Abstract. The analysis of the geologic record has revealed a question concerning how the Late Ordovician glaciation could have occurred simultaneously with ...
PALEOCEANOGRAPHY,
VOL. 14, NO. 4, PAGES 542-558, AUGUST 1999
Late Ordovician glaciationunder high atmosphericCO2: A coupled model analysis PascaleF. Poussart,Andrew J. Weaver, and ChristopherR. Barnes Schoolof Earth and OceanSciences,Universityof Victoria, Victoria, BritishColumbia,Canada
Abstract. The analysisof the geologicrecordhas revealeda questionconcerning how the Late Ordovicianglaciation couldhave occurredsimultaneously with high CO2 levels (10-18x). Sensitivity studiesusing a coupledatmosphereocean-sea ice modelshowthat it is possibleto maintaina permanentsnow cover(which corresponds to 60% of all the glacialdepositsfoundon Gondwana)under10x CO2 levels,warmfall/coolspringorbitalparameters, a 4.5% reductionin solar luminosity, a length of day of 21.5 hours, and an enhancedsnow/seaice albedoof 0.3. A cold summer orbit experimentwith 10x CO2 anda reducedsnow/seaice albedoof 0.1 alsosustainsa permanent(albeit less extensive)snow cover. The geographicconfigurationof the Late Ordovicianresultsin an up to -42% increasein the global ocean polewardheat transportin the SouthernHemisphererelativeto present-day anda significantasymmetryrelative to the equator.
1. Introduction
Analysis of the geological, chemical, and paleontological records of the Ordovician System has revealed a question concerning how the Late Ordovician glaciation could have occurred simultaneously with high atmospheric CO2 concentrations. The continental glaciation on Gondwanais now a well-documentedeventby the work of Beuf et al. [ 1971], Hambrey and Harland [1981], and others. In terms of its extent it is the third most important glacial period in the Phanerozoic (past 545 m.y.) and coincides with the second greatestmassextinctionevent for that time [Sepkoski,1995]. The chronologyand timing of the glaciation remain issues of contention. Frakes et al. [1992] argued from the spatial and temporal distribution of continental glacial deposits and glaciomarine sedimentsthat glacial inception was initiated during the Caradoc(464 - 443 Ma) and reachedits maximum during the Ashgill, more specifically during the Hirnantian Stage (-440 Ma). The major ice sheet then coveredat least North-central and West Africa, which was centered near the
SouthPole. This is also supportedby the stableisotoperecord [Middletonet al., 1991; Wanget al., 1997; Veizeret al., 1999] and eustaticcurves[Rossand Ross, 1992]. Event stratigraphy conductedin northwest Africa suggests three to four major growthphasesof the continentalice cap from the late Caradoc through the Ashgill [Barnes, 1986]. Whereasthe ice sheet probablybeganto retreatduringthe Llandovery(-438 Ma) and had disappearedcompletelyby the end of the Wenlock (-421 Ma), smaller glacial centersdevelopedin southernAfrica and southern Brazil, apparently following the motion of Gondwanaover the South Pole [Caputo and Crowell, 1985; Caputo, 1998]. On the basis of carbonandoxygen isotopic data derivedfrom an Ordovician-Silurian boundary interval, as well as bathymetricevidence,Brenchleyet al. [1995] arguedfor
Copyright 1999 by the AmericanGeophysicalUnion. Paper number 1999PA900021. 0883-8305/99/1999PA900021512.00
542
a short-livedglaciation(0.5 m.y.). Additionalsupportfor this alternativehypothesishas comefrom the precisedating of NorthAfrican glacialdepositsof Hirnantianageby Paris et al. [ 1995], Marshall et al. [1997], and Underwoodet al. [1997]. The spatialdistributionof LateOrdovicianglacial deposits on Gondwanasuggests thatland-basedice spreadoutwardto at least 40øS. The floating ice may have extended10ø farther equatorward [Brenchley and Newall, 1984]. Estimatesof the
sizeof theicesheet varyfrom6-8x 106km2[Beufetal., 1971; Hambrey,1985]to 11.8 x 106km2 whenaccounting for some erosion [Crowley and Baum, 1991]. Evidence for high atmosphericCO2 has stemmedmainly from the study of paleosols and geochemical modeling experiments. The analysis of goethites from an ironstone in the Upper Ordovician Neda Formation and potentially contemporaneous to the glaciationhas yielded an atmospheric pro 2 estimate of-20x preindustriallevels (pil = 280 ppm) [Yapp and Poths, 1992]. These estimates are also supported by geochemical modeling experiments that predicted atmospheric pro 2 levels -14 + 6x pil [Berner, 1994]. However, Berner's estimates may not be simultaneous to a glaciation confined to the Hirnantian since the model has a time step of 10 m.y. In addition, as pointed out by Kasting [1992], the large uncertainties associated with geochemical estimatesof pCO 2 challengesthe existenceof a CO2-glaciation paradox. The carbon isotope compilations provided by Brenchleyet al. [1995] and Paris et al. [1995] reveal a distinct positive shift at the end of the Ordovician, which suggestsa significant change in the mode of carbon cycling. The widespreaddistribution of this shift in the latest Ordovician suggeststhat it was of global significanceand that it may have had a common origin. Wang et al. [1993] proposedthat a decreasein pCO 2 in the ocean or in the atmosphere, together with a greater organic carbonburial rate during the Hirnantian, might have causedthe positive shift. Higher than usual organic carbon burial rates may be expected during glacial times. If CO2 levels wer-erelatively low for a short period of time simultaneousto the glacial maxima, as surmizedfrom the positive shift of the carbon isotope [Wang et al., 1993;
POUSSARTET AL.: ORDOVICIAN MODELING
Brenchleyet al., 1995], the climatic conditions that led to the glacial inception may still have includedCO: concentrations 8-20x pil. The occurrence of the Late Ordovician glaciation simultaneous with high CO2 levels has been addressed previouslythroughseveralmodeling analyses. Crowleyet al. [1987] showed through energy balance model (EBM) experiments that the unique geographic configuration of Gondwana(with the South Pole in a coastal location) could
543
section4 addresses the sensitivity of the coupledmodel to changesin atmospheric CO: (section4.1), orbitalparameters (section4.2), and sea-ice/snowalbedofeedback(section4.3); section 5 provides a discussionof the results; and section 6 containsconcludingremarks.
2. The Coupled Model
The atmospheric component of the coupled model is the EMBM developedby Fanning and Weaver [1996]. The model includes a representation of the thermodynamic and levels. Subsequent experiments,which includednonlinear icealbedo feedbacks,gave further support to the geographic hydrological componentsof the climate system and is based positioninghypothesis[Hyde et al., 1990; Crowley and Baum, on the vertically integrated energy-moistureequations, with advectionparameterized usingan eddy-diffusiveapproach.The 1991]. However, becausethe timescales involved in tectonic EMBM assumesno heat or water storage on land, so that processes are incompatible with a short-lived Hirnantian glaciation, it is evident that the particular geographic surfacelatent and heat fluxes are zero there. Precipitation configuration of Gondwanais a necessarybut not sufficient occurswhenever the relative humidity is >85%, falling as conditionfor the growthanddecayof the Gondwanaice cap. snow whenever the surface air temperature(SAT) is below
have allowedfor permanentsnow cover with higher CO2
Crowleyand Baum [1991] foundthat by prescribinga 3.55.0% decreasein solarluminosity[Endaland Sofia, 1981] and a realistic Ordovician geography, the EBM sustained permanentice when driven by a 7-13x CO2 forcing. As an extension to their previous analysis, Crowley and Baum [1995] used the Global Environmental and Ecological Simulation of Interactive Systems (GENESIS)atmospheric general circulation model (AGCM) 1.02A with a specified mixedlayer slab ocean [Thompsonand Pollard, 1995], CO: levels 14x pil, -4.5% solar luminosity, and an orbital configuration prescribed for minimum summer insolation receipt(cold summerorbit (CSO)). They focusedon the effects of differentcombinationsof ocean heat transport (Ix versus 1.5x meridionally symmetrizedheat transport based on a present-daycontrol experiment)andtopography(300-500 m) on high-latitudesnow cover over Gondwana. They showed that in the elevatedtopographycase(500 m), permanentsnow cover was found where the geological record indicated the Ordovicianice sheetwas present.Gibbset al. [1997] examined the sensitivity of the Late Ordovician climate to variations in atmospheric CO: using the GENESIS AGCM and found that a "runawayicehouse"occurredwith 8x CO2 levels, whereasfor the 10x experiment, a snow cover coincident with the
-10øC.
