GEOPHYSICAL RESEAIO,CH LETTEI•,S. VOl,. 28, NO. 2, PAGES 2813-286,.IANUARY 15,2001

CO2 levelsrequired for deglaciationof a "Near-Snowball"

Thomas J. Crowley, William T. Hyde Dept. of Oceanography,Texas A&M University,College Station,Texas

W. Richard Peltier

Dept. of Physics,University of Toronto,Toronto, Canada

Abstract. Geologic evidence suggeststhat in the Late Hoffman et al. [1988] invoked estimatesby Caldiera and Neoproterozoic,(-600 Ma) almost all land masses were Kasting[1992] that massivelevels of atmosphericCO 2 (--0.i glaciated, with sea-levelglaciation existing even at the bar) wererequired to escapefrom this state,and suchvalues equator. A recentmodeling study has shownthat it is possi- are comparableto thosecalculated by Hydeet al. [2000]. ble to simulate an ice-coveredEarth glaciation with a cou- A furtherresult of the Hyde et al. [2000] study is that pledclimate/ice-sheet model. However,separate general cir- stand-alonegeneral circulation model (GCM) experiments culation model experimentssuggest that a secondsolution with fixed ice sheetsin some casesyielded open water may exist with a substantialarea of ice free ocean in the conditionsin equatorialregions removed from the tropics.Although 0.1 to 0.3 of an atmosphereof CO2 (-•300 [cf. Baumand Crowley,2000]. The existenceof openwater to 1000 X) is requiredfor deglaciationof a "SnowballEarth," solutionsreflects standardlength-scale arguments discussed the "exit"CO 2 levelsfor an openwater solution could be in the climate literature[North, 1984; Manabe and Broccoli, significantly less. In this paper we utilize a coupled 1985], in which the thermalinfluence of an ice sheet(or other climate/ice sheet model to demonstratefour points: (1) the perturbation)diminishes by a scale which is determinedfrom openwater solution can be simulatedin the coupledmodel if characteristicradiative damping and heat transportvalues for the sea ice parameter is adjusted slightly; (2) a major the Earth. GCM sensitivityexperiments [Baum and Crowley, reductionin ice volume from the open water/equatorialice 2000] indicate that this open water solution is obtained for solutionoccurs at a CO2 levelof about4X presentvalues - CO2 valuesof about2.5X presentlevels. This studyalso abouttwo ordersof magnitudeless than required for exit from demonstratedthat a strong negativecloud feedbackplays a the "hard"snowball initial state;(3) additionalCO 2 increases role in offsetting the enhancedalbedo from expanded ice are requiredto get fuller meltbackof the ice;and (4) the open cover, as the decreasedair temperaturesare associatedwith water solution exhibits hysteresisproperties, such that lower water vapor levels [cf. Jenkins, 1993; Chandler and climateswith the same level of CO2 mayevolve into either the Sohl, 2000]. snowball,open water, or a warmerworld solution,with the The existenceof the open water solution has important trajectorydepending on initial conditions.These results set implicationsfor climate theory and the survival of metazoans, usefultargets for geochemicalcalculations of CO2 changes and it would provide a valuable constrainton geochemical associatedwith the open-watersolution. modelsof the late Neoproterozoic. The open water may also leadto significantlylower deglaciation CO 2 thresholdlevels than are necessary for a "hard" snowball solution. In this Introduction paper we use the samecoupled climate/ice sheet model as that The possibilityof a late Neoproterozoic(---800-600 Ma, in Hyde et al. [2000] to conduct a range of sensitivity million ago) "SnowballEarth" has stimulateda great experimentsto determinewhether open water solutionscan be deal of interest. Ice sheetsapparently reached the equatorat obtainedby suchcoupled models and, if so, the CO2 levels thistime [e.g,Hambrey and Hadand, 1985;Eries, 1993], and requiredfor deglaciationfrom sucha state. the first appearanceof multi-celledanimals (metazoans) may havepreceded [e.g, Hofmannet al., 1990;Bromham et al., 1998] the last phaseof the glaciation(-600 Ma). Although Climate Model and Boundary Conditions early modelingwork on this problemfailed to simulateequa- The climate/ice sheet model [Deblonde and Peltier, 1991; torial glaciation,recent studieswith a coupledclimate/ice Deblonde et al., 1991; Tarasov and Peltier, 1997] consistsof sheetmodel [Hyde et aL, 2000] indicatethat a SnowballEarth four submodels,which respectively predict ice flow, mass solutioncan be parsimoniouslyobtained if solar luminosity balance, temperature, and bedrock sinking. The latter is is reducedby the astrophysicallyreasonable value of 6% and assumedto occur with a time constantof 4,000 years [Peltier, CO2 levelsare near present values. Hyde et al. [2000]further 1998]. Ice flows subject to a temperature-independent demonstrated that the existence of the ice-covered Earth rheologybased on the Nye [1959] formulation,and massbal- solution is dependenton inclusionof the ice sheet, because ance is computedaccording to statisticalmodels [Reeh, 1990; the elevation and thermal inertia of the ice-sheet allows it to Huybrechts and T'Siobbel, 1995], which take as input "weather" Milankovitch orbital configurations. monthly temperatures generated from a nonlinear, two- dimensionaldiffusive seasonalEBM [Hyde et al., 1990]. The Copyright2001 by the American Geophysical Union. EBM has been validated against many different GCMs Papernumber 2000GL011836. [Crowley et al., 1991], while the coupled ice sheet model 0094-8276/01/2000GL011836505.00 reproducesmany featuresof both the last glacial cycle and the

