estimate of the contribution of the terrestrial biosphere and ocean of 89–283 GtC (Table 1).

Table 1: Reduction in twenty-first-century emissions and in 2100 atmospheric carbon burden in GtC.

The direct impact of solar geoengineering on the loss of carbon from permafrost soils is unexplored. We instead include estimates from recent intercomparisons of dynamic permafrost models that estimate a decrease in cumulative emissions under RCP8.5 to range from 27 to 122 GtC (ref. 6).

These rough estimates should be interpreted with caution. Caveats include: neglecting the differences between the Special Report on Emissions Scenarios (SRES)-A2 (ref. 11) and RCP8.5 (refs 4,5,6) scenarios; neglecting the fact that simulating carbon-cycle feedback by eliminating all climate change11 is, at best, a rough proxy for solar geoengineering; and ignoring more speculative carbon feedbacks such as sea-bed methane hydrates7. Moreover, our subsequent rough translation of carbon burden to emissions, and vice versa, does not account for changes in ocean buffering12.

For a given energy demand and fuel mix, carbon emissions will rise with temperature, as the efficiency of heat engines declines with rising ambient temperature. Warming will also decrease energy demand for heating and increase energy demand for cooling. We use a global estimate of energy demand response to warming in the residential sector13, roughly scaled to cover the commercial sector as well as transport, along with various estimates of the impact of climate changes on energy use, to yield a rough estimate that avoiding the warming in an RCP8.5 emissions scenario decreases cumulative emission by 24–54 GtC (see Supplementary Materials).


Risks, uncertainties, and inter-temporal trade-offs make simple cost-effectiveness estimates a poor measure of the overall utility of solar geoengineering. Narrow calculations of costs make solar geoengineering, in particular using stratospheric aerosols, appear 'too cheap'. Our analysis does not claim completeness. There are clearly unquantified and perhaps unquantifiable risks of solar geoengineering. Those may imply that the only relevant decision criterion for solar geoengineering deployment is one based on risk–risk tradeoffs, not one based on cost–benefit analysis. Increased albedo is not anti-CO2. But when considering solar geoengineering as a means of reducing carbon burdens, cost-effectiveness is relevant because the comparison is to other means of achieving the same result.

We estimate the cost-effectiveness of solar geoengineering's carbon cycle impact using our estimate of the equivalent twenty-first-century emissions reductions of 232–527 GtC (Table 1) converted roughly into 850–1,900 GtCO2. We assume a radiative forcing efficacy14 of 0.55 Wm−2, triple a prior engineering estimate of aircraft lofting costs15 for a cost of US$2 billion per Mt per year, use monitoring costs equal to the totality of the current annual US Global Change Research Program budget, rounded up to US$3 billion per year16, and use a central discount rate of 3% (Supplementary Materials).

Total costs come to approximately US$300 billion for the twenty-first century. That roughly equals estimated equivalent mitigation costs of US$0.2–0.4 per tCO2, dipping to US$0.1–0.2 per tCO2 for a 5% discount rate, and increasing slightly to US$0.2–0.5 per tCO2 for a 2.5% rate. Regardless of the specific range used, these numbers are far below current estimates of the costs of CDR9, which can go into the hundreds of dollars per tCO2.


If used to offset changes in twenty-first-century radiative forcing under an RCP8.5 emissions scenario, our rough estimates suggest that solar geoengineering could reduce the carbon burden in 2100 by around 160–370 GtC, roughly equivalent to reducing twenty-first-century emissions by 850–1,900 GtCO2 at a mitigation cost of US$0.2–0.4 per tCO2. Rather than having no impact on carbon, solar geoengineering may be among the most cost-effective methods of limiting the rise in CO2 concentrations and, therefore, the rise in ocean acidification.

Even with these carbon benefits, solar geoengineering cannot substitute for cutting emissions. For one, our rough estimates, using an extreme scenario, show a total emissions impact of 'only' around 12–26% of total twenty-first-century emissions under RCP8.5 (ref. 17).

Second, the two primary factors we identify here, carbon-cycle feedback and permafrost release, merely move carbon within the biosphere. Only the smaller third factor, via the energy sector, prevents moving carbon from the geosphere. Unlike some forms of CDR, no mechanism here removes carbon from the biosphere and puts it back into the geosphere. Thus, terminating solar geoengineering efforts would lead to a significant adverse carbon impact.

Third, none of this addresses an oft-cited, indirect link via societal responses, often under the heading of 'moral hazard'18. Solar geoengineering may have direct implications on nations' and jurisdictions' willingness to cut emissions. The phenomenon is important and empirically still understudied18,19. But there is a sharp distinction between political questions about the response to possible or actual deployment of solar geoengineering and technical questions about carbon cycle response. Both questions matter. Policy-relevant analysis must not confuse them.

The need for integrated research

We intend our rough estimate of solar geoengineering's potential to reduce carbon burden not as an answer, but as a spur for further research. That begins with a more detailed look at direct carbon burden and emissions impacts of solar geoengineering scenarios. RCP8.5 is but one such scenario. It must not end there, for a number of reasons.