Over land, snowfall is redistributed in order to cover
the entire grid cell, whereasover sea-ice a fractional area is permitted. An ice-albedofeedbackis includedwheneversea-ice is present and whenever the SAT over land is below -10øC through the local reduction of the latitudinal profile of the planetarycoalbedoby 0.1 [Graves et al., 1993] as in Fanning and Weaver [1996] and Weaver et al. [1998]. When atmospherictemperaturesrise above -10øC, the snow melt rate
linearlyincreases from0 to 5 cm d'• at 0øC. The EMBM was coupledto the GeophysicalFluid Dynamics Laboratorymodularoceanmodel [Pacanowski,1995] through latent,
sensible, and radiative heat transfers at the air-sea
interface (the ocean component is discussedin detail by Weaverand Hughes[1996]). A simplifiedthermodynamicseaice model based on Semtner [1976] and Hibler [1979] was also includedand is detailed by Fanning and Weaver [1996]. Over
sea-ice,surfacefluxesdueto longwaveand shortwaveradiative
forcing,aswellas sensibleandlatentheatexchanges, were balanced by theconduction of heatthrough theiceandthe heat absorbed because of ice melt.
The advantages gainedfromusingthe EMBM arisefrom its simplicity, computational efficiency, and freedom from explicit
flux adjustments.
The later allows for
the
of pastclimates differingfromthepresent-day. distributionof glacial depositssurvivedthroughthe summer. investigation Coupled modelexperiments providean adequate representation There was no permanentsnowcoverin the 18x experiment. of the present-dayclimate and comparefavorably with Here we conduct a series of sensitivity studies with an energy/moisturebalancemodel (EMBM)coupled to an ocean general circulation model (OGCM)and to a thermodynamic sea-icemodel. The work presentedhereis complementaryto previousAGCM experimentsas we now emphasizethe ocean's role in influencingthe Late Ordovicianclimate. Advantages stemmingfrom the use of the coupledOGCM-EMBM includea directestimateof oceanheat transportsas well as the inclusion of ocean-atmosphere feedbacks,which wouldbe ignoredin an ocean-onlymodel. The use of the computationallyefficient EMBM also allows for the assessmentof a wide range of parameters,which have not been previously addressed.These include the effect of orbital configurations intermediate betweenhot andcold summerorbits and the sensitivity to the magnitudeof the sea-ice/snowfeedbackparameter. The remainderof this paperis structuredas follows: section 2 provides an overview of the coupled model; section 3 presents the Late Ordovician control climatic response;
paleoclimatic data [Fanningand Weaver,1997; Weaveret al., 1998]. However,the simpleparameterization of the diffusive moisturetransportcausesglobal precipitationratesto be too
low compared to present-day observations.Precipitation patternsalsofail to reproduce the belt of high precipitation associatedwith the IntertropicalConvergenceZone. While the EMBM has numerous deficiencies [see Fanningand Weaver,1996],it is a usefultool for process-oriented coupled ocean-atmospherestudies such as the one conductedhere. In
the following subsectionswe describethe deviation of the
presentversionof this coupled modelfromthat described by Fanningand Weaver[1996]andWeaveret al. [1998]. 2.1.
Land-Sea
Distribution
The land-seadistribution is derivedfrom the one usedby CrowleyandBaum [1995]. Since there was no reconstruction availablefor the Late Ordovician, Crowley and Baum modified
544
POUSSART ET AL.: ORDOVICIAN MODELING Annual
mean
surface
currents
::....:.:::.?-::-...:: :::::::::::::::::::::::::::::::::: ::::::::::::::::::::::::::: :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: 5•:..:t•.½:'•5?:ZL'::X:L .&½'•..•>:5: L..•5: :.•: ............•;•.:•:-:.:•½•-•
6O N
•
• •
%•••
.....
'"•"'•:"':•*''•
..............
•
Ocean
::::::::::::::::::::::::: '
-
'
6O S
150 W
0
150 E
Longitude
30 cms-1
Figure 1. Annualmeansurfacevelocities(cm s") obtainedfrom Ordoviciancontrolexperiment.Ovals representocean gyres:(1) northPanthalassic convergence, (2) southPanthalassic convergence, (3) southPaleo-Tethys convergence,and (4) northPaleo-Tethysconvergence.Roundedrectanglesrepresentsurfacecurrents:Panthalassic CircumpolarCurrent(PCC), IapetusCurrent(IC), NorthandSouthEquatorialCurrents (NEC, andSEC),andAntarctica Current(AC). Vectorsare drawn every grid point for legibility.
the Scotese and Golonka [1992] reconstruction of the Silurian (Wenlock epoch, -433 Ma) by rotating the landmasses5ø to the north. Gondwana's position was then tangential to the
South Pole (Figure 1). For this study a linear interpolationwas conductedon the original mask to correspondto the model's 3.75 ø zonal
resolution
and
1.8555 ø meridional
resolution.
From lack of geological evidence a flat-bottom topography was prescribed,with constantdepth of 5400 m. 2.2.
Earth's
Estimates
Rotation
of the rotational
Rate
deceleration
of the Earth
have
et al., 1989]. As such, we increased the Earth's rotation rate from 7.29xl 0 '5 to 8.12x 10-5 rads '•. Orbital
[1989], changes in the Earth-moon distance, .the Earth's rotation rate, and its moment of inertia during the Phanerozoic could have induced -10% variations in precessional periods and-20% variations in obliquity periods. However, the
magnitudeof the changein forcingis unknown. Hencethe two extreme orbital configurations found for the Pleistocene were used [Berger, 1978]. 2.4.
been obtained from the paleontological analysis of stromatolites [Scrutton, 1978] and tidal rhythmites [Sonett et al., 1996; Williams et al., 1997]. The transfer of angular momentum to the moon's orbit causes the major axis of the moon's orbit to increaseat the expense of the Earth's rotation rate. This has resultedin an appreciableincrease in the length of day (LOD) throughgeologictime [Sonettet al., 1988]. Here the Early Silurian LOD estimateof 21.5 hourswas used[Berger
2.3.
2a). According to calculations conductedby Berger et al.
Atmospheric
CO2
Experimentswere conductedusing CO2 valuesof 10x, 14x, and 18x pil. These valuesare representativeof the range of CO2 estimates. The CO2 radiative forcing is calculatedas follows:
F=5.77x1031n( C02! k 350.0 )
This logarithmic relationship reflects the saturation of differentCO2absorptionbandsat higher concentrationsand
givesa radiative forcingof 4 W m'2for a doubling of CO2 [Ramanathan et al., 1987]anda climatesensitivityof 0.75øC W-•m• [Weaveret al., 1998].
Parameters
The sensitivity of the coupledmodel to orbital parameters was investigated through a series of experiments using two extreme and two intermediate configurations. These configurations are defined as follows: a "hot summer orbit" (HSO) with an obliquityof 24.5ø, an eccentricityof 0.06, and a longitude of perihelion of 270ø (Figure 2b), a CSO (22.0 ø, 0.06, and 90ø; Figure 2c), a "warm fall-cool spring orbit" (WFCS) (22.0ø, 0.06, and 0ø), and a "cool fall-warm spring orbit" (CFWS) (22.0 ø, 0.06, and 180ø). For comparison, the present-dayconfigurationis (23.47ø, 0.0167, and 282ø; Figure
2.5.
Solar Luminosity
Studieson solar evolution have suggesteda 25-40% increasein luminosityover the Earth'shistory [Newmanand Rood, 1977; Endal and Sofia, 1981; Bahcall, 1988]. Late. Ordovicianestimatesof solarluminosityimply a 3.5-5.0% decrease. These estimatesdependon the definition of the original components of the Sun'score [Crowleyand Baum,
1991]. Here,weuseda 4.5%decrease assuggested by Crowley and Baum [1995] 1.306x 103Wm '2.
and hence a
solar
constant of
POUSSART ETAL.:ORDOVICIAN MODELING
545
Present-day configuration (NH) 1
(NH)
Summersolstice(NH) "•:::.•i• •"_ •!i":•/•a nter)
June 21
Aphelion• _
•Perihelion '•Winter solstice (NH) - Dec. 21 i(ob1.=23.5')
'Fall)
(
Autumnal equinox (NH) Hot Summer Orbit (SH)
(Fall)
nalequinox (SH)
,••Summer) /•
Winter solstice(SH)- Aphelion
:.• ß
June27
•
X•(obl.=24.5')
•
•.'•Perihelion .Summer solstice (SH) -Dec. 27
•=270
(Spring)
Vernalequinox(SH)
C
ColdSummer Orbit(SH) Vernal equinox
•
nter)
,/
p_.
Summer solstic½(SH) -Aphelion,
.
Penhehon Wtnter
solstice (8H)June 13
Dec. 13
(Fa•l)
(Summer)
equinox(SH) Autumnal
Figure 2. Sketch oftheorbital configurations used inthemodeling experiments: (a)present-day orbit (PDO), (b)hotsummer orbit (HSO) fortheSouthern Hemisphere, (c)coldsummer orbit (CSO) fortheSouthern Hemisphere. Thelongitude ofthe perihelion (A)isrelative tothevernal equinox intheNorthern Hemisphere. Theeccentricity oforbit isnottoscale.