283 284 CROWLEYET AL.; CO 2 LEVELS REQUIRED FOR DEGLACIATION OFA NEAR-SNOWBALLEARTH

Carboniferousand Neoproterozoicglaciations [Tarasoy and Ice Sheet Peltier, 1999; Hyde et al., 1999, 2000]. 60' 60' Baselinepaleogeography for the Neoproterozoicis from Dalziel [1997], althoughsensitivity experiments [Hyde et al., 30' 30' 2000] indicate that the principal results are not strongly dependenton detailsof geography.We examineda rangeof elevationsfor the continentsranging from sea level to 500 m. In the extremelycold climatewe studyin this paperour equi- -30' -30'

librium results are relatively insensitiveto this parameter. -60' -60' The model is driven with Milankovitch insolation variations appropriateto the Pleistocene[see for example Hyde et. al 1989]. Other boundaryconditions include a solar luminosity I i I 6% belowpresent [Crowley and Baum,1993], CO 2 levels 1 1000 2000 3000 4000 5000 varying from -•0.5-10X present levels and a variable sea ice Mean Annual Temperature albedo. The sea ice albedo of the EBM in Pleistocene simulations [Deblonde and Peltier, 1991] is 0.5. This value 60' 60' is lower than employed in GCMs and reflects the fact that the 30' 30' EBM specificationis for the total surface-atmospherecolumn. Due to atmosphericreflectance and absorption,the column 0' - - 0' albedois about 10% lower than the surface[see Hyde et al., 1989]. Becausecloudiness decreases in our Neoproterozoic -30' -30' GCM runs[Baum and Crowley,2000], we mimic this effect on -60' -60' absorbedsolar radiationby loweringalbedo to 0.45 over sea ice. The net changesin radiationreceipt from the decreased ,,• ! . ! ! I I . albedo(---25 W/m 2) arestill somewhat less than the changes in I I I I I I I ! surface radiation receipt from a 20% decreasein cloudiness -50 -40 -30 -20 -10 0 10 20 [Baum and Crowley, 2000], but given all the uncertaintiesin Figure 2. Example of model open water solution: (a) the albedo calculations, we consider it a reasonableadjust- distribution of ice sheet (thickness in m)' (b) mean annual ment for the model. Further justification for the sea ice temperature. albedo reduction is basedon the suppositionthat the albedo should be lower at low zenith angles in the tropics [for example, the albedo of water varies from 0.02 to 0.06 as the zenith angle changesfrom 0 to 60 degrees,Peixoto and Oort, northern hemisphere mid/high-latitute land areas. Previous 1992, p. 103]. work [Hyde et al., 1999] indicates only a modest sensitivity Precipitation is highly parametertized;we employ a spa- to variations in this term in very cold climates, and for the tially uniform value of 0.6 mm/day, which declines with work presentedhere we considerthe choice satisfactoryfor an height and temperature. This value is characteristicof present initial examinationof the "openwater" deglaciationproblem. The numerical experimentswere conductedin two ways. In the first class of experimentsthe model was initialized as ice NeoproterozoicIce Age Phase Space free, and run until equilibrium under prescribed input parameters. Most such runs came to equilibrium after "-:•, o i is "- 100,000 years of integration, though some were extended to 10•.. " 4- O'O' .-O...... ' 'O'i-': O O O • +.... + i++ --. ..+ 200,000 years to determine the stability of the resulting climate against Milankovitch forcing. A secondset of runs e !%o .' ...... was started at the end of a 200,000 simulation that E 4o- eo yielded a stable open water solution. In this set of

