One proposal for the mitigation of ongoing global warming is the sequestration of carbon dioxide extracted at combustion sites or directly from the air1,2. Such sequestration could help avoid a large rise in atmospheric CO2 concentration from unchecked use of fossil fuels, and hence extreme warming in the near future3,4. However, it is not clear how effective different types of sequestration and associated leakage are in the long term, and what their consequences might be. Here I present projections over 100,000 years for five scenarios of carbon sequestration and leakage with an Earth system model5. Most of the investigated scenarios result in a large, delayed warming in the atmosphere as well as oxygen depletion, acidification and elevated CO2 concentrations in the ocean. Specifically, deep-ocean carbon storage leads to extreme acidification and CO2 concentrations in the deep ocean, together with a return to the adverse conditions of a business-as-usual projection with no sequestration over several thousand years. Geological storage may be more effective in delaying the return to the conditions of a business-as-usual projection, especially for storage in offshore sediments. However, leakage of 1% or less per thousand years from an underground stored reservoir, or continuous resequestration far into the future, would be required to maintain conditions close to those of a low-emission projection with no sequestration.
Massive application of carbon dioxide capture and storage (CCS) has become a prominent proposal for mitigating global warming1,2. Recent discussions of CCS have been centred on its motivation, feasibility and cost compared with other mitigation strategies2,6,7,8. Concerns have also been raised about possible consequences of storing CO2 in the ocean and of leakage of CO2 stored in geological formations1,9. Here I compare the effectiveness as well as Earth system consequences of CO2 leakage for sequestration in the ocean and in onshore and offshore geological formations. For this purpose I make long projections with the reduced-form, Danish Center for Earth System Science (DCESS) model5.
As a target for desirable future emissions, I designed an emission scenario for avoiding strong global warming (Fig. 1, AG scenario; see the Methods section). For a projection forced with AG emissions, atmospheric pCO2 peaks at 408 ppm by AD 2038 and mean atmosphere warming peaks at 2.03 °C by AD 2102 (as compared with AD 1765; Fig. 2a,c). As actual fossil-fuel emissions may continue to rise during the next decade, these and related results10 indicate that greater global emission reductions than those assumed here may be needed during the rest of this century to avoid warming greater than 2 °C above pre-industrial conditions. Still greater reductions would be required to meet this goal if actual climate sensitivity is greater than the model value of 3 °C. In the AG projection, atmosphere warming does not fall below 1 °C until AD 4100, owing mainly to slow ocean CO2 uptake.
As an example for possible future, ‘business-as-usual’ emissions, I use the extended A2 emission scenario11 (Fig. 1; see the Methods section). A projection forced with A2 emissions shows atmosphere warming above 5 °C by AD 2130 (Fig. 2c), ocean warming above 3 °C by AD 4010 (Fig. 2e) and tenfold expansion of ocean ‘dead zone’ volume by AD 3000 (Fig. 2f; ref. 11). In the A2 projection, greater warming from higher emissions induces positive carbon feedbacks, such as less CO2 dissolution in the ocean and more soil remineralization11, that further extend the duration of strong warming. As a result, atmosphere warming does not fall below 3 °C until AD 9550. Three million years ago, mean global temperature was about 2–3 °C warmer than pre-industrial conditions and sea level was about 20–30 m higher12. Such a large sea level rise can be explained only by partial melting of the Greenland and Antarctic ice sheets. Given the size and duration of global warming in the A2 projection, it is likely that large sea level rise from ice-sheet melting would also be a consequence of the A2 emissions. The extended atmospheric warming of the A2 projection also promotes greater ocean warming, leading to enhanced ‘dead zones’ from decreased oxygen solubility as well as greater sea level rise from thermal expansion. Furthermore, ocean heating would probably lead to methane release from methane hydrate melting in ocean sediments and still more long-term global warming not considered here13.