2.6.
for the shorterLOD of 21.5 hoursas follows:
Topography
Thereare currentlyno detailedtopographicmaps of the OrdovicianPeriod. Hencea flat topographyon the continents
f•a )=0.898(u•, (u.,,F,)=-Zø(u,, ,v,, , where Usandvs (Uuandvu ) represent thescaled (unscaled)
(at sealevel)wasprescribed. Thisis probablya realisticfirstcomponents of thewindspeed andfpdandfo arethe presentorder approximationsince duringthe Middle Ordovician, day and Ordovician Coriolis parameters. Thesewindspeeds extensive epicontinental seas were covering parts of were used in the calculation of latent and sensible heat fluxes Gondwana, Laurentia, Baltica, and Siberia.
In addition,
because riverdrainagebasinsareunknownfor thistime period, we defineda river maskin which rain falling and snow melt on
landis equallyredistributed to all coastallandpoints. 2.7.
Wind
Fields
and convertedto wind stress fields directly to drive the ocean.
As suggested by Jenkinset al. [1993], the position of the Hadley/Ferrel/Polarcells would have been dramatically differentduringthe Precambrianwhen the LOD was 14 hours. Ourgeostrophicmodificationof Gibbs windsis a good firstorder approximationfor small perturbationsin LOD from
For this study we usedmonthly averagedwind fields present. previously generated withtheGENESIS AGCM[Gibbset al., 2.8. Changes in Albedo 1997]. The wind fieldi wereobtainedusing an Ordovician
geography[Crowley and Baum, 1995], 10x or 18x atmosphericCO2 levels, CSO, 4.5% decreasein solar luminosity,andpresent-day LOD. For a completedescription of the global climatemodelGENESIS,see Thompson and Pollard [1995]. Fr0_m___..the geostrophic balancethe windspeed
The natureof the Earth'ssurface(vegetationcover,snow-ice
cover,etc.)stronglyinfluences 'theamount of solarradiation absorbed or reflectedat the surface.Hencethe albedoplays an
important rolein theenergy balance.Themodelpartially accountsfor this mechanismby allowing the planetary
fieldswerescaledeverywhereoutside10øNand10øSto account coalbedoto decreasewheneversnow/ice surfacesare present.
546
POUSSART ET AL.: ORDOVICIAN MODELING
No parameterizationis includedfor changesin albedo over
in this paper, as well as a description of their boundary
land becauseof variations in vegetation cover and soil types
conditions, are summarized in Table 1.
The modeledsurfacecirculationpatterns (Figure 1) compare or to the presenceof epicontinentalseas. The Ordovician Periodoccurredprior to the developmentof land plants, and favorably with the description provided by Wilde [1991]. knowledgeof the geographiclocationof lakesis negligible. Because of the absence of continental masses in the Northern The sensitivity of the coupledmodel to changesin the Hemisphere, the wind stress field between 30ø and 60øN magnitude of theice-snowalbedofeedback wastestedby using induces a strictly zonal west-to-east surface circulation. threepossiblelocalreductions to the planetarycoalbedo(0.1, Panthalassic waters confined to this latitudinal belt travel at
0.2, and 0.3) when sea ice/snowwas present.The solar zenith
speedscomparableto present-dayvaluescharacteristicof the
Current(average = 24 cm s'l), as they angleeffect,which is taken into accountin the present-day AntarcticCircumpolar estimateof 0.1 [Graveset al., 1993], dependson the obliquity circle the globe unimpeded (PanthalassaCircumpolarCurrent). of the Earth.This parameteris likely to havevariedthroughout The east-to-westsurfacecurrentsadjacentto the North Pole are the OrdovicianPeriod. In addition,the present-daycoalbedo weak as they are under persistent sea-ice. The presence of
reductionis probablyan underestimate for glaciatedclimates Laurentia, Baltica, Siberia, and northern Gondwana results in becausethere is a large differencein albedo betweenfresh the formation of meridional flows at equatorial latitudes. Strong westward flowing equatorial currents straddling the (0.75-0.95) and old (0.40-0.60) snow [Houghton, 1985]. equator exist between Laurentia and northern Gondwana. Laurentia and Siberia act as meridional barriers as they divert the North Equatorial Current northward and force the flow to All experimentswere conductedfrom resting state and join the North Paleo-Tethysocean gyre. The South Equatorial integrated to equilibrium, whichis definedasthetimewhenthe Current travels freely across the PanthalassicOcean until it oceanmeanheatfluxhasdropped below10'2W m-2. Initial reachesthe Australian-Antarcticshield and becomespart of the conditions for the ocean model included a simplified sea westernboundary Antarctica Current as it travels toward the surfacetemperature (SST)distributionfromLevitusand Boyer South Pole. The second segment of the South Equatorial [1994] and globallyconstantsalinity. The initial atmosphere Currentarisesfrom the SouthChina region and flows westward was0øC everywhereand containedno moisture. until the Avalonia-Baltica shieldblocks it. Two major oceanic high-pressuregyres are present in the southernmidlatitudes: the south Panthalassicgyre and the south Paleo-Tethys gyre. 3. Modeled Late Ordovician Control The two systems are linked through the IapetusCurrent. The Climatic Response modeledannual mean sea surfacesalinity distribution (Figure We startour analysisby presentingresultsfrom the control 3) revealsthe highestsalinity valueswithin 10ø-30ø north and experimentin whichthe orbital parameters(CFWS)describea southof the equator,where most of the evaporationtakes place midpointconfigurationbetweenthe HSO and CSO while the (not shown). Salinity values am minimum at polar latitudes obliquity is kept to a minimum. This setting providesan where melting of sea-ice becomesimportant. The SAT over Gondwanaremains above freezing during the appropriaterepresentation of the averageorbital forcingthat would occur on timescales representative of the Late monthof January(Figure 4b), whereasthe spatial distribution Ordovicianglaciation. Resultsfrom all experimentsdiscussed of freezingatmospherictemperaturesduring the month of July 2.9.
Initial
Conditions
Table 1. Late Ordovician Modeling Results Experiments
Int.
To
Ta
E(SH)
A,(SH)
A,(NH)
Aso
A>O.I
A>O.01
1.4 1.4
12.6 12.6
20.4 16.2
0.00 0.17
0.00 0.11
0.00 0.12
10.6 6.7
4.13 14.5
3000 3000 3000 4000 3000 4000
12.6 12.3 14.3 14.5 15.3 15.4
2645 139 0 260 0 210
17.3 21.4 18.6 14.5 16.2 12.3
0.54 0.00 0.00 0.00 0.00 0.00
1.45 13.70 0.41 0.00 0.00 0.00 0.00 0.00
3.88 14.30 0.42 0.00 0.00 0.00 0.00 0.00
lx_pdo.1
4000
10.5
2900
7.26 17.70 2.90 0.65 0.00 0.65 0.00 0.39 1.45
23.9 59.4
10x_cso.1 10x_hso.1 14x_cfws.1 14x_wfcs.1 18x_cfws.1 18x_wfcs.1
1.1 0.6 1.4 1.4 1.7 1.7 1.7 1.8 1.2
470 2530 73000 640500
0.93 2.83
10x_wfcs.2(aj0.2) 10x_wfcs.3(aj0.3)
3000 4000 2100 2100
Control(10x) 10x_wfcs.1(aj0.1)
45.7
.........
The name of each experimenthas three components: the first refers to the atmosphericCO2 concentration(relative to 280 ppm),the secondis the orbitalconfiguration, and the third represents the magnitudeof the ice-snowalbedoparameter. Control(10x) is equivalentto 10x_cfws.1. Int. representsthe integrationtime (years) required to reach equilibrium. All
experiments exceptfor the lx_pdo.1 havea prescribed 21.5 hourlengthof day (LOD) and-4.5% solarluminosity.To is the depth-averaged globaloceanequilibrium temperature (degrees Celsius);Ta is theannualmeanequilibrium SAT (degrees Celsius); V,(SH)represents the Januarymeansea-icevolumein the Southern Hemisphere (cubickilometers), Ai(SH ) is the January averagedsea-iceareain the SouthernHemisphere,andA,(NH) is the Julyaveragedsea-icearea in the NorthernHemisphere
(X106km2).Asais theJanuary averaged snowareaonGondwana andA>O.1 andA>O.01 are snowareasonGondwana with thickness greater than0.1and0.01m, respectively (x106km2).