(D ' : : experimentswe linearlyincreased the CO 2 concentrationfrom • 4e. ..-: years 220,000 to 270,000 and ran the model to equilibrium (•,- ::o.+.o. o . > o .... with the new forcing, generally until about year 350,000. In •' : -" '"' -(9 • i--' : * ..... this way we are able to test the stability of the open water (D .... ,• ...... oO. o solution to changesin forcing. •-- ..: .,... .-o . ..,., .-...... --..._ ...... + ...... '" :" . :" '. o.-' ' ...... "" :' -~' ,. ' ..... -50,,. --"*•" ..:'" '- •.,: "** '* ',-' ... 5 Results

ß '"' 0.75 -10 Sea Ice Albedo IR Forcing A summary of the principal results (Figure l) illustrates mean global temperatureas a function of both sea ice albedo Figure 1. "Phase-space"plot of global sea-level temperature vs. sea ice albedo and infrared radiative forcing perturbations andCO 2 radiativeforcing perturbations. With a seaice albedo of 0.5 or greater, there are two main solutions, one with a (a proxyfor CO2). Differentcolors represent the different basic statesof the model - "incomplete"glaciation (red), fully totally ice-covered planet (blue symbols), the other with ice-covered planet (blue), and ice-covered continents with glaciation extending only to 25-30ø paleolatitude (red). open water (green). Circles representexperiments in which However, as sea ice albedo is lowered slightly, a third average continental elevation is 400 m, crossesexperiments solution emerges (green) that has ice sheets on most of the with elevations at 200 m. continents,including ice at the equator,but also a substantial CROWLEYET AL.; CO2 LEVELSREQUIRED FOR DEGLACIATION OF A NEAR-SNOWBALLEARTH 285

NeoproterozoicIce Age HysteresisLoop ice-free land to occur is on isolated continents at some dis- tance from the main ice sheet mass. If these results can be verified by further work, they hav.e

• 10 significant implications for understandingNeoproterozoic glaciationand carboncycling. The CO2 levelsrequired to E 5 obtain open water deglaciationare about two ordersof magni- tude less than would be requiredfor a "hard"Snowball Earth -• 0 solution. This result suggeststhat the planet remained in a near-snowball state for a significantly shorter time than hypothesizedin Hoffman et al. [1998]. Although it would / only require about 100,000 years for degassing to raise o3 -15 atmosphericCO 2 levelsto this value,the reactiveocean iI carbonatereservoir must first be neutralized. This might take / ¸ •- Snowball Earth on the order of 0.5 million years [L. Derry, personal

-6 -4 -2 0 2 4 6 8 10 12 14 communication,2000]. As ice sheet calving would still be InfraredWarming (W m -2) providing ground carbonateto the ocean, this estimate would be extended by some indeterminate amount. The estimated Figure 3o Hysteresisplot of snowball earth runs. Planetary average temperature decreaseswith infrared forcing. The totaltime for atmosphericCO 2 buildupis in thelower half of upperbranch represents a climate statewithout equatorialice. the range of uncertaintyfor the durationof the Varanger The full snowballearth solutionis shownin the lower right glaciation based on paleomagneticdata [Sohl et al., 1999] corner. The open water solution follows the lower branch of and geochemical calculations [Jacobsen and Kauffman, the hysteresisplot (equivalentto about0.7-3.0X present CO 2 1999]. The requiredadditional rise in CO2 after equatorial levels),with deglacationcommencing at --3.5 - 4X CO2. deglaciationmight be explainable in terms of CO2 released Circles refer to runs startingfrom an ice-free initial state,X's fromcarbonate precipitation in Neoproterozoicpostglacial runsstarting from the open water initial state. caprocks. The simulatedglacial-interglacial fluctuations duringthis lattertime spanare consistentwith evidencefrom Neoproterozoicshelf deposits[Kennedy and Christie-Blick, areaof openwater (Figure 2) in very muchthe samearea as 1998]. simulatedby GCM studies[Baum and Crowley, 2000]. If the hysteresisresponse can be verifiedit maypresent an Simulationsstarting from the open water initial state additionalinteresting target for geochemicalmodel studies to indicateclear hysteresisproperties (Figure 3), whereinthe determinewhether there are any scenarios that might lead to a modeltends to remainin the full ice/openwater solutionat "CO2 attractor"in the •-0.7-3.0XCO2 range,thereby higher CO2 levelsthan would occurif the modelrun was preventingthe Earthfrom reachingthe hardsnowball state, startedfrom a no-ice initial state. The range of radiative with attendantimplications for the survivalof metazoansand forcingvalues for the hysteresiseffect is equivalentto about CO2 levelsrequired for deglaciation. 0.7-3.0Xthe present CO 2 levelin theatmosphere. At 3.5-4.0X present levels extensive deglaciation commencesand ice volumedecreases about an orderof magnitude.The duration of the deglaciation(20,000 - 40,000 years) is not nearly as "Exitfrom Oasis" experiment abrupt (Figure 4) as for the full-snowballresponse (2,000 -3W/m2 +3W/m2 years- notethat the latter did not occur until CO 2 levelswere - , • about two orders of magnitudehigher than in the present 1' runs). A near-instantaneousCO2 increaseyields a similar CO2 Increase time scalefor the deglaciation.Even after majordeglaciation begins ice volumes(Figure 4) are greaterthan Pleistoceneice sheets (about35 X 106km3). A progressiveincrease in radiative forcingto levelsas high as 10X CO2 resultsin a steady E decreasein ice volume, with glacial-interglacialfluctuations +7 W/m2 due to Milankovitch forcing (as the ice volume decreasesthis forcingbecomes more important). lOO

Discussion and Conclusions + 10W/m2 i i i i i i i To summarize,a new set of coupledEBM/ice sheetmodel IntegrationTime (kyr) simulationswith variable sea ice albedo provide additional support for a possible open water solution in a Figure4. Ice volumesfor differentlevels of CO2 values Neoproterozoic ice-covered continent scenario. Because of associatedwith the warmingof theopen water solution. More uncertaintiesin Neoproterozoic continental reconstructions, than 20,000 - 40,000 modelyears are requiredto get the particularland masseswe simulateas ice free may have deglaciationwith a rampincrease in CO 2 lasting50,000 years. The CO2 changesequivalent to the radiativeforcing been in different locationsthan specified. If locationsof the perturbationsare approximately---I.7X (3 W/m2),-•3.5X(•-7 real landmassescould be more confidentlydetermined, such W/m2),---6X(10 W/m2),and--10X (12 W/m2).The first data would enablea crucial test of our open water solutionto 100,000 years of the model run has been truncated becauseit be performed. Our resultssuggest that the mostlikely area for representsspinup time. 286 CROWLEYET AL.; CO 2 LEVELS REQUIRED FOR DEGLACIAT•ON OFA NEAR-SNOWBALLEARTH

Acknowledgments. This researchwas supportedby NSF grants Hyde, W. T., K.-Y. Kim, T. J. Crowley,and G. R. North,On the relation OPP-9615011 and ATM-9817560 and National Sciences and betweenpolar continentality and climate: Studieswith a nonlinear EngineeringResearch Council of Canadagrant A9627. We thank the energybalance model, J. Geophys.Res., Dll, 18653-18668,1990. reviewersand S. Baum, N. Christie-Blick,L. Derry, A. Knoll, and L. Hyde,W. T., T. J. Crowley,L. Tarasov,and W. R. Peltier,The Pangean Sohl for assistance and comments. : Studies with a coupled climate-ice sheet model, Clim. Dynam., 15, 619-629, 1999. Hyde, W. T., T. J. Crowley, S. K. Baum, and W. R. Peltier, References Neoproterozoic"Snowball Earth" simulationswith a coupled climate/icesheet model, Nature, 405, 425-430, 2000. Baum, S. K., and T. J. Crowley, GCM responseto late Precambrian Jacobsen,S. B., and A. J. Kaufman,The Sr, C, and O isotopicevolution (-600 Ma) ice-coveredcontinents, Geophys. Res. Letr (submitted, of Neoproterozoicseawater. Chem. Geol., 161, 37-57, 1999. 2000). Jenkins,G. S., A generalcirculation model study of the effectsof faster Bromham,L., A. Rambaut,R. Fortey,A. Cooper,and D. Penny,Testing rotationrate, enhancedCO 2 concentration,and reducedsolar the Cambrianexplosion hypothesis by using a moleculardating forcing: Implicationsfor the faint-youngsun paradox, J. Geophys. technique,Proc. Natl. Acad.Sci. USA,95, 12386-12389,1998. Res., 98, 20803-20811, 1993. Caldiera, K., and J. F. Kasting, Susceptibilityof the early Earth to Kennedy,M. J., and N. Christie-Blick,Sequence stratigraphy of late irreversibleglaciation caused by carbondioxide, Nature, 359, 226- Neoproterozoicglacial to post-glacialtransition: Implications for the 228, 1992. timingof 513Cexcursions, Soc. Sed. Geol. Conf On Strataand Chandler,M. A., and L. E. Sohl, Climate forcingsand the initiationof Sequenceson Shelvesand Slopes(abstract), 1998. low-latitudeice sheetsduring the NeoproterozoicVaranger glacial Manabe,S., and A. J. Broccoli,The influenceof continentalice sheets interval,d. Geophys.Res., 105, 20737-20750,2000. on the climateof an ice age,d. Geophys.Res., 90, 2167-2190, 1985. CrowIcy,T.J., and S. K. Baum,Effect of decreasedsolar luminosity on North,G. R., The smallice cap instability in diffusiveclimate models, d. late Precambrianice extent, d. Geophys.Res., 98, 16723-16732, Atmos.Sci., 32, 1301-1307, 1984. 1993. Nye,J. F., The motionof ice sheetsand , d. Glaciol.,3, 493-507, Crowley, T. J., S. K. Baum, and W. T. Hyde, Climate model 1959. comparisonsof Gondwananand Laurentideglaciations, ,/. Geophys. Peixoto, J.P. , and A. H. Oort, The Physics of Climate, Amer. Inst. Res., DS, 9217-9226, 1991. Physics,New York, 1992. Dalziel, I. W. D., Overview: Neoproterozoic-Paleozoicgeography and Peltier,W. R., Postglacialvariations in the levelof the sea:Implications tectonics: Review, hypothesis,environmental speculation. Geol. for climatedynamics and solidEarth geophysics, Rev. Geophysics, Soc. Am. Bull., 109, 16-42, 1997. 36, 603-689, 1998. Deblonde,G., and W. R. Peltier, Simulationsof continentalice sheet Reeh,N., Parameterizationof melt rate and surfacetemperature on the growthover the lastglacial-interglacial cycle: Experimentswith a GreenlandIce Sheet,Polarforsch.. 59(3), 113-128,1990. one-level seasonal energy balance model including realistic Sohl, L.E., N. Christie-Blick,and D.V. Kent, Palcomagneticpolarity geography,d. Geophys.Res., 86, 9189-9215,1991. reversalsin Marinoan (600 Ma) glacial depositsof Australia: Deblonde,G., W. R. Peltier,and W. T. Hyde,Simulations of continental Implicationsfor the durationof low-latitudeglaciation in the ice-sheetgrowth over the glacial-interglacialcycle: Experiments Neoproterozoic,Geol. Soc. Amer. Bull., 111, 1120-1139, 1999. with a one level seasonalenergy balance model including seasonal Tarsasov,L., and W. R. Peltier,Terminating the 100 kyr ice age cycle, ice albedofeedback, Glob. Plan. Change,98, 37-55, 1991. d. Geophys.Res., D18, 21665-21693,1997. Eyles,N., Earth'sglacial record and its tectonicsetting, Earth Sci.Rev., Tarasov,L. and W. R. Peltier,The impactof thermomechanicalice- 35, 1-248 (1993). sheetcoupling on a modelof the 100kyr ice-agecycle, d. Geophys. Hambrey,H. A., and W. B. Hadand,The late Proterozoicglacial era, Res., 104, 9517-9545, 1999. Palaeogeog.,Palaeoclim., Palaeoecol., 51, 255-272,1985. Hoffman, P.F., A.J. Kaufman, G.P. Halverson, and D.P. Schrag, A Neoproterozoicsnowball earth. Science,281, 1342-1346,1998. T.J. Crowley, Department of Oceanography, Texas A&M Holmann, H. J., G. M. Narbonne, and J. D. Aitken, Ediacaranremains University, College Station, TX 77843, USA. email: from intertillite beds in northwesternCanada. Geology, 18, 1199- tcrowley•ocean.tamu.edu 1202, 1990. W.T. Hyde, Departmentof Oceanography,Texas A&M University, Huybrechts,P. and S. T'Siobbel, Thermomechanicalmodelling of CollegeStation, TX 77843, USA. email:hyde•rossby.tamu.edu northern hemisphere ice sheets with a two-level mass-balance W.R. Peltier,Department of Physics,University of Toronto,Toronto, paramererization,Ann. Glaciol., 21, 111-116,1995. Ontario,M5S 1A7, Canada.mail: [email protected] Hyde, W. T., T. J. Crowley, K.-Y. Kim, and G. R. North, Comparisonof GCM and energybalance model simulations of seasonaltemperature (ReceivedMay 24, 2000; revisedSeptember 20, 2000; changesover the past 18,000years, d. Clim.,2. 864-887, 1989. acceptedSeptember 20, 2000.)