Global change in the Late Devonian: modelling the Frasnian^Famennian short-term carbon isotope excursions Yves Godde ¤ris a; , Michael M. Joachimski b a Laboratoire des Me ¤canismes et Transfert en Ge ¤ologie, UMR 5563, CNRS-Universite ¤ Paul Sabatier-IRD, Toulouse, France b Institute of Geology and Mineralogy, University of Erlangen-Nu «rnberg, Erlangen, Germany Received 7 June 2002; received in revised form 8 August 2003; accepted 12 September 2003 Abstract A model of the global biogeochemical cycles coupled to a energy-balance climate model (the COMBINE model) is used to identify the causes of two large N 13 C value excursions across the Frasnian^Famennian (F-F) boundary. We test a scenario that links the sea-level rise to stratification of the Proto-Tethys ocean through the formation of warm saline deep waters in extended epicontinental seas. Even though this scenario can produce dysoxia below 100 m depth, it fails to increase the global burial flux of organic carbon and thus seawater N 13 C values, since stratification of the ocean leads to decreased productivity in surface waters. Several scenarios postulating a continental origin of the perturbations in the Late Devonian biogeochemical cycles are then tested. We found that weathering of platform carbonates exposed during the Early Famennian sea-level fall can account for a maximum positive shift in N 13 C value of +0.7x at the end of the sea-level fall episode. Another +1.0x increase in N 13 C might originate from rapid spreading of vascular land plants near the F-F boundary, postulating that higher plants globally increased the weatherability of continental surface, and that colonized continental area increased by 30% across the F-F boundary. Finally, the N 13 C excursion observed at the base of Upper rhenana Zone and the rapid increase of the carbon isotope ratios at the F-F boundary require an increase of phosphorus delivery to the ocean by 40%, coeval with the sea-level rises. Once the calculated N 13 C values are in agreement with the measured data, the COMBINE model calculates a decrease in atmospheric pCO 2 from pre-perturbation 2925 ppmv in the Lower rhenana conodont Zone to 1560 ppmv in the Upper triangularis Zone. This decrease in pCO 2 is due to the increase in burial of organic matter during the Kellwasser events, and increased continental weatherability triggered by the spreading of continental vascular plants. These changes occur within 4 million years. The corresponding global climatic cooling reaches 4.4‡C at the pole, and 2.1‡C at the equator. ȣ 2003 Elsevier B.V. All rights reserved. Keywords: modelling; carbon dioxide; seawater; isotope ratios; Frasnian; Famennian 1. Introduction The Late Devonian faunal crisis represents one of the most prominent mass extinction events in 0031-0182 / 03 / $ ^ see front matter ȣ 2003 Elsevier B.V. All rights reserved. doi :10.1016/S0031-0182(03)00641-2 * Corresponding author. Tel.: +33-5-6155-6841; Fax: +33-5-61-55-61-13. E-mail addresses: [email protected](Y. Godde ¤ris), [email protected](M.M. Joachimski). Palaeogeography, Palaeoclimatology, Palaeoecology 202 (2004) 309^329 www.elsevier.com/locate/palaeo
Transcript
Global change in the Late Devonian:modelling the Frasnian^Famennian short-term
carbon isotope excursions
Yves Godde¤ris a;�, Michael M. Joachimski b
a Laboratoire des Me¤canismes et Transfert en Ge¤ologie, UMR 5563, CNRS-Universite¤ Paul Sabatier-IRD, Toulouse, Franceb Institute of Geology and Mineralogy, University of Erlangen-Nu«rnberg, Erlangen, Germany
Received 7 June 2002; received in revised form 8 August 2003; accepted 12 September 2003
Abstract
A model of the global biogeochemical cycles coupled to a energy-balance climate model (the COMBINE model)is used to identify the causes of two large N
13C value excursions across the Frasnian^Famennian (F-F) boundary. Wetest a scenario that links the sea-level rise to stratification of the Proto-Tethys ocean through the formation of warmsaline deep waters in extended epicontinental seas. Even though this scenario can produce dysoxia below 100 m depth,it fails to increase the global burial flux of organic carbon and thus seawater N
13C values, since stratification of theocean leads to decreased productivity in surface waters. Several scenarios postulating a continental origin of theperturbations in the Late Devonian biogeochemical cycles are then tested. We found that weathering of platformcarbonates exposed during the Early Famennian sea-level fall can account for a maximum positive shift in N
13C valueof +0.7x at the end of the sea-level fall episode. Another +1.0x increase in N
13C might originate from rapidspreading of vascular land plants near the F-F boundary, postulating that higher plants globally increased theweatherability of continental surface, and that colonized continental area increased by 30% across the F-F boundary.Finally, the N
13C excursion observed at the base of Upper rhenana Zone and the rapid increase of the carbon isotoperatios at the F-F boundary require an increase of phosphorus delivery to the ocean by 40%, coeval with the sea-levelrises. Once the calculated N
13C values are in agreement with the measured data, the COMBINE model calculates adecrease in atmospheric pCO2 from pre-perturbation 2925 ppmv in the Lower rhenana conodont Zone to 1560 ppmvin the Upper triangularis Zone. This decrease in pCO2 is due to the increase in burial of organic matter during theKellwasser events, and increased continental weatherability triggered by the spreading of continental vascular plants.These changes occur within 4 million years. The corresponding global climatic cooling reaches 4.4‡C at the pole, and2.1‡C at the equator.F 2003 Elsevier B.V. All rights reserved.
Earth history, a¡ecting preferentially low-latitudetropical shallow water ecosystems. This crisislasted 1^3 million years with a major extinctionpulse at the Frasnian^Famennian (F-F) bound-ary. The causes of this mass extinction havebeen widely discussed in the literature (for a re-view see Hallam and Wignall, 1997). Based onvarious qualitative scenarios, the extinction hasbeen attributed to global anoxia (Buggisch,1991; Joachimski and Buggisch, 1993; Murphyet al., 2000), climate warming (Thompson andNewton, 1988; Ormiston and Oglesby, 1995), cli-mate cooling (Copper, 1977, 1986; Joachimskiand Buggisch, 2002), the spreading of land plants(Algeo et al., 1995), intensi¢ed tectonic activity(Racki, 1998), and single or multiple bolide im-pacts (McLaren, 1970; McGhee, 2001).
Coincident with this crisis are two periods ofenhanced sedimentation of organic matter (Kell-wasser horizons) that accumulated under dysoxicto anoxic conditions (Joachimski and Buggisch,1993). Both periods of enhanced organic carbonburial correspond to major positive excursions of+3x in global surface water N
13C values (Joa-chimski et al., 2002). The ¢rst excursion coincideswith a minor short-term rise in sea level (Fig. 1).The onset of the second excursion is observedwithin the latest Frasnian and also correlateswith a sea-level rise. However, a prominent sea-level fall is observed in the earliest Famennianwith the N
13C record still displaying high valuesuntil the base of the Upper triangularis conodontZone, (Joachimski and Buggisch, 1993; Joachim-ski et al., 2001, 2002).
Fig. 1. Carbonate N13C value record measured across the F-F boundary in the Steinbruch Schmidt (open circles) and Benner
(squares) sections (Joachimski and Buggisch, 1993) in comparison to changes in sea level (Johnson et al., 1985). Lower andUpper Kellwasser horizons are represented by the gray bars. Minor di¡erence in the timing of the lower carbon isotope excur-sions is based on the assumption of constant sedimentation rate.
The duration of the positive carbon isotopeanomalies is estimated using the Devonian timescale of Tucker et al. (1998) and assumptions con-cerning the relative duration of individual cono-dont zones from Sandberg and Ziegler (1996).The durations of the Upper Palmatolepis rhenana,Palmatolepis linguiformis and Lower Palmatolepistriangularis conodont Zones are estimated to be850, 360 and 730 kyr, respectively. Knowing thethickness of sediments deposited during the re-spective conodont zones and assuming constantsedimentation rates, the durations of the N
13C val-ue anomalies are estimated to be around 0.5 Myr(base of the Upper rhenana Zone) and 1.2^1.8Myr (F-F transition). Given the response timeof carbon within the ocean^atmosphere system(V200 000 years, FrancPois and Godde¤ris, 1998),such big and short N13C value excursions probablycorrespond to major perturbations of the carbongeochemical cycle. Interestingly, oxygen isotoperatios measured on conodont apatite show twopositive excursions of +1.5x that parallel thepositive excursions in N
13C values (Joachimskiand Buggisch, 2002). Calculated paleotempera-tures indicate signi¢cant climatic cooling of thelow latitudes of 5^7‡C, potentially as a conse-quence of the perturbations in the carbon geo-chemical cycle.
The aim of this study is to quantify the mainbiogeochemical £uxes across the F-F boundaryand to model the imprint on global climate. Sincethe only available quantitative time series that re-corded environmental changes across the F-Fboundary are the measured carbonate N
13C andN18O values, any scenario of the evolution of the
global carbon^alkalinity cycles must be in agree-ment with those quantitative data. We thus buildand use a numerical model of the carbon, alka-linity, oxygen and phosphorus biogeochemicalcycles, coupled to a 1-D energy-balance climatemodel (the COMBINE model) to explore the val-idity of the various proposed scenarios for the F-F interval. The model calculates the N
13C valuesof all the exospheric carbon boxes and the outputis compared to measured values from one scenar-io to the other. Once agreement between calcu-lated and measured N
13C data has been reached,we have some con¢dence in the calculated evolu-
tion of the content of the various reservoirs (forinstance, the atmospheric pCO2) and in the calcu-lated evolution of global climate.
2. The COMBINE model
2.1. General structure
The COMBINE model (COupled Model of BI-ogeochemical cycles aNd climatE) (Fig. 2) isan atmosphere^ocean geochemical model fullycoupled to a simple 1-D climatic model. To repro-duce the Late Devonian geographic con¢guration,we divide the ocean into two major oceanic ba-sins, the Panthalassa and the Proto-Tethys oceans(Scotese and McKerrow, 1990). The respectiveareas of the two basins are estimated from Devo-nian paleogeographic maps to be 355U106 km2
for the Panthalassa and 53U106 km2 for the Pro-to-Tethys ocean. Another geographical con¢gura-tion might be computed for other continental con-¢gurations corresponding to other events in thegeological past. This geographical partitioning isrequired because most of the isotopic data weremeasured on sections deposited on the margins ofthe Proto-Tethys ocean and since the response toany perturbation of this oceanic basin might bepotentially di¡erent from the response of the Pan-thalassa ocean. Both oceanic basins are dividedinto ¢ve boxes. The open oceans include thephotic zone (0^100 m depth), a thermocline reser-voir (100^1000 m depth) and a deep sea reservoir(1000^5000 m depth). Two epicontinental reser-voirs are added for both basins: a surface epicon-tinental box, ranging from 0 to 100 m, and a deepepicontinental box (100^200 m depth). These twoboxes represent the shallow epicontinental seas.The atmosphere is described by one box. Ele-ments may be added to or removed from theocean^atmosphere system due to chemical weath-ering of the continents and the input by volcanicdegassing, and through deposition on the sea-£oor.
The geochemical model is coupled to an energy-balance climatic model developed by FrancPoisand Walker (1992) that has been previously usedand validated for several paleoclimatic studies
(FrancPois and Walker, 1992; Godde¤ris andFrancPois, 1995; Veizer et al., 2000). It calculatesat each time step the mean air temperature in 18latitude bands as a function of (1) the model at-mospheric CO2, (2) the latitudinal distribution ofDevonian continental masses, and (3) Devoniansolar luminosity (98% of the present-day solarluminosity; Endal and So¢a, 1981). The meridio-nal heat £ux is divided into the sensible, latent,and oceanic components. The sensible and ocean-ic heat transfer coe⁄cients are made dependenton latitude. Today, these coe⁄cients are lowerover the South Pole compared to their valuesaround the North Pole. This particular con¢gura-tion is the result of the existence of circumpolaratmospheric and oceanic currents along Antarcti-
ca. We choose to keep the heat transfer coe⁄-cients identical to modern values in order to re-produce a colder and drier climate at the Southpole, where the Gondwana continent was located,than at the North pole, itself free of continentalmass.
The water £uxes between the oceanic boxes areprescribed. Only a rough pattern of the Devonianoceanic circulation can be estimated. Based on thefact that the seaways between the Proto-Tethysand the Panthalassa basins were apparently re-stricted to continental platforms (Scotese andMcKerrow, 1990), we assume weak to no hori-zontal exchange between the thermocline anddeep reservoirs of both basins (respectively 1and 0U106 m3/s). The exchange through the
Fig. 2. A schematic view of the COMBINE model. The global ocean is divided into two oceanic basins. The Proto-Tethys basinis the smallest one. The carbon, alkalinity, oxygen and phosphorus contents are calculated for each box at each time step of theintegration. The black arrows represent the transfer of elements through geological processes: continental weathering and degas-sing through volcanoes and mid-ocean ridge systems.
m3/s. We furthermore assume that the verticalcirculation in the Panthalassa ocean can bedescribed by the upwelling of waters from thethermocline reservoir into the epicontinentalreservoirs, compensated by the sink of surfacewaters directly from the open ocean surface res-ervoir into the deep ocean reservoir. The input ofwell-oxygenated surface waters directly into thedeep sea reservoir results in a steady-state oxygenconcentration higher in the deep reservoir than inthe thermocline reservoir. The Panthalassa oceancan thus be compared to the modern globalocean. On the other hand, we assume a simplemixing between all the boxes of the Proto-Tethysreservoir, without strong upwelling, a hypothesisthat appears reasonable regarding the setting ofthe Proto-Tethys basin (Ormiston and Oglesby,1995). Furthermore, we assume no input of sur-face waters directly into the deep ocean, leadingto the location of an oxygen minimum zone in thedeep reservoir, below 1000 m. The Proto-Tethysbasin can thus be seen as a huge lake, relativelystrati¢ed, with less oxygenated conditions asdepth increases.
2.2. Geochemical cycles
The model includes geochemical cycles for car-bon, phosphorus (which is the only nutrient mod-elled), alkalinity and oxygen. The amount of car-bon in the atmospheric reservoir is calculated bysolving at each time step the di¡erential equationdescribing the atmospheric carbon budget. Theinputs are the subaerial degassing by volcanism(held constant at the 3.0U1012 mol/yr value),and the CO2 degassing of the surface ocean, whilethe outputs are the consumption of atmosphericCO2 by chemical weathering of carbonate andsilicate rocks, and the dissolution of CO2 in sur-face waters. The consumption of atmosphericCO2 by silicate weathering is a linear functionof continental runo¡ and land area, and a non-linear function (exponential) of the mean air tem-perature as well as of atmospheric CO2 content(FrancPois and Walker, 1992). Because of this non-linearity, the use of global mean air temperatureabove continental areas in order to calculate the
global silicate weathering rate is not rigorous. Wethus follow an approach similar to FrancPois andWalker (1992) and Godde¤ris and FrancPois (1995)by dividing the Earth into 18 latitude bands, andusing the mean zonal air temperature calculatedfor each latitude band by the climate model toestimate the zonal silicate weathering £ux. Allthe terms are then added up to provide the global£ux:
FwsilðtÞ ¼ ksilU
X18j¼1
areajðtÞUrunoff jðtÞU�
expTj3288:15
10:95
� ��UFCO2Uf e ð1Þ
where ksil is a constant, areaj(t) and runo¡j(t)are respectively the continental area and runo¡at any time step in the latitude band j, and Tj isthe mean air temperature in the latitude band j.The factor 10.95 assumes an activation energy ofabout 63 000 J/mol for the dissolution of silicates(Brady, 1991). The zonal continental runo¡ is es-timated using the parameterization developed byFrancPois and Walker (1992), as a function of zo-nal temperature, latitude and the fraction of con-tinental surface in each latitude band. FCO2 rep-resents the direct dependence of weathering onatmospheric pCO2. In the absence of vegetation,this factor equals the atmospheric CO2 (expressedin PAL) raised to the power 0.5 (Berner, 1994). Inthe presence of higher plants, this factor is calcu-lated as follows (Berner, 1994):
FCO2 ¼2pCO2
1þ pCO2
� �0:4 ð2Þ
We assume that, prior to the F-F perturbation,an arbitrary 10% of the continental area was cov-ered with higher vegetation. The mean FCO2 fac-tor for each latitude bands is calculated as a linearmixing between non-vegetated and vegetated areavalues, weighted by the areas. fe is a factor intro-duced by Berner (1994) that considers an expectedenhancement of chemical weathering due to thepresence of higher land plants (concentration oforganic acids in soil, presence of microbes, devel-opment of roots). This factor is normalized to itspresent-day value. Following Berner (1994), fe isset to 0.15 for non-vegetated areas, and to 0.75
where kcarb is a constant. FCO2 represents the at-mospheric CO2 dependence of the carbonateweathering, and is equal to the same FCO2 factorused in the silicate weathering calculation (Berner,1994).
The amount of total dissolved inorganic carbon(DIC) and alkalinity for each oceanic box is cal-culated at each time step by solving the di¡eren-tial equation describing its budget. Depending onthe box (Fig. 2), the inputs are the CO2 degassingat mid-ocean ridge systems into the deep ocean(global value kept constant at 1.5U1012 mol/yr),the supply of carbon and alkalinity by continentalweathering, the DIC and alkalinity £ux trans-ported by water £uxes into the ocean, the oxida-tion of the sinking organic matter from the photiczone into the deeper parts of the ocean, and thedissolution of atmospheric CO2 into the ocean.The outputs are represented by the £ow of DICand/or alkalinity out of the box by transport, bythe consumption of carbon by biological activitywithin the photic zone, by deposition of carbonateand of organic carbon, and degassing of CO2 tothe atmosphere.
The complete carbonate speciation, includingthe calculation of pH, [H2CO3], [HCO3
3 ], [CO233 ]
as a function of the salinity and temperature ofthe oceanic reservoir, and of the partial pressureof dissolved CO2, is then calculated in each oce-anic box, once their respective DIC and total al-kalinity content are calculated by solving the dif-ferential system. The intermediate depth and deepsea temperatures are ¢xed to 276 K, while thesurface temperature of the Panthalassa basin iscalculated as the mean global air temperature at
the surface estimated by the climate model. TheProto-Tethys sea surface temperature is calculatedas the mean air temperature between 10‡S and10‡N. The air^sea CO2 exchange between the sur-face oceanic box i and the atmosphere is esti-mated as follows:
F atm�iðCO2Þ ¼ K0U pCOatm23pCOi
2
� �Uareai ð4Þ
where K0 is a constant (Sarmiento et al., 1992),pCOatm
2 is the partial pressure of atmosphericCO2, pCOi
2 is the partial pressure of dissolvedCO2 in reservoir i, and areai is the reservoirarea at the surface.
The calcium budget calculated for each oceanicbox allows the determination of the concentrationof Ca2þ. The inputs to the ocean are the weath-ering of continental silicate and carbonate rocks,while the output is the accumulation of carbonateon the sea£oor. There is no convincing evidencefor the existence of calcareous phytoplankton inthe Paleozoic. For this reason, we assume that theonly precipitation of carbonate occurs on the con-tinental shelves as shallow water carbonates, de-posited above 100 m water depth. The carbonateaccumulation £ux is calculated following Opdykeand Wilkinson (1988):
Fdcarb ¼ kdUarea100iUð6 calcite31Þ1:7 ð5Þ
where 6calcite is the calculated saturation ratio forcalcite, kd is a constant and area100i is the hori-zontal area above 100 m depth for the epiconti-nental surface reservoir i, calculated by the hyp-sometric module of the model (see Section 2.3).
The production of organic carbon occurs with-in the photic zones of both oceanic basins. Thebiological productivity is made proportional tothe phosphorus input into the photic zones. Thisproductivity £ux feeds a particulate organic car-bon reservoir (Petsch and Berner, 1998). The C:Pvalue of the productivity £ux is ¢xed to 117:1(Anderson and Sarmiento, 1994). Dissolved phos-phate is removed from the surface reservoir andincorporated into a particulate phosphorus reser-voir. Particulate organic carbon and phosphorusthen sink within the thermocline reservoir if theyoriginated from the open ocean photic zone, ortowards the sediment or deep epicontinental res-ervoir if they originated from the surface epicon-
tinental reservoir. Recycling occurs within thewater column. The recycling rate is a linear func-tion of the dissolved oxygen content above5U1033
Wmol/kg O2. Below this value, anoxic re-cycling is assumed to occur at a constant rate.The model includes a simpli¢ed sediment modulein order to estimate the amount of organic carbonand phosphorus that is buried relative to theamount that reaches the sediment, the di¡erencebeing recycled (FrancPois and Walker, 1992; God-de¤ris, 1997). This model includes a sedimentmixed layer where organic matter is oxidized byoxygen, and a sulfate reduction layer below it.Since no sulfur cycle is included in this model,the concentration of sulfate within the deepestsediment layer is assumed to be constant. TheC:P ratio of the buried organic matter is madedependent on the dissolved oxygen level withinthe water in contact with the sediment, increasingif the degree of anoxia increases (Van Cappellenand Ingall, 1996).
Phosphorite mineral precipitation, P sorptionand iron coprecipitation are assumed to occur in-distinctly within all oceanic reservoirs in contactwith the sea£oor. The corresponding removal ofphosphorus is made proportional to the dissolvedphosphate content of the respective reservoir. Anadditional sink of phosphorus is represented bythe scavenging of phosphorus by ferric oxyhy-droxides formed within hydrothermal systems.This process occurs within the deep oceanic reser-voir of both the Panthalassa and Proto-Tethysoceans, and is assumed to be proportional tothe hydrothermal activity (kept constant in allsimulations), and to the dissolved phosphate con-centration of the deep reservoirs (Benitez-Nelson,2000). The supply of phosphorus from continentalrock weathering is modelled as a function of con-tinental runo¡ and vegetation in the same way ascontinental carbonate weathering. Under present-day conditions, the weathering input of P into theocean reaches 45.0U109 mol/yr (Petsch and Ber-ner, 1998).
At the time scale of the present study, thesource of oxygen is thus the burial of organicmatter, while the sink is the oxidation of old sedi-mentary reduced carbon (SRC) exposed on thecontinents. The oxidation of old SRC is assumed
to be dependent on the continental runo¡ in thesame way as the weathering of continental car-bonates (FrancPois and Walker, 1992):
where kSRC is a constant. The crustal carbon re-leased during the oxidation of the SRC is trans-ported to the ocean, and constitutes a source ofinorganic carbon for the ocean.
In order to increase the time step of the calcu-lation, the partial pressure of oxygen in surfacewaters is assumed to be instantaneously at equi-librium with the atmospheric O2 pressure. Thesurface water concentration of dissolved oxygenis then calculated as a function of the partial pres-sure in water, salinity and temperature of surfacewaters using the Wanninkof (1992) formalism.
The COMBINE model includes a 13C cycle.The exchange of CO2 between the various surfaceocean boxes and the atmosphere occurs with anisotope fractionation, through the exchange equa-tion derived by Munhoven (1997):
f oc�atm ¼ K0U P asUpCOatm23ð
N13Ci3N
13Catm þ P sa� �
UpCOi2ÞUareai ð7Þ
where foc�atm represents the net carbon isotopeexchange between the atmosphere and the oceanicsurface reservoir i (positive if the net £ux is fromthe ocean towards the atmosphere), N
13Ci is theN13C value of DIC in surface ocean reservoir i,
and N13Catm is the atmospheric N
13C value. Pas
and Psa equal respectively Kas31 and Ksa31,where Kas and Ksa are the one-way fractionationfactors. Pas and Psa are ¢xed to 30.002 and30.010 (Siegenthaler and Mu«nnich, 1981). TheN13C value of each DIC species is calculated for
each oceanic reservoir. The fractionation betweendissolved CO2 and bicarbonate, and between bi-carbonate and carbonate ion is calculated as afunction of temperature (Mook et al., 1974; Free-man and Hayes, 1992) in each oceanic box. Thebiological isotope fractionation occurring duringthe formation of particulate organic carbon in thephotic zone is calculated for each reservoir as afunction of dissolved CO2 and PO33
4 (Kump andArthur, 1999), and dissolved oxygen (Berner et
al., 2000). Carbonate precipitation is assumed tooccur without fractionation. The N
13C values ofcontinental SRC and continental carbonates arerespectively ¢xed at 328x (Hayes et al., 1999)and +1x PDB (excluding the carbonate plat-forms exposed during the Famennian sea-levelfall, see below). The N
13C of the carbon degassedalong mid-ocean ridges is set at the mantle valueof 35x PDB (Holser et al., 1988). The N
13Cvalue of subaerial volcanism is the result of themixing of mantle-originated carbon, carbon fromrecycled carbonates (including alteration productsof sea£oor basalts) and sedimentary organic car-bon at subduction zones (Wallmann, 1999). Therespective contributions of these three terms arenot known (Godde¤ris and Veizer, 2000; Deines,2002). The N
13C value of subaerial volcanism ishere set at a constant 31x PDB for calibrationreasons, thus assuming a mixing between mantlecarbon and carbon degassed from subducting car-bonates in agreement with Caldeira (1992) andGodde¤ris and FrancPois (1996).
Finally, the salinity of each reservoir is calcu-lated. Since isolated seas are assumed to be partof the model epicontinental reservoirs, these res-ervoirs should display a higher salinity than theopen ocean reservoir. We thus force an evapora-tion minus precipitation budget to positive valuesfor the epicontinental reservoirs, and to negativevalues for open ocean surface reservoirs in such away that the global water mass is held constant.This disequilibrium is a function of the epiconti-nental sea surfaces, increasing when the surfaceof epicontinental seas increases (during sea-levelrises). In the ocean, salinity is transported bythe water £uxes. At steady state, this procedureleads to a salinity ranging from 34.5x for theProto-Tethys open surface reservoir (PTES) to37x for the small surface epicontinental reser-voir of the Proto-Tethys basin (PTESE). Theglobal mean salinity is ¢xed to a constant valueof 35x.
2.3. Sea-level changes
The sea-level curve (Johnson et al., 1985) is aqualitative reconstruction (Fig. 1). The exact am-plitudes of the sea-level changes are not known.
However, these are second-order sea-levelchanges, and their amplitude most probably liesbetween 1 m and several tens of meters. Here, weassume two sea-level rises of 10 m each, while thesea-level fall during the Early Famennian is as-sumed to be 20 m. There are good indicatorsthat the two sea-level rises are not glacio-eustatic£uctuations, since there is no evidence for the ex-istence of a continental ice cap that could £uctu-ate in volume prior to the end of the Frasnian. Onthe other hand, these two transgressions are toorapid to be induced by variations in the volume ofthe mid-ocean ridge system. Since the mechanismfor these sea-level rises is uncertain (Johnson etal., 1985), and in order to maintain the amountof water constant in the system, we simply modelthe two sea-level rises by arti¢cially pushing upthe deep sea£oor. Regarding the Early Famen-nian sea-level fall, there is only vague evidencefor the development of a hypothetical ice sheet,responsible for the sea-level drop (Streel et al.,2000). We will not take this into account, simplyassuming an arti¢cial deepening of the deep sea-£oor, to produce a sea-level fall at constant watervolume. Since our goal is to study the impact ofthese £uctuations on the geochemical cycles, theexact mechanism that produces these sea-levelrises and falls is of secondary importance, aslong as the causes of these £uctuations do notimply directly geochemical £uxes (such as degas-sing of CO2, or changes in the weatherability ofcontinental surfaces following the development ofcontinental ice sheets). The lack of constraints onthe causes of the sea-level £uctuations leads us toneglect these possibilities. The expected impacts ofsea-level change on the geochemical cycles as de-scribed by the COMBINE model at the millionyear time scale are as follows:
(1) The consecutive change in continental areawill directly in£uence the discharge of elementsproduced by chemical weathering of continentalrocks. This e¡ect is quanti¢ed within the model,through the hypsometric module which estimatesthe continental area £ooded during sea-level riseor exposed during sea-level fall. This module cal-culates the volume and horizontal surfaces for alloceanic boxes, through simple geometric consid-erations, as a function of sea level.
(2) A sea-level rise or fall will change the vol-ume and surface of the oceanic reservoirs anda¡ect evaporation as well as continental precipi-tation and runo¡. We thus modi¢ed the FrancPoisand Walker (1992) parameterization for continen-tal runo¡ by adding a factor proportional to theratio of total oceanic versus continental area.Since the ocean basins increase during the sea-level rises (mainly through the extension of theepicontinental reservoirs), evaporation increases,and the model runo¡ is consequently enhanced.During sea-level fall episodes, the continental run-o¡ tends to decrease.
(3) Sea-level changes may also a¡ect the ocean-ic circulation. Sea-level rise episodes might resultin the enhanced formation of warm saline deepwaters in epicontinental seas, £owing back intothe ocean, and ¢nally stratifying it by reducingthe vertical mixing (Brass et al., 1982). In orderto test this hypothesis, we link the vertical mixingof the oceanic basins to sea level. The verticalmixing is assumed to decrease linearly with sealevel during the sea-level rise episodes. We chooseto cut completely the vertical mixing at the sea-level maxima to maximize the impacts of possibleocean stagnation during sea-level rises. In the sim-ulations described below, we only apply the re-duction of the vertical mixing to the Proto-Tethysbasin. We also performed simulations where thevertical mixing was reduced in both the Pantha-lassa and Proto-Tethys oceans. However, thisgenerally increases the disagreement betweendata and model output.
2.4. Calibration procedure
The weathering constants ksil, kcarb and kSRC
are estimated using the temperature and runo¡calculated for each latitude bands by the climatemodel under present-day conditions (1 PAL at-mospheric CO2, present-day continental con¢gu-ration and present-day solar luminosity). ksil isadjusted so that the present-day calculated con-sumption of CO2 by silicate weathering reaches11.7U1012 mol/yr (Gaillardet et al., 1999). kcarbis calibrated so that the global consumption ofatmospheric CO2 by carbonate weatheringreaches 20.6U1012 mol/yr (Petsch and Berner,
1998) under present-day conditions, and kSRC is¢xed assuming that about 4.1U1012 mol/yr of O2
are consumed by the oxidation of SRC (Petschand Berner, 1998). The calibration of the oceanicand sediment submodel constants requires the cal-culation of the oceanic primary productivity, therecycling of organic matter within the water col-umn, the deposition of carbonates. This calibra-tion procedure cannot be performed with theCOMBINE model in its Devonian con¢gurationfor two reasons. First, the con¢guration of theDevonian oceanic model di¡ers from the con¢gu-ration of the present-day oceans. Second, theCOMBINE model in its Devonian version doesnot take into account the formation and subse-quent deposition of carbonates in the open ocean.For these reasons, the ocean and sediment sub-model constants were calibrated using a version ofCOMBINE ¢tted for present-day conditions (in-cluding carbonate formation in the open ocean,deep sea carbonate accumulation, one oceanic ba-sin representing the global ocean, global circula-tion ¢tted to the Munhoven and FrancPois (1996)ocean box model). Once the ocean and sedimentsubmodel constants ¢t so that DIC, alkalinity andoxygen gradients from the surface to the bottomwaters are reproduced for the present-day condi-tions, these constants are exported into the COM-BINE model in its Devonian version. The mainparameters and calibration constants used in theCOMBINE model and not listed in the text canbe found in Table 1.
The COMBINE model (in its Devonian con¢g-uration) is then run for 20 million years with aglobal CO2 degassing held constant at 4.5U1012
mol/yr (volcanic+mid-ocean ridge), and reaches asteady state that will be used as the pre-perturba-tion value. The atmospheric pCO2 and pO2 stabi-lize at 2925 ppm (10.5 times the present-day par-tial pressure) and 0.165 bar (0.66 times thepresent-day partial pressure), respectively. Giventhe range of uncertainties of the available recon-structions for these two variables, these values fallwithin an acceptable range. Global carbonate ac-cumulation reaches 14.2U1012 mol/yr, 65% of itbeing deposited in the Proto-Tethys epicontinen-tal area, the remaining 35% accumulating on thePanthalassa shelves. The burial of organic carbon
equals 3.1U1012 mol/yr. The most favorable set-ting for the burial of organic carbon is the deepProto-Tethys ocean sea£oor, where 45% of thetotal burial of organic carbon occurs. The burialof organic carbon within the Panthalassa epicon-tinental reservoirs accounts for 42% of the totalburial £ux. The rest is buried on the deep Pan-thalassa sea£oor, and within the epicontinentalrealm of the Proto-Tethys ocean.
3. Plausible scenarios and results
Given the number of processes and parametersincluded in the COMBINE model, the number of
sensitivity tests that can be performed is almostunlimited. We choose to present only the simula-tions that illustrate the general behavior of themodel, in response to the applied perturbations.All the following results show general trends thatare not dependent on the values of the parametersat the ¢rst order, particularly on the pre-pertur-bation oceanic circulation. However, it should bekept in mind that the exact values of the modeloutput are somewhat dependent on those param-eters, particularly the regional distribution of theburial £uxes. But the general trends, spreadingover several 105 to 106 years are not. Fig. 3 dis-plays a brief schematic description of the testedhypotheses and the performed simulations.
Table 1Main parameters and calibration constants of the COMBINE model
Constant Value Reference
Continental silicate weathering constant ksil 7.59U1035 mol/l calibrated from Gaillardet et al., 1999Continental carbonate weathering kcarb 2.77U1033 mol/l calibrated from Petsch and Berner, 1998SRC oxidation constant kSRC 5.53U1034 mol/l calibrated from Petsch and Berner, 1998Biotic carbonate accumulation constant kd 0.172 mol/yr/m2 FrancPois and Walker, 1992C:P of buried organic matter under full oxic conditions 200 mol/mol Van Cappellen and Ingall, 1996C:P of buried organic matter under full anoxic conditions 4000 mol/mol Van Cappellen and Ingall, 1996Panthalassa surface 355U106 km2 Estimated from Scotese and McKerrow, 1990Proto-Tethys surface 53U106 km2 Estimated from Scotese and McKerrow, 1990Total continental surface 101.7U106 km2 Estimated from Scotese and McKerrow, 1990CO2 air^sea exchange constant K0 0.0572 mol/m2/yr/Watm Sarmiento et al., 1992Devonian solar constant 1342.6 W/m2 Endal and So¢a, 1981N13C value of continental carbonate +1x PDB Holser et al., 1988N13C value of continental SRC 328x PDB Hayes et al., 1999N13C value of mantle carbon 35x PDB Holser et al., 1988N13C value of aerial volcanic CO2 31x PDB Mean value between the mid-ocean ridge and
carbonate recycling, following the suggestionof Caldeira (1992)
Fig. 3. A schematic view of the tested hypotheses and performed simulations.
This ¢rst test is performed assuming all theabove e¡ects of sea-level change: continentalarea changes, modi¢cations in the volume andsurface of the oceanic basins, intensi¢cation ofcontinental runo¡ during sea-level rise episodes,stagnation at the sea-level maxima within the Pro-to-Tethys ocean only, and decrease of continentalruno¡ during sea-level fall. The calculated carbonisotope composition of the PTESE reservoir
shows little change (Fig. 4e). During the sea-levelrises, the PTESE N
13C value remains roughly con-stant, while the carbon isotope composition of thePTESE decreases by about 0.15x during theEarly Famennian fall in sea level. The model out-put disagrees with the observations. These minorchanges in the carbon isotope signal are not driv-en by continental processes, since continentalweathering £uxes remain almost unchanged dur-ing the simulation with increasing runo¡ compen-sating for the reduction in continental area duringthe sea-level rises and the decreasing runo¡ com-pensating for the increasing continental area dur-ing the Early Famennian sea-level fall. On theother hand, oceanic processes are heavily a¡ectedby the change in sea level. During both sea-levelrises, the oxygen level in the Proto-Tethys deepepicontinental reservoir (PTEDE) declines withinthe dysoxic domain (Fig. 4c). The C:P value ofburied organic matter below the PTEDE reservoirincreases from 1000:1 to 2300:1 during the ¢rstsea-level rise, even reaching 3000:1 during the sec-ond sea-level fall. As a result, PO33
4 concentrationincreases signi¢cantly within the PTEDE reservoirduring the sea-level rises (Fig. 5).
However, the coeval calculated primary pro-ductivity within the Proto-Tethys surface reser-voirs is reduced during the sea-level rises, sincethe upwelling of nutrients within the photic zonealmost stops at the sea-level maximum. Despite afew percent increase in the primary productivityin the Panthalassa epicontinental surface reservoir
Fig. 4. Results of the SEA simulation (description in thetext). (a) Global continental silicate weathering. (b) Atmo-spheric partial pressure of oxygen (in PAL). (c) Dissolvedoxygen concentration; solid line: deep epicontinental Proto-Tethys reservoir (PTEDE), dotted line: deep epicontinentalPanthalassa reservoir (PTADE); gray box: dysoxic domain.(d) Burial of organic carbon; dotted line: burial within thePTADE, dashed line: burial within the deep Proto-Tethysreservoir, solid line: total burial. (e) Calculated N
13C valuesof dissolved CO23
3 within the PTESE reservoir, compared tothe available data. The shaded area includes 68% of the dataaround the moving average ( V standard deviation around themean for a Gaussian distribution of the data points). Themoving average was run through the dataset with a step of100 000 years, and a window width of 200 000 years. (f) Cal-culated atmospheric CO2 partial pressure.
(PTASE), the global productivity decreases by al-most 10%. Globally, the amount of organic car-bon reaching the sediment is thus reduced, but theconditions of preservation are improved by thedeveloping dysoxia. As a result, the global £uxof organic matter to the sediment displays a com-plex pattern, £uctuating rapidly during the sea-level rises, between 2.6U1012 and 3.3U1012 mol/year (starting from a pre-perturbation value of3.1U1012 mol/year; Fig. 4d). This rather complexbehavior is the result of the combination of thechanges in the organic carbon burial £ux withinthe di¡erent oceanic reservoirs. Regarding themain contributions to the global £ux, the organiccarbon burial £ux increases by about 60% in thedeep epicontinental area of the Panthalassa ocean(PTADE), but declines by 60% in the deep Proto-Tethys open reservoir (Fig. 4d). The combinationof these changes is ¢rst a sudden but limited de-crease of the total organic carbon burial, followedby a slow recovery back to pre-perturbation val-ues covering the one million years of the EarlyFamennian sea-level fall (Fig. 4d).
Continental weathering £uxes are weakly per-turbed by the Famennian sea-level fall (Fig. 4a forsilicate weathering). The slow recovery of the or-ganic carbon deposition £ux together with a more
or less constant SRC oxidation on continents re-sults in a slow increase of the calculated seawaterN13C value. Furthermore, the persisting disequilib-
rium between SRC oxidation and organic carbondeposition leads to an increase in atmosphericpCO2, reaching 3150 ppmv at the end of thesea-level fall episode, and to a decrease in the at-mospheric oxygen level by about 0.0027 PAL(Fig. 4f).
As a preliminary conclusion, the simulationshows that the impact of sea-level changes mayhave resulted in changes in oceanic circulation,runo¡ and geometry of the epicontinental reser-voirs. However, this scenario cannot explain thevariations in the carbon isotope composition re-corded across the F-F boundary. The model sug-gests that the Early Famennian climate is warmedup due to increasing atmospheric pCO2, and thatatmospheric oxygen content declines while pre-vious reconstructions suggest a rapid increasefrom the Late Devonian towards the Carbonifer-ous (Berner, 2001).
3.2. Weathering of carbonate platforms
We modi¢ed the SEA scenario by introducingexpected consequences of carbonate platformgrowth and erosion during sea-level rises andfalls, respectively (PLAT simulation). The PLATscenario can be seen as an application of sedimen-tary rapid recycling (Berner and Can¢eld, 1989)applied within the context of short-term isotopicexcursions. During the two Frasnian sea-levelrises, the model calculates the amount of carbon-ates deposited on the £ooded part of the conti-nents, and estimates their subsequent erosion onceexposed to the atmosphere proportionally to theremaining mass of these platforms as a functionof modelled climate (Eq. 3), in addition to thebackground continental £ux. In terms of the car-bon and alkalinity budgets, the sea-level rise epi-sodes have very little impact on the geochemicalcycles, because they are short and limited in am-plitude. The maximum carbonate mass depositedduring the sea-level rises is calculated to be0.5U1018 mol carbon. The impact on seawaterN13C value is extremely small, since the carbonates
were deposited in isotopic equilibrium with coeval
Fig. 5. PO334 concentration within the PTEDE, as calculated
seawater, their subsequent erosion having a veryminor impact on the oceanic N
13C budget. On theother hand, the Famennian sea-level fall mighthave had a major impact on the seawater N
13Ccomposition, since older carbonates with N
13Cvalues di¡erent from the Late Frasnian mean oce-anic value may have been exposed during a rela-tively long period (1.5 million years). Because theuppermost part of these platform carbonates willbe of Devonian age, we assume that the meanN13C value of carbon released by chemical weath-
ering of these carbonates can be approximatedby an average carbon isotopic composition of+1.5x PDB (Veizer et al., 1999). Estimatingthe actual surface of exposed carbonates duringthe sea-level fall is not possible. Again, in orderto identify the maximum possible impact, we willassume that all continental margins exposed dur-ing the sea-level fall (below the pre-perturbationsea level) are covered by Devonian carbonates.This assumption implies that the increase of con-tinental area during the sea-level fall a¡ected nei-ther silicate nor kerogen carbon weathering.
As expected, the response of the ocean^atmo-sphere system is similar to the results of the SEAsimulation for the Frasnian (Fig. 6). The sea-levelfall observed during the earliest Famennian re-sults in the exposure and weathering of Devoniancarbonates consuming atmospheric CO2 at a cal-culated maximum rate of 4.7U1012 mol/year. Themodel assumes that carbonates contain P at a C:Pratio of 1000:1 (Froelich et al., 1982). The weath-ering of the carbonate platforms exposed by thesea-level fall thus implies an increase in the deliv-ery of continental phosphorus to the ocean by4.7U109 mol/year, representing a 17% increaseof the continental phosphorus £ux relative tothe coeval SEA scenario value. The consequenceis an enhanced primary productivity within theProto-Tethys photic zone and consequently anenhanced organic carbon burial £ux in the Pro-to-Tethys basin, reaching up to 2.0U1012 mol/year (compared to the pre-perturbation value of1.5U1012 mol/year; Fig. 6d). On the other hand,oxygenation of the epicontinental deep reservoirs(due to the decrease in size of these reservoirsduring the sea-level fall) leads to a decreased pres-ervation of organic matter within the epicontinen-
tal sedimentary layer. This e¡ect is comparable inmagnitude to the SEA simulation (Fig. 6c). Thenet e¡ect is a global burial £ux of organic carbonwhich is in excess by about 0.4U1012 mol/yearcompared to the production of carbon by conti-nental kerogen weathering for about one millionyears (covering the sea-level fall episode).
As a result of this disequilibrium, atmospheric
Fig. 6. Results of the PLAT simulation (description in thetext). See caption of Fig. 4 for more details.
oxygen increases from 0.665 PAL at the F-Fboundary to 0.676 PAL at the end of the sea-levelfall episode (Fig. 6b). Atmospheric CO2 is weaklyin£uenced, despite the fact that the organic car-bon subcycle acts as a global sink of carbon (car-bon sink being larger than the source; Fig. 6f).However, the additional continental area exposedduring the sea-level fall episode is covered by car-bonates in this simulation. As a consequence, thedecrease of the consumption of atmospheric CO2
by continental silicate weathering due to the de-crease in runo¡ (linked to a drier climate duringthe sea-level fall episode) is not compensated byan increase in the area of exposed silicate rocks.The decrease in silicate weathering (Fig. 6a) leadsto an increasing pCO2 at assumed constant mid-ocean ridge and volcanic CO2 degassing rates,counteracting the sink due to the enhanced burialof organic matter.
In addition, the modelled PTESE N13C data dis-
play a comparable pattern (minor negative excur-sion in N
13C value) during the two sea-level riseevents. The Early Famennian sea-level fall is char-acterized by a +0.7x excursion lasting about1 million years, which is in better agreementwith measured N
13C values (Fig. 6e). The causesof this excursion are twofold: (1) the calculatedweathering of the carbonate platform directly in-jects relatively heavy carbon (N13C= 1.5x) intothe ocean; (2) the phosphorus supplied by theweathering of carbonates enhances the productiv-ity in surface waters, and the organic carbon de-position £ux. It should be noted that in terms ofN13C values the response of the various oceanic
surface reservoirs (epicontinental or open ocean)is almost identical. The carbon reservoir of eachoceanic surface reservoir ‘interacts’ with atmo-spheric CO2 through gas exchange, tending tohomogenize the isotopic ratios of surface waters.Small di¡erences (0.4x maximum) are due todi¡erences in temperature in the various surfacereservoirs, and to changes in productivity. How-ever, even if the agreement between the calculatedand measured N
13C data is improved, the modelsuggests that the maximum e¡ect of carbonateplatform weathering is not su⁄cient to accountfor the high N
13C values measured during theEarly Famennian.
Carbonate platform weathering has been in-voked by Kump et al. (1999) in order to explaina +6x excursion in carbonate carbon isotopecomposition during the Late Ordovician. In theirstudy, the authors argue that such a prominentexcursion in the N
13C values might be producedby changing the relative contribution of carbonateversus organic carbon weathering. Increasing thefraction of carbon derived from carbonate weath-ering in the total riverine carbon £ux from 72% to96% is thought to be su⁄cient to increase sea-water N
13C value by 6x. However, Kump et al.(1999) keep the global £ux of riverine carbon con-stant, so that their suggested increase in the car-bonate contribution implies a proportional de-crease in the carbon £ux from the oxidation ofsedimentary organic carbon, which has an impor-tant impact on the seawater N
13C value. In thepresent work, the 50% enhanced carbonate weath-ering increases seawater N
13C value by only0.7x. Indeed, the global riverine carbon £uxfrom the oxidation of continental sedimentary or-ganic carbon together with the contribution fromcarbonate weathering increases widely during thesea-level fall. As a consequence, the impact ofcarbonate weathering on the carbon isotopic bud-get of the ocean might be much smaller than sug-gested by Kump et al. (1999).
3.3. Colonization of continental surfaces by landplants
Algeo et al. (1995) and Algeo and Scheckler(1998) suggested that the rapid colonization ofdrier upland areas by higher plants during theLate Devonian might explain the sedimentationof organic carbon-rich Late Devonian blackshales and the observed excursions in carbonateN13C values. The development of soils and micro-
bial activities as a consequence of the colonizationby land plants may have a¡ected local weather-ing, resulting in a global enhancement of theweatherability of continental surfaces (weather-ability is de¢ned as the product of factors thata¡ect chemical erosion other than climate change;Kump and Arthur, 1997). Mathematically, thisenhanced weatherability translates into the in-crease of the fe factor in Eqs. 1 and 3. fe increases
by a factor of 5, if a non-vegetated area becomescovered by vascular plants (gymnosperms; Ber-ner, 1994). Furthermore, the dependence of chem-ical weathering on atmospheric CO2 will change(Eq. 2), due to the concentration of atmosphericCO2 in soils by plant roots and microbes (Berner,1994).
The overall e¡ect of the colonization of conti-nental surfaces by land plants is thus an increasein global chemical weathering. As a result, pCO2
will decrease, since the enhanced weathering willconsume CO2. After a few million years (severaltimes the residence time of carbon in the ocean^atmosphere system), the atmospheric pCO2 willreach a new steady-state value, which will be low-er than the initial concentration. If volcanic de-gassing was constant during the transition, thedelivery of nutrients by continental chemicalweathering will be the same prior to and afterthe colonization event. However, during the tran-sition from a high to a low steady-state pCO2, the£ux of dissolved materials transferred from thecontinents to the ocean will ¢rst increase rapidly,and then decrease towards its pre-perturbationvalue. This non-equilibrium situation will resultin an enhanced delivery of nutrients to the ocean,stimulate oceanic productivity, and potentially re-inforce anoxic conditions in the deeper part of theepicontinental and open oceans.
However, two questions remain to be an-swered: (1) Does the spreading of vascular plantsincrease global weatherability? This question isfully debated in a recent review paper by Boucotand Gray (2001). The authors emphasize that,even if the development of higher plants initiallyintensi¢ed chemical weathering, the e⁄ciency ofthis process will decrease through time as soilsaccumulate over the bedrock, limiting the circula-tion of water. The e¡ect of higher plants on chem-ical weathering might only be e⁄cient on a longtime scale if mechanical erosion regularly removesthe soil cover. In the particular case of the LateDevonian time period, we feel that the coloniza-tion of the continents by land plants will result inan enhancement of the weatherability of the con-tinental surface for at least several 105 to 106
years, prior to the development of thick soils.As an example, Tardy (1993) calculated that it
takes about 330 000 years to develop a 1 m thicksoil horizon in a tropical environment character-ized by a precipitation of 800 mm/yr. Further-more, as will be discussed below from model sim-ulations, the Early Famennian is a period of long-term climatic cooling, an evolution that will re-duce the extension of warm tropical climatic belts,and ¢nally enhance mechanical weathering. Forthese two reasons, we argue that the hypothesisof Algeo and Scheckler (1998) might be a relevantprocess to increase the delivery of nutrients to theoceans. (2) The second question is one of timing.Is there any connection between the Late Fras-nian sea-level rises and the colonization of conti-nental area by vascular plants? Is it possible thata sea-level rise leading to increased moisture overcontinents triggered a rapid colonization of dryuplands by land plants during this period of in-tense and rapid plant evolution?
We tested such a hypothetical scenario (VEGsimulation). Starting from a vegetation (vascularplants) that covered 10% of the continental areaduring the Frasnian in the PLAT scenario, wesuperimpose a 5% increase of the continental ve-getated area triggered by the ¢rst sea-level rise,and a further 25% increase during the secondsea-level rise. This colonization is assumed to beequally distributed between all latitude bands.These numbers were chosen ¢rstly to allow us toexplore the impact of minor and major spreadingof vascular plants, and secondly since they repre-sent the best a posteriori combination to ¢t thecalculated seawater N
13C value with the data.Both colonization events start at the onset ofthe sea-level rises. The colonization events occurinstantaneously (at geological time scales), but feand FCO2 for the new vegetated areas change lin-early over 50 000 years from non-vegetated values(fe = 0.15) to fully vegetated values (fe = 0.75;Berner, 1994). This delay roughly accounts forthe time required for the development of newsoils.
The impact on the calculated evolution of at-mospheric CO2 pressure is drastic (Fig. 7f). Dur-ing the ¢rst sea-level rise, pCO2 ¢rst declines from2925 to about 2730 ppmv. A further decrease toabout 2540 ppmv is observed just prior to theonset of the second sea-level rise (this change cor-
responds to an overall change in mean globaltemperature of about 0.5‡C). The e¡ect of thesecond sea-level rise, combined with a 25% in-crease in plant cover, is a global pCO2 decreaseof about 1000 ppmv, which corresponds to anadditional global cooling of more than 2‡C,reached about 1 million years after the F-Fboundary. pCO2 then remains constant, having
reached a new steady state. The reason for theenduring decrease in atmospheric CO2 well afterthe sea-level rises is due to the increase of thecontinental phosphorus delivery as a consequenceof the spreading of land plants. Since this £ux isassumed to be not directly dependent on temper-ature (Eq. 3), it is less sensitive to the ongoingglobal climatic cooling than the global silicateweathering £ux. Due to the increase of nutrientsupply, productivity in surface waters and conse-quently organic carbon burial are strongly en-hanced (Fig. 7d) during more than 1 Myr afterthe colonization events.
The two sea-level rises and colonization eventsresult in an increase of the PTESE N
13C value by0.4x and 1.4x, respectively (Fig. 7e). The Fa-mennian N
13C value plateau is now well repro-duced by combining the e¡ect of carbonate plat-form weathering and spreading of land plants.However, there is a 500 000-year delay betweenthe onset of the colonization event and the max-imum excursion in the PTESE carbon isotope val-ues, because of the residence time of carbon in theocean^atmosphere system. Finally, the atmo-spheric oxygen concentration increases by about0.05 PAL (2.0U1018 mol) over 3 million years dueto the increasing burial of organic matter (Fig.7b). Atmospheric oxygen has not yet reached a
Fig. 7. Results of the VEG simulation (description in thetext). See caption of Fig. 4 for more details.
Fig. 8. Total carbonate deposition calculated by the VEGand PIN simulations.
steady state at the end of the simulation, andwill probably increase for an additional 2 millionyears. The carbonate accumulation £ux displays amajor increase during the Early Famennian, fol-lowing the increased delivery of alkalinity into theoceans related to the increased weatherability ofthe continents and weathering of carbonates dur-ing the sea-level fall (Fig. 8).
3.4. Continental phosphorus required
Finally, we used the COMBINE model in areverse mode. The two main disagreements be-tween calculated seawater N
13C values by theVEG simulation and measured N
13C values areobserved in the Upper rhenana conodont Zone(the model almost does not produce an isotopicexcursion), and at the F-F boundary (the calcu-lated rate of increase in N
13C value is too gradualin comparison to the measured data; Fig. 7e). Byadding to the VEG simulation an arbitrary pro-cess in order to reproduce the £uctuations in sea-water N
13C value, the model will calculate an at-mospheric pCO2 and a climate in agreement withthe available N
13C isotope data. However, the re-verse mode will not give any information aboutthe cause of these perturbations.
A convenient way to produce excursions in sea-water N13C value is to perturb the organic carbonsubcycle, since the N
13C value of the organic mat-ter is largely di¡erent from the mean seawaterN13C value. Assuming that the Frasnian excur-
sions are linked to the sea-level rises, we simplyadd a continental phosphorus £ux proportionallyto the change in sea level (PIN simulation). Suchadditional P £ux facilitates the onset of dysoxicconditions during the sea-level rises by increasingthe primary productivity within epicontinentalsurface waters. Starting from the VEG simulation,we found that an additional £ux of 8.0U109 mol/year of phosphorus (40% increase compared tothe pre-perturbation Devonian £ux) to the VEGscenario is required at the sea-level maxima toaccount for the observed shift in the carbon iso-tope composition, and to match the sharp in-crease in the N
13C values at the F-F boundary(Fig. 9e). Since the COMBINE model is nowworking in a reverse mode, the origin of this
£ux cannot be deduced from the simulation itself.We can only speculate on the origin of this £uxwhich could be an in£ux of nutrients from£ooded lagoons as a consequence of the rise insea level (Racki, 1993). However, once the PTESEN13C values are reproduced, we have some con¢-
dence in the output of the model describing the
Fig. 9. Results of the PIN simulation (description in thetext). See caption of Fig. 4 for more details.
geochemical cycles of carbon, oxygen and phos-phorus.
With this model, atmospheric pCO2 declinesduring both sea-level rises by 530 ppmv and 940ppmv, respectively. Atmospheric CO2 concentra-tions remain constant during the Early Famen-nian (Fig. 9f). The overall decline in globalmean temperature across the F-F boundaryequals 2.7‡C (4.4‡C at the south pole, 2.1‡C atthe equator; Fig. 10). The Panthalassa and Pro-to-Tethys epicontinental seas experience dysoxicconditions during the sea-level rise, essentiallydue to the intensi¢ed productivity in surfacewaters driven by the enhanced continental nu-trient supply and a coeval sluggish vertical mix-ing. As a consequence, the organic carbon burialincreases during the sea-level rise events reachingmore than 9U1012 mol/year at the sea-level max-ima, a value three times higher than the pre-per-turbation Frasnian value (Fig. 9d). Atmosphericoxygen now displays two stepwise increases dur-ing the Kellwasser event (about 4U1017 mol oxy-gen are added to the atmospheric reservoir duringboth events), followed by a slower increase relatedto the spreading of land plants (VEG simulation;Fig. 9b). The model predicts a rapid and short-lived increase in the carbonate deposition rateduring the sea-level rise events, essentially relatedto decreasing atmospheric pCO2, leading to lessacidic conditions in epicontinental seas (Fig. 8).
4. Conclusions
This study demonstrates the complexity of in-terpreting the short-term excursions in N
13C val-ues observed in the Late Frasnian and earliestFamennian. Various scenarios were tested in or-der to reproduce the measured N
13C values acrossthe F-F boundary using the COMBINE model,which is a model of the global biogeochemicalcycles fully coupled to an energy-balance climatemodel. A scenario exclusively based on oceanicprocesses fails to generate dysoxic conditions be-low 100 m in epicontinental seas and an increasein organic carbon deposition. There is an unre-solved paradox between the requirement of astrati¢ed ocean to produce anoxia, and the re-quirement of a sustained biological productivitywithin the photic zone of the ocean to ensurethe production and subsequent preservation oflarge quantities of organic carbon in the sedi-ments (strong productivity in the photic zone re-quires the existence of strong upwelling). By strat-ifying the ocean during the sea-level rises as aresult of the perturbation of the ocean verticalmixing through the formation of warm salinedeep waters in extended epicontinental seas, weproduced widespread anoxia, but no excursionin seawater N
13C values as well as no signi¢cantchange in atmospheric pCO2.
However, we demonstrate the importance ofthe three continental processes in generating theshort-term carbon isotope excursions in the LateFrasnian and Early Famennian.
(1) The weathering of carbonate platform ex-posed during the Early Famennian sea-level fallmight account for a maximum +0.7x positiveshift in seawater N
13C values.(2) The potential rapid spreading of land plants
across the F-F transition, through the coloniza-tion of about 30% of the continental surface, ac-counts for a further +1.0x positive shift in N
13Cduring the Early Famennian. The change in con-tinental weatherability results in the consumptionof 1400 ppmv of atmospheric CO2 across theboundary. The Early Famennian climate is calcu-lated to be 2.7‡C cooler than the Late Frasnianclimate (4.4‡C at the pole, and 2.1‡C at the equa-tor), in agreement with various sedimentological
Fig. 10. Di¡erence in surface temperature between the Uppertriangularis and Lower rhenana conodont Zones as a func-tion of latitude, as calculated by the PIN simulation.
and isotopic constraints (Streel et al., 2000; Joa-chimski and Buggisch, 2002). However, becauseof residence time consideration, such a scenariocannot explain the short-lived excursion in thecarbon isotope composition in the Upper rhenanaconodont Zone and the rapid increase in the N
13Cvalues in the latest Frasnian.
(3) These rapid positive shifts in the carbonisotope composition require an increase in thephosphorus £ux from the continent by about40% during the sea-level rises and coeval oceanstagnation, to sustain productivity in the photiczone. This additional phosphorus £ux possiblyoriginated from £ooded lagoons (Racki, 1993).As a result, the total burial of organic carbonincreases by a factor of 2 to 3 during the sea-levelrises. This organic carbon is accumulated underdysoxic conditions, and results in the formation ofthe two Kellwasser horizons. The oxygen contentof the atmosphere consequently increases by morethan 2.20U1018 mol (0.06 PAL) across the F-Fboundary, leading to a better oxygenation of theoceanic reservoirs in the Famennian compared tothe Frasnian.
Acknowledgements
Guy Munhoven and Louis M. FrancPois at theLPAP (Lie'ge) are greatly acknowledged for stim-ulating discussions about the model structure. Wethank Kevin Faure and an anonymous reviewerfor constructive comments and for improving thereadability of this contribution. We thank FinnSurlyk for valuable comments.
References
Algeo, T.J., Berner, R.A., Maynard, J.B., Scheckler, S.E.,1995. Late Devonian oceanic anoxic events and biotic crises:‘rooted’ in the evolution of vascular land plants? Geol. Soc.Am. Today 5, 63^66.
Algeo, T.J., Scheckler, S.E., 1998. Terrestrial-marine telecon-nections in the Devonian: links between the evolution ofland plants, weathering processes, and marine anoxic events.Phil. Trans. R. Soc. London B 353, 113^130.
Anderson, L.A., Sarmiento, J.L., 1994. Red¢eld ratio of re-mineralization determined by nutrient data analysis. GlobalBiogeochem. Cycles 8, 65^80.
Benitez-Nelson, C.R., 2000. The biogeochemical cycling ofphosphorus in marine systems. Earth Sci. Rev. 51, 109^135.
Berner, R.A., 1994. GEOCARB II: a revised model for atmo-spheric CO2 over Phanerozoic time. Am. J. Sci. 294, 56^91.
Boucot, A.J., Gray, J., 2001. A critique of Phanerozoic cli-matic models involving changes in the CO2 content of theatmosphere. Earth Sci. Rev. 56, 1^159.
Brady, P.V., 1991. The e¡ect of silicate weathering on globaltemperature and atmospheric CO2. J. Geophys. Res. 96,18101^18106.
Brass, G.W., Southam, J.R., Peterson, W.H., 1982. Warm sa-line bottom water in the ancient ocean. Nature 296, 620^623.
Buggisch, W., 1991. The global Frasnian-Famennian ‘Kell-wasser event’. Geol. Rundsch. 80, 49^72.
Caldeira, K., 1992. Enhanced Cenozoic chemical weatheringand the subduction of pelagic carbonates. Nature 357,578^581.
Copper, P., 1977. Paleolatitudes in the Devonian of Brazil andthe Frasnian-Famennian extinction event. Palaeogeogr. Pa-laeoclimatol. Palaeoecol. 21, 165^207.
Deines, P., 2002. The carbon isotope geochemistry of mantlexenoliths. Earth Sci. Rev. 58, 247^278.
Endal, A.S., So¢a, S., 1981. Rotation in solar-type stars, I,Evolutionary models for the spin-down of the Sun. Astro-phys. J. 243, 625^640.
FrancPois, L.M., Walker, J.C.G., 1992. Modelling the Phaner-ozoic carbon cycle and climate: constraints from the 87Sr/86Sr isotopic ratio of seawater. Am. J. Sci. 292, 81^135.
FrancPois, L.M., Godde¤ris, Y., 1998. Isotopic constraints onthe Cenozoic evolution of the carbon cycle. Chem. Geol.145, 177^212.
Freeman, K.H., Hayes, J.M., 1992. Fractionation of carbonisotopes by phytoplankton and estimates of ancient CO2
levels. Global Biogeochem. Cycles 6, 185^198.Froelich, P.N., Bender, M.L., Luedtke, N.A., 1982. The ma-
of the strontium and carbon cycles: relative importance ofcontinental erosion and mantle exchange. Chem. Geol. 126,169^190.
Godde¤ris, Y., FrancPois, L.M., 1996. Balancing the Cenozoiccarbon and alkalinity cycles: constraints from isotopic re-cords. Geophys. Res. Lett. 23, 3743^3746.
Godde¤ris, Y., Veizer, J., 2000. Tectonic control of chemicaland isotopic composition of ancient oceans: the impact ofcontinental growth. Am. J. Sci. 300, 434^461.
Hallam, A., Wignall, P.B., 1997. Mass Extinctions and TheirAftermath. Oxford University Press, Oxford.
Hayes, J.M., Strauss, H., Kaufman, A.J., 1999. The abundanceof 13C in marine organic matter and isotopic fractionationin the global biogeochemical cycle of the carbon during thepast 800 Ma. Chem. Geol. 161, 103^125.
Holser, W.T., Schidlowski, M., McKenzie, F.T., Maynard,J.B., 1988. Geochemical cycles of carbon and sulfur. In:Gregor, C.B., Garrels, R.M., McKenzie, F.T., Maynard,J.B. (Eds.), Chemical Cycles in the Evolution of the Earth.Wiley, New York, pp. 105^173.
Joachimski, M., Buggisch, W., 1993. Anoxic events in the LateFrasnian ^ causes of the Frasnian^Famennian faunal crisis?Geology 21, 675^678.
Joachimski, M., Ostertag-Henning, C., Pancost, R.D., Strauss,H., Freeman, K.H., Littke, R., Sinninghe Damste¤, J.S.,Racki, G., 2001. Water column anoxia, enhanced productiv-ity and concomitant changes in N
Joachimski, M.M., Pancost, R.D., Freeman, K.H., Ostertag-Henning, C., Buggisch, W., 2002. Carbon isotope geochem-istry of the Frasnian^Famennian transition. Palaeogeogr.Palaeoclimatol. Palaeoecol. 181, 91^109.
Joachimski, M.M., Buggisch, W., 2002. Conodont apatite N18Osignatures indicate climatic cooling as a trigger of the LateDevonian (F-F) mass extinction. Geology 30, 711^714.
Kump, L.R., Arthur, M.A., 1997. Global chemical erosionduring the Cenozoic: weatherability balances the budgets.In: Ruddiman, W.F. (Ed.), Tectonic Uplift and ClimateChange. Plenum Press, New York, pp. 399^426.
Kump, L.R., Arthur, M.A., Patzkowsky, M.E., Gibbs, M.T.,Pinkus, D.S., Sheehan, P.M., 1999. A weathering hypothesisfor glaciation at high atmospheric pCO2 during the LateOrdovician. Palaeogeogr. Palaeoclimatol. Palaeoecol. 152,173^187.
McGhee, G.R., 2001. The ‘multiple impacts hypothesis’ formass exctinction: a comparison of the Late Devonian andthe Late Eocene. Palaeogeogr. Palaeoclimatol. Palaeoecol.176, 47^58.
McLaren, D.J., 1970. Time, life, and boundaries. J. Paleontol.44, 801^815.
Mook, W.G., Bommerson, J.C., Staverman, W.H., 1974. Car-bon isotope fractionation between dissolved and gaseouscarbon dioxide. Earth Planet. Sci. Lett. 22, 169^176.
Munhoven, G., 1997. Modeling Glacial-Interglacial Atmo-spheric CO2 Variations: The Role of Continental Weather-ing. PhD Thesis, University of Lie'ge, Lie'ge.
Munhoven, G., FrancPois, L.M., 1996. Glacial-interglacial var-iability of atmospheric CO2 due to changing continentalsilicate rock weathering: a model study. J. Geophys. Res.101, 21423^21437.
Murphy, A.E., Sageman, B.B., Hollander D.J., Lyons, T.W.,2000. Black shale deposition and faunal overturn in the De-vonian Appalachian basin: clastic starvation, seasonalwater-column mixing, and e⁄cient biolimiting nutrient recy-cling. In: Huber, B.T., MacLeod, K.G., Wing, S.L. (Eds.),Warm Climates in Earth History. Cambridge UniversityPress, Cambridge, pp. 351^385.
Opdyke, B.N., Wilkinson, 1988. Surface area control of shal-low cratonic to deep marine carbonate accumulation. Pale-oceanography 3, 685^703.
Ormiston, A.R., Oglesby, R.J., 1995. E¡ect of Late Devonianpaleoclimate on source rock quality and location. Am. As-soc. Pet. Geol. Stud. Geol. 40, 105^132.
Petsch, S., Berner, R.A., 1998. Coupling the geochemicalcycles of C, P, Fe, and S: the e¡ect on atmospheric O2
and the isotopic records of carbon and sulfur. Am. J. Sci.298, 246^262.
Racki, G., 1993. Evolution of the bank to reef complex in theDevonian of the Holy Cross mountains. Acta Palaeontol.Pol. 37, 87^182.
Sandberg, C.S., Ziegler, W., 1996. Devonian conodont bio-chronology in geologic time calibration. Senckenb. Lethaea76, 259^265.
Sarmiento, J.L., Orr, J.C., Siegenthaler, U., 1992. A perturba-tion simulation of CO2 uptake in an ocean general circula-tion model. J. Geophys. Res. 97, 3621^3645.
Siegenthaler, U., Mu«nnich, K.O., 1981. 13C/12C fractionationduring CO2 transfer from air to sea. In: Bolin, B. (Ed.),Carbon Cycle Modeling. John Wiley and Sons, Chichester,pp. 249^257.
Streel, M., Caputo, M.V., Loboziak, S., Melo, J.H.G., 2000.Late Frasnian^Famennian climates based on palynomorphanalyses and the question of the Late Devonian glaciations.Earth Sci. Rev. 52, 121^173.
Tardy, Y., 1993. Climats, pale¤oclimats et bioge¤odynamique dupaysage tropical. In: Paquet, H., Clauer, N. (Eds.), Compte-rendu du colloque ‘Se¤dimentologie et ge¤ochimie de la sur-face’ a' la me¤moire de G. Millot. Colloque de l’Acade¤mie desSciences et du CADAS, Paris.
Thompson, J.B., Newton, C.R., 1988. Late Devonian mass
extinction: Episodic climatic cooling or warming. In:McMillian, N.J., Embry, A., Glass, D.J. (Eds.), Devonianof the World. Can. Soc. Petrol. Mineral. Mem. 14, pp. 29^34.
Tucker, R.D., Bradley, D.C., Ver Straeten, C.A., Harris, A.G.,Ebert, J.R., McCutcheon, S.R., 1998. New U^Pb zircon agesand the duration and division of Devonian time. Earth Plan-et. Sci. Lett. 158, 175^186.
Van Cappellen, P., Ingall, E.D., 1996. Redox stabilization ofthe atmosphere and oceans by phosphorus-limited marineproductivity. Science 271, 493^496.
Veizer, J., Ala, D., Azmy, K., Bruckschen, P., Buhl, D.,Bruhn, F., Carden, G.A.F., Diener, A., Ebneth, S., God-
de¤ris, Y., Jasper, T., Korte, C., Pawellek, F., Podlaha,O.G., Strauss, H., 1999. 87Sr/86Sr, N13C and N