POUSSART ET AL.: ORDOVICIAN
MODELING
547
Global distributionof the annual mean sea surface salinity
60 N
34.75 _
.--------- 34.50
_--35.50 ______••___._•'('(E•T-,' •
---35.50 • '• -
•
•'--• 35.50 •
35.25
•
35.00
/I/
•
"-
35.50 -• L•
35ß0½ -
60 S
o• 150 W
0
150 E
Longitude
Figure3. Globaldistribution of theannual meanseasurface salinityforcontrol experiment in partsperthousand.
a
C
January ssT
July SST
b
d
January SAT
July SAT
Figure 4. ModeledJanuaryand July sea surfacetemperature(SST) and surfaceair temperature(SAT) for control experiment (southpolarorthographic projection).Asterisks represent anapproximate distribution of glacialdeposits [Pariset al., 1995]. Contourintervalsare4øC.The heavylineovertheoceanfor SSTgraphsrepresents the maximumsea-iceextent, whereasoverGondwanait represents the locationof the-10.0øC contouror the snowextent(whicheverextendsthe farthest
to lowlatitudes). Theiceedgedelimitsthemaximum snow/sea-ice-covered areapresent duringthemonths of January or July. Shadingindicatestemperatures below0øC.
548
POUSSART ET AL.' ORDOVICIAN MODELING Global Poleward Heat Transport' 10X
5 _
_
_
o
-5
9os
6os
5os
o
50N
90N
60N
Latitude
Global Ocean Poleward Heat Transport' 10X /
\
/
\
iI
1
,,-'"",, II /
0
..
F F •
'..
/
'•'" '•"•.--
/
•
,
x
t t
/
•
t': •. / ",
:' ."
-1
-2
_
,
,
_
_
_
_
_
_
_
_
_
_
-4
-
90S
,
•
I
60S
I
I
30S
r
0
•
I
50N
•
•
I
60N
i
i
9( N
Latitude
Figure5. (a) Annualmeanpoleward planetary heattransport in Petawatts (PW) withocean(dashed line)andatmosphere (dotted)components fortheOrdovician controlexperiment, and(b) globaloceanpolewardheattransport for the Ordovician controlexperiment in PW. The threecurvesrepresent total(solid),advective(dottedline), and diffusive(dashedline) transports (1 PW=10•w).
fully covers glacial deposits found on Gondwana(Figure 4d). SSTs cooler than -1.8øC, as depictedby the ice edge (Figures 4a and 4c), are located in the vicinity of the South Pole and coincide with the majority of the coastal glacial deposits. A permanent sea-ice cover exists in this control experiment, although permanent snow cover is not supported over
where Qin and Qout are the zonally and annually averaged incomingshortwaveand outgoinglongwaveradiation,a is the Earth'sradius,and Osis the latitudeof the southernboundary. The atmosphericheat transportis obtainedby subtractingthe
oceanheattransport fromTpl(CP).
In the Northern Hemispherethe atmosphericcontribution dominatesthe planetary transport at most latitudes, with a The energy balance requirementsimply that at equilibrium maximum of-4.5 Petawatts (PW)near 30øN (Figure 5a). The the annual and global mean incoming shortwave radiation oceanheat transportis important in the SouthernHemisphere balancesthe outgoinglongwave radiation. The planetary heat and at midlatitudes in the Northern Hemisphere. The sole transport (Tpl(CP)), averaged across a line of latitude0, is then region for which the oceandominates(with a maximumof Gondwana (Table 1).
calculated
-2.1
as follows:
½)=
PW at 10øS)is
at low latitudes in the Southern
Hemisphere.Theselow latitudesare regions of maximumtopto-bottom ocean potential temperature differences. In addition, the presenceof a meridionalland barrier allows for
POUSSART ET AL.: ORDOVICIAN MODELING
a
Planetary HeatTransport
2
gos
60S
30S
, , , , , ..... 0
30N
60N
gON
Lotitude
Atmosphere Poleward Heat Tronsport
i •
i
I
i
_
0 -2
-4
-6 ,
90S
,
I
60S
I
30S
I
0
I
•
•
30N
t
60N
,
,
90N
Lotitude
Global Oceon Poleward Heat Transport
-2
90S
60S
30S
0
30N
60N
90N
Lotitude
Figure 6. Heat transport comparisonbetween the present-day run (dashed line) and the Ordovician control run (solid line) in PW: (a) planetaryheat transport,(b) atmospherepoleward heat transport,and (c) globaloceanpolewardheattransport.
bulk of the heat transport is done by advection (Figure 5b, dottedline) via the Antarctica Current. Between the equator and 25øN the net oceanic heat transport is small as the northward advective Ekman transport counters the southward subsurfacediffusive transport. As one moves northward of 25øN, most poleward heat transport is done by diffusion (Figure 5b, dashedline), whereassouthwardtransport is done by advective Ekman transport. For comparison purposes a "present-day"experiment was conductedusing present-day geography, CO2 levels, orbital parameters, solar luminosity, ice-snow albedo feedbacks, and LOD [WiebeandWeaver, 1999]. The planetary heat transport peaks in midlatitudeswith a maximum of -5.0 PW in the Northern Hemisphere for both the present-dayand Ordovician control experiments (Figure 6a). In the SouthernHemisphere the present-daypeak transport(-5.5 PW) is slightly less than the Ordovician control experiment transport (-6.0 PW, Figure 6a). The atmospheric and oceanic components of the planetary heat transportfor the present-dayrun (Figures 6b and 6c) agree favorably with observations [Hartmann, 1994], although differences are apparent. Because of the coarse resolution of the coupled model, the maximum ocean heat transportsobtained from the present-dayexperiment (-0.8 PW in Northern Hemisphereand ~1.4 PW in SouthernHemisphere) are significantly smaller than that derived from observations [Fanning and Weaver, 1996]. The global oceanpolewardheat transportincreasedby up to -42% in the SouthernHemispherebetween the present-dayrun and Ordoviciancontrol experimentand was shifted polewardin the Northern Hemisphere. Previous modeling experiments with the GENESIS AGCM coupled to a mixed layer slab ocean used a prescribed zonally symmetric ocean heat transport [Crowley and Baum, 1995; Gibbs et al., 1997]. The two ocean heat transport schemesused were based on estimated presentday values (with a maximum transport of -0.45 PW) and on a 50% increase (with a maximum transport of -0.68 PW) to accountfor the potential effect of increasedatmospheric CO2 [Crowley, 1993]. The ocean heat transport values used by Crowleyand Baum [1995] were significantly smaller than the valuescomputedby the OGCM usedin this study. In addition, the poleward heat transport averaged at each latitude used in Crowley and Baum [1995] was symmetrized and unchanged from the present-day run. The much different geographic configuration characteristicof the Late Ordovician resultedin a spatial redistribution of the heat transport and a significant asymmetryrelative to the equator(Figure 6c). These findings question the validity of the use of symmetric (about the equator) prescribed ocean heat transports in previous uncoupledatmosphere/mixedlayer oceanmodeling studies.
4. Coupled Model Responseto Ordovician Radiative Forcing 4.1.
the existenceof an intensewesternboundarycurrentflowing alongeasternGondwana,the AntarcticaCurrent(Figure1). The asymmetry of the total ocean heat transport curve indicates a more significant transport in the Southern Hemispherethan in its northern counterpart. In addition, the
549
Sensitivity
to Atmospheric
CO•
The sensitivity of the coupled model to changes in atmospheric CO2 levels within the Late Ordovician context was investigated by conducting three experiments with concentrations10x (control), 14x (14x_cfws.1), and 18x pil (18x_cfws.1). These values span the range of uncertainty associatedwith Berner's [1994] geochemicalestimates.
550
POUSSART ETAL.:ORDOVICIAN MODELING Annual mean surface air temperature difference: 18x - 10x CO2 I
....
'i
60 N
,
'• '' I
,
'
2.5
6O S
I
150 W
0
150 E
Longitude
Annualmean surfaceair temperaturedifference:HSO- CSO (10x)
ß:::{:i.!:::: :•;. :•½'::!:: !'-: •:i:!•-::: :;:::::::::::::::::::::::::::::::: - -.!!.':i :::!.i--::'!:•i:':ii-:::: •:!. :!ii;::-!!::: :•:!i•::-:'...:•i: ••;•:.-'•!i!:::i:•i!' :!?:!: ::::'::::::'::'0. '"
.-
_•---o.s
1.0 -----------.--1.0 ----------1.0
150 W
0 Longitude
150 E
Annualmeansurfaceairtemperature difference: aj03- aj01 ::.i ............... !-.:-..-.i..:: '.-;'.;;:-.:.:::.:'-.; -• ' '.-..""..-'":""'.'.'.".:'.'.: ....... .:.'.' .'.'.. :'.'::':!:' . ..... : .. ;-'-:-' ' . .......
: ....... :-' •........... •--ß
60 N
6O S
150 W
0
150 E
Longitude
Figure7. Annual meanSAT(degrees Celsius) difference for(a) 18x- 10xCO2,(b) hotsummer orbit- coldsummer orbit, and(c)aj0.3- aj0.1.Contour intervals areforFigures 7aand7b0.5øC and1.0øC forFigure 7c. Theheavylineoverthe
oceanrepresents themaximumsea-iceextent,whereasoverGondwana it represents the locationof the-10.0øCcontouror
thesnowextent(whichever extends thefarthest to lowlatitudes). Theice edgedelimits themaximum snow/sea-ice-covered areapresent duringa I yearperiod.Shading indicates temperatures below0øC.
Thespatialdistribution of theannualmeanSAT response to changesin atmosphericCO2 (18x-10x, Figure7a) indicates that atmospheric temperatures are quitesensitiveto variations in CO2forcing.Theminimumwarmingis foundin the tropics (-2.5øC), whereas the maximum warming occurs over Gondwana (~5øC) and in the polar latitudes of both hemispheres.The amplificationof the differenceat high
latitudes is due to snow and sea-ice albedo feedbacks. The
presenceof a large zonal belt of increasedatmospheric temperatures bordering the south coast of Gondwana is
attributed to the contraction of the Southern Hemisphere seaice cover. Boththe insulatingeffectof sea-iceandthe albedo. feedbacksassociatedwith its reflective surfacecontribute in
thedampening of the response.A similarresponse is visible
POUSSART ET AL.: ORDOVICIAN
in the Northern Hemispherealong the ice edge. The large a 90N temperatureincreaseover Gondwanais connectedto snowalbedo-temperature feedbacks. The effect of atmospheric COt on seasonality is 60N investigatedthrough a latitude versustime distributionof the differencein atmospherictemperaturebetweenthe 18x and 10x 3ON experiments (Figure 8a). The seasonal cycle within the
coupledmodelis drivenby changesin solarinsolationandby monthlyaveraged winds. The atmospheric thermalresponse is
.•
MODELING
551
Atmospherictemperature difference (18x-10x)
0
accentuated in the polar regionsduringthe winter seasonwhen ice-albedofeedbackdifferencewould be greatest(Figure 8a). In the Southern Hemisphere, maximum warming is achievedin mid-August,which representsthe end of the australwinter and also the time when the ice albedo difference is largest. The annualglobal mean SAT increasedby 2.7øC between the 10x and 18x experiments,whereasthe globally averaged oceantemperature warmedby 0.4øC(Table 1). When compared to an experimentusing present-dayradiative forcing, orbital
MAR
MAY
JUL
SEP
NOV
Months
forcing, and length ofday, these represent anincrease of2.1 ob
Atmospherictemperature difference (HSO - CSO)
and 4.8øC for the 10x and 18x experiments, respectively (Table 2). The globally averagedoceantemperaturesimilarly
shows anincrease of0.2øand0.5øC relative tothepresent-day experimentfor the 10x and 18x experiments,respectively. The snowcover relative to changesin COt concentrations is assessedthrough the evaluation of the January average snow
cover onGondwana andsea-ice extent.Themonth of January.•
o
is analyzedbecauseit coincideswith the end of the austral summer. We assumethat snow presentduring that period has
survived through themeltseasonandis likely to become part of a permanentice sheet. Januarysnowcoveron Gondwanadid not survivefor any of the CFWS CO2 sensitivity experiments conductedin this study (Table 1). However, a parallel set of
CO2 sensitivity experiments wasperformed by changingonly the orbital forcing to a WFCS configuration (Table 1' 10x_wfcs.1, 14x_wfcs.1, and 18x_wfcs.1). Although a
OAk
of Gondwana.
MAY
JUL
SEP
NOV
Months
permanentsnow cover did not subsistfor the 14x and 18x C 90N WFCS experiments,the 10x_wfcs.1 did have a small area of snow (0.17 x 106 kin2), which survivedin the southern latitudes
MAR
Atmospherictemperaturedifference (aj05 - aj01)
60N
Estimates of Southern Hemisphere sea-ice extent represent
January averages, whereas for the Northern Hemisphere, estimatesrepresentJuly averagesas they coincidewith the end of the australand boreal summer,respectively. Sea-iceextent
3ON
showsa strongdependence on atmospheric COtlevels. For the CFWS experiments,SouthernHemispheresea-iceonly survived for the 10x scenario with a January average area of
3OS
0.93 x 106 km2 (Table 1). However,for the WFCS CO2 sensitivityexperiments,Januarysea-icecover in the Southern
Hemisphere survived for the10x(2.83 x 106kin2),14x (0.65 x 106km2),and18x(0.39x 106km2) cases.A largeNorthern Hemispheresea-iceextentis presentin all experiments(CFWS and WFCS) and reachesa maximumextentof 20.4 x 106km2 in the 10x_cfwscontrol experiment(Table 1). This sea-icecover extends down to ~60øN.
60S
90S
JAN
MAR
MAY
JUL
SEP
NOV
Months
Figure 8. Latitudeversustime distribution of the differencein zonal meanSAT (degreesCelsius)between(a) 18x- 10x, (b) hot summer
There are145melting days peryear forthe10xexperimen t, 0.5øC orbit= cold summer orbit, and (c)aj03- ajO. t. Contour intervals are for Figure8a, 1.0øCfor Figure8b, and2.0øCfor Figure8c.
whereasfor the 18x experiment, melting takes place for
170d y-•(Table3). Assuming a maximum meltrateof 5 cm d-l, this implies that for an ice cap to develop on Gondwanathe theseamountsof snowduringthe winter months,precipitation minimum winter snow depth needsto be > 7.3 m for the 10x ratesshouldnear3.7 (10x) and4.9 m yr'• (18x). Present-day (rain equivalent= 2.2 m) and8.5 m for the 18x (rain equivalent precipitation ratesrarely exceed-5 m yr'• exceptalong = 2.6 m) in order to exceed the summer melt. To accumulate mountain ranges or coastal areas. It is unlikely that
552
POUSSART
ET AL.: ORDOVICIAN
MODELING
Table 2. Climate Sensitivity Parameters
CO2Relative to pil,ppm
AForcing, W m-2 ACO 2 -4.5%sl
NetForcing, W m'2
ATaRelative to lx_pdo. 1,øC
280 (Ix) 2800 (10x) 3920 (14x)
-1.3 12.0 13.9
-10.8 -10.8 -10.8
-12.1 1.2 3.1
...... 2.1 3.8
5040 (18x)
15.4
-10.8
4.6
4.8
AToRelative to lx_pdo. 1,øC 0.2 0.5 0.5
For the lx_pdo.1 experimentwe usedpresent-dayradiativeforcing, orbital forcing, and lengthof day and prescribed a LateOrdoviciangeography; pil is preindustrial levels,andsl is solarluminosity.
insolationreceipt is not alteredsignificantly betweenthe two configurations. The theory of the ice ages relies on the idea that summer
precipitation rates would be this high in the interior of Gondwana.
The
maximum difference in
atmospheric
temperatures occursduringthe australwinter when sea-iceand snowcoversare at their maximum extent (up to 8.5øC, Figure 8a). The annualmeanpolartemperature for the 10x is -14.7øC and is-10.0øC for the 18x experiment(Table 3). 4.2.
Model
Sensitivity
to Orbital
temperaturesplay the critical role in glacial inception. Whereas high-latitude winter temperaturesare always cold
Parameters
We now presentan analysisof the sensitivityof the coupled model to changes in orbital parameters within the Late Ordovician
context.
To this
end we consider
two
extreme
orbital configurations CSO and HSO. The annual mean insolationreceipt at the top of the atmospheredependson the eccentricityand hencedoes not vary between the CSO and the HSO experiments. However, the latitudinal profile of the zonally averagedannual insolation shows marked differences between the two configurations. While polar latitudes receive
anannualsurplus of solarinsolation by asmuchas 16.5 W m'2
enough to ensure precipitation to fall as snow, summer temperatures can be warm enough so as to prevent the development of a permanent snow cover. In fact, Crowell [1978] argued that summer temperatures play a more significant role in glacial inception than the supply of
moistureø Model results show that the intensity of the seasonalcycle is accentuatedin the HSO experiment and diminishedin the CSO experiment(Figure8b). An analysisof the seasonalcycle provides a means for determining the criticalsummertemperaturefor whichglacial inception is still possible. When summer temperaturesare too hot, the developmentof a permanentice cap is not possible. October polar SATs are -9.5øcooler for the HSO experiment, whereas
FebruarySATs are-7.5øC warmerfor the SouthernHemisphere (Figure 8b). Although temperatureextremesare greater for the CSO(upto 4 W m-2). The climaticresponse to the spatial HSO configuration, the annually averagedpolar temperatureis variationsin insolation(Figure 7b) showsthat the annual SAT -IøC cooler for the CSO becauseof the presence of a field strongly depends on the solar insolation forcing. permanent snow cover (Table 3). In addition, summer Equatorialand tropical SATs are -0.5øC cooler, whereaspolar temperatures never rise above freezing for the CSO temperaturesare-IøC warmer for the HSO configuration. experiment. Following the argumentpresentedby Crowleyet Maximum warming occurs in polar latitudes of both al. [1987], with annual precipitationratesas low as 10 cm and hemisphereswhere the differencein annual solar radiation is below lYeezing summer temperatures,a 1000 m thick ice sheet the largest. Although changesin the orbital forcing influence could be producedwithin the relatively short time of 10,000 the spatial distribution of atmospheric temperatures, results years. On the other hand, extremely high snow precipitation show no major changeson annuallyaveragedglobal ocean and duringthe winter seasonwouldnot survivethrougha long and intense melt season. atmospherictemperatures(Table 1). The absenceof a global thermal response is explained by the fact that the annual The modeled permanent snow/sea-ice cover, relative to for the HSO compared to CSO, the equatorial and tropical latitudes receive less insolation in the HSO comparedto the
Table
3.
Climatic Indicatorsfor Glacial Inception on the South Pole
Experiment
Annually Averaged Polar Temperature, øC
Lengthof Melt Season,days
10x
-14.7
18x
-10.0
145 170
Aj0.1
-15.0
135
Aj0.3
-31.6
HSO
-13.4
150
0
CSO
-14.7
145
Experimentnamesrepresentthe following runs' 10x, control; 18x, 18x_cfws.l' Aj0.1, 10x_wfcs.1; Aj0.3, 10x_wfcs.3;hot summerorbit (HSO), 10x_hso.l' and cold summerorbit (CSO), 10x_cso.1.
POUSSARTET AL.: ORDOVICIAN MODELING
553
Annualmean sea surfacetemperature(aj01)
a 6O N
6O S
150 W
0
150 E
Longitude
Annual mean sea surfacetemperature(aj03)
b 6O N
o
6O S
150 W
0
150 E
Longitude
Annual mean sea surfacetemperaturedifference:aj03 - aj01 -
i
i
i
60 N
.•.
0
.... •--•• '•-'1
60 S
150 W
0 Longitude
150 E
Figure9. AnnualmeanSST(degrees Celsius)distributions for (a)'aj0.1and(b) aj0.3experiments. Contourintervalsare 5øC. The heavy line over the oceanrepresentsthe maximum sea-iceextent, whereas over Gondwana it representsthe location of the -10.0øC contouror the snow extent (whicheverextendsthe farthestto low latitudes).The ice edge delimitsthe maximum snow/sea-ice-coveredarea presentduring a 1 year period. (c) Annual mean SST (degrees Celsius) difference between the aj0.3 and aj0.1 experiments. The heavy line is the ice edge for the aj0.3 experiment, and the contour intervals are IøC. Shadingindicatestemperatures below 0øC in Figures9a and 9b.
from0.65 x 106km2forthe HSOto changes in orbitalforcing,reflected thedecrease in incoming (Januaryaverage)increased solar radiation at high latitudesfor the CSO experiment and 2.90 x 106 km2 for the CSOconfiguration.Therewasno resulted in an increase of the snow cover and sea-ice extent.
The spatial increaseof highly reflective surfacesproduceda furtherdecreasein absorbedsolar radiation, resulting in a furthercooling. This explains why the thermalresponsewas amplified at high latitudes through this positive feedback loop. The extent of sea-ice in the southern high latitudes
permanentsnow cover for the HSO experiment,whereasfor the CSO the modeled snow cover was 0.54 x 106 km 2. The extent
of the NorthernHemispheresea-icecover was maximumfor the
HSO experiment(21.4 x 106 km2) as the configuration corresponds to a CSO configuration for the Northern Hemisphere.
554
POUSSART ETAL.:ORDOVICIAN MODELING
(-
0
8
o
8
,•
õ
8
o
8
se•lePl•
leAe'l
8
leAel
8
õ
8
8
8 sa•lelhlu! qldeo
8
8
POUSSART ET AL.: ORDOVICIAN MODELING Albedo jump = 0.1
a ..ß
.. ß
.
ß
:
..
.. .
.
ß
Albedo jump = 0.2
b ß
..
ß
ß
.
.
:.:
555
et al., 1983; Graves et al., 1993]. Results show a strong response of the climate model to changes in the ice-snow albedo parameter (Table 1). The annual global mean ocean temperaturedecreasedby 0.8øC and the SAT decreasedby 5.9øC between the aj0.1 and aj0.3 experiments. The cooling occursthroughout the year and is enhancedat high latitudes where snow/ice-covered surfaces are present (Figure 8c). Maximum cooling in the southern high latitudes takes place during the winter months(May-June-July) and summermonths (November-December-January)when differences in snow/icecovered areas are the largest and solar insolation reaches its extremes (Figure 8c). SATs remain below -10.0øC (the temperature above which melting occurs in the model) throughout the year for the aj0.3 experiment, whereas the aj0.1 experimenthas a melting seasonof 135 days (Table 3). The averagedtemperatureduring the melt seasonfor the aj0.1 experiment is -2.2øC, whereas it is -3.6øC for the CSO experiment. The fact that summertemperaturesare slightly warmerfor the aj0.1 experiment than for the CSO may in part explain why the snow and sea-ice accumulation is more important in the latter. The ice edge in the northern high latitudes expanded 10ø equatorwardbetween the aj0.1 and aj0.3 experiments because of the enhanced snow/ice albedo effect (Figures 9a and 9b, heavy line). The difference in the annualmean SST between the
aj0.3 andaj0.1 experiments(Figure9c) showsa coolingover all latitudesexceptfor the NorthernHemispherehigh latitudes where permanentsea-ice is present in both experiments. EquatorialandsubtropicalSSTscooledby -2øC. The largest
...
cooling is found along the southeasterncoast of Gondwana.
.. ß
ß ß
.. ß
.,
%-.'
ß
.•.
ß .
..,
This region, which was warmedby the westernboundary AntarcticaCurrentin the aj0.1 experiment,is permanently coveredby sea-icein the aj0.3 experiment. The oceancirculationalso showsa strong sensitivity to
changes in the ice-snow albedo parameter(Figure 10).
Albedo jump = 0.3
Experiments reveal an equatorwardshift of the locations of
deepwaterformationwith a higher ice-snowalbedoparameter (Figures10a and 10c). In addition,the overturningshowsa markeddeepeningand strengtheningbelow 2000 m with an increasedice-snowalbedoparameter. However,a comparison of the global ocean heat transport between the two sensitivities (Figure 10d) showsa decreasein the poleward heat transportin the SouthernHemisphereand north of 35øN
-..
with an increasedice-snow albedoparameter. In the Southern Hemisphere,where there exists a meridional land barrier, the overturning is the most dominant mechanism for oceanic heat transport. In general, ocean heat transport is calculated
followingthe temperature differencebetweenpolewardsurface
ß
ß ß ß
ß
ß .
waters and equatorwardflowing subsurfacewaters.
It is
possible that the observed reduction in heat transport originates from the reduction of this vertical temperature Figure ll. Annuallyaveragedice/snowconcentrationover sea/land, respectively, for ice-snowalbedoparameters of (a) 0.1, (b) 0.2, and (c) 0.3. Contourintervalsare 0.3, and asterisksrepresentan approximate distribution of glacialdeposits[Paris et al., 1995].
gradient(as surmizedfrom Figures9a and9b). The sea-iceextentadjacentto the South Pole increasedfrom
2.83x 106km2in theaj0.1runto 17.70x l06km2in theaj0.3 run, whereasthe January avesagesnow cover on Gondwana
increased from0.17 x 106to 14.50x 106km2 (Table1). This 4.3.
Model
Feedback
Sensitivity
to Sea-ice/Snow Albedo
Parameter
Sensitivityof the coupledmodel to changesin the ice-snow albedofeedbackparameterwas testedfor three possible local reductionsto the planetary coalbedo(0.1, 0.2, and 0.3) [North
value alone,excludingthe sea-icearea,represents123% of the estimatedice sheet. Clearly, the resultsfrom the increasedicesnow albedo parameter support the notion of sustained
glaciated conditions in the southern high latitudes of Gondwanaunder 10x atmosphericCO2 concentrations. In terms of annually averaged sea-ice and land snow
556
POUSSARTET AL.' ORDOVICIAN MODELING
concentrations (Figure 11) the area inclusiveof all glacial deposits is snowcovered60% of the yearfor aj0.3,whereasfor the aj0.1 experimentthis areais only fully covered30% of the
to that calculatedfrom Ramanathan[1987]. As this question
year.
is clearly still unresolved,further studiesare neededin order to
5. Discussion
improve the radiative estimates. The latitudinal planetary albedoprofile may also have differedfrom the present-day
There is no direct evidencefor a Northern Hemisphereice cap contemporary to the Late Ordovician glaciation [Brenchley, 1988], partly since no cratonic areaswere located in the northern high latitudes. However, because of the absence of land masses in the Northern Hemisphere, the modeled meridional circulation is very weak thereby enhancing the isolation of the northern polar latitudes from the equatorialsourceof heat. Consequently,results from all sensitivity experiments consistently reveal the presence of permanentsea-icecover extendingfrom the pole to 60ø-70øN.