A future pathway with continued high fossil-fuel use but with fully developed CCS can be formulated as gross CO2 emissions from the A2 scenario combined with carbon capture at power plants, direct extraction of CO2 from the atmosphere14 and subsequent sequestration such as to reduce net CO2 emissions to those of the AG scenario (Fig. 1). CCS implementation requires considerable energy use1 such that less net energy is available for society in this pathway than in the A2 scenario although gross CO2 emissions are the same for both. All emissions other than CO2, for example those of methane, nitrous oxide and black carbon, are taken to follow the A2 scenario. In this pathway, total gross and net CO2 emissions are 3,721 GtC and 822 GtC such that 2,899 GtC is sequestered over two centuries. Such immense sequestration probably does not exceed the storage capacity in geological formations15 nor in the ocean and may be technologically feasible over the timescale considered16. Below I consider long projections for this sequestration pathway applied to five different storage/leakage cases.
An ocean storage projection, OS, assumes CO2 injection at depths between 2,500 and 3,500 m of the low–mid latitude ocean and total dissolution at the depths injected. Such sequestration mitigates much of the initial warming but this warming still exceeds that of the AG projection by about 1 °C because of greater net radiative forcing from non-CO2 greenhouse gases (Fig. 2c). Ocean storage leads to a mean pH decrease at 3,000 m depth of more than one unit but surface-ocean pH decreases much less than in the A2 projection. In addition, this injection causes a mean pCO2 increase at 3,000 m depth of almost 7,000 ppm (Table 1). Both increased acidity and pCO2 can adversely affect deep-sea life17,18. Atmosphere warming drops slightly after the initial warming peak but rises again to 3.5 °C by AD 4000 as some sequestered CO2 leaks to the atmosphere over ocean circulation and mixing timescales (Fig. 2c). Subsequently, the OS projection follows closely the A2 projection and suffers similar large ‘dead zone’ volume increase (Fig. 2f) and probable large sea level rise. Ocean CO2 injection also leads to very low carbonate ion concentration (<15 μmol kg−1) at depth and enhanced CaCO3 dissolution in sediments that acts to suppress oceanic, and thereby atmospheric, CO2. However, ocean circulation and mixing limits sediment exposure to such low carbonate ion concentrations to times much shorter than CaCO3 dissolution timescales and not so much extra CaCO3 is dissolved. By AD 10000 this has led to an extra ocean carbon inventory of 115 GtC and an extra atmospheric pCO2 drawdown of only 5 ppm compared with the A2 projection (Fig. 2a,d).
Geological storage sites for CO2 vary greatly in type and location whereby deep saline formations provide by far the largest potential storage capacity1,15. Great heterogeneity in storage sites and conditions at them makes direct modelling of the total leakage of geologically stored CO2 very difficult. Thus, I base my projections on the Intergovernmental Panel on Climate Change (IPCC) conclusion1 that the fraction retained in appropriately selected and managed geological reservoirs is very likely to exceed 99% over 100 years and is likely to exceed 99% over 1,000 years. In addition, as a ‘worst case’ leakage scenario I also consider 99% retention over 10 years. The time evolution of the geologically stored, CO2 reservoir, G(t), consistent with these assumptions may be expressed as
where S(t) is the sequestration rate (Fig. 1) and a G(t) is the leakage rate, whereby a is an inverse-leakage timescale. I consider three onshore geological storage projections, GS1, GS2 and GS3, with leakage to the atmosphere and a equal to 1×10−3, 1×10−4 and 1×10−5 yr−1 from the above (Fig. 2b).
As for the ocean storage projection, all onshore geological storage projections show mitigation of much of the initial warming of the A2 projection (Fig. 2). Later on, ‘rapidly leaking’ projection GS1 has higher pCO2 and even more atmosphere and ocean warming and ocean oxygen depletion than the OS projection. Indeed, after several thousand years, pCO2 and atmosphere warming of the GS1 projection even exceed those of the A2 projection with no sequestration. This is consistent with and extends earlier results for shorter projections with faster leakage4. The extra, long-term warming extends far into the future, fostering more ocean warming and ocean oxygen depletion than in the A2 projection then. ‘Moderately leaking’ projection GS2 shows greatly reduced pCO2, atmosphere and ocean warming, ocean acidification and ocean ‘dead zone’ volume for more than 10,000 years, compared with the A2 projection (Fig. 2). Subsequently, these properties increase to exceed those of the A2 projection. For ‘weakly leaking’ projection GS3, atmospheric pCO2 drops below 350 ppm by AD 2890 and remains below this level (Fig. 2a). Likewise, atmospheric warming decreases rapidly to levels slightly above present-day global warming and initial evolutions of ocean warming, acidification and oxygen depletion are similar to the AG projection. Subsequently, the GS3 projection diverges towards the A2 projection, exhibiting slow enhancement of ‘dead zone’ volume during the last two-thirds of the projection. A ‘best case’ projection with no leakage from geological storage would converge to very near the AG projection after several hundred years.
