The millennial atmospheric lifetime of anthropogenic CO2AbstractIntroductionAn informal model intercomparison on the fate of fossil fuel CO2The CO2 peakCO2 partition between the atmosphere and the oceanLong-term buffering by sediments and weatheringImplicationsAvoiding dangerous climate changeSea levelLong-term positive feedbacks in the carbon cycleGeoengineering climateDiscussionA widespread misconceptionAnalogous issuesConclusionReferencesThe millennial atmospheric lifetimeof anthropogenic CO2David Archer & Victor BrovkinReceived: 19 December 2006 / Published online: 4 June 2008#The Author(s) 2008Abstract The notion is pervasive in the climate science community and in the public atlarge that the climate impacts of fossil fuel CO2release will only persist for a few centuries.This conclusion has no basis in theory or models of the atmosphere/ocean carbon cycle,which we review here. The largest fraction of the CO2recovery will take place on timescales of centuries, as CO2invades the ocean, but a significant fraction of the fossil fuelCO2, ranging in published models in the literature from 20–60%, remains airborne for athousand years or longer. Ultimate recovery takes place on time scales of hundreds ofthousands of years, a geologic longevity typically associated in public perceptions withnuclear waste. The glacial/interglacial climate cycles demonstrate that ice sheets and sealevel respond dramatically to millennial-timescale changes in climate forcing. There arealso potential positive feedbacks in the carbon cycle, including methane hydrates in theocean, and peat frozen in permafrost, that are most sensitive to the long tail of the fossil fuelCO2in the atmosphere.1 IntroductionThe ocean contains 50 times more dissolved oxidized carbon than the atmosphere does, and70% of the surface of the earth is covered by ocean. For these reasons, the prevalentopinion among earth scientists in the early twentieth century was that the oceans wouldprevent industrial activity from increasing the pCO2of the atmosphere. This view prevaileduntil precise measurements of free-atmosphere pCO2values showed an increasing trend of(at that time) 0.8 ppm yr−1(Keeling 1961).Climatic Change (2008) 90:283–297DOI 10.1007/s10584-008-9413-1D. Archer (*)Department of the Geophysical Sciences, University of Chicago, 5712 S Ellis, Chicago, IL, USAe-mail: [email protected]. BrovkinPotsdam Institute for Climate Impact Research, P. O. Box 601203, Potsdam, GermanyPresent address:V. BrovkinMax Planck Institute for Meteorology, Bundesstr. 55, 20146 Hamburg, GermanyAt about the same time as the first accurate pCO2measurements, Revelle and Suess(1957) showed that the uptake of CO2into seawater is enhanced by carbonate bufferchemistry, but only one tenth as strongly as might be naively inferred from the relativeconcentrations of carbon in the air and in the water. Most of the carbon dissolved inseawater is in the form of bicarbonate, rather than in the form of carbonate ion, which reactsto buffer CO2invasion. In the model results presented below, we will see that the CO2uptake capacity of the ocean diminishes with increasing CO2release, because of thedepletion of the carbonate ion content of the ocean.To make matters worse, the rate of CO2uptake by the oceans is much slower than might beinferred from the large surface area of the oceans. Only a small area of the oceancommunicates with the largest “pool” of water, the deep sea. Therefore the equilibration timebetween the atmosphere and the ocean is several centuries, much longer than one mightnaively expect by simply looking at a globe, or at a “blue planet” photograph from space.Carbon cycle models respond to a release of new CO2into the atmosphere in a series ofseveral well-defined stages lasting for many millennia (Fig. 1; Archer et al. 1998; Broeckerand Takahashi 1978; Caldeira 1995; Caldeira and Rau 2000; Sundquist 1990; Sundquist1991; Tans and Bakwin 1995; Walker and Kasting 1992). The CO2in the atmospherethrough this time will not consist of the exact same CO2molecules emitted from fossil fuelcombustion, because of the copious exchange of carbon with the ocean and the landsurface. However, the CO2concentration in the air remains higher than it would have been,because of the larger inventory of CO2in the atmosphere/ocean/land carbon cycle.We present a summary of long-term carbon cycle models from the recently publishedliterature (Table 1). Some of the models are more detailed than others, and the models makedifferent assumptions about the responses of the terrestrial biosphere, and the oceancirculation, to climate changes. We also show new results from the CLIMBER model (Fig. 2).The model results are sorted according to the amount of CO2released, into “Moderate”and “Large” CO2slugs. Moderate is 1,000–2,000 gigatons of carbon (Gton C), while Largeis 4,000–5,000 Gton C. For comparison, the IPCC business-as-usual scenario (SRES A1B)0%10%20%30%40%50%60%05000 35000Reaction with Igneous Rocks10000400003000015000Year A.D.2500020000Reaction with CaCO3Ocean InvasionAirborne Fraction of the Total CO2 ReleaseFig. 1 Schematic breakdown of the atmospheric lifetime of fossil fuel CO2into various long-term naturalsinks. Model results from Archer (2005)284 Climatic Change (2008) 90:283–297winds up releasing about 1,600 Gton C to the atmosphere by the year 2100. Business-as-usual to 2100 is generally considered to be a large carbon release, rather than moderate, butit is actually moderate relative to the available coal deposits, which total about 5,000 GtonC (Rogner 1997). Frozen methane hydrate deposits in the ocean may contain another4008001200160020002000 4000 6000 8000 10000 12000024682000 2200 2400 2600 2800 30000102030Yr.AD°C ppm Gton C / yrSRES B1SRES A21000 Gton C5000 Gton CabcFig. 2 A response ofCLIMBER-2 model (Brovkin etal. 2002; Brovkin et al. 2007;Ganopolski et al. 1998)toModerate (1,000 Gton C) andLarge (5,000 Gton C) fossil fuelslugs. The equilibrium climatesensitivity of the model is 2.6°C.Temperatures were smoothedwith a 250 filter to eliminate aspurious fluctuation of Antarcticsea ice caused by the low modelresolution. The land carbon cyclewas neglected in these simula-tions while deep sea sedimentswere explicitly simulated using asediment diagenesis model(Archer 1991). a Emissionsscenarios and reference IPCCSRES scenarios (B1 and A2).b Simulated atmospheric CO2(ppmv). c Simulated changes inglobal annual mean air surfacetemperature