[1995] estimatewas basedon Kiehl and Dickinson [1987] and
represents, for the 10xCO2case,a -3.0 W m'2increase relative
estimate because of changes in land-sea distribution, land surface cover, and cloud cover [Jenkins et al., 1993]. At presentthe albedocurvesgeneratedthroughOrdovicianAGCM experiments account for land-sea effects but do not include particular land surface cover effects such as variations in soil typesand presenceof large water masseson the continents. In addition,they do not accountfor potential changesin cloud coverdue to the shorterlengthof day or differentgeography. 6.
Conclusion
A minimum sea-ice extent (at the end of boreal summer) is
Results from the modeling experiments show that it is possible to maintain a substantial permanentsnow cover in 106 km2, whereasa maximumis achievedin the 10x_wfcs.3 the southernhigh latitudes of Gondwanagiven 10x CO2 experiment with a valueof 59.4 x 106 km2. For comparison, concentrations, WFCS forcing, a 4.5% reduction in solar present-day Southern Oceanmaximumsea-iceareais -15 x 106 luminosity, a length of day of 21.5 hours, and an ice-snow km2 [Hartmann, 1994]. The presenceof an extensive albedoparameterof 0.3. The single most important internal permanent sea-ice cover must have affected the biological parameter,which determineswhether a permanentsnow cover production in polar latitudes, but the lack of land in these existed or not, is the magnitude of the ice-snow albedo latitudesrendersthe assessment of its implicationdifficult. feedback.In addition,the CSO experimentwith 10x CO2, a Geochemical isotopic excursions derived from the decreased solarluminosityand shortened lengthof day, andan Ordoviciangeologic recordsuggestan arrayof possible time ice-snowalbedoparameterof 0.1 also sustaineda permanent framesfor the glaciation, the shortestone being -1 m.y. (albeit less extensive) snow cover. These findings reinforce Clearly, if the glaciation was indeed brief, several orbital conclusionspreviously derived from AGCM experimentsas configurations could still have been involved during the their permanent snow cover is sustained under similar length of the event. Hence our use of an intermediate boundary conditions. configurationto examine the averageclimate producedunder The geographicconfigurationof the Late Ordovician, which the specifiedsetof parametersis justified. It is also possible was drasticallydifferentthan that of present-day,resultedin an that an extreme orbital configurationcould have played a up to-42% increase in the global ocean poleward heat preconditioning role for the creation of a glacial climate transport in the Southern Hemisphere, a latitudinal nearing the end of the Ordovician. redistributionof the ocean heat transport, and a significant As previously argued by Crowell [1978], summer asymmetry relative to the equator. Future Ordovician AGCM temperatures may play a morecritical role in glacial inception modelingstudiesshouldthereforeincludeoceanheat transport thanthe supplyof moisture.Modelresultsshowedthat values directly calculated by an OGCM, such as the ones whereashigh-latitude winter temperatureswere always cold obtained in this study. In light of the restraints and enough to ensure precipitation to fall as snow, summer uncertaintiesassociatedwith this studythe coupledmodel was temperatures were in some instances, such as for the HSO, able to producea permanent snow cover on Gondwana,which warm enoughso as to preventthe developmentof a permanent included 60% of all the glacial deposits found on the snow cover. Perhaps sustained ice albedo feedbacks, supercontinent. achievedin the 18x_wfcs.1 experiment with a value of 12.3 x
concurrent with the existence of an ice sheet, would be
sufficient to inhibit the termination of the glaciation when Acknowledgments.We thankT. Crowleyand S. Baumfor providing orbital forcing would enter an HSO configuration. The the land-sea mask and M. Gibbs for the wind fields used in the inclusionof a thermomechanical ice sheetmodelto a coupled experiments.We are gratefulto them as well as M. Eby and A. Fanning OGCM-EMBM would allow this to be tested further by for discussionsand suggestionslinked to this work. C. Poulsen,M. avoidingthe shortcomingsof the presentlyusedsimple snow Gibbs, and an anonymousreviewer are thanked for their constructive parameterization. There are a number of uncertainties
associated
with
this
study. The CO2/radiativeeffect relationship usedhere is based on present-day estimates and may not be applicable for concentrationsas high as 10-18x pil. Crowley and Baum's
reviews. This researchwas supportedby an FCAR Fellowshipawarded to PFPandby NSERC, CSHD, AES/CICS, and Steacie operatinggrants, as well as an IBM SUR grant awarded to AJW. All numerical computationswere conducted locally on a suite of IBM RS6000s including two SP2s. Infrastructure support from the University of Victoria is gratefully acknowledged.
POUSSART ET AL.: ORDOVICIAN
MODELING
557
References Bahcall,
J.
N.,
Solar
models,
neutrino
experiments, and helioseismology, Rev. Mod. Phys., 60, 297-372, 1988. Barnes, C. R., The faunal extinction event
near the Ordovician-Silurian boundary' A climatically induced crisis, in Global BioEvents,editedby O. Walliser, pp. 121-126, Springer-Verlag,New York, 1986. Berger, A. L., Long-term variation of caloric insolationresultingfrom the Earth's orbital elements,Quat. Res., 9, 139-167, 1978. Berger, A. L., M. F. Loutre and V. Dehant, Influence of the changing lunar orbit on the astronomical frequencies of preQuaternary insolation patterns, Paleoceanography,4, 555-564, 1989. Berner, R. A., 3GEOCARB
model
of
II:
A
revised
atmospheric CO2
over
Fanning,A. F. and A. J. Weaver, Temporal- Paris, F., Z. Elaouad-Debbaj, J. C. Jaglin, D. Massa and L. Oulebsir, Chitinozoans and geographicalmeltwater influences on the North Atlantic conveyor: Implications for Late Ordovician glacial events on the Younger Dryas, Paleoceanography, Gondwana, in Ordovician Odyssey: Short 12, 307-320, 1997. Papersfor the 7th International Symposium Frakes, L. A., J. E. Francis and J. I. Syktes, on the Ordovician System,edited by J. D. ClimateModesof the Phanerozoic,274 pp., Cooper,M. L. Droserand S. L. Finney, pp. CambridgeUniv. Press,New York, 1992. 171-176, Pacific Section Society for Gibbs, M. T., E. J. Barron and L. R. Kump, An Sedimentary Geology, Fullerton, Calif., 1995. atmospheric pCO: thresholdfor glaciation Ramanathan, V., The role of Earth radiation in the Late Ordovician, Geology, 25, 447450, 1997. budget studies in climate and general Graves, C. E., W. Lee and G. R. North, New circulation research,J. Geophys.Res., 92, 4075-4095, 1987. parameterizations and sensitivities for Ramanathan, V., et al., Climate-chemical simple climate models, J. Geophys. Res., 98, 5025-5036, 1993. interactions and effects of changing Hambrey, M. J., The Late Ordovician-Early atmospherictrace gases, Rev. Geophys., 25, 1441-1482, 1987. Silurian Glacial Period, Palaeogeogr.,
Phanerozoic time, Am. J. Sci., 294, 56-91,
Palaeoclimatol.,Palaeoecol.,
1994.
1985.
Beuf, S., B. Biju-Duval, O. de Carpal, P. Rognon,O. Gariel and A. Bennacef, Les gr•s du Pa16ozoiqueinf6rieur au Sahara: S6dimentation et discontinuit6s, 6volution structurale d'un craton, Publ. Inst. Fr. Pet. Tech., 18, 1-464, 1971.
Brenchley,P. J., Environmentalchangesclose to the Ordovician-Silurian boundary, Bull. Br. Mus. Nat. Hist., 43, 377-385, 1988.
Brenchley, P. J. and G. Newall, Late Ordovician environmental changes and their effect on faunas, in Aspects of the OrdovicianSystem,editedby D. L. Bruton, ppo65-79 , Univ. of Oslo, Oslo, Norway, 1984.
Brenchley,P. J., G. A. F. Carden and J. D. Marshall, Environmental changes associated with the "first strike" of the Late
Ordovician mass extinction, Mod.
Geol.,
20, 69-82, 1995.
Caputo,M V., Ordovician-Silurianglaciations and global sea-level changes, in Silurian Cycles: Linkagesof Dynamic Stratigraphy With Atmospheric, Oceanic, and Tectonic Changes,edited by E. Landing and M.E. Johnson,pp. 15-26, New York State Educ. Dep., New York, 1998.
Caputo,M. V. andJ. CoCrowell,Migrationof glacialcentersacrossGondwanaduring Paleozoic Era, Geol. Soc. Am. Bull., 96, 1020-1036, 1985.
51, 273-289,
Ross,J. R. P. and C. A. Ross, Ordovician sea-
level fluctuations, in Global Perspectives Hambrey, M. J. and W. B. Harland, Earth's on Ordovician Geology, edited by B. D. Pre-Pleistocene Glacial Record, 1004 pp., Webby and J. R. Laurie, pp. 327-335 , A. A. Balkema, Brookfield, Vt., 1992. CambridgeUniv. Press,New York, 1981. Scotese, C. R. and J. Golonka, Hartmann,D. L., Global Physical Climatology, PaleogeographicAtlas, Paleomap Project, 411 pp., Academic, San Diego, Calif., 1994. Univ. of Tex., Arlington, 1992. Hibler, W. D., A dynamic thermodynamicsea Scrutton, C. T., Periodic growth features in ice model, J. Phys. Oceanogr., 9, 815-846, fossil organismsand the length of the day 1979.
and the month, in Tidal Friction and the
Houghton,H. G., Physical Meteorology, 442 pp., Mass. Inst. of Technol., Cambridge, 1985.
Earth's Rotation, edited by P. Brosche and J. Sundermann,pp. 154-196 , SpringerVerlag, New York, 1978.
model for the Hyde, W. T.,' K. Kim and T. J. Crowley, On Semtner, A. J., A the relation between polar continentality thermodynamic growth of sea ice in and climate: Studies with a nonlinear numerical investigations of climate, J. seasonal energy balance model, J. Phys.Oceanogr.,6, 379-389, 1976. Geophys.Res.,95, 18,653-18,668,1990. Sepkoski, J., Jr., Patterns of Phanerozoic Jenkins, G. S., H. G. Marshall and W. R. extinction:A perspectivefrom global data Kuhn, Precambrian climate: The effects of land area and Earth's rotation rate, J.
Geophys.Res., 98, 8785-8791, 1993. Kasting,J. F., Paradoxlost and paradox found, Nature, 355, 676-677, 1992. Kiehl, J. T. and R. E. Dickinson, A studyof the radiative effects of enhancedatmospheric
CO: and CH4 on early Earth surface temperatures,J. Geophys. Res., 92, 29912998, 1987.
Levitus, S. and T. P. Boyer, World Ocean Atlas 1994: Temperature, 117 pp., Washington, D.C., NOAA, 1994.
Crowell, J. C., Gondwana glaciations, Marshall,J. D., P. J. Brenchley, P. Mason, G. A. Wolff, R. A. Astini, L. Hints and T. cyclotherms,continentalpositioningand climatic change, Am. J. Sci., 278, 1345-
Meidla, Global carbon isotopic events
1372, 1978.
associated
with
mass
extinction
and
bases,
in
Global
Events
and
Event
Stratigraphyin the Phanerozoic,edited by O. H. Walliser, pp. 35-51 , SpringerVerlag, New York, 1995. Sonett,C. P., S. A. Finney, and C. R. Williams, The
lunar
orbit in the Late
Precambrian
and the Elatina sandstone laminae, Nature, 335, 806-808, 1988. Sonett, C. P., E. P. Kvale, A. Zakharian, M. A. Chan and T. M. Demko, Late Proterozoic and Paleozoic tides, retreat of the moon, and rotation of the Earth, Science, 273, 100-104, 1996.
Thompson,S. L. and D. Pollard, A global climate model (GENESIS) with a LandSurface Transfer Scheme (LSX), I, Present climate simulation, J. Clim., 8, 732-
761, 1995. glaciation in the Late Ordovician, greenhouse effect, Bull. Am. Meteorol. Palaeogeogr.,Palaeoclimatol.,Palaeoecol., Underwood, C. J., S. F. Crowley, J. Do Marshall and P. J. Brenchley, HighSoc., 74, 2363-2373, 1993. 132, 195-210, 1997. Crowley,T. J., J. G. Mengel and D. A. Short, Middleton, P. D., J. D. Marshall and P. J. resolutioncarbon isotope stratigraphyof Brenchley, Evidence for isotopicchange Gondwanaland'sseasonal cycle, Nature, the basal Silurian stratotype(Dob's Linn, 329, 803-807, 1987. associatedwith Late Ordovicianglaciation, Scotland) and its .global correlation, J. Geol. Soc., 154, 709-718, 1997. Crowley, T. J. and S. K. Baum, Toward from brachiopodsand marine cements of central Sweden, in Advances in Ordovician Veizer, J., et al., 87SR/86SR, •5180evolutionof reconciliationof Late Ordovician (-440 Geology,editedby C. R. Barnesand S.H. Phanerozoic seawater, Chem. Geol., in Ma) glaciationwith very highCO2 levels, J. Geophys.Res.,96, 22,597-22,610, 1991. Williams, Pap. Geol. Surv. Can. 90-9, 313press, 1999. 323,1991. Wang, K., B. D. E. Chatterton,M. Attrep and Crowley,T. J. and S. K. Baum,Reconciling C. J. Orth, Late Ordovician mass extinction Newman, M. J. and R. T. Rood, Implications Late Ordovician(440 Ma) glaciationwith of solar evolution for the Earth's early in the Selwyn Basin, northwesternCanada: very high(14X) CO2 levels,J. Geophys. Res., 100, 1093-1101, 1995. atmosphere, Science, 198, 1035-1037, Geochemical, sedimentological, and 1977. Endal, A. S. and S. Sofia, Rotation in solarpaleontological evidence, Can. J. Earth Scien., 30, 1870-1880, 1993. North, G. R., J. G. Mengel and D. A. Short, ty• stars,I, Evolutionarymodelsfor the Simpleenergybalancemodel resolvingthe Wang, K., B. D. E. Chattertonand Y. Wang, spin-downof the sun, Astrophys.Z, 243, 625-640, 1981. seasonsand the continents:Applicationto An organic carbon isotoperecord of Late Fanning, A. F. and A. J. Weaver, An the astronomicaltheory of the ice ages,J. Ordovician to Early Silurian marine Geophys.Res., 88, 6576-6586, 1983. atmospheric energy-moisture balance sedimentaryrocks, Yangtze Sea, South model:Climatology,interpentadalclimate Pacanowski,R., MOM 2 documentation user's China: Implications for CO2 changes guide and reference manual, in technical change,and couplingto an oceangeneral during the Hirnantian glaciation, circulationmodel,J. Geophys.Res., 101, report, pp. 232, Geophys.Fluid Dyn. Lab. Palaeogeogr., Palaeoclimatol.,Palaeoecol., 15,111-15,128, 1996. 132, 147-158, 1997. Ocean Group, Princeton,N.J., 1995.
Crowley, T. J., Geological assessment of the
558
Weaver, A. J., M. Eby, A. F. Fanningand E. C. Wiebe, Simulated iafluence of carbon dioxide, orbital forcing and ice sheetson the climate of the Last Glacial Maximum,
Nature, 394, 847-853, 1998.
POUSSART ET AL.: ORDOVICIAN
MODELING
Wilde, P., Oceanography in the Ordovician,in Advances in Ordovician Geology, edited by C. R. Barnesand S.H. Williams,Pap. Geol. Surv. Can. 90-9, 283-298,1991. Williams, D. M., J. Harkin and A. H. N. Rice, Umbers, ocean crust and the Irish
Weaver, A. J. and T. M. C. Hughes, On the incompatibilityof ocean and atmosphere Caledonides:terrane transpressionand the models and the need for flux adjustments, morphologyof the Laurentian margin, J. Clim. Dyn., 12, 141-170, 1996. Geol. Soc. London, 154, 829-838, 1997. Wiebe, E. C. and A. J. Weaver, On the sensitivityof global warming experiments Yapp, C. J. and H. Poths,Ancient atmospheric CO2 pressures inferred from natural to the parameterisationof sub-grid scale goethites,Nature,355, 342-344, 1992. oceanmixing, Clim. Dyn., in press,1999.
C. R. Barnes, P. F. Poussart,and A. J. Weaver, School of Earth and Ocean Sciences, University of Victoria, P.O. Box 3055, Victoria, British Columbia, Canada V8W 3P6. (crbarnes•uvic.ca;
poussart @ocean.seos.uvic.ca; weaver@ocean. seos.uvic.ca)
(Received December 15, 1998; revised April 1, 1999; acceptedApril 4, 1999.)