I also consider a projection, GS1O, for storage in deep-ocean sediments, with a as for GS1 but with leakage to the ocean, distributed as in the OS projection. The GS1O projection also shows initial mitigation (Fig. 2). In this ‘rapidly leaking’ projection, atmosphere warming and surface-ocean acidification are significantly decreased and delayed compared with the GS1 projection, although the GS1O projection also approaches the A2 projection later on (Fig. 2; Table 1). Furthermore, as the timescale of this leakage is much greater than that of sequestration, the impact of the GS1O projection on deep-ocean conditions is much less than for the OS projection (Table 1). For these and other reasons such as potential CO2 immobilization in deep-sea sediments19, CO2 storage in offshore sediments may be preferable to onshore geological storage or ocean storage. However, the storage capacity of suitable offshore sediment reservoirs is uncertain20.
CO2 storage would have to last for tens of thousands of years to avoid strong, delayed global warming and marked Earth system changes such as vast expansion of ocean ‘dead zones’. This is much longer than 4,000 years claimed earlier20 from an apparent misinterpretation of carbon–climate models results21. Deep-ocean sequestration cannot meet this storage criterion and brings with it serious ecological side effects and hardly any advantage from enhanced CaCO3 dissolution. It is not clear whether geological sequestration could meet the above criterion, although storage in offshore sediments could help if suitable storage sites could be found. Even if massive, ‘safe’ sequestration could be implemented very soon, great reductions in net radiative forcing from non-CO2 greenhouse gases and black carbon, as in the AG scenario, would also be needed to avoid warming greater than 2 °C above pre-industrial levels in this century.
The carbon sequestration pathway presented here posits great human effort in the next two hundred years but subsequent human inaction in the spirit of ‘out of sight, out of mind’. Alternatively, long-term leakage from the ocean or geological reservoirs could be actively countered by resequestration to stabilize climate at some desired level. However, there are serious concerns connected with this. First, it would be difficult to gauge the global leakage rate that would have to be matched by the resequestration rate. Long-term monitoring of atmospheric CO2 concentrations would probably be the best way to address this but natural carbon-cycle fluctuations would complicate this approach. Second, resequestration would have to be carried out over many thousands of years, a burden for future society not unlike that of long-term management of nuclear waste. By greatly limiting carbon emissions in our time, we could reduce the need for massive CO2 sequestration and thus reduce unwanted consequences and burdens over many future generations from the leakage of sequestered CO2.
The DCESS model5 includes atmosphere, ocean, ocean sediment, land biosphere and lithosphere modules. Sea ice and snow cover are diagnosed from the atmospheric temperature distribution but there is no ice-sheet module. Earth system data and complex model results were used to calibrate the model to a pre-industrial (AD 1765), steady-state climate with a long-term climate sensitivity of 3.0 °C for a doubling of atmospheric pCO2. (An instantaneous pCO2 change from 278 to 556 ppm from this steady state forces mean atmospheric warming of 1.53, 2.58, 2.87 and 3.00 °C after 10, 100, 1,000 and 10,000 years of model integration.) The model is well suited for long projections because it includes CO2–climate feedbacks in the land biosphere and ocean, organic and carbonate carbon cycling in the ocean sediment and climate-dependent rock weathering.
All 100,000-year projections start with initial conditions from the pre-industrial, steady state and are forced initially by prescribed anthropogenic emissions of CO2, CH4 and N2O and prescribed radiative forcing from other (minor) greenhouse gases, aerosols and solar-radiation changes5. Fossil-fuel CO2 emissions from AD 1765 to 2004 are from ref. 22. Land-use-change CO2 emissions for AD 1860–1950 are from ref. 23. Linear interpolation from a zero value at AD 1765 is used before AD 1860. For AD 1950–1980, values from ref. 23 are multiplied by a factor varying linearly from 1 to 0.73 over this period and, for AD 1980–2000, are multiplied by 0.73 to yield mean values of about 1.4 and 1.6 GtC yr−1 for the 1980s and 1990s, respectively, in agreement with the recent IPCC assessment24. Anthropogenic CH4 emissions from AD 1860 to 1994 are from ref. 25 minus a value of 0.04 Gt(CH4) yr−1, an estimate for anthropogenic, pre-industrial emissions included in the model land-biosphere emissions5. Linear interpolation from a zero value at AD 1765 is used before AD 1860. Values for AD 1995–2000 are assigned the ‘corrected’ AD 1994 value. We take anthropogenic N2O emissions to be proportional to those of CH4 (in units mol s−1) with a proportionality constant of 0.007 from a model fit to N2O observations. Net radiative forcing from AD 1765 to 2000 resulting from other (minor) greenhouse gases, aerosols and solar-radiation changes is calculated from results in ref. 26 by subtracting radiative forcing resulting from CO2, CH4 and N2O from total radiative forcing. For the total radiative forcing described above, the projections reproduce observed evolutions up to AD 2000 of key climate metrics including atmosphere and ocean warming, atmospheric gas contents and ocean and land biosphere CO2 uptakes5.
The ‘business-as-usual’ A2 scenario and all carbon sequestration scenarios use fossil-fuel and land-use-change CO2 emissions for the periods AD 2004–2100 and AD 2000–2100, respectively, from the IPCC SRES A2 scenario27. After AD 2100, total CO2 emissions are extrapolated using polynomial functions to tail off by the early twenty-third century. It is assumed that by then accessible fossil-fuel resources will have become limited and/or alternative energy sources will have been adopted. Anthropogenic CH4 emissions for the period AD 2000–2100 are taken from the SRES A2 scenario and after AD 2100 are extrapolated as above. Anthropogenic N2O emissions from AD 2000 onward are estimated from the CH4 emissions as described above. Radiative forcings from other (minor) greenhouse gases and aerosols for the period AD 2000–2100 were compiled from the SRES A2 scenario and after AD 2100 are extrapolated as above. With the exception of fossil-fuel emissions, the low-emission, AG scenario uses emissions and radiative forcings after AD 2000 as described above but based on the SRES B1 scenario27. For the AG scenario, fossil-fuel emissions are taken to rise after AD 2004, level off by AD 2010 and subsequently fall to meet global emission reduction targets of 20% by 2020 and 60% by AD 2050 (compared with AD 1990). For simplicity, solar forcing is maintained constant after AD 2000 in all scenarios. This is motivated in part by the fact that orbital forcing will be weaker than normal for the next 100,000 years28.
Matlab programs for running the DCESS model can be downloaded at www.dcess.dk and include forcing functions and plotting routines needed to reproduce the results reported here.
IPCC, Metz, B., Davidson, O., de Coninck, H., Loos, M. & Meyer, L. (eds) in Special Report on Carbon Dioxide Capture and Storage 431 (Cambridge Univ. Press, 2005).
Hazeldine, R. S. Carbon capture and storage: How green can black be? Science 325, 1647–1652 (2009).
Kheshgi, H. S. & Archer, D. A nonlinear convolution model for the evasion of CO2 injected into the deep ocean. J. Geophys. Res. 109, C02007 (2004).
Haugan, P. & Joos, F. Metrics to assess the mitigation of global warming by carbon capture and storage in the ocean and in geological reservoirs. Geophys. Res. Lett. 31, L18202 (2004).
Shaffer, G., Olsen, S. M. & Pedersen, J. O. P. Presentation, calibration and validation of the low-order, DCESS Earth System model. Geosci. Model Dev. 1, 17–51 (2008).
Keller, K., McInerney, D. & Bradford, D. F. Carbon dioxide sequestration: How much and when? Clim. Change 88, 267–291 (2008).
Keith, D. W. Why capture CO2 from the atmosphere? Science 325, 1654–1655 (2009).
Rochelle, G. T. Amine scrubbing for CO2 capture. Science 325, 1652–1654 (2009).
Caldeira, K. & Wickett, M. E. Ocean model predictions of chemistry changes from carbon dioxide emissions to the atmosphere and ocean. J. Geophys. Res. 110, C09S04 (2005).
Meinshausen, M. et al. Greenhouse-gas emission targets for limiting global warming to 2 °C. Nature 458, 1158–1162 (2009).
Shaffer, G., Olsen, S. M. & Pedersen, J. O. P. Long-term ocean oxygen depletion in response to carbon dioxide emissions from fossil fuels. Nature Geosci. 2, 105–109 (2009).
Dowsett, H. et al. Joint investigations of the Middle Pliocene climate I: PRISM paleoclimate reconstructions. Glob. Planet. Change 9, 169–195 (1994).
Archer, D., Buffett, B. & Brovkin, V. Ocean methane hydrates as a slow tipping point in the global carbon cycle. Proc. Natl Acad. Sci. 106, 20596–20601 (2009).
Jones, N. Sucking it up. Nature 458, 1094–1097 (2009).
Orr, F. M. Jr Onshore geologic storage of CO2 . Science 325, 1656–1658 (2009).
Grübler, A., Nakicenovic, N. & Victor, D. G. Dynamics of energy technologies and global change. Energy Policy 27, 247–280 (1999).
Pörtner, H. O., Langenbuch, M. & Michaelidis, B. Effects of CO2 on marine animals: Interactions with temperature and hypoxia regimes. J. Geophys. Res. 110, C09S10 (2005).
Brewer, P. G. & Peltzer, E. T. Limits to marine life. Science 324, 347–348 (2009).
House, K. Z., Schrag, D. P., Harvey, C. F. & Lackner, K. S. Permanent carbon dioxide storage in deep-sea sediments. Proc. Natl Acad. Sci 103, 12291–12295 (2006).
Schrag, D. P. Storage of carbon dioxide in offshore sediments. Science 325, 1658–1659 (2009).
Archer, D. & Brovkin, V. The millennial atmospheric lifetime of anthropogenic CO2 . Clim. Change 90, 283–297 (2008).
Marland, G., Boden, T. A. & Andres, R. J. Trends: A Compendium of Data on Global Change (Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, US Department of Energy, 2007).
Houghton, R. A. Magnitude, distribution and causes of terrestrial carbon sinks and some implications for policy. Clim. Policy 2, 71–88 (2002).
IPCC, Solomon, S. D. et al. (eds) in Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (Cambridge Univ. Press, 2007).
Stern, D. I. & Kaufmann, R. K. Trends: A Compendium of Data on Global Change (Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, US Department of Energy, 1998).
Crowley, T. J. Causes of climate change over the past 1,000 years. Science 289, 270–277 (2000).
Nakicenovic, N. & Swart, R. (eds) A Special Report of Working Group III of the Intergovernmental Panel on Climate Change (Cambridge Univ. Press, 2000).
Loutre, M. F. & Berger, A. Future climatic changes: Are we entering an exceptionally long interglacial? Clim. Change 46, 61–90 (2000).
I thank P. Brewer, R. Garreaud, S. M. Olsen and J. O. P. Pedersen for comments. Part of this research was carried out at the Department of Geophysics, University of Chile, Santiago.
The author declares no competing financial interests.
About this article
Cite this article
Shaffer, G. Long-term effectiveness and consequences of carbon dioxide sequestration. Nature Geosci 3, 464–467 (2010). https://doi.org/10.1038/ngeo896
Reproducing natural variations in CO2 concentration in vadose zone wells with observed atmospheric pressure and groundwater data
Journal of Environmental Management (2020)
CO2-brine relative permeability and capillary pressure of Tuscaloosa sandstone: Effect of anisotropy
Advances in Water Resources (2020)
The value of hydrogen and carbon capture, storage and utilisation in decarbonising energy: Insights from integrated value chain optimisation
Applied Energy (2020)
Time-dependent observation of a cage-specific guest exchange in sI hydrates for CH4 recovery and CO2 sequestration
Chemical Engineering Journal (2020)
Effects of soil moisture content on CO2 triggered soil physicochemical properties in a near-surface environment
Journal of Soils and Sediments (2020)