First, if solar geoengineering is used to stabilize radiative forcing under a scenario with stronger climate policy and lower carbon emissions, then the reduction in carbon burden will be correspondingly smaller.

Second, the amount of solar geoengineering is a policy choice. While climate-modelling studies often assume that solar geoengineering will be used to offset all warming, a moderate scenario with less solar geoengineering is likely a better policy10. For example, using solar geoengineering to halve the rate of radiative forcing growth might better balance the risks and benefits4,10. Under such a scenario, everything else staying equal, the reduction in carbon burden would likely be roughly halved as well relative to the calculations here. Some carbon impacts also depend on the type of solar geoengineering and specific materials used. Our rough cost calculation, in particular, assumes using sulfate aerosols. Resulting stratospheric ozone depletion may lead to small increases in ocean acidification20. Using different compounds21,22 may have lower effects or even the opposite effect. Marine cloud brightening similarly has direct implications on emissions, carbon burden23, and, thus, also ocean acidification24.

The third reason is moral hazard: the need to consider social and societal responses beyond the technical calculations.

Sensible policy decisions about both emissions mitigation and solar geoengineering will be aided by better estimates of the carbon-cycle benefits of solar geoengineering and of the way the reduction in carbon burden scales with the amount of solar geoengineering and mitigation. A coordinated research effort should aim to understand the coupling between solar geoengineering, CDR, the energy system, and the 'natural' carbon cycle. Policymakers cannot make sound choices without a sustained, integrated research programme.


  1. 1.

    National Research Council. Climate Intervention: Reflecting Sunlight to Cool Earth (National Academies Press, 2015).

  2. 2.

    If all else fails. The Economist (26 November 2015).

  3. 3.

    Bull. At. Sci. 64, 14–18 (2008).

  4. 4.

    , & Nat. Commun. 5, 3304 (2014).

  5. 5.

    , & J. Geophys. Res. Biogeosci. 121, 2–27 (2016).

  6. 6.

    et al. Nature 520, 171–179 (2015).

  7. 7.

    , , , & Earth Planet. Sci. Lett. 367, 105–115 (2013).

  8. 8.

    , , & Atmos. Chem. Phys. 16, 1479–1489 (2016).

  9. 9.

    National Research Council. Climate Intervention: Carbon Dioxide Removal and Reliable Sequestration (National Academies Press, 2015).

  10. 10.

    & Earth's Future 4, 549–559 (2016).

  11. 11.

    et al. J. Clim. 19, 3337–3353 (2006).

  12. 12.

    , & Glob. Biogeochem. Cycles 24, GB1002 (2010).

  13. 13.

    & Energy Policy 37, 507–521 (2009).

  14. 14.

    , , , & Geophys. Res. Lett. 37, L18805 (2010).

  15. 15.

    , & Environ. Res. Lett. 7, 034019 (2012).

  16. 16.

    Our Changing Planet: the US Global Change Research Program for Fiscal Year 2017 (USGCRP, 2016).

  17. 17.

    et al. Climatic Change 109, 33 (2011).

  18. 18.

    et al. Earth's Future 4, 536–542 (2016).

  19. 19.

    & Earth's Future 5, 136–143 (2017).

  20. 20.

    et al. Geophys. Res. Lett. 36, L12606 (2009).

  21. 21.

    , & Atmos. Chem. Phys. 15, 11835–11859 (2015).

  22. 22.

    , , & Proc. Natl Acad. Sci. USA 113, 14910–14914 (2016).

  23. 23.

    , & Geophys. Res. Lett. 42, 2951–2960 (2015).

  24. 24.

    , , & Geophys. Res. Lett. 43, 7600–7608 (2016).

  25. 25.

    et al. J. Clim. 26, 4398–4413 (2013).

Download references


The authors thank K. Caldeira for discussion and feedback.

Author information


  1. David W. Keith and Gernot Wagner are at the Harvard John A. Paulson School of Engineering and Applied Sciences, 12 Oxford Street, Cambridge, Massachusetts 02138, USA, and Harvard Kennedy School, 79 John F. Kennedy Street, Cambridge, Massachusetts 02138, USA

    • David W. Keith
    •  & Gernot Wagner
  2. Claire L. Zabel is at The Open Philanthropy Project, 182 Howard Street #225, San Francisco, California 94105, USA

    • Claire L. Zabel


  1. Search for David W. Keith in:

  2. Search for Gernot Wagner in:

  3. Search for Claire L. Zabel in:

Competing interests

C.L.Z. began work on this analysis while a researcher at Harvard. She now works for the Open Philanthropy Project, which subsequently became a funder of Harvard's Solar Geoengineering Research Project, co-directed by D.W.K. and G.W.

Corresponding author

Correspondence to David W. Keith.

Supplementary information

PDF files

  1. 1.

    Supplementary Information

    Solar geoengineering reduces atmospheric carbon burden

About this article

Publication history



Further reading

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing