Phanerozoic Atmospheric CO Change

Earth-Science Reviews 54Ž. 2001 349–392 www.elsevier.comrlocaterearscirev

Phanerozoic atmospheric CO2 change: evaluating geochemical and paleobiological approaches

Dana L. Royer a,), Robert A. Berner a,1, David J. Beerling b,2 a Kline Geology Laboratory, Yale UniÕersity, P.O. Box 208109, New HaÕen, CT 06520-8109, USA b Department of Animal and Plant Sciences, UniÕersity of Sheffield, Sheffield S10 2TN, UK

Accepted 8 November 2000

Abstract

The theory and use of geochemical modeling of the long-term carbon cycle and four paleo-PCO2 proxies are reviewed and discussed in order to discern the best applications for each method. Geochemical models provide PCO2 predictions for the entire Phanerozoic, but most existing models present 5–10 m.y. means, and so often do not resolve short-term excursions. Error estimates based on sensitivity analyses range from "75–200 ppmV for the Tertiary to as much as "3000 ppmV during the early Paleozoic. 13 The d C of pedogenic carbonates provide the best proxy-based PCO2 estimates for the pre-Tertiary, with error estimates ranging from "500–1000 ppmV. Pre-Devonian estimates should be treated cautiously. Error estimates for Tertiary reconstructions via this proxy are higher than other proxies and models Ž."400–500 ppmV , and should not be solely relied upon. We also show the importance of measuring the d13C of coexisting organic matter instead of inferring its value from marine carbonates. The d13C of the organic remains of phytoplankton from sediment cores provide high temporal resolutionŽ up to 1034 –10 " " year.Ž , high precision 25–100 ppmV for the Tertiary to 150–200 ppmV for the Cretaceous . PCO2 estimates that can be near continuous for most of the Tertiary. Its high temporal resolution and availability of continuous sequences is advantageous for studies aiming to discern short-term excursions. This method, however, must correct for changes in growth ; rate and oxygen level. At elevated PCO2 Ž.750–1250 ppmV , this proxy loses its sensitivity and is not useful. The stomatal density and stomatal index of land plants also provide high temporal resolution Ž-102 year. , high " precision Ž.10–40 ppmV for the Tertiary and possibly Cretaceous PCO2 estimates, and so also is ideal for discerning short-term excursions. Unfortunately, this proxy also loses sensitivity at some level of PCO2 above 350 ppmVŽ which, currently, is largely undetermined. . Our analysis of the recently developed d11B technique shows that it currently is not yet well constrained. Most importantly, it requires the assumption that the boron isotopic composition of the ocean remains nearly constant through

) Corresponding author. Fax: q1-203-432-3134. E-mail addresses: [email protected]Ž. D.L. Royer , [email protected] Ž R.A. Berner . , [email protected] Ž.D.J. Beerling . 1 Fax: q1-203-432-3134. 2 Fax: q44-0114-222-0002.

0012-8252r01r$ - see front matter q 2001 Elsevier Science B.V. All rights reserved. PII: S0012-8252Ž. 00 00042-8 350 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 time. In addition, it assumes that there are no biological or temperature effects and that diagenetic alteration of the boron isotopic composition does not occur.

Keywords: carbon dioxide; paleoatmosphere; paleoclimatology; indicators

1. Introduction mination of the composition of air trapped in glacial iceŽ. e.g., Friedli et al., 1986; Petit et al., 1999 . During the last decade, numerous methods for However, this method is only useful for the past 400 evaluating past concentration of atmospheric carbon ka because of the absence of ice older than this. r dioxideŽ. PCO2 have been developed and or re- Thus, other methods have been applied to the older fined. The most reliable method has been the deter- geologic record. Quantifying paleo-CO2 is vital for

Nomenclature:

Symbols commonly used in text ) Ca2PCO ) 3 Cs2soil COŽ. mol cm C37:2 diunsaturated long-chained alkenones ) 2 y1 Ds2diffusion coefficient for soil COŽ. cm s r piap ratio of internal:ambient CO 2 S CO2 contributed by biological respiration z depth in a soil profileŽ. cm

z characteristic CO2 production depthŽ. cm adiffand ´ diff fractionation factor for diffusion during carbon fixationŽ. ‰ affand ´ kinetic fractionation factor for carbon fixationŽ. ‰ a PPand ´ combined fractionation for carbon fixationŽ. ‰ 13 dda C of the atmosphereŽ. ‰ 13 ddcc C of pedogenic carbonateŽ. ‰ 13 ddi2C of CO at the site of carbon fixationŽ. ‰ 13 ddocc C of marine carbonateŽ. ‰ 13 ddom C of organic matterŽ. ‰ 13 ddp C of photosynthateŽ. ‰ 13 dds2C of soil COŽ. ‰ 13 ddf C of soil-respired CO2 Ž. ‰ 11 ddÝB B of total dissolved BŽ. ‰ ´ free air porosity in soil

´a ambient atmospheric partial pressure of water vapor ´i intercellular partial pressure of water vapor ) y1 y3 fs2CO production rateŽ. mol s cm m growth rateŽ. dy1 r tortuosity D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 351 understanding climate dynamics, most notably its basis is dominated by the exchange of carbon be- effect on global temperatureŽ e.g., Sloan and Rea, tween rocks and the surficial reservoir consisting of 1995; Kothavala et al., 1999. through the so-called the atmosphereqbiosphereqoceansqsoils, and the

‘greenhouse effect’. PCO2 affects many other as- processes involved in the long-term cycle are differ- pects of the biosphere including particularly the ent from those that are involved in the short-term physiology, productivity and distribution of terres- cycle, which considers only the exchange of carbon trial vegetation. This in turn influences our interpre- within the surficial reservoirŽ. see Berner, 1999 . tation of the plant fossil recordŽ. Beerling, 1998b and Because the amount of CO2 in the atmosphere is exerts an important impact on the feedback between exceedingly small compared to the carbon fluxes vegetation and climate by changing the exchange of into and out of it, it is very difficult to calculate energy and water vapor between the land surface and changes in PCO2 due to imbalances between these the overlying atmosphereŽ. e.g., Bounoua et al., 1999 . fluxes over millions of yearsŽ Berner and Caldeira,

Quantifying the PCO2 of the ancient atmosphere, 1997. . Calculations of PCO2 from imbalances via therefore, provides a firmer basis for assessing the time-dependent modeling have been madeŽ Berner et linkages between PCO2 and the biosphere that are al., 1983; Lasaga et al., 1985. , but they require urgently required for understanding the geologic past treatment of fluxes of additional elements such as Ž.Beerling, 2000 . Ca, Mg, and S, as well as fluxes of carbon between Here, we critically consider the underlying theory, the ocean and atmosphere as well as between rocks practice, and applicability of geochemical modeling and the surficial reservoir. Because of their complex- of the long-termŽ. multimillion year carbon cycle ity, these models will not be covered here. Instead, and four paleo-PCO2 proxies. Several of the proxies simpler models dealing only with carbon, principally were developed for and first applied to the Quater- the GEOCARB and similar modelsŽ Budyko et al., nary, but their application to the pre-Quaternary 1987; Berner, 1991, 1994; Caldeira and Kasting, record will be emphasized here. The four proxies are 1992; Franc¸ois and Walker, 1992; Kump and Arthur, the d13C of phytoplankton, the d13C of pedogenic 1997; Tajika, 1998; Berner and Kothavala, 2001; carbonatesŽ. including the goethite method , stomatal Wallman, 2001. , will be discussed. Several other density and stomatal index, and the d11 B of marine modeling approaches to the long-term carbon cycle calcium carbonate. In considering the applicability of have been made, but they do not actually calculate the various approaches to the geologic record, their CO2 concentrations and will not be discussed here associated temporal resolutions and precision of ŽWalker et al., 1981; Caldeira, 1992; Raymo and

PCO2 estimates will be emphasized. Ruddiman, 1992; Godderis and Franc¸ois, 1995, 1996; Derry and France-Lanord, 1996; France-Lanord and Derry, 1997; Gibbs et al., 1997, 1999; Kump and 2. Geochemical modeling of the long-term carbon Arthur, 1999; McCauley and DePaolo, 1997; Raymo, cycle 1997; Franc¸ois and Godderis, 1998. . The GEOCARB model considers the input of

The level of atmospheric CO2 over geologic time CO2 to the atmosphere from the thermal breakdown can be estimated by constructing mass balance ex- of carbon in rocks resulting in volcanic, metamor- pressions for all the processes that bring about inputs phic, and diagenetic degassing. Additional CO2 in- and outputs of CO2 to and from the atmosphere. This put comes from the weathering of organic matter in is a daunting task at any timescale. The best one can rocks exposed on the continents. Carbon dioxide is do is to construct theoretical models and attempt to removed from the atmosphere by the weathering of devise rates and rate laws for processes involved in Ca–Mg silicate and carbonate minerals to form dis- the carbon cycle and how these rates may have solved bicarbonate in groundwater, followed by y changed over time. Since this review is concerned transport of the HCO3 by rivers to the oceans where y with pre-Quaternary PCO2 , only models that treat the HCO3 is removed as Ca–Mg carbonates in CO2 changes over millions of years will be dis- bottom sediments. Thus, carbon is transferred from cussed. The carbon cycle on a multimillion year the atmosphere to carbonate rocks. In some models 352 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

s Ž.e.g., Wallman, 2001 , direct weathering of Ca–Mg where: Fwc, F wg rate of release of carbon to the silicates to carbonates on the seafloor is also consid- atmosphereqbiosphereqoceansqsoils system via ered. The burial of organic matter in both marine and the weathering of carbonatesŽ. c and organic matter s non-marine sediments also removes CO2 fixed by Ž.g; Fmc, F mg rate of release of carbon to the photosynthesis from the atmosphere. atmosphereqbiosphereqoceansqsoils; system via Because the mass of carbon exchanging with the metamorphicrvolcanicrdiagenetic breakdown of s rocks via volcanism, weathering, etc., over millions carbonatesŽ. c and organic matter Ž. g ; Fbc, F bg of years is so much larger than that present at any burial rate of carbon as carbonatesŽ. c and organic q s 13 given time in the surficial reservoirŽ atmosphere matterŽ. g in sediments; dcg, d d C valueŽ per biosphereqoceansqsoils. , it can be assumed that mil.Ž.Ž. for carbonates c and organic matter g under- s 13 for each large time step in the calculation the sum of going weathering; d bc d C valueŽ. per mil for s fluxes from rocks to the surficial system are essen- carbonates undergoing burial in sediments; ac tially equal to the sum of fluxes back to the rocks. carbon isotope fractionationŽ. per mil between or- This is what is done in GEOCARB modeling and it ganic matter and carbonates both undergoing burial is this assumption of quasi-steady state that allows in sediments. These equations represent the quasi- the calculation of PCO2 over geologic time. It repre- steady state equivalence of the sum of CO2 inputs sents the reasonable idea that extremely large masses and outputs for each time step in the modelingŽ for of carbon, that would otherwise result if the fluxes GEOCARB the time step is 1 million year. . did not balance, cannot be stored in any portion of From these equations and additional algebraic the surficial reservoirŽ for example, the continents expressions for each of the input fluxesŽ terms on the can accommodate just so much biomass, and excess left-hand side of the equations. , and how they change y HCO3 in the oceans will eventually precipitate as with time, one can solve the two equations for each CaCO3 due to excessive supersaturation—see Berner flux F as a function of time. This requires data on and Caldeira, 1997 for further discussion. . the following subjects as they affect continental weathering: total land area and mean continental elevationŽ. from paleogeographic maps , relative ar- 2.1. Method of calculation eas of exposure of silicate vs. carbonate rocksŽ from paleogeologic reconstructions. , global river runoff Carbon cycle models treat a large variety of pro- ŽŽ.from general circulation climate models GCMs cesses and, therefore, involve mathematical treat- applied to ancient paleogeographies. , land tempera- ment that is far more complex than that used for the tureŽ. from GCMs , the rise of large vascular land other paleo-CO2 proxies discussed elsewhere in the plants and their quantitative effect on weathering present review. For this reason, the entire GEO- Ž.from modern weathering studies , the effect of solar CARB model is not presented here but only sum- evolution on mean global temperature and river marized qualitatively with a minimum of mathema- runoffŽ. from GCMs , the effect of changes in PCO2 tical expressions. Further details of the modeling on climate and river runoffŽ atmospheric greenhouse can be found in the original papersŽ Berner, 1991, effect from GCMs. , and the effect of changes in

1994; Berner and Kothavala, 2001; variants of GEO- PCO2 on the rate of growth and weathering by land CARB-type modeling are presented by others refer- plantsŽ. from modern plant experiments . enced above, but they rest on similar reasoning. . The Inputs of CO2 to the atmosphere are parameter- fundamental basis of the GEOCARB model is two ized in the GEOCARB model in terms of: seafloor carbon mass balance equations: spreading rate as a guide to global tectonic degassing Žfrom seafloor arearage relations and paleo-sea level q q q s q FwcF mcF wgF mgF bcF bg Ž.1 data. , and the rise of calcareous plankton as they affect the amounts of CaCO in deep sea sediments d q qd q 3 cwcmcgwgmgŽ.F F Ž.F F subjected to heating and degassing during plate subduction. Organic carbon and calcium carbonate burial sd q d ya bcF bcŽ. bc cF bg Ž.2 rates in sediments are derived from the carbon iso- D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 353

s topic record of carbonates being buried at each time biological activity due to land plants ŽŽ.ftE 1at s r step and Eqs.Ž. 1 and Ž. 2 above. A summary of these present.Ž. ; ftD river runoff Ž.t river runoff at pre- s factors considered by the GEOCARB model is shown sent due to changes in paleogeography; ftBŽ. in Table 1. dimensionless feedback factor representing the ef-

The calculation of PCO2 in GEOCARB modeling fects of CO2 and temperature on weathering rate. Ž.sensu strictu is derived by determining the rate of The value of FwsiŽ.0 can be estimated from mod- weathering of Ca–Mg silicates on land to Ca–Mg ern river water data and values of ftRHŽ., ft Ž., y 0 carbonates in the oceans. This is equivalent to Fbc ftEDŽ., and ft Ž. are obtained via the methods out- Fwc from Eq.Ž. 1 . The expression for silicate weath- lined above. With these results Eq.Ž. 3 can then be ering is: solved for ftBBŽ.. The parameter ft Ž .represents the effects on silicate mineral weathering of changes in FtsF yF wsiŽ. bc wc temperature and runoff as they are, in turn, affected s 0.65 by changes in PCO and solar radiation. Also in- ftftftftftBHREDŽ. Ž. Ž. Ž. Ž. F wsi Ž.0 2 cluded is the effect of PCO on plant-mediated Ž.3 2 weathering. The parameter ftBŽ.represents negative s where: FtwsiŽ. rate of Ca–Mg silicate weathering feedback against a runaway greenhouse or icehouse with ultimate conversion to Ca–Mg carbonates; the climate. Derivation of the resulting complex alge- value for the present is designated as FwsiŽ.0; ft H Ž. braic expression for ftBŽ.allows solution of it for s effect on weathering of global mean land tempera- PCO2B once the actual value of ftŽ.has been deter- ture at some past time Ž.t rpresent global mean land mined. s temperature; ftRŽ. effect on weathering rate of mean continental relief at time Ž.t rmean continental 2.2. Problems with the method s relief at present; ftEŽ. dimensionless parameter expressing the dependence of weathering rate on soil A large number of assumptions go into GEO- CARB and the similar models of Caldeira and Kast- ingŽ. 1992 and Tajika Ž. 1998 . Simple error analysis Table 1 is impossible but a rough idea of error can be Ž Outline of processes in the GEOCARB II model after Berner, determined by sensitivity analysis. This was done in 1991, 1994. the GEOCARB modelingŽ Berner, 1991, 1994; Weathering of silicates, carbonates, and organic matter . Ž.1 Topographic relief as affected by mountain uplift Berner and Kothavala, 2001 by varying each of the Ž.silicates and organic matter values for ftREŽ., ft Ž., etc., from no change from the Ž.2 Global land area Ž carbonates . present to geologically extreme maximum and mini- Ž.3 Global river runoff and land temperature as affected by mum values and examining the effect on CO2 . For changes in continental area and position example, varying the effect of different groups of Ž.4 Relative areas of land underlain by silicates vs. carbonates Ž.5 Rise and evolution of vascular land plants land plants on chemical weathering of Ca–Mg sili- Ž.6 Enhancement of weathering flux by changes in cates results in large changes in calculated CO2 level global temperature and runoff Ž.Berner, 1991, 1994; Berner and Kothavala, 2001 . Ž.a Due to evolution of the sun Varying groups of related parameters, such as the Ž. Ž . b Due to changes in atmospheric CO2 greenhouse effect assumption of no change in tectonic degassing, no Ž.7 Enhancement of plant-mediated weathering due to fertilization by atmospheric CO change in paleogeography, no evolution of land 2 plants, etc., has also been done. Such sensitivity Thermal degassing of CO2 from Õolcanism, diagenesis, analysis has allowed the construction of crude error and metamorphism A B Ž.1 Changes in global seafloor spreading rate margins for the best estimate of CO2 concentration over Phanerozoic time. The error margins are Ž.2 Transfer of CaCO3 between platforms and the deep sea large and show that the Abest estimatesB are proba- Burial of organic matter and carbonates in sediments Ž. Ž.1 Calculated via mass balance from sum of input fluxes bly good to only a factor of about two see Fig. 13 . Ž.2 Relative proportions derived from carbon isotopic data Another problem lies with the limits of temporal resolution. Data are input into the GEOCARB model 354 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 every 10 million yearsŽ every 5 million years for amenable to modification and improvement. The some Cenozoic isotopic data. . As a result, the model method is best evaluated by its agreement or lack of cannot delineate short-lived events such as the agreement with the other methods. As an example, Permo-Triassic and Cretaceous-TertiaryŽ. K–T ex- there is surprisingly good agreement, within error tinction, the late Ordovician glaciation, and the Pale- ranges for each method, between the PCO2 values ocene–EoceneŽ. P–E d13C excursion. The GEO- calculated via GEOCARB modeling and those de- CARB model is intended to treat only more gradual rived from the study of carbonate paleosolsŽ see Fig. changes over many millions of years. However, this 14. . does not preclude the application of similar carbon cycle mass balance modeling to shorter timescales as has been done by Gibbs et al.Ž. 1997 for the late 3. d13C of phytoplankton Ordovician glaciation. Even with sensitivity analysis the model calcula- The carbon isotopic composition of biomass is a tions are no better than the various assumptions function of the carbon source, the carbon assimila- made by the GEOCARB model. Some of the larger tion pathway, and the biosynthesis and metabolism problems, ranked in order of more-to-less serious, of the assimilated organic carbon. In sedimentary are: organic matter, diagenetic processes may also be important. In the case of autotrophs, the carbon 1. The content of CaCO3 in subducting sediments assimilation pathway often strongly alters the carbon is virtually unknown before the JurassicŽ 150 isotopic signature relative to the carbon source. For Ma. . example, the equilibrium and kinetic isotope effects 2. The estimate of paleorelief, and its effect on associated with photosynthesis consistently fraction- weathering, is poorly known. ate strongly against 13C. One factor that affects these 3. Seafloor spreading rates before the Jurassic are fractionations in phytoplankton is the concentration not represented simply by changes in sea level, wx of CO22 dissolved in waterŽ COŽaq. .Ž Degens et al., as is assumed by the model. 1968; McCabe, 1985. . This relationship is often 4. River runoff and land temperature calculations inverted and applied to the geologic past to estimate are based on GCM calculations for flat conti- paleoatmospheric CO2 , ranging from the late Trias- nents. Recalculation based on realistic topogra- sic to QuaternaryŽ Popp et al., 1989; Jasper and phy is needed. Hayes, 1990; Freeman and Hayes, 1992; Pagani et 5. The effect of large shifts in the proportions of al., 1999a,b. . basalt vs. granite weathering has not been evaluated. 3.1. Photosynthetic fractionation of 13C 6. Little is known of how changes from gym- nosperms to angiosperms affected the rate of Two processes fractionate the carbon pool during plant-mediated weathering. photosynthesis: the equilibrium isotope effect of CO2 7. Global degassing may not scale linearly with diffusion at the boundary layer of the photosynthe- seafloor spreading rate as assumed. sizer, and the kinetic isotope effect of CO2 fixation. 8. Other factors have been ignoredŽ midplate su- The difference in diffusivity for two isotopes in a perplume degassing, seafloor basalt weather- given gas is proportional to the square root of their ing, etc.. . reduced massesŽ. Mason and Marrero, 1970 ; in air, 13 12 CO22 is 4.4‰ less diffusive than COŽ Craig, 2.3. Summary 1953.Ž , and for water the difference is 0.7‰ O’Leary, 1984. . This fractionation factor can be expressed as The carbon cycle modeling method has many the following: inherent problems, but it at least considers and dis- q 1000 da cusses the processes that actually bring about changes a s 4 diff q Ž. 1000 di in atmospheric CO2 over geologic time. It is always D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 355

a s d s where difffractionation due to diffusion, a For phytoplankton, the following values for ´diff d13 d sd13 C of the atmosphere, and i C at the site of and ´f can be substitutedŽ. see above : fixation. The kinetic effect associated with carbon ´ ' q r fixation is usually driven by rubiscoŽ ribulose-1,5-bi- Pia0.7 24.3 p p .8Ž. phosphate carboxylase oxygenase. , the enzyme that In most vascular plants, stomataŽ pores through catalyzes the reaction between RuBPŽ ribulose-1,5- which plants exchange gases and other constituents biphosphate.Ž and CO to form PGA 3-phosphog- 2 with the atmosphere. help regulate CO within their lyceric acid. , a 3-carbon acid. In vitro, there is a 2 mesophyll. If the CO external to a leaf rises or 29‰ fractionation against13 C at 258CŽ Roeske and 2 carbon assimilation rates drop, stomata-bearing plants O’Leary, 1985. . In vivo, however, the value for respond by reducing their stomatal pore area, and vascular plants is closer to 27‰Ž Evans et al., 1986; vice-versa, so that p rp usually remains at approxi- Farquhar et al., 1982. and 25‰ for phytoplankton ia mately 0.7Ž Polley et al., 1993; Ehleringer and Cer- ŽLaws et al., 1995; Bidigare et al., 1997; Popp et al., ling, 1995; Beerling, 1996; Bettarini et al., 1997; 1998; Hayes et al., 1999. . These discrepancies are Arens et al., 2000. , particularly for unstressed plants. likely due to both a drop in CO from the intercellu- 2 Phytoplanktons lack stomata and consequently have lar spaces to the site of fixation and a f10% of less control over p rp . Furthermore, the diffusion of total carbon fixation via PEP carboxylase, an enzyme ia CO in water is about 10 4 times slower than in air, that only mildly fractionates against13 CŽ Farquhar 2 suggesting again less internal control over p rp . and Richards, 1984. . The kinetic fractionation factor ia One might, therefore, expect a stronger correlation in can be expressed as the following: wx phytoplankton between CO2Žaq. and ´ P. q 1000 di a s 5 f q Ž. 3.2. Correlation between[] CO and ´ 1000 dp 2(aq) P

s McCabeŽ. 1985 experimentally quantified the re- where af kinetic fractionation associated with car- s 13 lationship betweenwx CO and ´ using mixed bon fixation, and dp d C of the photosynthate. 2Žaq. P Although the fractionations associated with photo- algal populations from several New Zealand lakes: synthesis are highly invariant, for a given d , d can ap ´ s Ž.17.0"2.2 log CO y3.4 vary by as much as 15‰, particularly for photosyn- P2Žaq. Ž thesizers in stressed environments e.g., water stress 2F CO F74 mM9Ž. in land plants, nutrient stress.Ž. . Farquhar et al. 1982 2Žaq. proposed that this variation is principally a function where the uncertainty represents the 95‰ confidence of the ratio of internal CO to ambient CO Ž.p rp . 22iainterval. Rau et al.Ž. 1989, 1991b measured dp and In essence, the d13C of inorganic carbon within wx r calculated CO2Žaq. for the South Atlantic Weddell photosynthesizers follows a Rayleigh distillation pro- Sea and the Drake Passage, and found the following r cess, with piap controlling how closed the system relationships: is. This relationship can be expressed as the follow- sy y ingŽ. after Farquhar et al., 1982; Popp et al., 1989 : Ž.S. Atlantic da20.8 CO Žaq. 12.6

s q y r 8- CO -24 mM10 a Pdifffdiffiaa Ž.a a p p Ž.6 2Žaq. Ž. d sy y s Ž.Drake Passage a20.90 CO Žaq. 9.40 where a P fractionation factor integrating the two isotopic fractionations described aboveŽŽ. Eqs. 4 and 15- CO -23 mM.Ž. 11 Ž..5 . This can be recast in terms of ´ values, where 2Žaq. ´' ay = 3 Ž.1 10 : Jasper and HayesŽ. 1990 established a similar relationship using reconstructed ´ values from a ´ '´ q ´ y´ r P PdifffdiffiaŽ.p p .7 Ž.late Quaternary hemipelagic sediment core and cor- 356 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

wx Fig. 1. Reported values for the relationship between ´ P2and COŽaq. using various techniques and settingsŽ. oceanic and lacustrine . Equations given in textŽŽ.Ž.Ž.. Eqs. 9 , 12 – 17 .

wx responding CO2Žaq. values calculated from the Vos- Hinga et al.Ž. 1994 experimentally grew the di- tok ice core: atom Skeletonema costatum at three temperatures. s y After accounting for possible pH effects, the follow- ´ P232.9log CO Žaq. 14.3 ing relationships were observed: - - 5 CO2Žaq. 8 mM.Ž. 12 Ž.98C ´ s11.21log CO q8.99 Hollander and McKenzieŽ. 1991 also generated an P2Žaq. wx - - estimate of this relationship, calculating CO2Žaq. 10 CO2Žaq. 50 mM15Ž. and ´ over an annual cycle in Lake Greifen, P 8 ´ s q Switzerland: Ž.15 C P213.06log CO Žaq. 8.58 ´ s y - - P211.64log CO Žaq. 3.56 5 CO2Žaq. 125mM16Ž. 10- CO -90 mM.Ž. 13 s q 2Žaq. Ž.258C ´ P29.09log CO Žaq. 1.89 Freeman and HayesŽ. 1992 compiled GEOSECS 17- CO -115mM.Ž. 17 data from several ocean basins, and found the fol- 2Žaq. wx lowing relationship between calculated CO2Žaq. and The results of the five studies solving for ´ P are ´ P: shown in Fig. 1. Significant variations among the s q estimates exist. Using Henry’s law to convert ´ P212.03log CO Žaq. 1.19 wxCO to PCOŽ after Freeman and Hayes, 1992; - - 2Žaq. 2 8 CO2Žaq. 25mM.Ž. 14 see Section 3.5. , Miocene ´ P values of 15.8‰ D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 357

Ž.Freeman and Hayes, 1992 yield a PCO2 estimate likely follow the phytoplankton trend because their of 400 ppmV using the regression of McCabeŽ. 1985 carbon is, at best, several steps removed from pri- but 1400 ppmV using Hollander and McKenzie mary marine photosynthate. Ž.1991 ; early Cretaceous values of 21.2‰ Ž Freeman In response to this limitation, specific biomarkers and Hayes, 1992. yield an estimate of 1000 ppmV for photosynthesis have been identified and mea- using McCabeŽ. 1985 but 4900 ppmV using Hollan- sured in lieu of bulk organic matterŽ Hayes et al., der and McKenzieŽ. 1991 . Beyond the differences 1987, 1989b; Jasper and Hayes, 1990. . Geopor- among these logarithmic relationships, it is not clear phyrins, derived from the aromatic nucleus of based on the above discussion whether the relation- chlorophyllŽ. Ekstrom et al., 1983; Hayes, 1993 , are wx ship between CO2Žaq. and ´ P should be logarithmic one such biomarker used in paleoatmospheric CO2 or linearŽ Rau et al., 1991a; Freeman and Hayes, reconstructionsŽ Popp et al., 1989; Freeman and 1992.Ž . This will be discussed in detail below Sec- Hayes, 1992. . While terrestrial plants are a potential tion 3.4. . source of geoporphyrins, the presence of geoporphyrins in marine settings is probably negligible

3.3. Determining paleo-´P Ž.Hayes et al., 1989b . The appropriate isotopic path- ways are shown in Fig. 2. In modern plants, geopor-

As defined aboveŽŽ.. Eq. 7 , ´ P represents the phyrin precursorsŽ. e.g., chlorophyllide are enriched fractionation of 13C due to photosynthesis. For most in13 C relative to bulk biomass by f0.5‰Ž Hayes et modern phytoplankton, ´ P ranges between 10‰ and al., 1987. . Based on theoretical considerations, the 20‰. Calculating this value is not straightforward, conversion of this precursor to geoporphyrins proba- r particularly in fossil material where piap is un- bly does not involve any isotopic fractionation. If a known. Instead, modern relationships between what fractionation does existŽ e.g., preferential decomposi- can be measuredŽ. e.g., carbonates, organic carbon tion of12 C. it may, for example, be expected to be and ´ PPare used to infer paleo-´ . Fig. 2 shows the large in highly oxygenated environmentsŽ Hayes et modern isotopic relationships between ´ P and car- al., 1989b. . The effects of thermal diagenesis have bonates and organic carbon. also been shown to be negligible, and can always be 13 The use of bulk organic carbon to estimate ´ P verified by comparing geoporphyrin d C values ŽArthur et al., 1985; Dean et al., 1986; Kump et al., with those of another class of compounds, for exam- 1999. likely leads to erroneous estimates of paleo- ple geolipidsŽ. Hayes et al., 1989a .

CO2 Ž. Hayes et al., 1989b; Pagani et al., 2000 due Diunsaturated long-chained alkenonesŽ. C37:2 are both to the input of terrigenous organic matter and also used as a biomarkerŽ Jasper and Hayes, 1990; non-primary marine photosynthate. As discussed Jasper et al., 1994; Pagani et al., 1999a,b. , but are aboveŽ. Section 3.1 , terrestrial plants are better preserved in sediments back only to the Cenomanian r equipped to regulate piap , and as a consequence Ž.Žca. 95 Ma Farrimond et al., 1986 . , and are scarce their carbon isotopic variability over geologic time in pre-Neogene sediments. Unlike geoporphyrins,

Ž.where large PCO2 fluctuations likely occurred is C37:2 alkenones are specific to certain haptophytic much smaller than that for marine organic carbon algaeŽ.Ž. Prymnesiophyceae Conte et al., 1994 , en- ŽDean et al., 1986; but see Bocherens et al., 1993; suring that their isotopic signatures are not diluted by Jones, 1994. . Likewise, carbon isotopic trends in terrestrial contaminants. Thus, when using these wx non-photosynthetic organic material will also not alkenones, one can eliminate the CO2Žaq. –´ P rela-

Fig. 2. Isotopic relationships between preserved carbonates and organic matter. Values are for d13 C. Modified after Hayes et al.Ž. 1989b . 358 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

tionships not based solely on these haptophytic algae Given the above guidelines, ´ P can be calculated Ž.see Fig. 1 . In the modern open-ocean, C 37:2 as follows: alkenones are only synthesized by Emiliania huxleyi q dd 1000 and its relative Gephyrocapsa oceanica ŽMarlowe et ´ ' y1 =1000fd yd 20 Pdpq Ž. al., 1990; Conte et al., 1994. , both of which are most dp 1000 common at mid-latitudes. Laboratory analyses find f where d sd13C of CO , and d sd13C of the C37:2 alkenones 4‰ lighter than bulk haptophytic d2Žaq. p biomassŽ Jasper and Hayes, 1990; Bidigare et al., bulk primary photosynthate. Assuming no carbonate d 1997; Popp et al., 1998. . In addition, the unsatura- diagenesis, d can be calculated from the organic tion ratios of C alkenonesŽwxw C r C qC x. matter’s associated calcite. Assuming no secondary 37 37:2 37:2 37:3 d are correlated with temperatureŽ Brassell et al., 1986; processing of the organic biomarker, p can be Prahl and Wakeham, 1987. , and have been useful for calculated from the biomarker. For geoporphyrins, f alkenone-based paleo-CO work in the Quaternary the corresponding fractionation factor is 0.5%, 2 f Ž.Jasper et al., 1994 , but not the Miocene Ž Pagani et and for C37:2 alkenones, 4‰. Thus, using the al., 1999a. . alkenone biomarker for sediments with a shallow As shown in Fig. 2, at least two fractionations water paleotemperature of 258C and the fractionation factors in Eqs.Ž. 18 and Ž. 19 , ´ can be estimated exist between CO2Žaq. and preserved carbonates. P 13 y by: CO2Žaq. is enriched in d C when converted to HCO3 and again further to calcite. A common value as- q dd 990 signed to these combined fractionationsŽ. at 258Cis ´ sy1 =1000.Ž. 21 p d q 10.8‰Ž. e.g., Popp et al., 1989; see Fig. 2 . This p 996 value can vary among species, for example, 9.3‰ for Globigerinoides ruber at 248CŽ Jasper and Hayes, ´ 1990. . The fractionations are also temperature de- 3.4. Other confounding factors in determining P pendent, and recent paleo-CO2 studies have used 18 shallow water d O reconstructed temperatures to As discussed aboveŽŽ.. see Eq. 7 , ´ P is a function r wx solve for these fractionation factorsŽ Freeman and of piap , where p arepresents CO 2Žaq. just outside Hayes, 1992; Pagani et al., 1999a,b. . One quantifica- the boundary layer of the cell. This CO2 need not tion of this dependency isŽ after Romanek et al., equal the global mean concentration, e.g., in areas of 1992; Mook et al., 1974, respectively. : upwelling or in stagnant, stratified watersŽ Popp et al., 1989; Freeman and Hayes, 1992; Goericke et al., wx y s y 1994. . The global mean CO2Žaq. is required to ´calciteCO 2Žg. 11.98 0.12T Ž.8C18 Ž. accurately estimate PCO2 Ž. see Section 3.5 . One ´ yCO s0.19y373rT Ž.K19 Ž . approach to minimize this potential bias is sampling CO2Žaq. 2Žg. sites associated with low productivityŽ Pagani et al., 1999a,b. . so that at 258C, ´s10.0‰. Different species may fractionate carbon differ- Diagenetic processes such as reprecipitation and ently during photosynthesis. Hinga et al.Ž. 1994 found differential dissolution can also affect isotopic com- ´ P values 8–10‰ lighter in E. huxleyi Ža hapto- positions. The presence of recognizable nannofossils phytic alga.Ž. than in S. costatum a diatom under and frequency of bioturbation and incomplete ce- identical conditions. This problem can be largely mentation are typically used to gauge the severity of removed by using C37:2 alkenones, since their pro- diagenesisŽ. Hayes et al., 1989b; Pagani et al., 1999a . duction is restricted to one taxon. r wx Comparison of Sr Ca ratios with that in unaltered At low CO2Žaq. , some types of phytoplankton y contemporaneous calcite can also be usedŽ Hayes et appear to actively transport HCO3 Ž Hinga et al., al., 1989b. since SrrCa ratios are lower in non-bio- 1994; Laws et al., 1995, 1997; Popp et al., 1998; but genic calcite than in biogenic calciteŽ Baker et al., see Bidigare et al., 1997.Ž. . Laws et al. 1995 con- 1982. . cluded that the diatom Phaeodactylum tricornutum D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 359

y wx probably actively transports HCO32 when CO Žaq. Growth rateŽ. Fry and Wainwright, 1991 and cell - 10 mM. Bidigare et al.Ž. 1997 and Laws et al. geometryŽ. Popp et al., 1998 can also influence ´ P. Ž.1997 report a similar threshold for other species. If If pi is purely supplied by diffusion in phytoplank- the conversion from bicarbonate to carbon dioxide tonŽ. i.e., no active transport , the growth rate m y1 occurs intracellularly, the subsequent enhanced Žd. is proportional to the difference between pa growth rateŽ. see Fig. 3 will likely offset the heavy and pi: carbon supplied by the HCOy Ž. Laws et al., 1995 . If 3 ms y the conversion occurs extracellularly, the resulting kp1akp 2i.22Ž. elevated growth rate is still expected but thewx CO 2Žaq. If this is solved for p and substituted into Eq.Ž. 7 , supplied by the bicarbonate will probably be similar i the following relationship is found where growth rate to the bulk CO2P . This, in turn, may affect ´ ŽLaws y is negatively correlated with ´ ŽRau et al., 1992; et al., 1995. . This extracellular conversion of HCO P 3 Franc¸ois et al., 1993; Goericke et al., 1994; Laws et to CO is probably the most common active 2Žaq. al., 1995; Bidigare et al., 1997, 1999b; but see Hinga pathwayŽ. Laws et al., 1997 . Additionally, many y et al., 1994; Popp et al., 1998, 1999. : types of phytoplankton fix HCO3 with PEP carboxylase, which lead to lower ´ valuesŽ see Section s q y y r r P ´ Pdifffdiff1´ Ž.Ž.´ ´ k m p a2k .23 Ž. 3.1. . Most importantly, this rate of fixation may wx depend on CO2Žaq. , and thus covary with PCO2 This equation provides a mechanism for the loga- Ž.Hinga et al., 1994; Reinfelder et al., 2000 . rithmic behavior previously observed when ´ P was

rwx s y1 Fig. 3. Relationship between ´ P2and m COŽaq. , where m growth rateŽ d. . Data include mean for the modern ocean and experimental results under light:dark cycles of 24:0 and 12:12 h. Regression equation for 24:0-h cycle: ysy0.015xq0.371 Žns5; r 2 s0.97. . Data from Laws et al.Ž. 1995 . 360 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

wx plotted against pa2Ž COŽaq. .Ž Fig. 1 . . In contrast, late, changes in cell membrane diffusivity could also ´ mr a linear relationship exists between P and affect ´ P Ž.Goericke et al., 1994 . Finally, the diatom wx CO2Žaq. Ž. Fig. 3 . Since experimental ´ P values are P. tricornutum was recently found to discriminate f rwxs 13 consistently 25‰ when m CO2Žaq. 0Ž Laws et more strongly against CŽ. 1.4‰ in a 32.5% vs. al., 1995; Bidigare et al., 1997; Popp et al., 1998; see 22% O2 atmosphereŽ. Berner et al., 2000 . The effect Fig. 3. , k12and k must be proportional to each is ascribed to carbon recycling via photorespiration, other. The slopes, however, need not be equal for which is partially a function of the O22 to CO ratio. different species. Popp et al.Ž. 1998 calculated, how- This ratio was probably higher than today during the ever, that a 20-fold difference in slope among four late Paleozoic and lower during, for example, the species when experimentally plotting ´ P vs. early and mid-MesozoicŽ Berner and Canfield, 1989; rwx m CO2Žaq. was removed when cell geometryŽ i.e., Berner et al., 2000. . Since photorespiration in mod- cellular carbon content to surface area ratio. was ern phytoplankton is lowŽ. Burns and Beardall, 1987 , considered. Popp et al.Ž. 1999 observed similar be- this effect may only be important for late Paleozoic r havior in the modern Southern Ocean. Thus, k12k reconstructions. may largely be a function of the carbon content to cell surface area ratio. 3.5. ConÕerting[] CO to PCO Fossil studies are generally limited in quantifying 2(aq) 2 growth rate and cell geometry. Bidigare et al.Ž 1997, Oncewx CO is calculated, it must be converted 1999a.Ž observed a striking correlation r 2 s0.82. in 2Žaq. y to PCO . According to Henry’s law, this relationship modern oceans betweenw PO3 x and ´ , and as- 2 4P is a partial function of temperature. For example, the cribed the covariation to growth rates. Assuming a wx ´ increase in CO2Žaq. solubility with decreasing tem- P maximum of 25‰, this relationship is as follows: wx perature explains the anomalously high CO2Žaq. and y corresponding ´ values at high latitudesŽ Rau et Ž.25y´ CO s158 PO3 q49. Ž. 24 P P2Žaq. 4 al., 1989, 1991b; Popp et al., 1999; Andrusevich et al., 2000. . Salinity also influences this relationship, Pagani et al.Ž. 1999a,b assumed k sk , solved 12 but is generally disregarded in paleostudiesŽ e.g., Eq.Ž. 23 for m, then substituted Eq. Ž. 24 for m in Eq. Pagani et al., 1999a,b. . One also assumes equilib- Ž.23 . Fossil sites were then carefully selected to rium between CO and atmospheric COŽ e.g., correspond with stable, low productivity areas in the 2Žaq. 2 y Freeman and Hayes, 1992; Pagani et al., 1999a,b. . present-day so that modernw PO3 x values for these 4 Large sections of the modern ocean are up to "50 areas could be applied to the fossil sites. Given these y ppmV out of equilibrium with the atmosphereŽ Tans wPO3 x estimates, ´ could then be calculated. Bidi- 4P et al., 1990. , and slightly larger values have been gare et al.Ž. 1997 suggested CdrCa ratios might also y calculated for the QuaternaryŽ. Jasper et al., 1994 , be useful as aw PO3 x proxy. Cell geometry is more 4 making this a dangerous assumption. Pagani et al. difficult to quantify in fossil studies, where in mod- Ž.1999a,b selected low productivity sites, which typi- ern plankton can be approximated by their volume to cally are closest to equilibrium with respect to CO . surface area ratioŽ. Popp et al., 1998 . This variable 2 may be insignificant in C37:2 alkenone studies, however, since they are restricted to a few species that in 3.6. Summary modern oceans show little volume to surface area variationŽ. Pagani et al., 1999a . Thus, assuming k1 Several studies have estimated paleoatmospheric s 13 k2 in these studies may not introduce large errors. CO2 from the d C of marine phytoplankton, both In addition, ´ P in alkenone-producing species ap- for the QuaternaryŽ Jasper and Hayes, 1990; Rau et pears insensitive to growth rateŽ. Popp et al., 1998 . al., 1991a; Jasper et al., 1994. and pre-Quaternary Hinga et al.Ž. 1994 demonstrated that a shift in Ž.Freeman and Hayes, 1992; Pagani et al., 1999a,b . pH from 7.9 to 8.3 induced a ´ P change of 10‰. Interestingly, White et al.Ž. 1994 used an analogous Little other evidence concerning this factor exists. technique for mosses, which lack foliar stomata, Although virtually impossible to observe or calcu- spanning the Holocene. D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 361

Many relationships and assumptions are required ple integrated is less in most cases. This temporal 13 to estimate PCO2 from the d C of phytoplankton. resolution exceeds both geochemical modelsŽ Sec- Pagani et al.Ž. 1999a propagated errors from their tion 2. and the method of pedogenic carbonates w 3yx " estimates of PO4 Ž.0.1 mM , temperature Ž.Section 4 , but is generally less than the method of " " " Ž.28C , salinity Ž 35 1‰ .Ž , ´f 26 1‰ . , slope stomatal parametersŽ. Section 5 . The precision of w 3yx " and intercept of PO4 –m relationshipŽ 11%, repre- individual PCO2 estimates is also quite high Ž 8– senting the 95% confidence interval in the modern 15% in the cases of Freeman and Hayes, 1992; 13 " dataset. , and analytical error for d C 37:2 Ž.0.5‰ Pagani et al., 1999a,b. . Again, this level of precision 13 " and d C calcite Ž.0.2‰ , and calculated an error exceeds that of models and pedogenic carbonates, envelope for PCO2 of 15%. This error envelope is but is generally exceeded by stomatal parameters. not likely to decrease substantially in future studies since the ´ –wx CO relationship for C P2Žaq. 37:2 13 alkenones is currently the most reliable. As discussed 4. d C of pedogenic carbonates aboveŽ. Section 3.4 , its reliability stems from the narrow range of species producing the alkenones, the 4.1. The model species’ narrow range of cell geometry, and their In regions receiving less than approximately 800 observed insensitivityŽ. in terms of ´ to growth P mm annual precipitation, pedogenic carbonatesŽ i.e., rate. Pagani et al.Ž. 1999a,b also selected stable, low authigenic carbonates in soils.Ž are common Royer, productivity sites, which help minimize non-climati- 1999. . In the case of CaCO3 , the principal source of cally driven fluctuations ofwx CO due to processes q 2Žaq. calcium is wind-blown dust and dissolved Ca2 in such as upwelling. Such settings also minimize varia- rainwaterŽ. Gile et al., 1979 . The carbonate ion is tion in growth rate and cell geometryŽ Popp et al., typically inherited from biological respired COŽ e.g., 1998. .wx CO in regions of low productivity are 2 2Žaq. organic decomposition, root respiration. , not carbon- also more likely to be in equilibrium with PCO . 2 ate weathering or groundwater COŽ Cerling et al., Pagani et al.Ž. 1999a,b estimated PCO from a site 2 2 1989; Quade et al., 1989. . This is because the rate of Ž.ODP 588 with relatively high paleo-surface sea pedogenic carbonate formation is 10 2 to 10 3 times temperaturesŽ. 15–228C . Due to the higher concen- slower than the rate of soil respirationŽ Cerling, trations of CO in colder water, ´ in cold water 2Žaq. P 1984, 1999. . Since this biological CO flux domi- phytoplankton is less sensitive to PCOŽ Franc¸ois et 2 2 natesŽ. hereafter referred to as soil-respired CO , soil al., 1993; Popp et al., 1999. , and thus warm water 2 CO in well-aerated soils can be modeled as a sites are preferable. 2 standard diffusion–production equationŽ Baver et al., One drawback to the use of C alkenones is 37:2 1972; Kirkham and Powers, 1972; Cerling, 1984. : their relative scarcity in many sediments, particularly E ) E2 ) at higher latitudes and in pre-Neogene sediments CssC sD))qf Ž.z Ž25 . Ž.Marlowe et al., 1990 . For pre-Cenomanian sites, Et ssEz 2 where C37:2 alkenones are unknown, another ) s y3 s where Cs2soil COŽ mol cm. , z depth in the biomarker is required, which at present would de- ) s soil profileŽ. cm , and fs2CO production rate as a crease the precision of PCO2 estimates. A general y1 y3 ) s function of depthŽ mol s cm. . Ds diffusion drawback to the use of marine phytoplankton as a 2 y1 coefficient for CO2 in the soilŽ cm s. , and is PCO2 indicator is its decreased precision at high calculated as follows: PCO , where ´ asymptotically approaches 25‰. 2p ) s ´r This break in slope usually occurs between 750 and DsairD Ž.26 s 1250 ppmV CO2 Ž. Kump and Arthur, 1999 . where Dairdiffusion of CO 2 in the atmosphere, The temporal resolution of this proxy can be quite which is a function of pressure and temperature, high, particularly in densely sampled ocean sediment ´sfree air porosity in the soilŽ. 0-´F1 , and coresŽ. e.g., Pagani et al., 1999a,b . In the case of rstortuosity factor Žfpermeability, where 0-rF ) Pagani et al.Ž. 1999a , the average sampling density 1.. Ds is assumed constant with depthŽ Cerling, was approximately 200 ka. The time that each sam- 1991. . 362 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

) 13r 12 The CO2s production term Žf . is modeled to ets in Eq.Ž. 30 represent the C C ratio of soil exponentially decay with depthŽ Dorr¨¨ and Munnich, CO2sŽ.Žd Cerling, 1999 . . 1990. , such that: Eqs.Ž. 28 and Ž. 30 can be transformed and simpli- ) )) fied to solve for C , the concentration of atmo- f Ž.z sf Ž.0 exp Žyzrz . Ž27 . a ss spheric COŽ. Davidson, 1995; Cerling, 1999 . The s 2 where z characteristic CO2 production depthŽ. cm resulting expression is: Ž.Cerling, 1991 . The model is insensitive whether ) d y1.0044d y4.4 f Ž. ) s f s z exponentially decays, linearly decays, or re- C sSz 31 a Ž.y Ž . mains constant, so long as the characteristic depth dasd Ž.z is correct Ž Solomon and Cerling, 1987 . . z is not d d sd13 highly constrained in modern soilsŽ. Cerling, 1991 ; where f and a2C of soil-respired CO and atmospheric CO , respectively. however, Dorr¨¨ and MunnichŽ. 1990 report values of 2 10–20 cm for forested soils; deeper values can be expected in grasslandsŽ. Cerling, 1991 . 4.2. Patterns in d13C of pedogenic carbonate in Under steady-state conditions, Eq.Ž. 25 equals modern soil profiles zero. Given the following boundary conditions: ) s ))s )s CsaŽ.0 C where C a2PCO , and CL sŽ. k If pedogenic carbonates form in equilibrium with where L is the depth of an impermeable boundary 13 soil CO2cc , their d C values Ž.d can be used as a Ž.e.g., groundwater table and k is a constant, the proxy for dscc. d measurements in modern soils general solution to Eq.Ž. 25 is Ž Cerling, 1991 . : Ž.formed during the Holocene validate this assump- C ) Ž.z sSz Ž.qC ) Ž28 . tionŽ. Quade et al., 1989; Cerling et al., 1991a . Fig. sa 4 shows the striking correlation between field mea- where SzŽ.represents the concentration of CO2 con- surements from one soil profile from NevadaŽ Quade tributed by biological respiration, and is given as: et al., 1989. and the equilibrium-based model predic- f ) Ž.0 z 2 tionŽ. Cerling, 1984 . Fig. 4 also verifies the underly- s s y y r SzŽ.) 1 exp Žz z ..29 Ž . ing assumption that dsccŽ.and, by extension, d is Ds determined by the diffusion of atmospheric and soil- respired CO , with little influence from groundwa- Thus, soil CO2 represents a mixture of atmo- 2 ter, parent material, or kinetic or Rayleigh distillation spheric and soil-respired CO2 . Eq.Ž. 28 can be recast 13 ) fractionations. If this condition were not met, one in terms of the d CofCs Ž.Cerling, 1984, 1991 . This transformation involves the assumption that would expect heavier dcc values and more variation soil-respired CO has the same isotopic composition at depth. Cerling et al.Ž. 1989 measured dcc and the 2 13 as soil organic matter. The resulting expression is: d C of coexisting organic matter Ž.dom in 10 modern soils at depth, where the contribution of da under today’s PCO2cc levels is minimal. d was enriched 14–16‰ relative to d , consistent with the equilib- dsŽ.z om rium fractionations between these two components. ) R DRsaf ) SzŽ. qC The sharp decrease in dcc at shallow soil depths is 13 q a q 1 Ds ž/1 Rf ž/1 Ra d sy1 typical, and represents mixing between heavy a and ))R RPDB DCsaf light d . Ed rEz approaches zero by 50 cm soil SzŽ.1yq f s 13 1qR 1qR 0ž/Ds ž/f ž/a depth in most pedogenic carbonate-producing soils Ž.Cerling, 1984; Ekart et al., 1999 . Thus, in paleosol = 1000 Ž.30 work where z is not known precisely, one must s 13 13 s where ds2sd C of soil CO , D diffusion coeffi- control for this variability by ensuring that measured 13 2 y1 cient of CO2 in soilŽ cm s. , and Rf , Ra, and carbonates are at least 50 cm from the paleosurface. s13 r12 RPDBC C ratios of soil-respired CO 2 , atmo- At times of high PCO2cc , this asymptotic value of d spheric CO2 , and the PDB standard, respectively. will shift to heavier values, and vice-versa for times Note that the terms enclosed within the inner brack- of low PCO2 . D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 363

13 Fig. 4. Measurements of d C of pedogenic carbonate Ž.dcc as a function of depth from one soil profile. This soil profile developed on ) limestone parent material. Solid line represents model prediction based on best estimates of soil temperature, ´, r, z, fs Ž.z , df , and da. Figure redrawn from Quade et al.Ž. 1989 .

4.3. Estimating S() z ppmV, but increases to 5500 ppmV for a SzŽ.of 9000 ppmVŽ. see Fig. 5 .

The estimation of atmospheric CO2 from Eq.Ž. 30 Carbonates can form at depth in waterlogged hinges on a number of relationships and associated gleyed soils. Such soils can have values of SzŽ. assumptions. The components of SzŽ.must be esti- exceeding 25,000 ppmV, causing overestimation of mated, namely soil temperature, pressure, porosity PCO2 Ž. Cerling, 1991 . Thus, soils with gleyed fea- Ž.´ , tortuosity Ž.r , characteristic CO2 production tures should be avoided. Pedogenic carbonates also ) depth Ž.z , and soil CO2s respiration rate Žf Ž..0. form in desert soils, which typically have values of ) The model is particularly sensitive to ´, z, and fs SzŽ.below 3000 ppmV Ž Cerling, 1991 . and should Ž.0 . For example, using a typical modern pedogenic also be avoided. Desert soils usually have neither carbonate-forming soilŽ. Cerling, 1991 with a dcc of pedogenic carbonates at depths exceeding 50 cm y 7‰, a z of 10 cm yields a PCO2 estimate of 2500 Ž.Royer, 1999 nor well-developed leached horizons. ppmV, but increases to 5000 ppmV for a z of 20 Conversely, non-desert soils rarely have pedogenic cm. Fortunately, excursions in one or more of these carbonates within the upper 20 cm of the profile six variables is often balanced by shifts in one or Ž.Cerling, 1991 . more of the other variablesŽ. Cerling, 1999 . There- No pattern in the current literature emerges before, it is not unreasonable to combine these factors tween the SzŽ.chosen for model calculations and and estimate SzŽ.directly. Sz Ž., as presented above, the environmental context of the paleosol. Values for ) is the component of Cs from soil respiration. This SzŽ.range from 3000 to 5000 Ž Ghosh et al., 1995; value is not well constrained in modern soils, but Lee, 1999.Ž , 3000 to 7000 Mora et al., 1996 . , 5000 current data suggest values of SzŽ.at depths )50 to 10,000Ž Mora et al., 1991; Cerling, 1992b; Lee cm for pedogenic carbonate-producing soils of and Hisada, 1999.Ž , and 10,000 ppmV Andrews et 3000–5000 and 5000–9000 ppmV for low and high al., 1995. . As discussed above, pertinent values of productivity soils, respectivelyŽ Brook et al., 1983; SzŽ.for pedogenic carbonate-forming soils are not Solomon and Cerling, 1987; Cerling, 1999. . Again, well constrained, but highly productive soils have y using a typical modern soil with a dcc of 7‰, a values closer to 9000 ppmV while less productive SzŽ.of 3000 ppmV yields a PCO2 estimate of 2000 soils lie closer to 3000 ppmV. Given the sensitivity 364 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

Fig. 5. Sensitivity of PCO2 estimates to SzŽ.. The three plotted values of Sz Ž.span the expected range for pedogenic carbonate-forming 13 soils. Note that sensitivity increases with heavy pedogenic carbonate d C values Ž.Ždcci.e., high PCO 2 . . Lines drawn from Eq. Ž. 31 s sy sy assuming Dcc – s9.0‰, d a 6.5‰, and df 26%. of the model to this variable, more research is re- plants or other unusually 13C-enriched organic matter quired. is the cause for heavy dccvalues, not elevated PCO 2 Recently, Ekart et al.Ž. 1999 compiled published Ž.Wright and Vanstone, 1991, 1992 . Unfortunately, and unpublished dccdata and calculated PCO 2 for organic matter is often rare in well-aerated soils 68 paleosol sites. Many variables were standardized pertinent to this methodŽ Cerling, 1991, 1999; Mora to facilitate more direct comparison among the sites. and Driese, 1999. , and some studies simply apply

A value for SzŽ.of 5000 ppmV was uniformly today’s mean domfor C 3 plants to their fossil sites applied, which given our poor understanding of SzŽ. Ž.e.g., Lee, 1999 . 13 is not unreasonable, but surely significant variation The d C of the atmosphere Ž.da is required by in SzŽ.existed among the 68 sites. the model, but is one of the least sensitive variables Ž.Cerling, 1992b . Typically, values are derived from

4.4. Estimating df , das, and d the marine carbonate record Ž.Ždocc e.g., Veizer et al., 1999. , assuming a constant fractionation and According to Eq.Ž. 30 , the d13C of soil-respired ocean–atmosphere equilibriumŽ. but see Section 3.5 .

CO2 Ž.df must also be estimated. This value differs The assumption of equilibrium may not be valid from soil CO2sŽ.d by both the 4.4‰ diffusional during intervals of rapid global change such as the 13 12 fractionation between C and CŽ Craig, 1953; P–E boundaryŽ. Ekart et al., 1999 . da has in turn Cerling et al., 1991b. and the relative input of heavy been used as a dom proxy. Ekart et al.Ž. 1999 used a 13 daom. The d C of soil organic matter Ž.d appears smoothed version of the docc curve of Veizer et al. roughly equal to its associated df Ž.Cerling, 1992b , Ž.1999 to infer da, and then assumed a constant and can be taken as a direct proxyŽ Cerling, 1992b; fractionation of 18‰ between the atmosphere and

Mora et al., 1991, 1996. . It is possible, however, for organic matter Ž.Da–om . While this procedure was differential decomposition, burial diagenesis, or con- required to compare studies that had not measured tamination by modern organic matter to jeopardize dom with those that had, it nonetheless potentially this equalityŽ. Ekart et al., 1999 . It is important to introduces error, both because Da–om may not be measure dom at each site, and not assume a constant constant and the smoothed Veizer et al.Ž. 1999 curve y valueŽ. e.g., 24‰ , because excursions occur in the does not accommodate for short-term da excursions. geologic recordŽ. Bocherens et al., 1993; Jones, 1994 . In addition, plants growing in the seasonably dry to

Accounting for domeliminates the argument that C 4 semi-arid regions conducive to pedogenic carbonate D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 365 formation are likely to be water stressed, which may 308C can differ by 700 to 1500 ppmVŽ Cerling, reduce Da – om. Calculating d omfrom d a also greatly 1999; Ekart et al., 1999. . increases the sensitivity of the model to da Ž.Fig. 6 . This may be important for times where docc is not 4.5. C4 plants well constrained, such as the Permo-Carboniferous. f PCO2 estimates of Ekart et al.Ž. 1999 during this The domof C 4 plants is enriched 14‰ relative f interval decrease by 700 to 2500 ppmV when docc to C3 plants, having a modern mean value of values from Popp et al.Ž. 1986 are used instead of y12.5‰Ž. Deines, 1980 . This large isotopic differ-

Veizer et al.Ž.Ž 1999 Berner, unpublished data . . When ence is reflected in dcc , and reinforces the need to possible, dom should always be measured directly. measure coexisting domŽ.Cerling, 1992a,b . If d om The last parameter required to calculate PCO2 is values indicative of C4 vegetation are found, how- 13 the d C of the soil CO2sŽ.d . As discussed above ever, such sites should not be used for estimating 13 Ž.Section 4.2 , since equilibrium is common between paleo-CO2a4 . The d C gradient between d and C - dsccand d , and no other soil component significantly dominated df Ž.ca. 6‰ is much smaller than that 13 contributes to dcc , the d C of pedogenic carbonate between da3and C -dominated df Ž.ca. 20‰ , and is used as a proxy. This conversion involves a tem- thus the precision of PCO24 estimates based on C - perature-dependent fractionationŽŽ.. Eq. 18 , and so dominated sites is much poorerŽ Cerling, 1984; Fig. soil temperature must be estimated. This variable is 7. . Evidence for C4 plants in pre-Neogene deposits not well known, as it depends on air temperature, is sparseŽ Cerling, 1991; but see Bocherens et al., time of year of carbonate formation, soil texture, and 1993; Jones, 1994. , and so is not a critical issue for depth in profile. Seasonal temperature fluctuations older sites. are dampened in soils at depth, and most pedogenic carbonates form in the low-to-mid-latitudes. An esti- 4.6. Choosing appropriate paleosols mate of 258C is commonly usedŽ Cerling, 1991; Mora et al., 1996; Ekart et al., 1999. ; however, one Substantial literature exists pertaining to identify- could expect considerable variation. Calculations of ing paleosols in the fieldŽ e.g., Retallack, 1988,

PCO2 based on soil temperatures of 208C versus 1990. . Common characteristics include clayskin-

13 13 Fig. 6. Sensitivity of PCO2aa estimates to the d C of the atmospheric Ž.d when d is used to infer the d C of coexisting organic matter 13 Ž.Ždom e.g., Ekart et al., 1999 . . The two d C values of pedogenic carbonate Ž.dcc span most of the range currently observed in the fossil record. Note that the sensitivity increases with heavier dccand lighter d aŽi.e., at higher PCO 2 . . Lines drawn from Eqs. Ž. 18 and Ž. 31 s s s s assuming SzŽ. 5000 ppmV, soil temperature 258C, Docc – a8‰, and D a – om18‰. Assuming a constant D occ – a, differences in published da values for the Permo-Carboniferous exceed 2‰Ž. see Section 4.4 for discussion . 366 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

Fig. 7. Difference in sensitivity of PCO2 estimates between pedogenic carbonate formed in pure C34 and C biomass systems. Lines drawn s s sy sy y from Eq.Ž. 31 assuming Sz Ž. 4500 ppmV, Dcc – s9.0‰, d a 6.5‰, and df 26% and 12‰ for C34 and C biomass, respectively. bounded peds, root traces, and evidence of bioturba- more likely to form abioticallyŽ Mora et al., 1991; tionŽ. Cerling, 1992a, 1999 . Further analysis is re- Mora and Driese, 1999. . quired, however, to determine if a given paleosol is As discussed aboveŽ. Section 4.2 , it is crucial for appropriate for this method. Although analyses of soils with moderate to high respiration rates to mea- ) dcc from modern soils show little input from detrital sure dcc from 50 cm depth. Analyzing soils at too carbonate, even for soil profiles developed on car- shallow of a depth will lead to overestimation of bonate parent materialŽ. Quade et al., 1989 , it is wise PCO2 . Mora et al.Ž. 1996 found that in paleo-verti- to avoid such soilsŽ Cerling, 1984; Ekart et al., sols, which seasonally formed deep cracks, the dcc 1999. . This is particularly important for paleosols of nodules were consistently heavier than the dcc of lacking well-developed leached horizons above the rhizoliths deeper in the profile. This may also high- pedogenic carbonates, as the ds in such soils are light the need to sample deeper in vertic paleosols often influenced by detrital carbonateŽ Cerling, due to their associated greater penetration of atmo-

1992a, 1999. . Likewise, potential paleosols resting spheric CO2 Ž. Mora and Driese, 1999 . on marine or lacustrine carbonates should be avoided Burial diagenesis can potentially alter the dcc ŽMora et al., 1991; Driese et al., 1992; Cerling, signature. Most studies find no evidence for burial 1999; Ekart et al., 1999. . Groundwater calcretes are diagenesis, however, even in recrystallized carbon- always to be avoided, as they do not form in a ates with large intra-site d18 O variabilityŽ Cerling, diffusion-dominated system. Such soils may show 1991; Mora et al., 1996. . Nevertheless, micritic car- signs of gleying, or contain massive carbonate de- bonate in the form of distinct nodules or rhizoliths is posits, but can be difficult to distinguish from pedo- preferable. Veins, coalballs, septarian nodules, and genic carbonates formed in unsaturated zonesŽ Cer- radiating crystals are to be avoidedŽ Ekart et al., ling, 1999. . 1999. . Carbonates that formed in soils with little biological activityŽ e.g., before the Devonian colonization 4.7. Goethite method of vascular land plants. should be treated with caution, since their associated respiration rates were An independent PCO2 proxy has been developed undoubtedly very lowŽ Mora et al., 1991; Ekart et involving trace pedogenic carbonates contained al., 1999. , leading to inflated estimates of PCO2 . In within goethitesŽŽ Fe CO3 . OH .Ž Yapp and Poths, addition, under such conditions carbonates are much 1992, 1996. . In brief, the concentration and d13Cof D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 367

) d ) this goethite is a function of Cssand , respec- Cs2is small, leading to imprecise PCO estimates tively, so that the atmosphere–goethite–organic mat- Ž.Cerling, 1992b . For example, Ekart et al. Ž. 1999 " ter system can be modeled as simple mixing between estimated PCO2 from 15 modern soils to be 433 the two end membersŽ. Yapp and Poths, 1992 : 385 ppmVŽ. 1s . Thus, this proxy should not be -; s y r q relied upon when PCO2 1000 ppmV. This model dggŽ.d Ža. dgŽom. XagX d Žom. Ž.32 also loses precision at high PCO2 Ž. see Fig. 5 , but s 13 s where dggd C of goethite, d Ža. and dgŽom. the the rate of loss is less than the other proxies, and d theoretical g if only influenced by the atmosphere should be the proxy of choice when PCO2 is high and organic matter, respectively, Xsmole fraction Ž.);1500 ppmV . The temporal range of this s of FeŽ. CO3a OH in the goethite, and X the theoret- method also exceeds the other proxies, and should " ical X if only influenced by the atmosphere. yield reasonable estimates of PCO2 Ž 500 to 1000 Thus, in a preserved soil profile, dg will decrease ppmV. back to the Devonian, the time of rapid and X increase with depth, and they are inversely vascular land plant colonization. Pre-Devonian esti- related to one anotherŽ. Yapp and Poths, 1992 . This mates should be considered carefully since terrestrial relationship is insensitive to variations in soil pro- biological productivity was undoubtedly low. r ductivity. If one plots 1 X vs. dg , the slope term Unfortunately, the temporal resolution of this wxy Ž dgŽa. dgŽom. Xa2. is related to PCO . If multiple method is the poorest of the proxies described here. measurements within a given profile are possible, the Pedogenic carbonates typically form on the timescale 34 intercept term should be dgŽom. ŽYapp and Poths, of 10 –10 yearsŽ. Cerling, 1984, 1999 , which 1992. . Multiple measurements are often not possible, should be considered the upper limit of the method’s however, in which case the d13C of coexisting or- temporal resolution. Studies wishing to quantify ganic matter can be taken as a direct estimateŽ Yapp PCO2 over shorter time intervals should use other and Poths, 1996. . As in the analogous case of Ekart proxies. In addition, this proxy should not be used y et al.Ž. 1999 , dgŽa. dgŽom. is assumed invariant during intervals with sharp isotopic excursions due through timeŽŽ.. Yapp and Poths 1996 use 16‰ , to potential disequilibria among the isotopic reser- allowing for the calculation of Xa. As discussed voirsŽ. Ekart et al., 1999 . aboveŽ. Section 4.4 , this assumption can be dangerous. PCO , in turn, is related to X via Henry’s law. 2a5. Stomatal density and stomatal index Although this model is rather sensitive to temper- Ž. ature Yapp and Poths, 1992 , published error esti- Terrestrial plants obtain CO2 from the atmosphere "; mates for PCO2 Ž.1200 ppmV only account for for growth, and thus necessarily lose water vapor to variability in dgŽom. Ž.Yapp and Poths, 1996 . As with an unsaturated atmosphere. The classical dilemma the model of CerlingŽ. 1984, 1991 , carbonate-con- between carbon acquisition and water loss results in taining soils that developed without a strong biologi- the concept of plant water-use efficiencyŽ. WUE , cal componentŽ e.g., before the rise of vascular land which can be defined in a number of different ways. plants. are prone to contain a non-biogenic fraction, In the short-term of minutes, WUE is calculated as and should be treated carefully. the ratio of assimilation of CO2 by photosynthesis to loss of water by transpirationŽ. Stanhill, 1986 , de- 4.8. Summary scribed by: = y gsaiŽ.p p In the absence of C plants, SzŽ., soil tempera- 4 1.6=P A ture, and PCO are the most sensitive variables in WUEss Ž.33 2 g = e ye Cerling’s modelŽ. Cerling, 1991, 1999 . Since soil siaŽ.E temperature and, in particular, SzŽ.are not well P constrained, it is advisable to include as many soils where gs is the stomatal conductance to water vapor, from geographically diverse sites as possibleŽ Cer- which is 1.6 times lower when considered as a ling, 1991, 1992; Mora and Driese, 1999. . When conductance to CO2a , p and e aare the partial pres- - PCO2 is lowŽŽ.. e.g., 10% of Sz , its influence on sures of CO2 and water vapor, respectively, in the 368 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

air outside the leaf, piiand e are the partial pres- densityŽ. SD at ambient CO2 Ž Beerling and Wood- sures of CO2 and water vapor, respectively, in the ward, 1996a. , but plants tend to optimize carbon leaf air spaces and P is the atmospheric pressure. gain per unit of water lossŽ. i.e., WUE through the This definition simplifies to: fine control offered by regulating of stomatal open- p yp ing and closingŽ. Cowan, 1977 . In a CO -rich atmo- s ai 2 WUE Ž.34 sphere, leaf WUE increases because CO induced 1.6= Ž.e ye 2 ia partial stomatal closure reduces water vapor losses and emphasizes the importance of stomatal behavior without reducing the CO2 -related stimulation of and morphology on the bi-directional control of CO2 photosynthesisŽ. Morison, 1985 . An additional im- and water vapor fluxes from the leaf. It should be provement in WUE is possible in an elevated CO2 noted in passing that timescales are of crucial impor- atmosphere by reducing the number of stomatal pores tance to plant WUE. On the longer timescale of a Ž.stomatal density, SD on the leaf surface Ž Wood- day, net assimilation will be less than expected by ward, 1987; Woodward and Bazzaz, 1988. . Changes just integrating assimilation rates during the whole in SD can have an important effect on plant fitness day, because respiratory loss of CO2 by non-photo- by influencing the growth response of whole plants synthetic organs, and by leaves in darkness, must be ŽKundu and Tigerstedt, 1999; Hovenden and Schi- accounted forŽ Farquhar et al., 1982; Farquhar and manski, 2000.Ž. . Woodward 1987 and Woodward Richards, 1984. . In addition, water loss may occur and BazzazŽ. 1988 demonstrated that atmospheric through the leaf cuticleŽ. Jones, 1992 , a feature that CO2 can regulate stomatal formationŽ Beerling and is rarely measured. A more complete definition of Chaloner, 1992. and subsequent studies have shown WUE is therefore given by: the inverse relationship between stomatal density y = y Ž.Ž.paip 1 w r Ž.SD and PCO2 to be widespread across different WUEs Ž.35 Ž. 1.6= Ž.Ž.e ye = 1qw taxonomic groups Beerling and Woodward, 1996b ia w pointing to its potential to offer a plant-based paleo- w where r is the fraction of assimilates respired away CO sensor using fossil leaves that extends back qw 2 by the plant and the termŽ. 1 w allows for tran- through geological time. We emphasize that at the spiration through the cuticle. present the underlying genetic mechanismŽ. s linking The WUE of plants over an entire growing season CO2 and SD remains to be identified and so the includes plant respiratory carbon losses through relationship must be regarded as correlative, not maintenance and synthetic respirationŽ. Jones, 1992 causative. and whole-canopy transpiration calculated from This section describes important considerations growth analyses: regarding the use of stomatal measurements on fossil y We0W leaves to detect paleo-CO2 signals. For details of k= t yt fossil leaf cuticle preparation, and methods for stan- WUEs e0 Ž.36 dardizing stomatal counts made on leaf surfaces, HEdt etc., see Jones and RoweŽ. 1999 . 5.1. The concept and use of stomatal index where W0eand W are the total dry weights of the plant at the beginning, t0e, and end, t , of the growing To interpret paleo-CO2 signals from fossil leaves, season, HEdt is the sum of daily transpiration and k there is need to control for other features of the converts changes in dry matter to CO2 equivalents. environment that might influence their stomatal char- This version of WUE is under variable control by actersŽ. Beerling, 1999 . Given that plants tends to stomatal behavior, depending on the coupling of the optimize their WUE, it follows that water stress canopy with the over-lying airŽ Jarvis and Mc- would be expected to reduce transpirational water Naugthon, 1985. . losses, and this is usually brought about through Plant WUE, as defined in timescales of minutes partial stomatal closure in response to droughtŽ e.g., ŽŽ.Ž..Eqs. 34 and 35 , and for whole growing seasons Clifford et al., 1995. . It is interesting to note that, ŽŽ..Eq. 36 , has been correlated with leaf stomatal paradoxically, SD typically increases under D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 369 droughted conditionsŽ Salisbury, 1927; Beerling, though, is insensitive to changes in the mole fraction

1999; see Ticha,´ 1982; Royer, 2001 for reviews. of CO2 Ž. Woodward and Bazzaz, 1988 , indicating because of a reduction in epidermal cell size causing that the mechanism controlling SI and SD is sensi- the stomata to be packed closer together. However, tive to CO2 partial pressure and not mole fractionŽ or under these circumstances, the stomata are nearly concentration. . Since CO2 partial pressure falls with closed and the leaf angle relative to the sun is lower, increasing altitude, but mole fraction is constant both of which reduce transpiration water losses. Ž.Gale, 1972 , these experimental observations largely SalisburyŽ. 1927 introduced the concept of stom- explain the general trend of increasing SD of trees atal indexŽ. SI to remove the effects of drought, and shrubs with altitudeŽ Korner¨ et al., 1986; Wood- calculated as follows: ward, 1986. . From this, therefore, it follows that r SD fossil leaves showing changes in SD SI though time SIŽ. % s =100 Ž 37 . are actually recording ancient variations in atmo- SDqED spheric CO2 partial pressure. For leaves obtained where SD and ED are, respectively, stomatal density from sites at altitudes close to sea level, the partial and epidermal cell density. Eq.Ž. 37 shows that pressure of CO22 is approximately equivalent to CO effectively SI is the proportion of epidermal cells concentration; quantitative CO2 partial pressure re- that are stomataŽ where stomata are defined as the constructions from SDrSI measurements can there- stomatal pore and two flanking guard cells. . It is fore be readily converted to units of CO2 concentra- therefore independent of epidermal cell size and tion. However, for paleo-CO2 reconstructions using measures stomatal initiation and development, and is fossils from sites at significant altitude, such a con- not affected by subsequent cell expansion. Unlike version requires the difficult-to-test assumption that

CO2 Ž Woodward, 1987; Woodward and Bazzaz, total atmospheric pressure has not changedŽ cf. 1988. , there is no experimental data demonstrating Rundgren and Beerling, 1999. . Clearly, it is advis- that water stress influences stomatal initiationŽ Salis- able to control for paleoaltitude during the design of bury, 1927; Sharma and Dunn, 1968, 1969; Estiarte investigations aiming to reconstruct paleo-CO2 varia- et al., 1994; Clifford et al., 1995. . tions with this technique. A recent review of the effects of irradiance, temperature, vapor pressure deficit and water supply on 5.2. Canopy-scale considerations leaf SD and SI concluded that SI has the fortuitous property of being relatively insensitive to all of these Due to canopy photosynthesis and plant and soil but sensitive to atmospheric CO2 levels, whereas SD respiration, the CO2 concentration experienced by can be influenced by each factor separately or in leaves within a canopy can deviate from the global combinationŽ. Beerling, 1999 . Such considerations ambient valueŽ Bazzaz and Williams, 1991; Buch- clearly indicate that SI rather than SD should be mann et al., 1996. . This, in turn, may introduce bias r more secure for interpreting paleo-CO2 variations in PCO2 estimates from the SD SI measurements from fossils. We recognize, however, that epidermal made on fossil leaves if derived from a dense cells can become difficult or impossible to accurately closed-canopy where the PCO2 values at the time of count on poorly preserved fossil cuticles and this in stomatal development in the leaf were very different turn requires the use of SD, but with a suitable from ambient. Atmospheric data from Harvard For- measure of cautionŽ Beerling et al., 1993; McElwain est in north central Massachusetts, USA, indicate and Chaloner, 1996. . that CO2 concentration differences between the am- An important and often over-looked consideration bient atmosphere and close to the ground surface r in the relationship between SD SI and CO2 is what Ž.4.5 m can be as large as 35 ppmV, especially stomatal developmentrinitiation is responding to. during the summer when soil and plant respiration Careful experimental work has shown that plants rates are highŽ. Fig. 8 . However, since bud formation have the capacity to respond to a reduction in the and bud-burst by deciduous trees typically occurs in partial pressure of CO2 by an increase in leaf SI the spring and fall when CO2 differences between Ž.Woodward and Bazzaz, 1988 . The response, the ambient atmosphere and canopy are -10 ppmV 370 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

Fig. 8. Canopy CO22 relative to ambient CO for four heights within a tree canopy in 1996. Canopy height is ca. 24 m. Ordinate represents 7-day running average of daily averages of hourly measurements at each height Ž.ns5311 for each height . Measurements at 29.0 m height taken as ambient valueŽ. mean for time interval at this heights370 ppmV . Raw data available at http:rrwww.as.harvard.edur chemistryrhfrprofilerprofile.html, and used with permission of S. Wofsy.

Ž.Fig. 8 , this effect by itself is unlikely to strongly where I0 is the photosynthetically active irradiance bias paleo-CO2 trends. incident on a canopy, I is the irradiance beneath a For tropical forests, it is important to note that the leaf area index L, and k is the extinction coefficient potential for a forest canopy to become uncoupled for irradianceŽ. typically 0.5 . This effect results in from the over-lying atmosphere, and therefore alter the development of so-called sun and shade leaves the local CO2 environment, is strongly dependent Ž.Givnish, 1988 with differing SD values; SI values upon the regional meteorology and geographical lo- remain largely conservativeŽ Salisbury, 1927; cationŽ. cf. Grace et al., 1995 . Leaf life spans and Kurschner,¨ 1997; Wagner, 1998. . It might be envis- primary production in tropical evergreen trees tends aged therefore that SDŽ. and possibly SI determina- to occur continuously so that a single canopy com- tions from fossil leaf assemblages constituting a prises of a range of cohortsŽ e.g., Reich et al., 1991; mixture of sun and shade leaves might introduce a Clark et al., 1992; Williams-Linera, 1997. that po- bias resulting from this effectŽ. Poole et al., 1996 . tentially developed at different CO2 concentrations. There is some hope that such a bias is minimized in This suggests that the use of fossil leaves from the fossil record. Studies on the transport processes stratified ancient Mesozoic and Tertiary tropical responsible for sorting leaf material prior to fossiliza- forests for paleo-CO2 signals may be more suscepti- tion indicate a preferential selection of sun morpho- ble to errors arising from within canopy PCO2 fluc- typesŽ. Spicer, 1981 and in this context, careful tuations; for tropical Amazonian forests, within consideration of the depositional environment is re- canopy PCO2 values are 30% higher than the global quired. valueŽ. Grace et al., 1995 . It is interesting to speculate that these two fea-

A more important effect of canopy development turesŽ. CO2 and irradiance of the environment al- is the logarithmic attenuation of irradiance with tered by canopy development might be expected to canopy depth, described by Beers’ Law as: lead to evolutionary differences in the absolute stomatal density of understory herbs and trees, as pro- s = ykL I I0 e Ž.38 posed nearly 70 years ago by SalisburyŽ. 1927 . A D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 371 re-analysis of his data, accounting for the effects of remaining 5% showing an increaseŽ. Royer, 2001 . taxonomic relatedness, revealed that in fact tree SD responses followed a similar pattern. In these species typically have higher stomatal densities than studies, the median length of exposure to an instanta- herbaceous speciesŽ. Kelly and Beerling, 1995 , al- neous step increase in atmospheric CO2 concentra- though as expected marginal herbs had significantly tion was 45 days. However, analysis of the response higher stomatal densities than understory herbs. In of SD and SI on fossil leaves spanning the last 400 terms of their responsiveness to CO2 increases, both m.y. to CO2 concentrations above ambient, as deter- recent historical and experimental, no strong correla- mined from independent PCO2 proxies and long-term tions between stomatal density and growth form geochemical models, indicates an inverse linear rela- Ž.woody vs. non-woody; trees vs. shrubs , habitat tionshipŽ.Ž Fig. 10 Beerling and Woodward, 1997; Ž.cool vs. warm , or taxonomic relatedness have been Royer, 2001. . The data from fossils therefore inti- reportedŽ Woodward and Kelly, 1995; Beerling and mate that the apparent ‘ceiling’ CO2 concentration to Kelly, 1997. . which SDrSI respond is mutable and that given sufficient time to adapt, plants have the capacity to

5.3. Temporal sensitiÕity of stomatal responses to respond to above ambient CO2 concentrations. The CO2 change implication is that the stomatal method of paleo-CO2 estimation could potentially detect CO2 increases The inverse relationship between SI and atmo- above ambient despite the apparent non-linear re- spheric CO22 has been reported as non-linear at CO sponse shown by modern genotypes. concentrations above ambientŽ Woodward and Baz- The reason for the apparent insensitivity in over zaz, 1988; Kurschner¨ et al., 1997. , although between half the species studied to date in short-duration the range 250–370 ppmV most species typically experiments is not known, but may be related to the show a linear responseŽ. Fig. 9 . Indeed, a literature ‘ceiling’ to which modern genotypes have become survey of 65 SI responsesŽ. from a pool of 35 species adapted over the past ;2 m.y. of glacial–intergla- = to experimental CO2 enrichmentŽ usually 2 ambi- cial CO2 fluctuationsŽ Woodward, 1987; Beerling ent. reported that 29% of cases showed a reduction, and Chaloner, 1993. . If the linkage between PCO2 with 66% showing no significant response, and the and stomatal initiation is genetically based, it might

Fig. 9. Stomatal index response of five species to PCO2 . Data compiled from herbarium sheets and altitudinal transects. Sources are as follows: Ginkgo biloba and Metasequoia glyptostroboides Ž.Ž.Royer, unpublished data ; Betula pendula Wagner et al., 1996 ; Quercus petraea Ž.ŽKurschner¨ et al., 1996 ; Salix herbacea Rundgren and Beerling, 1999 . . 372 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

r s Fig. 10. Response of stomatal density measured on fossil leaves axes to atmospheric CO2 variations over the Phanerozoic Ž.r 0.74 . Redrawn from Beerling and WoodwardŽ. 1997 with additional data from Cleal et al. Ž. 1999 and Edwards Ž. 1998 . CO2 data from Berner Ž.1994 .

involve a time lag. The possibility of a genetic basis compared to their ambient CO2 grown counter parts is intimated by transplant experiments with the up- Ž.Fernandez´ et al., 1998 . land grass Nardus stricta ŽWoodward and Bazzaz, Whatever the mechanismŽ. s , we note that the 1988. and analysis of plants growing near natural unknown duration of exposure required for a plant to r high CO2 springs for many generationsŽ Beerling adjust its SD SI value to a high CO2 regime would and Woodward, 1997; Bettarini et al., 1997, 1998; appear to define the stomatal paleo-CO2 method’s Fernandez´ et al., 1998; Paoletti et al., 1998; Tognetti temporal resolution. For most studies on fossil plants et al., 2000. . In a reciprocal transplant experiment, encompassing tens of thousands to millions of years, plants of the herbaceous species Tussilgao farfara and therefore many generations of even long-lived collected close to a geothermal spring in Italy, known trees, it seems reasonable to assume any genetically to have grown under elevated CO2 concentrations determined ‘ceiling’ could be altered. for many decades, and those from control ambient A further important issue relating to this method

CO2 sites, were grown in controlled environments at is the strength of response, i.e., the magnitude of SI ambientŽ. 350 ppmV and elevated Ž. 700 ppmV CO2 change for a given unit of CO2 change. Two analy- partial pressures for two yearsŽ Beerling and Wood- ses on a life form basis have concluded that the ward, 1997. . The results indicated greater reductions degree of reduction in stomatal density to CO2 en- in SI under full irradiance conditions by plants from richment increases with the initial stomatal density the high CO2 source, compared to those from the ŽWoodward and Kelly, 1995; Beerling and Kelly, ambient CO2 source, providing some evidence for 1997. , defined by the relationship: genetic adaptation between different plant popula- SD sSD yexp 178=ln SDk 39 tions. The response is not, however, seen universally EAŽ.Ž. A Ž. in plants from high CO2 springsŽ Bettarini et al., where SDAE and SD are the stomatal densities at 1998. , although in an extreme case, plants of Spati- ambient and elevated CO2 concentrations, respec- phylum cannifolium and Bauhinia multinerÕia grow- tively, and k is the slope of the major axis least ing at CO2 concentrations of up to 27,000–35,000 squares regression resulting from independent con- ppmV in a natural cold CO2 spring in Venezuela trastsŽ see Woodward and Kelly, 1995; Beerling and were reported to have lower SI values by 70–80% Kelly, 1997. . The relationship is presented with the D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 373 caveat that it was derived mainly from short-term SD change predicted from analysis of results from studies, with a few samples from studies spanning herbarium studies and CO2 enrichment experiments decades to centuries. As yet, there are insufficient Ž.Ž.ks0.348 , 2 the change if only results from data for similar analyses on SI responses. experimental CO2 enrichment studies are included The response of SD predicted by Eq.Ž. 39 is Ž.Žks0.417 Woodward and Kelly, 1995 .Ž. and 3 the shown graphically in Fig. 11 for three cases:Ž. 1 the change predicted from a study of SD changes in

Fig. 11. Generalized relationship between the initial stomatal density of a leaf grown at ambient CO2 Ž. amb and Ž. a the loss of stomata from that leaf after growth with CO2 enrichment andŽ. b expressed as the proportion of stomata lost at high CO2 . Calculated from Eq. Ž 39 . based s s on analyses of herbarium and CO22 enrichment studies Ž.k 0.348 , CO enrichment studies only Ž.k 0.417 and the response of a s woodland flora to the past 70 years of CO2 increase Ž.k 0.260 . 374 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

British woodlands to the past 70 years of CO2 ‘calibration set’ to estimate PCO2 from the SI of the increase Ž.Žks0.26 Beerling and Kelly, 1997 . . It same species from the fossil record. Standard curves can be seen that across all cases, plants tend to generated in this way have to date been derived from dispense with between 10 and 50 stomatarmm2 in a mixture of herbarium data and data from plants an elevated CO2 atmosphereŽ. Fig. 11a , constituting growing across altitudinal gradientsŽ e.g., Rundgren a loss of between 30% and 10% of their original and Beerling, 1999; see Fig. 9. . Experimental data densityŽ. Fig. 11b . The relationships imply that there could also be used to develop training sets for use is a maximum number of stomata that can be lost; with pre-Quaternary fossil leaves but in many cases leaves cannot function effectively without stomata, experimental growth conditions are very different and this has led to the suggestion that in the short- from those experienced by plants in the field and term response curve that is sigmoidalŽ van der Burgh some cross calibration is required, e.g., by compar- et al., 1993; Kurschner¨ et al., 1997. . The exceptions ing the SI values of leaves grown at ambient CO2 to this statement are some amphibious species such partial pressures experimentally and naturally, before as Littorella uniflora Ž.Nielsen et al., 1991 and combining the datasets. Lobelia dortmanna Ž.Pedersen and Sand-Jensen, 1992 A detailed quantitative paleo-reconstruction of that function with no stomata at all. changes in the partial pressure of atmospheric CO2 These responsesŽ. Fig. 11 relate to the study of over the last 9000 ka of the Holocene using fossil fossil materials because extant groups of plants may leaves of Salix herbacea has been made by Rund- r have SD SI values reflecting the atmospheric CO2 gren and BeerlingŽ. 1999 . These authors recon- partial pressure at their time of originŽ Beerling and structed CO2 values from SI measurements made on Woodward, 1996b. . For example, gymnosperm lin- fossil leaves by developing a modern training set of eages evolved during times of perceived high PCO2 Salix herbacea SI values plotted against reference and in general have low SI values today; conse- CO2 partial pressures that were largely independent quently this minimizes potential for further reduc- of ice core data. Their resultsŽ. Fig. 12a closely tions in SI above present-day PCO2 . On this basis, matched the pattern of changes in atmospheric CO2 we suggest it is important to consider the absolute concentration documented from the ice core record SD and SI values for a species targeted for study as Ž.ŽFig. 12b Indermuhle¨ et al., 1999b . , providing in- this gives a likely assessment of its potential sensitiv- dependent support for the notion that small changes ity of paleo-CO2 change. This suggestion probably in atmospheric CO2 can be sustained despite the holds regardless of duration of exposure because of relative stability of the global climatic conditions functional considerations required for effective pho- Ž.Ciais, 1999 . This detailed, well-dated study pro- tosynthetic CO2 uptake and the generation of a vides the strongest validation for the SI-based CO2 transpiration stream. proxy to date and points the way for obtaining

quantitative CO2 reconstructions for the Cretaceous 5.4. Applications of stomatal method to detection of and Paleogene, if a taxon has a sufficiently long paleo-CO2 change fossil record, and has closely related extant represen- tatives amenable to CO2 -enrichment experiments. Application of stomatal indices as a PCO2 proxy There are probably no extant stomata-bearing plant involves many fewer assumptions and calculations species that extend back to the pre-Cretaceous. Par- than the other proxies described hereŽ Sections 3, 4, ticularly useful though are so called ‘living fossil’ and 6. . One simple approach involves tracking rela- plants, e.g., Metasequoia glyptostroboides and tive temporal trends in SD or SI of a single species Ginkgo biloba, a term coined by Darwin to describe ŽBeerling, 1993; Beerling et al., 1993; van de Water extant taxa belonging to lineages characterized by et al., 1994; McElwain et al., 1995; Cleal et al., little or no phenotypic change since the Mesozoic

1999. . Only qualitative trends in PCO2 can be de- Ž.Beerling, 1998a . One possible problem in using duced in this manner. Quantitative paleo-CO2 recon- these taxa to reconstruct paleo-CO2 changes in the structions can be made by establishing an SI–CO2 past is that the calibration dataset will necessarily relationship with known PCO2 , and then using this have been obtained from herbarium studies and ex- D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 375

Fig. 12. Holocene reconstructed variations inŽ. a the partial pressure of atmospheric CO2 using fossil Salix herbacea leaves Ž after Rundgren and Beerling, 1999.Ž. and b measurements of atmospheric CO2 from ice cores Ž after Indermuhle¨ et al., 1999b . . periments with modern genotypes that might show a environmentsŽ. Birks et al., 1999 . A recent study of non-linear response at CO2 concentrations above the early Holocene illustrates the difficulties that can ambient. This would limit the ability of the method arise when these considerations are neglected. Wag- to accurately reconstruct high PCO2 values. ner et al.Ž. 1999 presented a paleo-CO2 reconstruc- Considerable care is required in developing train- tion for the early Holocene, using a training set ing sets for the purpose of paleo-CO2 reconstruc- based on historical SI changes from birch Ž Betula tions, and every effort should be made to obtain pubescens, B. pendula. trees growing at a single site, materials from a wide range of plants with different and the approach yielded values of 240–360 ppmV. genotypes and growing in a variety of different These values were consistently 50–100 ppmV higher 376 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 than concentrations measured in ice cores from, the Nearest Living Relative approach. Taken Ž.Indermuhle¨ et al., 1999a or reconstructed from as absolute ratios, the picture that emerges from fossil Salix herbacea leavesŽ Beerling et al., 1995; detailed analyses of Devonian, Carboniferous, Per- Rundgren and Beerling, 1999. . Independent studies mian, Jurassic and Eocene fossil plants, and their of the response of SI in B. pubescens and B. pen- NLEsŽ McElwain and Chaloner, 1995, 1996; McEl- dula indicate that these taxa are rather insensitive to wain, 1998. , is that the SR shows a pattern of changes in CO2 partial pressure with altitude, and response through geological time that mirrors increases in the atmospheric CO2 concentration be- Phanerozoic CO2 trends predicted from mass bal- tween 1877 and 1978Ž. Birks et al., 1999 . The ance considerations of the long-term carbon cycle discrepancy appears to relate to the construction of a Ž.Fig. 13 . In an effort to make the SR method modern calibration dataset without paying sufficient semi-quantitative, calculated values have been lin- attention to the natural variability of leaf cellular early scaled with the output of Berner’s GEOCARB properties of Betula trees. II modelŽ e.g., an SR value of 2 converts to a PCO2 A qualitative approach to estimating changes in of 2=pre-industrial value.Ž McElwain, 1998 . . Al- paleo-CO2 in the distant past, independently of the though useful to a certain extent, the problem with need for calibration datasets, is the use of the stom- this development is that it fails to provide a method atal ratioŽ. SR concept Ž McElwain and Chaloner, of paleo-CO2 estimation independent of the models. 1995. . SR is defined as the ratio of the SI of the Nearest Living EquivalentŽ. NLEs taxon to the fossil 5.5. Summary to that of the fossil plant under investigationŽ Mc-

Elwain and Chaloner, 1995. . NLEs are defined as The precision in PCO2 estimates via stomatal the nearest ecological and morphological equivalent indices is the highest of any current proxy. When in modern floras to the fossil plant under considera- herbarium-based standard curves are inverted, the tion. Note that the concept is related to, but distinct 95% confidence intervals for PCO2 predictions fall

Fig. 13. Comparison of model predictions of atmospheric CO2 over the PhanerozoicŽ. from Berner and Kothavala, 2001 and stomatal-based estimates. Stomatal ratio dataŽ. unmarked filled boxes from McElwain and Chaloner Ž.Ž. 1995, 1996 and McElwain 1998 . Stomatal index dataŽ. ‘v’ from van der Burgh et al. Ž 1993 . . RCO22 units Ž ratio of atmospheric CO in the past relative to the pre-industrial value . of Berner and KothavalaŽ. 2001 converted to PCO2 assuming a time-averaged pre-industrial value of 250 pmmV, which is roughly the mean PCO2 over at least the last 400 kaŽ. Petit et al., 1999 . Middle line represents the Abest estimateB predictions of Berner and KothavalaŽ. 2001 , while the two straddling lines represent error estimates based on sensitivity analyses. D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 377 between "10 and "40 ppmVŽ van der Burgh et al., little isotopic discrimination. Because the relative 1993; Kurschner¨ et al., 1996; Rundgren and Beer- proportions of the two dissolved boron species vary ling, 1999; Wagner et al., 1999; Royer, unpublished with pH, and the degree of isotopic fractionation data. . Thus, this method is preferable for time peri- between the two species is known, the 11 Br10 Bof y ods when paleoatmospheric CO2 was roughly similar BŽ. OH4 also varies with pHŽ Hemming and Hanson, to present-day levels. When combined with experi- 1992; Sanyal et al., 1996. . If the boron incorporated mental data, the method can be applied to time into fossil calcareous organisms faithfully records y periods with elevated PCO2 . At some level of ele- the isotopic composition of BŽ. OH4 in paleo- vated PCO2 , however, SI responses of modern geno- seawater, it is possible to calculate the value of pH types may have lowered sensitivityŽ. see Section 5.3 , for ancient oceans, once proper corrections for tem- resulting in larger error envelopes. This CO2 satura- perature have been madeŽ Spivack et al., 1993; Sanyal tion level has been observed at around 340 ppmV et al., 1995, 1996; Palmer et al., 1998; Pearson and from some speciesŽ. Woodward and Bazzaz, 1988 , Palmer, 1999, 2000. . From paleo-pH, making certain but also frequently to concentrations similar to the assumptions about the behavior of dissolved carbon phytoplankton proxyŽ.Ž 750–1250 ppmV Woodward in seawater, one can calculate paleoatmospheric CO2 and Kelly, 1995; Kurschner¨ et al., 1997. . from paleo-pH and an assumed value for total dis- This method probably cannot be applied to pre- solved inorganic carbonŽ. DIC in seawater. This is Cretaceous sites due to its species-specific nature the basis for the boron isotopic method for estimat-

Ž.Section 5.4 . For pre-Cretaceous sites, the qualita- ing paleo-CO22 . Estimates of CO levels ranging tive stomatal ratio technique of McElwain and from the Paleocene to the Pleistocene using this ChalonerŽ. 1995 can be applied. Relative changes in method have been made by Pearson and Palmer SD and SI within a given sequence, which also does Ž.1999, 2000 . not require extant species, can be illuminatingŽ e.g., Cleal et al., 1999. . The temporal resolution of the SI 6.1. Method of calculation method can be very high. While multi-million year high-resolution data are not possible as with the Calculation of the value of pH from boron iso- phytoplankton method, short-term high-resolution topic data rests on a mass balance expression for data are widely available. Thus, this method is ideal boron isotopes and an equilibrium expression for for resolving potential rapid PCO changesŽ e.g., y 2 BŽ. OH and B Ž. OH . For isotopic mass balance: K–T boundary, P–E boundary. . For example, using 34 q y s the stomatal ratio method, McElwain et al.Ž. 1999 d B4X d B3Ž.1 X d SB Ž.40 documented a large PCO2 spike across the sw yxr w yxwq x where X BOHŽ.44Ž BOHŽ. BOHŽ. 3B4. , d Triassic–Jurassic boundary at two fossil sites. Addi- s 11 y s 11 s d BofBOHŽ.4B3 ,d d BofBOH,Ž. 3d SB tionally, in contrast to the stable isotope-based prox- y d11B of total dissolved boronŽw BŽ. OH x q ies, the SI method is insensitive to isotopic disequi- 4 wBŽ. OHx. , and brackets represent molar concentra- libria among the biospheric reservoirs, a potential 3 tions. The equilibrium expression for borate chemi- factor during times of rapid global change. cal equilibrium isŽ. in terms of X : wxwq xwr y xs XH 1 X K B Ž.41 d11 6. B of marine calcium carbonate s where K B the equilibrium constant in seawater. Combining Eqs.Ž. 40 and Ž. 41 to eliminate Ž.X : Dissolved boron in the oceans exists primarily as Ž. Ž.y w q x s y r y y B OH34 and B OH , and these two species differ in H K Ž.Ž.d S d Ž.D d S d 10 11 B BB4B BB4 their ratio of the boron isotopes B and B. Field Ž.42 observations and experimental studiesŽ Hemming and s y Hanson, 1992; Sanyal et al., 1996. have shown that where DBB3B4d d Žfractionation between dis- the uptake of boron into biogenic calcium carbonate solved species. . The method, in its simplest form, y records the isotopic composition of BŽ. OH4 with assumes that d SBBand D are constants, equal to 378 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 present values, and that K B can be calculated for mate of the mean residence time of B in seawater is different paleo-temperaturesŽ the small salinity varia- 14 million years. could alter oceanic B isotope com- tion for the open ocean should have a minimal effect position to an extent greater than that resulting from on K B3.. Also, it is assumed that CaCO contains variations in pH. Even holding the isotopic composi- y only BŽ. OH4 that is taken up with either no fraction of riverine input constant, Lemarchand et al. tionationŽ Hemming and Hanson, 1992; Palmer et Ž.2000 show that Cenozoic changes in d SB could be al., 1998. or a known constant degree of fractiona- as much as 2‰ in 20 million years. A difference of 11 tion Dc3so that measurement of d B of CaCO is 2‰ is equivalent to a change in pH of 0.3 units q equal to DcB4d Ž.Sanyal et al., 1996 . Ž.Palmer et al., 1998 , which means a change in To convert paleo-pH to paleo-CO2 further as- PCO2 of a factor of 2 to 4. sumptions are necessary. A common assumption is Another problem arises from the assumption of that the total dissolved inorganic carbonŽ. DIC of the the degree of fractionation of isotopes during the oceans has not changed with timeŽ Pearson and incorporation of boron in CaCO3 . Since fractionation Palmer, 1999. . If true, then PCO2 is calculated from during biological uptake has been demonstrated to the appropriate equilibrium expressionŽ corrected for vary among species due to vital effects, it is impor- changes in temperature. relating DIC, pH, and PCO2 tant to ascertain how this fractionation might vary and using the modern value for DIC. However, if with different organisms. Hemming and Hanson DIC has changed over time, then the calculated Ž.1992 found little evidence for fractionation for a result for PCO2 can be greatly affected. To achieve variety of modern calcareous organisms, but this was more constrained PCO2 estimates, Pearson and not verified by Vengosh et al.Ž. 1991 who found PalmerŽ. 2000 modeled changes in sea surface alka- much larger variations. Sanyal et al.Ž. 1995, 1996 , linity through time, which can also be used to con- based on experimental and field studies, found a vert pH to PCO2 . They predicted an overall decrease constant fractionation of about 4‰ for one foram in alkalinity from 60 Ma to present day. This trend species, Orbulina uniÕersa, but found no fractiona- has been independently corroborated using the d44Ca tion for another foram, Globigerinoides sacculifer. of marine carbonates as a proxy forw Ca2qx , which In the paleo-pH studies of Palmer et al.Ž. 1998 and should inversely relate to alkalinityŽ De La Rocha Pearson and PalmerŽ. 1999, 2000 , it is assumed that and De Paolo, 2000. . there was no fractionation for a variety of different foram species; in other words, the boron isotopic y 6.2. Problems with the method composition of calcite represents that for BŽ. OH4 in the original seawaterŽ in the above notation they D s Values of K BBand are functions of tempera- assume that Dc 0. . In the study of Pearson and ture, so that in estimating paleo-pH correction for the PalmerŽ. 1999 , support for this assumption is pro- effect of changes in temperatures on these parame- vided by the finding of essentially the same d11 B for ters needs to be made. This can be done for ancient different foram species from the same depthŽ temper- environments where paleoceanographic data are ature. range. However, the variability for forams was availableŽ e.g., Palmer et al., 1998; Pearson and found to be greaterŽ. up to 4‰ for a larger number Palmer, 1999, 2000. . However, a more serious prob- of samples in the study by Palmer et al.Ž. 1998 . lem arises in assuming that d SB has remained nearly Diagenesis can bring about appreciable changes in constant over time. Lemarchand et al.Ž. 2000 have the boron isotopic composition of CaCO3 during shown that the boron isotopic composition of the burial. Spivack and YouŽ. 1997 have convincingly ocean most likely has varied considerably over the demonstrated this from the analyses of bulk carbon- Cenozoic due to changes in the inputs and outputs of ate in an ODP core from the eastern equatorial boron to and from the oceans. For example, they Pacific Ocean. Here, they found that the d11 Bof show that the boron isotopic composition of present- carbonates ranged from y5.5‰ to 23‰ with the day rivers varies by over 40‰ in d11 B. Thus, changes most negative values representing a totally recrystal- in the relative proportions of riverine input from lized end member. In the same sediments, interstitial different sources over millions of yearsŽ their esti- waters showed much less variation in d11 B with D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 379 depth because of diffusional and advectional ex- within the upper 600 m of the oceans does not change with seawater. change with time;Ž. 5 the activity coefficient of Ce in Finally, there is the problem of calculating paleo- sedimentary minerals remains constant over time.

CO2 from paleo-pH. First of all, one must assume All of these assumptionsŽ and several others not equilibrium of CO2 between that dissolved in the discussed here. have serious problems. Moffett oceans and that present as a gas in the atmosphere Ž.1990 has shown that Ce oxidation is microbially Ž.see Section 3.5 . More important, it is not likely that affected and that the redox couple in seawater is not the oceans have exhibited a constant level of DIC explained simply by equilibrium with O22 and H O. over millions of years, as assumed in the calculations Cerium is a highly charged trace elementŽ about of Pearson and PalmerŽ.Ž 1999 but not in those of 10y12 M. in seawater and could easily be complexed Pearson and Palmer, 2000. . Carbon is continuously with many other speciesŽ such as dissolved organic supplied to the atmosphere and ocean by degassing matter. other than carbonates. As is stated in the from metamorphism and magmatism, and by the section on the boron isotope methodŽ. Section 6.2 , it weathering of carbonate minerals and organic car- is likely thatŽ. DIC and alkalinity have varied con- bon, and is continuously consumed by the production siderably over time. During several times in the of carbonate and organic carbon sedimentsŽ Berner, geologic past, the average O2 level in upper portions 1999. . Hence, the total dissolved inorganic carbon of the oceans could have varied considerablyŽ e.g., load of the ocean could be expected to vary over anoxic oceanic events during the Mesozoic. . Finally, time. it is not clear in what mineral form Ce was originally present in any given sediment sampleŽ Liu and 6.3. Summary Schmidt analyzed carbonates. and what were its solid solution properties; so it is difficult to assume The boron isotope method provides an indepen- that its thermodynamic behavior in the solid state has dent check on other paleoatmospheric CO2 methods. been constant over time. However, it is based on many assumptions that, if not correct, can lead to incorrect results. Before this method can be put into general use, it will be 8. Phanerozoic CO2 trends: a comparison of necessary to gain a better idea of the d11 B of ancient methods oceans, how boron isotopes are fractionated during uptake by various calcareous organisms, and how From this review of paleo-CO2 proxies, it is clear one can estimate the level of dissolved inorganic that different methods have different temporal reso- carbon in ancient oceans. lutionsŽ i.e., the amount of time a single sample integrates. , and could be usefully viewed in a hierar- chical series with the long-term carbon cycle models 7. Redox chemistry of marine cerium providing the overarching set of predictions on a

timescale of millions of years. Next, paleosol CO2 Liu and SchmittŽ. 1996 have devised a rather proxies provide coverage on a timescale of 10 3 –104 ingenious method, involving many equilibrium steps, yearsŽ the minimum time required for soil carbonates for deducing paleo-CO2 levels from the concentra- to form. , and these offer a broad crosscheck on the tion of cerium in sedimentary rocks. Their method, mass balance model results. The two ocean CO2 however, rests upon many assumptions. Among oth- proxies, phytoplankton carbon isotopes and boron 3qq q ers, these include:Ž. 1 the reaction 4Ce O2 isotopes, are limited by the resolution of the marine qs 4qq 4H 4Ce H2 O is at chemical equilibrium at records that depend on factors such as burial rate, all times;Ž. 2 total dissolved Ce in seawater is ac- productivity, and intensity of bioturbation. A reason- counted for almost entirely by carbonate-complexed able limit in temporal resolution for organic carbon species;Ž. 3 the total dissolved inorganic carbon DIC in pelagic sediments is 1034 –10 yearŽ Kennett, and alkalinity of seawater remains constant over 1982. . The terrestrial stomatal-based PCO2 indicator geological time;Ž. 4 the level of dissolved oxygen probably offers the best temporal resolution. In this 380 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 case, the temporal resolution of pre-Quaternary leaves tinuous. sequences. Again, very good sequences ex- will typically be limited by the response time for a ist in marine settings. For example, the average given species, which can range anywhere between sampling density in Pagani et al.’sŽ. 1999a,b se- several months to 10 2 year. This gives the following quence was ;200 k.y., and in some settings it may sequence of temporal sensitivity: mass balance- be possible to sample at or near the single-sample paleosols f phytoplankton s boron isotopes - integration time Ž;1034 –10 year. . Sequences of stomata. This emphasis on timescales is important paleosols are less common, and generally more tem- because for a given method to detect a change in the porally discontinuous. Sequences of plant assem- concentration of atmospheric CO2 , its response time blages vary in temporal quality. For example, lacus- needs to be less than an individual component of the trine sequences can have annual resolution, but rarely global carbon cycle, i.e., turnover times of organic persist for great lengths of time. and inorganic carbon in the oceanic, atmospheric, Paleo-CO2 proxies also vary in their precision of marine and terrestrial components. PCO2 estimates. For geochemical models, the preci- Other important temporal considerations are how sion of PCO2 predictions can only be assessed via well an individual PCO2 estimate can be dated, and qualitative sensitivity analysesŽ. Section 2.2 . In the the availability of continuous sequences. In most case of Berner and KothavalaŽ. 2001 , CO2 sensitiv- cases, the errors in absolute dating far exceed tempo- ity ranges from "75–200 ppmV for the Tertiary to " ral resolutions, and so the ability to compare PCO2 1500–3000 ppmV for the early PaleozoicŽ see estimates from different localities is diminished. In Figs. 13–16. . The pedogenic carbonate proxy ranges general, however, marine-based proxies can be better in error from "400–500 ppmV for the Tertiary to dated than terrestrial-based ones. Temporal trends in "500–1000 ppmV for the mid-Paleozoic through

PCO2 are best captured in continuousŽ or near con- MesozoicŽ. see Fig. 14 . PCO2 error estimates from

Fig. 14. Comparison of model predictions of atmospheric CO2 over the PhanerozoicŽ. from Berner, 1994 and pedogenic carbonate-based estimatesŽ. including the goethite method . The shaded region represents the error estimates of Berner and Kothavala Ž. 2001 . RCO2 units of Berner and KothavalaŽ. 2001 converted to PCO2 as in Fig. 13. Vertical lines represent 78 pedogenic carbonate-based PCO2 estimatesŽ data from Suchecky et al., 1988; Platt, 1989; Cerling, 1991, 1992a,b; Koch et al., 1992; Muchez et al., 1993; Sinha and Stott, 1994; Andrews et al., 1995; Ghosh et al., 1995; Mora et al., 1996; Yapp and Poths, 1996; Ekart et al., 1999; Elick et al., 1999; Lee, 1999; Lee and Hisada,

1999. . The solid line is a five-point running average of the mean PCO2 of every estimate. This approach smoothes short-term CO2 fluctuations and is more directly comparable with the model of Berner and KothavalaŽ. 2001 . The dashed line is a five-point running average incorporating a recalculation of Ekart et al.Ž. 1999 data during the late Carboniferous and early Permian using the marine carbonate d13 C data of Popp et al.Ž.Ž 1986 see Sections 4.4 and 8 for details . . D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 381

Fig. 15. Comparison of model predictions of atmospheric CO2 over the last 60 MaŽ. gray shaded area; from Berner and Kothavala, 2001 11 and marine d B-based estimatesŽ.Ž. black shaded area; from Pearson and Palmer, 2000 . RCO2 units of Berner and Kothavala 2001 converted to PCO2 as in Fig. 13. the d13C of phytoplankton range from "25–100 Ž.Royer, unpublished data; see Fig. 13 . Depending ppmV for the Tertiary to "150–200 ppmV for the on the plant species chosen, this method’s sensitivity ; CretaceousŽ. see Fig. 16 . Above 1000 ppmV, this to PCO2 diminishes at some point above 350 ppmV. method loses its sensitivity. For stomatal indices, Additionally, quantitative pre-Cretaceous estimates error estimates for the Tertiary are "10–40 ppmV are probably not possible due to the lack of long-

Fig. 16. Comparison of model predictions of atmospheric CO2 over the last 160 MaŽ. from Berner and Kothavala, 2001 and marine organic matter d13 C-based estimates. Unmarked filled boxes from Freeman and HayesŽ. 1992 ; geoporphyrins used as a biomarker. Open box from A B 13 Pagani et al.Ž. 1999a,b ; C37:2 alkenones used as a biomarker. Box labeled s from Stott Ž.Ž 1992 does not include estimates from P–E d C excursion. . Middle line represents the Abest estimateB predictions of Berner and KothavalaŽ. 2001 , while the two straddling lines represent error estimates based on sensitivity analyses. RCO2 units of Berner and KothavalaŽ. 2001 converted to PCO2 as in Fig. 13. 382 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 ranging taxa. Thus, the phytoplankton and stomatal Despite the wide variety of assumptions made for proxies should be relied upon for TertiaryŽ and pos- each method, a surprisingly coherent picture is sibly Cretaceous. reconstructions, while the pedo- emerging of Phanerozoic PCO2 changeŽ see Figs. genic carbonate proxy yields the most useful CO2 13–16. . The largest discrepancy exists between some estimates for the pre-Tertiary. of the Paleogene estimates of Pearson and Palmer

Fig. 17. Changes in theŽ. a stomatal index and stomatal ratio of fossil Ginkgo and cycad taxa from Greenland,Ž. b ratio of atmospheric CO2 in the past relative to the pre-industrial valueŽ. RCO2 and calculated change in global temperature, and Ž. c calculated changes in leaf width with and without the reconstructed changes in CO2 and temperature and observed changes in leaf width across the Triassic–Jurassic boundary from the Greenland fossilsŽ. redrawn from McElwain et al., 1999 . D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 383

Fig. 17 Ž.continued .

Ž.2000 using the boron isotope method and the other envisage, for example, the complementary use of methodsŽ. compare Fig. 15 with Figs. 13, 14 and 16 . stomata and phytoplankton to detect CO2 changes As discussed aboveŽ. Section 6.2 , however, these across the K–T boundary or the P–E boundary. boron-derived estimates are probably not well con- Similarly, the stomatal proxy might usefully be ap- strained. Another discrepancy exists between the pe- plied across the Cenomanian–Turonian boundary dogenic carbonate estimates of Ekart et al.Ž. 1999 where phytoplankton and terrestrial plant d13C shifts and the geochemical predictions of GEOCARB for have indirectly identified a possible abrupt fall in the late Carboniferous and early Permian, i.e., the CO2 Ž. Kuypers et al., 1999 . later stages of the Permo-Carboniferous glaciation Ž.see Fig. 14 . Ekart et al. Ž. 1999 used the marine carbonate d13C record of Veizer et al.Ž. 1999 to 13 predict the d C of the pedogenic carbonate’s associ- 9. Independent testing of paleo-CO2 estimates ated organic matter. This procedure decreases the precision of PCO2 estimatesŽ. see Section 4.4 . The An important feature of the all of the methods of PCO2occ discrepancy largely disappears if the d from paleo-CO2 reconstructions reviewed here is the need Popp et al.Ž. 1986 are used instead of Veizer et al. to seek independent evidence for an associated Ž.Ž1999 see Fig. 14 . . Alternatively, the discrepancy change in climate. For the geochemical carbon cycle may be an artifact of the differing temporal resolu- model predictions and paleosol estimates, times of tions of the two methods. Unlike the GEOCARB low global CO2 concentrations over the Phanerozoic model, the pedogenic carbonate proxy has the poten- generally coincide with episodes of major glaciations 34 tial to resolve short-livedŽ 10 –10 year. PCO2 Ž.Berner, 1998; Crowley, 2000 , with the exception of excursions. If true, this example highlights the im- the Ordovician when the solar constant was some portance of considering temporal resolution when 4–5% lower than now. Other indicators of global comparing PCO2 estimates from multiple methods. climate could also be deployed for this purpose Nevertheless, it is important to apply multiple Ž.Parrish, 1998 but are only rarely used. Instead, the

PCO2 proxies for a given time interval. We might recourse typically taken has been to compare PCO2 384 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 estimates between different methodsŽ Berner, 1998; tions across the T–J boundary that were mirrored Ekart et al., 1999; Pearson and Palmer, 1999. . when stomatal ratios were calculated from SI mea-

Perhaps because the stomatal paleo-CO2 proxy surements on their nearest living equivalentsŽ Fig. approach is still in its infancy relative to most of the 17a. . Conversion of the SR values to PCO2 values other CO2 proxies, and regarded as somewhat con- indicated that the atmospheric CO2 concentration ; ; troversial, where changes in CO2 have been detected increased massively from 700 to 1600 ppmV using this method more effort has been made to seek across the boundary with associated global independent evidence of global change from the ‘greenhouse’ warmingŽ. Fig. 17b . This CO2 increase geologic record. For example, Neogene oscillations coincided with independent evidence for extensive in atmospheric CO2 inferred from changes in the SI volcanic activity during the breakup of Pangea of fossil oak leaves were found to correlate with Ž.Marzoli et al., 1999 , and moreover, the quantity of climatic records determined from fossil pollen records CO2 required to increase the atmospheric concentra- in the Lower Rhynie EmbaymentŽ van der Burgh et tion by this amount is well within that estimated to al., 1993. . Low SI values, i.e., high CO2 episodes, be have been released during the production of flood correlated with warm climatic periods indicated from basalts of the Central Atlantic Magmatic Province the fossil pollen assemblages. Reaching back further Ž.CAMP . The timing of the CAMP formation and its into the geologic record, Cleal et al.Ž. 1999 tracked a temporal brevity, determined from 40Arr39Ar dating, decrease in the SD and SI of pteridiosperm Neu- support its possible involvement in the T–J boundary ropteris oÕata fronds through the Westphalian and extinctionŽ Marzoli et al., 1999; McElwain et al., into the Cantabrian of the Upper Carboniferous with 1999. . the inference atmospheric CO2 partial pressure rose The fossil plants themselves also provide evi- through this time. These authors then found secure dence for global climatic warming across the T–J evidence that the postulated CO2 increase correlated boundaryŽ. McElwain et al., 1999 . Analysis of with a decrease in the extent of glacial deposits in changes in the morphologies of fossil leaves down Gondwana and the reduced burial of terrestrial or- through the different plant beds in Greenland show ganic matter resulting from the loss of tropical wet- that the major floral turnover indicated by pollen land forests. This example therefore provides a case analysis was linked to the extinction of broad-leaved where changes in fossil plant morphology yield a taxa and the survival of those groups possessing global signal of CO2 change of considerable antiq- finely divided or dissected leaf morphologiesŽ Fig. uity. 17c. . Energy budget calculations for individual Recent work has shown that it is possible to leaves, based on several key physiological attributes, utilize the relatively rapid response time of the stom- indicate that wide leaves with limited convective atal CO22 method to detect abrupt CO changes, in cooling capacity, because of a low boundary layer particular across the Triassic–JurassicŽ. T–J bound- conductance, would have quickly reached lethal tem- aryŽ.Ž 205.7 Ma McElwain et al., 1999 . . The T–J peratures for plants distributed at low paleolaltitudes, boundary is of interest because an absence of secure if the CO2 increase and global warming had actually marine geochemical records of climate change at this occurredŽ. Fig. 17c . In contrast, if the CO2 and time means that the causes of this, the third largest global temperatures increases implied by the stom- extinction in the Phanerozoic, continue to remain atal data were simply artifacts of the datasets, and uncertain. In this context, it is interesting to note that both had remained constant, then no selective pres- the timing of all of the major animal extinction sures would have operated for any observed reduc- events over the past 500 m.y. differs from that of tion in leaf width to maintain their temperature be- plant extinction events. Therefore, fossil plants are low lethal thresholdsŽ. Fig. 17c . This integration of potentially able to provide enlightening records of stomatal-based paleo-CO2 estimates with indepen- environmental change during major faunal mass ex- dent data sources reinforces the view that global tinction events. atmospheric signals can be derived from plant fossils McElwain et al.Ž. 1999 reported that the SI of and strengthens our confidence in the approach. As fossil Gingko and Cycad taxa showed marked reduc- new approaches for interpreting changes in different D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 385 components of the global carbon cycle emergeŽ e.g., Beerling, D.J., 1998a. Atmospheric carbon dioxide, past climates Haug et al., 1999. then it should be possible to and the plant fossil record. Botanical Journal of Scotland 51, 49–68. devise alternative means of testing proxy estimates Beerling, D.J., 1998b. The future as the key to the past. Trends in of PCO2 change against independent geologic data Ecology and Evolution 13, 311–316. sources. Beerling, D.J., 1999. Stomatal density and index: theory and application. In: Jones, T.P., Rowe, N.P.Ž. Eds. , Fossil Plants and Spores: Modern Techniques. Geological Society, London, pp. 251–256. Acknowledgements Beerling, D.J., 2000. Global terrestrial productivity in the Meso- zoic. In: Hart, M.B.Ž. Ed. , Past, Present and Future Climates. DLR and RAB were partially supported by DOE Geological Society of London Special Publication 181, pp. 17–32. grant DE-FGO2 -95ER14522. DLR also acknowledges support from a NSF Graduate Research Fel- Beerling, D.J., Chaloner, W.G., 1992. Stomatal density as an indicator of atmospheric CO2 concentration. Holocene 2, 71– lowship. DJB gratefully acknowledges funding 78. through a Royal Society University Research Fel- Beerling, D.J., Chaloner, W.G., 1993. Evolutionary responses of lowship and the Natural Environment Research stomatal density to global CO2 change. Biological Journal of Council, UKŽ. award no. GR3r11900 . We thank S. the Linnean Society 48, 343–353. Wofsy, B. Munger, M. Goulden, and C. Barford for Beerling, D.J., Kelly, C.K., 1997. Stomatal density responses of temperate woodland plants over the past seven decades of CO2 providing canopy CO2 data from Harvard Forest. increase: a comparison of SalisburyŽ. 1927 with contemporary We also thank L. Kump and W. Chaloner for con- data. American Journal of Botany 84, 1572–1583. structive comments. Beerling, D.J., Woodward, F.I., 1996a. Stomatal density responses to global change. Advances in Bioclimatology 4, 171–221. Beerling, D.J., Woodward, F.I., 1996b. Palaeo-ecophysiological perspectives on plant responses to global change. Trends in References Ecology and Evolution 11, 20–23. Beerling, D.J., Woodward, F.I., 1997. Changes in land plant Andrews, J.E., Tandon, S.K., Dennis, P.F., 1995. Concentration of function over the Phanerozoic: reconstructions based on the carbon dioxide in the Late Cretaceous atmosphere. Journal of fossil record. Botanical Journal of the Linnean Society 124, the Geological SocietyŽ. London 152, 1–3. 137–153. Andrusevich, V.E., Engel, M.H., Zumberge, J.E., 2000. Effects of Beerling, D.J., Chaloner, W.G., Huntley, B., Pearson, J.A., Too- paleolatitude on the stable carbon isotope composition of ley, M.J., 1993. Stomatal density responds to the glacial cycle crude oils. Geology 28, 847–850. of environmental change. Proceedings of the Royal Society of

Arens, N.C., Jahren, A.H., Amundson, R., 2000. Can C3 plants London, Series B 251, 133–138. faithfully record the carbon isotopic composition of atmo- Beerling, D.J., Birks, H.H., Woodward, F.I., 1995. Rapid late-gla-

spheric carbon dioxide? Paleobiology 26, 137–164. cial atmospheric CO2 changes reconstructed from the stomatal Arthur, M.A., Dean, W.E., Claypool, G.E., 1985. Anomalous 13 C density record of fossil leaves. Journal of Quaternary Science enrichment in modern marine organic carbon. Nature 315, 10, 379–384.

216–218. Berner, R.A., 1991. A model for atmospheric CO2 over Phanero- Baker, P.A., Gieskes, J.M., Elderfield, H., 1982. Diagenesis of zoic time. American Journal of Science 291, 339–376. carbonates in deep-sea sediments—evidence from SrrCa ra- Berner, R.A., 1994. GEOCARB II: a revised model of atmo- 2q tios and interstitial dissolved Sr data. Journal of Sedimen- spheric CO2 over Phanerozoic time. American Journal of tary Petrology 52, 71–82. Science 294, 56–91.

Baver, L.D., Gardner, W.H., Gardner, W.R., 1972. Soil Physics. Berner, R.A., 1998. The carbon cycle and CO2 over Phanerozoic Wiley, New York. time: the role of land plants. Philosophical Transactions of the

Bazzaz, F.A., Williams, W.E., 1991. Atmospheric CO2 concentra- Royal Society of London, Series B 353, 75–82. tions within a mixed forest: implications for seedling growth. Berner, R.A., 1999. A new look at the long-term carbon cycle. Ecology 72, 12–16. GSA Today 9Ž. 11 , 1–6. Beerling, D.J., 1993. Changes in the stomatal density of Betula Berner, R.A., Caldeira, K., 1997. The need for mass balance and nana leaves in response to increases in atmospheric carbon feedback in the geochemical carbon cycle. Geology 25, 955– dioxide concentration since the late-glacial. Special Papers in 956. Palaeontology 49, 181–187. Berner, R.A., Canfield, D., 1989. A new model for atmospheric Beerling, D.J., 1996. 13 C discrimination by fossil leaves during oxygen over Phanerozoic time. American Journal of Science the late-glacial climate oscillation 12–10 ka BP: measure- 289, 333–361. ments and physiological controls. Oecologia 108, 29–37. Berner, R.A., Kothavala, Z., 2001. GEOCARB III: a revised 386 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

model of atmospheric CO2 over Phanerozoic time. American stand structure and vegetation type. Global Change Biology 2, Journal of ScienceŽ. in review . 421–433. Berner, R.A., Lasaga, A.C., Garrels, R.M., 1983. The carbonate– Budyko, M.I., Ronov, A.B., Yanshin, A.L., 1987. History of the silicate geochemical cycle and its effect on atmospheric car- Earth’s Atmosphere. Springer, Berlin. bon dioxide over the past 100 millions years. American Jour- Burns, B.D., Beardall, J., 1987. Utilization of inorganic carbon by nal of Science 283, 641–683. marine microalgae. Journal of Experimental Marine Biology Berner, R.A., Petsch, S.T., Lake, J.A., Beerling, D.J., Popp, B.N., and Ecology 107, 75–86. Lane, R.S., Laws, E.A., Westley, M.B., Cassar, N., Wood- Caldeira, K., 1992. Enhanced Cenozoic chemical weathering and ward, F.I., Quick, W.P., 2000. Isotope fractionation and atmo- the subduction of pelagic carbonate. Nature 357, 578–581.

spheric oxygen: implications for Phanerozoic O2 evolution. Caldeira, K., Kasting, J.F., 1992. The life-span of the biosphere Science 287, 1630–1633. revisited. Nature 360, 721–723. Bettarini, I., Miglietta, F., Raschi, A., 1997. Studying morpho- Cerling, T.E., 1984. The stable isotopic composition of modern physiological responses of Scirpus lacustris from naturally soil carbonate and its relationship to climate. Earth and Plane-

CO2 -enriched environments. In: Rachi, A., Miglietta, F., tary Science Letters 71, 229–240. Tognetti, R., van Gardingen, P.R.Ž. Eds. , Plant Responses to Cerling, T.E., 1991. Carbon dioxide in the atmosphere: evidence

Elevated CO2 . Cambridge Univ. Press, Cambridge, pp. 134– from Cenozoic and Mesozoic paleosols. American Journal of 147. Science 291, 377–400.

Bettarini, I., Vaccari, F.P., Miglietta, F., 1998. Elevated CO2 Cerling, T.E., 1992a. Further comments on using carbon isotopes concentrations and stomatal density: observations from 17 in palaeosols to estimate the CO2 content of the paleo-atmo- plant species growing in a CO2 spring in central Italy. Global sphere. Journal of the Geological SocietyŽ. London 149, 673– Change Biology 4, 17–22. 675. Bidigare, R.R., Fluegge, A., Freeman, K.H., Hanson, K.L., Hayes, Cerling, T.E., 1992b. Use of carbon isotopes in paleosols as an

J.M., Hollander, D., Jasper, J.P., King, L.L., Laws, E.A., indicator of the PŽ. CO2 of the paleoatmosphere. Global Bio- Milder, J., Millero, F.J., Pancost, R., Popp, B.N., Steinberg, geochemical Cycles 6, 307–314. P.A., Wakeham, S.G., 1997. Consistent fractionation of 13 Cin Cerling, T.E., 1999. Stable carbon isotopes in palaeosol carbon- nature and in the laboratory: growth-rate effects in some ates. Special Publications of the International Association of haptophyte algae. Global Biogeochemical Cycles 11, 279–292. Sedimentologists 27, 43–60. Bidigare, R.R., Fluegge, A., Freeman, K.H., Hanson, K.L., Hayes, Cerling, T.E., Quade, J., Wang, Y., Bowman, J.R., 1989. Carbon J.M., Hollander, D., Jasper, J.P., King, L.L., Laws, E.A., isotopes in soils and palaeosols as ecology and palaeoecology Milder, J., Millero, F.J., Pancost, R., Popp, B.N., Steinberg, indicators. Nature 341, 138–139. P.A., Wakeham, S.G., 1999a. Correction to AConsistent frac- Cerling, T.E., Quade, J., Ambrose, S.D., Sikes, N.E., 1991a. tionation of 13 C in nature and in the laboratory: growth-rate Fossil soils, grasses, and carbon isotopes from Fort Ternan, effects in some haptophyte algaeB by R.R. Bidigare et al. Kenya: grassland or woodland? Journal of Human Evolution Global Biogeochemical Cycles 13, 251–252. 21, 295–306. Bidigare, R.R., Hanson, K.L., Buesseler, K.O., Wakeham, S.G., Cerling, T.E., Solomon, D.K., Quade, J., Bowman, J.R., 1991b. Freeman, K.H., Pancost, R.D., Millero, F.J., Steinberg, P., On the isotopic composition of carbon in soil carbon dioxide. Popp, B.N., Latasa, M., Landry, M.R., Laws, E.A., 1999b. Geochimica et Cosmochimica Acta 55, 3403–3405. Iron-stimulated changes in 13 C fractionation and export by Ciais, P., 1999. Restless carbon pools. Nature 398, 111–112. equatorial Pacific phytoplankton: toward a paleogrowth rate Clark, D.B., Clark, D.A., Grayum, M.H., 1992. Leaf demography proxy. Paleoceanography 14, 589–595. of a neotropical rain forest cycad, Zamia skinneri Ž.Zamiaceae . Birks, H.H., Eide, W., Birks, H.J.B., 1999. Early Holocene atmo- American Journal of Botany 79, 28–33.

spheric CO2 concentrations. Science 286, 1815a. Cleal, C.J., James, R.M., Zodrow, E.L., 1999. Variation in stom- Bocherens, H., Friis, E.M., Mariotti, A., Pedersen, K.R., 1993. atal density in the Late Carboniferous gymnosperm frond Carbon isotopic abundances in Mesozoic and Cenozoic fossil Neuropteris oÕata. Palaios 14, 180–185. plants: palaeoecological implications. Lethaia 26, 347–358. Clifford, S.C., Black, C.R., Roberts, J.A., Stronach, I.M., Single- Bounoua, L., Collatz, G.J., Sellers, P.J., Randall, D.A., Dazlich, ton-Jones, P.R., Mohamed, A.D., Azam-Ali, S.N., 1995. The

D.A., Los, S.O., Berry, J.A., Fung, I., Tucker, C.J., Field, effect of elevated atmospheric CO2 and drought on stomatal C.B., Jensen, T.G., 1999. Interactions between vegetation and frequency in groundnut ŽŽ..Arachis hypogaea L. . Journal of climate: radiative and physiological effects of a doubled atmo- Experimental Botany 46, 847–852.

spheric CO2 . Journal of Climate 12, 309–324. Conte, M.H., Volman, J.K., Eglinton, G., 1994. Lipid biomarkers Brassell, S.C., Eglinton, G., Marlowe, I.T., Pfaumann, U., Sarn- of the Haptophyta. In: Green, J.C., Leadbeater, B.S.C.Ž. Eds. , thein, M., 1986. Molecular stratigraphy: a new tool for cli- The Haptophyte Algae. Clarendon Press, Oxford, pp. 351–377. matic assessment. Nature 320, 129–133. Cowan, I.R., 1977. Stomatal behaviour and environment. Ad- Brook, G.A., Folkoff, M.E., Box, E.O., 1983. A world model of vances in Botanical Research 77, 1176–1227. soil carbon dioxide. Earth Surface Landforms 8, 79–88. Craig, H., 1953. The geochemistry of stable carbon isotopes. Buchmann, N., Kao, W.Y., Ehleringer, J.R., 1996. Carbon dioxide Geochimica et Cosmochimica Acta 3, 53–92. concentrations within forest canopies: variations with time, Crowley, T.J., 2000. Carbon dioxide and Phanerozoic climate. In: D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 387

Huber, B.T., MacLeod, K.G., Wing, S.L.Ž. Eds. , Warm Cli- Farquhar, G.D., Richards, R.A., 1984. Isotopic composition of mates in Earth History. Cambridge Univ. Press, Cambridge, plant carbon correlates with water-use efficiency of wheat pp. 425–444. genotypes. Australian Journal of Plant Physiology 11, 539– Davidson, G.R., 1995. The stable isotopic composition and mea- 552. surement of carbon in soil CO2. Geochimica et Cosmochimica Farquhar, G.D., O’Leary, M.H., Berry, J.A., 1982. On the rela- Acta 59, 2485–2489. tionship between carbon isotope discrimination and the inter- Dean, W.E., Arthur, M.A., Claypool, G.E., 1986. Depletion of cellular carbon dioxide concentration in leaves. Australian 13 C in Cretaceous marine organic matter: source, diagenetic, Journal of Plant Physiology 9, 121–137. or environmental signal? Marine Geology 70, 119–157. Farrimond, P., Eglinton, G., Brassell, S.C., 1986. Alkenones in Degens, E.T., Guillard, R.R.L., Sackett, W.M., Hellebust, J.A., Cretaceous black shales, Blake-Bahama Basin, western North 1968. Metabolic fractionation of carbon isotopes in marine Atlantic. Organic Geochemistry 10, 897–903. plankton: I. Temperature and respiration experiments. Deep- Fernandez,´ M.D., Pieter, A., Donso, C., Tezara, W., Azkue, M., Sea Research 15, 1–9. Herrera, C., Rengifo, E., Herrera, A., 1998. Effects of a

Deines, P., 1980. The isotopic composition of reduced organic natural source of very high CO2 concentration on the leaf gas carbon. In: Fritz, P., Fontes, J.Ch.Ž. Eds. , Handbook of Envi- exchange, xylem water potential and stomatal characteristics ronmental Geochemistry, vol. 1. Elsevier, Amsterdam, pp. of Spatiphylum cannifolium and Bauhinia multiÕerÕia. New 329–406. Phytologist 138, 689–697. De La Rocha, C.L., De Paolo, D.J., 2000. Isotopic evidence for France-Lanord, C., Derry, L.A., 1997. Organic carbon burial variations in the marine calcium cycle over the Cenozoic. forcing of the carbon cycle from Himalayan erosion. Nature Science 289, 1176–1178. 390, 65–67. Derry, L.A., France-Lanord, C., 1996. Neogene growth of the François, L.M., Godderis, Y., 1998. Isotopic constraints on the sedimentary organic carbon reservoir. Paleoceanography 11, Cenozoic evolution of the carbon cycle. Chemical Geology 267–275. 145, 177–212. Dorr,¨¨ H., Munnich, K.O., 1990. 222 Rn flux and soil air concentra- François, L.M., Walker, J.C.G., 1992. Modeling the Phanerozoic tion profiles in West Germany. Soil 222 Rn as tracer for gas carbon cycle and climate constraints from the 87Srr86 Sr iso- transport in the unsaturated soil zone. Tellus 42B, 20–28. topic ratio of seawater. American Journal of Science 292, Driese, S.G., Mora, C.I., Cotter, E., Foreman, J.L., 1992. Paleope- 81–135. dology and stable isotope chemistry of Late Silurian vertic François, R., Altabet, M.A., Goericke, R., McCorkle, D.C., Brunet, paleosols, Bloomsburg Formation, central Pennsylvania. Jour- C., Poisson, A., 1993. Changes in the d13C of surface water nal of Sedimentary Petrology 62, 825–841. particulate organic matter across the subtropical convergence Edwards, D.E., 1998. Climate signals in Palaeozoic land plants. in the S.W. Indian Ocean. Global Biogeochemical Cycles 7, Philosophical Transactions of the Royal Society of London, 627–644. Series B 353, 141–157. Freeman, K.H., Hayes, J.M., 1992. Fractionation of carbon iso-

Ehleringer, J.R., Cerling, T.E., 1995. Atmospheric CO2 and the topes by phytoplankton and estimates of ancient CO2 levels. ratio of intercellular to ambient CO2 concentrations in plants. Global Biogeochemical Cycles 6, 185–198. Tree Physiology 15, 105–111. Friedli, H., Lotscher,¨ H., Oeschger, H., Siegenthaler, U., Stauffer, Ekart, D.D., Cerling, T.E., Montanez,˜ I.P., Tabor, N.J., 1999. A B., 1986. Ice core record of the 13 Cr12 C ratio of atmospheric 400 million year carbon isotope record of pedogenic carbon- CO2 in the past two centuries. Nature 324, 237–238. ate: implications for paleoatmospheric carbon dioxide. Ameri- Fry, B., Wainwright, S.C., 1991. Diatom sources of 13 C-rich can Journal of Science 299, 805–827. carbon in marine food webs. Marine Ecology: Progress Series Ekstrom, A., Fookes, C.J.R., Hambley, T., Loehh, J., Miller, S.A., 76, 149–157. Taylor, J.C., 1983. Determination of the crystal structure of a Gale, J., 1972. Availability of carbon dioxide for photosynthesis at petroporphyrin isolated from oil shale. Nature 306, 173–174. high altitudes: theoretical considerations. Ecology 53, 494– Elick, J.M., Mora, C.I., Driese, S.G., 1999. Elevated atmospheric 497.

CO2 levels and expansion of early vascular land plants: stable Ghosh, P., Bhattacharya, S.K., Jani, R.A., 1995. Palaeoclimate isotope evidence from the Battery Point Fm.Ž Early to Middle and palaeovegetation in central India during the Upper Creta- Devonian. , Gaspe´ Bay, Canada. GSA Abstracts with Programs ceous based on stable isotope composition of the palaeosol 31, A-159. carbonates. Palaeogeography, Palaeoclimatology, Palaeoecol- Estiarte, M., Penuelas,˜ J., Kimball, B.A., Idso, S.B., LaMorte, ogy 114, 285–296. R.L., Pinter, P.J., Wall, G.W., Garcia, R.L., 1994. Elevated Gibbs, M.T., Barron, E.J., Kump, L.R., 1997. An atmospheric

CO2 effects on stomatal density of wheat and sour orange pCO2 threshold for glaciation in the Late Ordovician. Geology trees. Journal of Experimental Botany 45, 1665–1668. 25, 447–450. Evans, J.R., Sharkey, T.D., Berry, J.A., Farquhar, G.D., 1986. Gibbs, M.T., Bluth, G.J.S., Fawcett, P.J., Kump, L.R., 1999. Carbon isotope discrimination measured concurrently with gas Global chemical erosion over the last 250 my: variations due

exchange to investigate CO2 diffusion in leaves of higher to changes in paleogeography, paleoclimate, and paleogeol- plants. Australian Journal of Plant Physiology 13, 281–292. ogy. American Journal of Science 299, 611–651. 388 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

Gile, L.H., Peterson, F.F., Grossman, R.B., 1979. The Desert Indermuhle,¨ A., Stauffer, B., Stocker, T.F., Raynaud, D., Barnola, Project Soil Monograph. U.S. Soil Conservation Service, J.M., 1999a. Early Holocene atmospheric CO2 concentrations. Washington. Science 286, 1815a. Givnish, T.J., 1988. Adaptation to sun and shade: a whole plant Indermuhle,¨ A., Stocker, T.F., Joos, F., Fischer, H., Smith, H.J., perspective. Australian Journal of Plant Physiology 15, 63–92. Wahlen, M., Deck, B., Mastroianni, D., Tschumi, J., Blunier, Godderis, Y., Franc¸ois, L.M., 1995. The Cenozoic evolution of T., Meyer, R., Stauffer, B., 1999b. Holocene carbon-cycle the strontium and carbon cycles: relative importance of conti- dynamics based on CO2 trapped in ice at Taylor Dome, nental erosion and mantle exchanges. Chemical Geology 126, Antarctica. Nature 398, 121–126. 169–190. Jarvis, P.G., McNaughton, K.G., 1985. Stomatal control of tran- Godderis, Y., Franc¸ois, L.M., 1996. Balancing the Cenozoic spiration: scaling up from leaf to region. Advances in Ecologi- carbon and alkalinity cycles: constraints from isotopic records. cal Research 15, 1–49.

Geophysical Research Letters 23, 3743–3746. Jasper, J.P., Hayes, J.M., 1990. A carbon isotope record of CO2 Goericke, R., Montoya, J.P., Fry, B., 1994. Physiology of isotope levels during the late Quaternary. Nature 347, 462–464. fractionation in algae and cyanobacteria. In: Lajtha, K., Mich- Jasper, J.P., Hayes, J.M., Mix, A.C., Prahl, F.G., 1994. Photosyn- ener, R.H.Ž. Eds. , Stable Isotopes in Ecology. Blackwell Sci- thetic fractionation of 13 C and concentrations of dissolved

ence, Oxford, pp. 187–221. CO2 in the central equatorial Pacific during the last 255,000 Grace, J., Lloyd, J., McIntyre, J., Miranda, A., Meir, P., Miranda, years. Paleoceanography 9, 781–798. H., Moncrieff, J., Massheder, J., Wright, I., Gash, J., 1995. Jones, H.G., 1992. Plants and Microclimate. Cambridge Univ. Fluxes of carbon dioxide and water vapour over an undis- Press, Cambridge. turbed tropical forest in south-west Amazonia. Global Change Jones, T.P., 1994. 13 C enriched Lower Carboniferous fossil plants Biology 1, 1–13. from Donegal, Ireland: carbon isotope constraints on taphon- Haug, G.H., Sigman, D.M., Tiedemann, R., Pedersen, T.F., Sarn- omy, diagenesis and palaeoenvironment. Review of Palaeob- thein, M., 1999. Onset of permanent stratification in the otany and Palynology 81, 53–64. subarctic Pacific Ocean. Nature 401, 779–782. Jones, T.P., Rowe, N.P.Ž. Eds. 1999. Fossil Plants and Spores: Hayes, J.M., 1993. Factors controlling 13 C contents of sedimen- Modern Techniques. Geological Society, London. tary organic compounds: principles and evidence. Marine Ge- Kelly, C.K., Beerling, D.J., 1995. Plant life form, stomatal density ology 113, 111–125. and taxonomic relatedness: a reanalysis of SalisburyŽ. 1927 . Hayes, J.M., Takigiku, R., Ocampo, R., Callot, H.J., Albrecht, P., Functional Ecology 9, 422–431. 1987. Isotopic compositions and probable origins of organic Kennett, J.P., 1982. Marine Geology. Prentice-Hall, Englewood molecules in the Eocene Messel shale. Nature 329, 48–51. Cliffs. Hayes, J.M., Freeman, K.H., Popp, B.N., Hoham, C.H., 1989a. Kirkham, D., Powers, W.L., 1972. Advanced Soil Physics. Compound-specific isotopic analyses: a novel tool for recon- Wiley-Interscience, New York. struction of ancient biogeochemical processes. Organic Geo- Koch, P.L., Zachos, J.C., Gingerich, P.D., 1992. Correlation chemistry 16, 1115–1128. between isotope records in marine and continental carbon Hayes, J.M., Popp, B.N., Takigiku, R., Johnson, M.W., 1989b. An reservoirs near the PalaeocenerEocene boundary. Nature 358, isotopic study of biogeochemical relationships between car- 319–322. bonates and organic carbon in the Greenhorn Formation. Korner,¨ Ch., Bannister, P., Mark, A.F., 1986. Altitudinal variation Geochimica et Cosmochimica Acta 53, 2961–2972. in stomatal conductance, nitrogen content and leaf anatomy in Hayes, J.M., Strauss, H., Kaufman, A.J., 1999. The abundance of different plant life forms in New Zealand. Oecologia 69, 13 C in marine organic matter and isotopic fractionation in the 577–588. global biogeochemical cycle of carbon during the past 800 Kothavala, Z., Oglesby, R.J., Saltzman, B., 1999. Sensitivity of Ma. Chemical Geology 161, 103–125. equilibrium surface temperature of CCM3 to systematic

Hemming, N.G., Hanson, G.N., 1992. Boron isotopic composition changes in atmospheric CO2 . Geophysical Research Letters and concentration in modern marine carbonates. Geochemica 26, 209–212. et Cosmochimica Acta 56, 537–543. Kump, L.R., Arthur, M.A., 1997. Global chemical erosion during Hinga, K.R., Arthur, M.A., Pilson, M.E.Q., Whitaker, D., 1994. the Cenozoic: weatherability balances the budgets. In: Ruddi- Carbon isotope fractionation by marine phytoplankton in cul- man, W.F.Ž. Ed. , Tectonic Uplift and Climate Change. Plenum,

ture: the effects of CO2 concentration, pH, temperature, and New York, pp. 399–426. species. Global Biogeochemical Cycles 8, 91–102. Kump, L.R., Arthur, M.A., 1999. Interpreting carbon-isotope ex-

Hollander, D.J., McKenzie, J.A., 1991. CO2 control on carbon- cursions: carbonates and organic matter. Chemical Geology isotope fractionation during aqueous photosynthesis: a paleo- 161, 181–198.

pCO2 barometer. Geology 19, 929–932. Kump, L.R., Arthur, M.A., Patzkowsky, M.E., Gibbs, M.T., Hovenden, M.J., Schimanski, L.J., 2000. Genotypic differences in Pinkus, D.S., Sheenan, P.M., 1999. A weathering hypothesis

growth and stomatal morphology of southern beech, Nothofa- for glaciation at high atmospheric pCO2 during the Late gus cunninghamii, exposed to depleted CO2 concentrations. Ordovician. Palaeoclimatology, Palaeogeography, Palaeoecol- Australian Journal of Plant Physiology 27, 281–287. ogy 152, 173–187. D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 389

Kundu, S.K., Tigerstedt, P.M.A., 1999. Variations in net photo- nental flood basalts of the central atlantic magmatice province. synthesis, stomatal characteristics, leaf area and whole-plant Science 284, 616–618. phytomass production among ten provenances of neem Mason, E.A., Marrero, T.R., 1970. The diffusion of atoms and Ž.Azadirachata indica . Tree Physiology 19, 47–52. molecules. In: Bates, D.R., Esterman, I.Ž. Eds. , Advances in Kurschner,¨ W.M., 1997. The anatomical diversity of recent and Atomic and Molecular Physics, vol. 6. Academic Press, New fossil leaves of the durmast oak ŽQuercus petraea LiebleinrQ. York, pp. 155–226. pseudocastanea Goeppert. —implications for their use as McCabe, B., 1985. The Dynamics of 13 C in Several New Zealand

biosensors of palaeoatmospheric CO2 levels. Review of Lakes. PhD Thesis. University of Waikato, New Zealand. Palaeobotany and Palynology 96, 1–30. McCauley, S.E., DePaolo, D.J., 1997. The marine 87Srr86 Sr and Kurschner,¨ W.M., van der Burgh, J., Visscher, H., Dilcher, D.L., d18 O records, Himalayan alkalinity fluxes and Cenozoic cli- 1996. Oak leaves as biosensors of late Neogene and early mate models. In: Ruddiman, W.F.Ž. Ed. , Tectonic Uplift and

Pleistocene paleoatmospheric CO2 concentrations. Marine Mi- Climate Change. Plenum, New York, pp. 427–467. cropaleontology 27, 299–312. McElwain, J.C., 1998. Do fossil plants signal palaeoatmospheric

Kurschner,¨ W.M., Wagner, F., Visscher, E.H., Visscher, H., 1997. CO2 concentration in the geological past? Philosophical Predicting the response of leaf stomatal frequency to a future Transactions of the Royal Society of London, Series B 353,

CO2 -enriched atmosphere: constraints from historical observa- 83–96. tions. Geologische Rundschau 86, 512–517. McElwain, J.C., Chaloner, W.G., 1995. Stomatal density and Kuypers, M.M.M., Pancost, R.D., Sinninghe Damste, J.S.S., 1999. index of fossil plants track atmospheric carbon dioxide in the

A large and abrupt fall in atmospheric CO2 concentration Palaeozoic. Annals of Botany 76, 389–395. during Cretaceous times. Nature 399, 342–345. McElwain, J.C., Chaloner, W.G., 1996. The fossil cuticle as a Lasaga, A.C., Berner, R.A., Garrels, R.M., 1985. An improved skeletal record of environmental change. Palaios 11, 376–388.

geochemical model of atmospheric CO2 fluctuations over the McElwain, J., Mitchell, F.J.G., Jones, M.B., 1995. Relationship of past 100 million years. In: Sunquist, E.T., Broecker, W.S. stomatal density and index of Salix cinerea to atmospheric

Ž.Eds. , The Carbon Cycle and Atmospheric CO2 : Natural carbon dioxide concentrations in the Holocene. Holocene 5, Variations Archean to Present. Geophysical Monographs vol. 216–219. 32. American Geophysical Union, 397. McElwain, J.C., Beerling, D.J., Woodward, F.I., 1999. Fossil Laws, E.A., Popp, B.N., Bidigare, R.R., Kennicutt, M.C., Macko, plants and global warming at the Triassic–Jurassic boundary. S.A., 1995. Dependence of phytoplankton carbon isotopic Science 285, 1386–1390. wx composition on growth rate and CO2aq : theoretical consider- Moffett, J.W., 1990. Microbially mediated cerium oxidation in sea ations and experimental results. Geochimica et Cosmochimica water. Nature 345, 421–423. Acta 59, 1131–1138. Mook, W.G., Bommerson, J.C., Staberman, W.H., 1974. Carbon Laws, E.A., Bidigare, R.R., Popp, B.N., 1997. Effect of growth isotope fractionation between dissolved bicarbonate and

rate and CO2 concentration on carbon isotopic fractionation by gaseous carbon dioxide. Earth and Planetary Science Letters the marine diatom Phaeodactylum tricornutum. Limnology 22, 169–176. and Oceanography 42, 1552–1560. Mora, C.I., Driese, S.G., 1999. Palaeoenvironment, palaeoclimate Lee, Y.I., 1999. Stable isotopic composition of calcic paleosols of and stable carbon isotopes of Palaeozoic red-bed palaeosols, the Early Cretaceous Hasandong Formation, southeastern Ko- Appalachian Basin, USA and Canada. Special Publications of rea. Palaeogeography, Palaeoclimatology, Palaeoecology 150, the International Association of Sedimentologists 27, 61–84. 123–133. Mora, C.I., Driese, S.G., Seager, P.G., 1991. Carbon dioxide in Lee, Y.I., Hisada, K., 1999. Stable isotopic composition of pedo- the Paleozoic atmosphere: evidence from carbon-isotope com- genic carbonates of the Early Cretaceous Shimonoseki Sub- positions of pedogenic carbonate. Geology 19, 1017–1020. group, western Honshu, Japan. Palaeogeography, Palaeoclima- Mora, C.I., Driese, S.G., Colarusso, L.A., 1996. Middle and Late

tology, Palaeoecology 153, 127–138. Paleozoic atmospheric CO2 levels from soil carbonate and Lemarchand, D., Gaillardet, J., Lewin, E.,´ Allegre,` C.J., 2000. organic matter. Science 271, 1105–1107. Boron isotope systematics in large rivers: implications for Morison, J.I.L., 1985. Sensitivity of stomata and water use effi- 11 oceanic d B secular variations. NatureŽ. in review . ciency to high CO2 . Plant, Cell and Environment 8, 467–474. Liu, Y.-G., Schmitt, R.A., 1996. Cretaceous Tertiary phenomena Muchez, P., Peeters, C., Keppens, E., Viaene, W.A., 1993. Stable

in the context of seafloor rearrangements and PŽ.CO2 fluctua- isotopic composition of paleosols in the Lower Visan of tions over the past 100 m.y. Geochimica et Cosmochimica eastern Belgium: evidence of evaporation and soil–gas CO2 . Acta 60, 973–994. Chemical Geology 106, 389–396. Marlowe, I.T., Brassell, S.C., Eglinton, G., Green, J.C., 1990. Nielsen, S.L., Gacia, E., Sand-Jensen, K., 1991. Land plants of Long-chain alkenones and alkyl alkenoates and the fossil amphibious Littorella uniflora Ž.L. Aschers. maintain utiliza-

coccolith record of marine sediments. Chemical Geology 88, tion of CO2 from the sediment. Oecologia 88, 258–262. 349–375. O’Leary, M.H., 1984. Measurement of the isotope fractionation Marzoli, A., Renne, P.R., Piccirillo, E.M., Ernesto, M., Belllieni, associated with diffusion of carbon dioxide in aqueous solu- G., De Min, A., 1999. Extensive 200-million-year-old conti- tion. Journal of Physical Chemistry 88, 823–825. 390 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

Pagani, M., Arthur, M.A., Freeman, K.H., 1999a. Miocene evolu- Tilbrook, B., Griffiths, F.B., Wright, S.W., Marchant, H.J., tion of atmospheric carbon dioxide. Paleoceanography 14, Bidigare, R.R., Laws, E.A., 1999. Controls on the carbon 273–292. isotopic composition of Southern Ocean phytoplankton. Global Pagani, M., Freeman, K.H., Arthur, M.A., 1999b. Late Miocene Biogeochemical Cycles 13, 827–843.

atmospheric CO24 concentrations and the expansion of C Prahl, F.G., Wakeham, S.G., 1987. Calibration of unsaturation grasses. Science 285, 876–879. patterns in long-chain ketone compositions for palaeotempera- Pagani, M., Freeman, K.H., Arthur, M.A., 2000. Isotope analyses ture assessment. Nature 330, 367–369. of molecular and total organic carbon from Miocene sedi- Quade, J., Cerling, T.E., Bowman, J.R., 1989. Systematic varia- ments. Geochimica et Cosmochimica Acta 64, 37–49. tions in the carbon and oxygen isotopic composition of pedo- Palmer, M.R., Pearson, P.N., Cobb, S.J., 1998. Reconstructing genic carbonate along elevation transects in the southern Great past ocean pH-depth profiles. Science 282, 1468–1471. Basin, United States. Geological Society of America Bulletin Paoletti, E., Nourrisson, G., Garrec, J.P., Raschi, A., 1998. Modi- 101, 464–475. fications of the leaf surface structures of Quercus ilex L. in Rau, G.H., Takahashi, T., Des Mariais, D.J., 1989. Latitudinal 13 open, naturally CO2 -enriched environments. Plant, Cell and variations in plankton d C: implications for CO2 and produc- Environment 21, 1071–1075. tivity in past oceans. Nature 341, 516–518. Parrish, J.T., 1998. Interpreting Pre-Quaternary Climate from the Rau, G.H., Froelich, P.N., Takahasi, T., Des Marais, D.J., 1991a. Geologic Record. Columbia Univ. Press, New York. Does sedimentary organic d13C record variations in Quater- wx Pearson, P.N., Palmer, M.R., 1999. Middle Eocene seawater pH nary ocean CO2Žaq. ? Paleoceanography 6, 335–347. and atmospheric carbon dioxide concentrations. Science 284, Rau, G.H., Takahashi, T., Des Marais, D.J., Sullivan, C.W., 1824–1826. 1991b. Particulate organic matter d13C variations across the Pearson, P.N., Palmer, M.R., 2000. Atmospheric carbon dioxide Drake Passage. Journal of Geophysical Research 96, 15131– concentrations over the past 60 million years. Nature 406, 15135. 695–699. Rau, G.H., Takahasi, T., Des Marais, D.J., Repeta, D.J., Martin, Pedersen, O., Sand-Jensen, K., 1992. Adaptations of submerged J.H., 1992. The relationship between d13C of organic matter wx Lobelia dortmanna to aerial life form: morphology, carbon and CO2Žaq. in ocean surface water: data from a JGOFS site sources and oxygen dynamics. Oikos 65, 89–96. in the northeast Atlantic Ocean and a model. Geochimica et Petit, J.R., Jouzel, J., Raynaud, D., Barkov, N.I., Barnola, J.-M., Cosmochimica Acta 56, 1413–1419. Basile, I., Bender, M., Chappellaz, J., Davis, M., Delaygue, Raymo, M.E., 1997. Carbon cycle models—how strong are the G., Delmotte, M., Kotlyakov, V.M., Legrand, M., Lipenkov, constraints? In: Ruddiman, W.F.Ž. Ed. , Tectonic Uplift and V.Y., Lorius, C., Pepin,´ L., Ritz, C., Saltzman, E., Stievenard, Climate Change. Plenum, New York, pp. 367–381. M., 1999. Climate and atmospheric history of the past 420,000 Raymo, M.E., Ruddiman, W.F., 1992. Tectonic forcing of late years from the Vostok ice core, Antarctica. Nature 399, 429– Cenozoic climate. Nature 359, 117–122. 436. Reich, P.B., Uhl, C., Walters, M.B., Ellsworth, D.S., 1991. Leaf Platt, N.H., 1989. Lacustrine carbonates and pedogenesis: sedi- lifespan as a determinant of leaf structure and function among mentology and origin of palustrine deposits from the Early 23 amazonian tree species. Oecologia 86, 16–24. Cretaceous Rupelo Formation, W. Cameros Basin, N. Spain. Reinfelder, J.R., Kraepiel, A.M.L., Morel, F.M.M., 2000. Unicel-

Sedimentology 36, 665–684. lular C4 photosynthesis in a marine diatom. Nature 407, Polley, H.W., Johnson, H.B., Marino, B.D., Mayeux, H.S., 1993. 996–999.

Increase in C3 plant water-use efficiency and biomass over Retallack, G.J., 1988. Field recognition of paleosols. Geological Glacial to present CO2 concentrations. Nature 361, 61–64. Society of America Special Papers 216, 1–20. Poole, I., Weyers, J.D.B., Lawson, T., Raven, J.A., 1996. Varia- Retallack, G.J., 1990. Soils of the Past. Unwin-Hyman, London. tions in stomatal density and index: implications for palaeocli- Roeske, C.A., O’Leary, M.H., 1985. Carbon isotope effect on mate reconstructions. Plant, Cell and Environment 19, 705– carboxylation of ribulose biphosphate catalyzed by ribulose- 712. biphosphate carboxylase from Rhodosprillum rubrum. Bio- Popp, B.N., Anderson, T.F., Sandberg, P.A., 1986. Brachiopods chemistry 24, 1603–1607. as indicators of original isotopic compositions in some Paleo- Romanek, C.S., Grossman, E.L., Morse, J.W., 1992. Carbon zoic limestones. Bulletin of the Geological Society of America isotopic fractionation in synthetic aragonite and calcite: effects 97, 1262–1269. of temperature and precipitation rate. Geochimica et Cos- Popp, B.N., Takigiku, R., Hayes, J.M., Louda, J.W., Baker, E.W., mochimica Acta 56, 419–430. 1989. The post-Paleozoic chronology and mechanism of 13 C Royer, D.L., 1999. Depth to pedogenic carbonate horizon as a depletion in primary marine organic matter. American Journal paleoprecipitation indicator? Geology 27, 1123–1126. of Science 289, 436–454. Royer, D.L., 2001. Stomatal density and stomatal index as indica-

Popp, B.N., Laws, E.A., Bidigare, R.R., Dore, J.E., Hanson, K.L., tors of atmospheric CO2 concentration. Review of Palaeob- Wakeham, S.G., 1998. Effect of phytoplankton cell geometry otany and PalynologyŽ. in press .

on carbon isotopic fractionation. Geochimica et Cosmochimica Rundgren, M., Beerling, D., 1999. A Holocene CO2 record from Acta 62, 69–77. the stomatal index of sub-fossil Salix herbacea L. leaves from Popp, B.N., Trull, T., Kenig, F., Wakeham, S.G., Rust, T.M., northern Sweden. Holocene 9, 509–513. D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392 391

Salisbury, E.J., 1927. On the causes and ecological significance of 2000. Comparative field water relations of three Mediter-

stomatal frequency, with special reference to the woodland ranean shrub species co-occurring in a natural CO2 vent. flora. Philosophical Transactions of the Royal Society of Journal of Experimental Botany 51, 1135–1146. London, Series B 216, 1–65. van de Water, P.K., Leavitt, S.W., Betancourt, J.L., 1994. Trends Sanyal, A., Hemming, N.G., Hanson, G.N., Broecker, W.S., 1995. in stomatal density and 13 Cr12 C ratios of Pinus flexilis Evidence for a higher pH in the glacial ocean from boron needles during last glacial–interglacial cycle. Science 264, isotopes in foraminifera. Nature 373, 234–236. 239–243. Sanyal, A., Hemming, N.G., Broecker, W.S., Lea, D.W., Spero, van der Burgh, J., Visscher, H., Dilcher, D.L., Kurschner,¨ W.M., H.J., Hanson, G.N., 1996. Oceanic pH control on the boron 1993. Paleoatmospheric signatures in Neogene fossil leaves. isotopic composition of foraminifera: evidence from culture Science 260, 1788–1790. experiments. Paleoceanography 11, 513–517. Veizer, J., Ala, D., Azmy, K., Brucjschen, P., Buhl, D., Bruhn, F., Sharma, G.K., Dunn, D.B., 1968. Effect of environment on the Carden, G.A.F., Diener, A., Ebneth, S., Godderis, Y., Jasper, cuticular features in Kalanchoe fedschenkoi. Bulletin of the T., Korte, C., Pawellek, F., Podlaha, O.G., Strauss, H., 1999. Torrey Botanical Club 95, 464–473. 87Srr86 Sr, d13C and d18 O evolution of Phanerozoic seawater. Sharma, G.K., Dunn, D.B., 1969. Environmental modifications of Chemical Geology 161, 59–88. leaf surface traits in Datura stramonium. Canadian Journal of Vengosh, A., Kolodny, Y., Starinsky, A., Chivas, A.R., McCul- Botany 47, 1211–1216. loch, M.T., 1991. Coprecipitation and isotopic fractionation of

Sinha, A., Stott, L.D., 1994. New atmospheric pCO2 estimates boron in modern biogenic carbonates. Geochimica et Cos- from paleosols during the late Paleocenerearly Eocene global mochimica Acta 55, 2901–2910. warming interval. Global and Planetary Change 9, 297–307. Wagner, F., 1998. The influence of environment on the stomatal Sloan, L.C., Rea, D.K., 1995. Atmospheric carbon dioxide and frequency in Betula. LPP Contributions Series 9, 1–102. early Eocene climate: a general circulation modeling sensitiv- Wagner, F., Below, R., De Klerk, P., Dilcher, D.L., Joosten, H., ity study. Palaeogeography, Palaeoclimatology, Palaeoecology Kurschner,¨ W.M., Visscher, H., 1996. A natural experiment 119, 275–292. on plant acclimation: lifetime stomatal frequency response of

Solomon, D.K., Cerling, T.E., 1987. The annual carbon dioxide an individual tree to annual atmospheric CO2 increase. Pro- cycle in a montane soil: observations, modeling, and implica- ceedings of the National Academy of Sciences of the United tions for weathering. Water Resources Research 23, 2257– States of America 93, 11705–11708. 2265. Wagner, F., Bohncke, S.J.P., Dilcher, D.L., Kurschner,¨ W.M., van Spicer, R.A., 1981. The sorting and deposition of allochthonous Geel, B., Visscher, H., 1999. Century-scale shifts in early

plant material in a modern environment at Silwood Lake, Holocene atmospheric CO2 concentration. Science 284, 1971– Silwood Park, Berkshire, England. U.S. Geological Survey 1973. Professional Paper 1143, 1–77. Walker, J.C.G., Hays, P.B., Kasting, J.F., 1981. A negative Spivack, A.J., You, C.-F., 1997. Boron isotopic geochemistry of feedback mechanism for the long-term stabilization of Earth’s carbonates and pore waters, Ocean Drilling Program Site 851. surface temperature. Journal of Geophysical Research 86, Earth and Planetary Science Letters 152, 113–122. 9776–9782. Spivack, A.J., You, C.-F., Smith, H.J., 1993. Foraminiferal boron Wallman, K., 2001. Controls on the Cretaceous and Cenozoic

isotope ratios as a proxy for surface ocean pH over the past 21 evolution of seawater composition, atmospheric CO2 and cli- Myr. Nature 363, 149–151. mate. Geochimica et Cosmochimica ActaŽ. in review . Stanhill, G., 1986. Water use efficiency. Advances in Agronomy White, J.W.C., Ciais, P., Figge, R.A., Kenny, R., Markgraf, V.,

39, 53–85. 1994. A high-resolution record of atmospheric CO2 content Stott, L.D., 1992. Higher temperatures and lower oceanic pCO2 : a from carbon isotopes in peat. Nature 367, 153–156. climate enigma at the end of the Paleocene epoch. Paleo- Williams-Linera, G., 1997. Phenology of deciduous and ceanography 7, 395–404. broadleaved evergreen tree species in a Mexican tropical Suchecky, R.K., Hubert, J.F., Birney de Wit, C.C., 1988. Isotopic lower montane forest. Global Ecology and Biogeography Let- imprint of climate and hydrogeochemistry on terrestrial strata ters 6, 115–127. of the Triassic–Jurassic Hartford and Fundy Rift Basins. Woodward, F.I., 1986. Ecophysiological studies on the shrub Journal of Sedimentary Petrology 58, 801–811. Vaccinium myrtillus L. taken from a wide altitudinal range. Tajika, E., 1998. Climate change during the last 150 million years: Oecologia 70, 580–586. reconstruction from a carbon cycle model. Earth and Planetary Woodward, F.I., 1987. Stomatal numbers are sensitive to in-

Science Letters 160, 695–707. creases in CO2 from pre-industrial levels. Nature 327, 617– Tans, P.P., Fung, I.Y., Takahasi, T., 1990. Observational con- 618.

straints on the global atmospheric CO2 budget. Science 247, Woodward, F.I., Bazzaz, F.A., 1988. The responses of stomatal 1431–1438. density to CO2 partial pressure. Journal of Experimental Ticha,´ I., 1982. Photosynthetic characteristics during ontogenesis Botany 39, 1771–1781. of leaves. 7. Stomata density and sizes. Photosynthetica 16, Woodward, F.I., Kelly, C.K., 1995. The influence of CO2 concen- 375–471. tration on stomatal density. New Phytologist 131, 311–327. Tognetti, R., Minnocci, A., Penuelas,˜ J., Raschi, A., Jones, M.B., Wright, V.P., Vanstone, S.D., 1991. Assessing the carbon dioxide 392 D.L. Royer et al.rEarth-Science ReÕiews 54() 2001 349–392

content of ancient atmospheres using palaeocalretes: theoreti- Yapp, C.J., Poths, H., 1992. Ancient atmospheric CO2 pressures cal and empirical constraints. Journal of the Geological Soci- inferred from natural goethites. Nature 355, 342–344. etyŽ. London 148, 945–947. Yapp, C.J., Poths, H., 1996. Carbon isotopes in continental weath-

Wright, V.P., Vanstone, S.D., 1992. Further comments on using ering environments and variations in ancient atmospheric CO2 carbon isotopes in palaeosols to estimate the CO2 content of pressure. Earth and Planetary Science Letters 137, 71–82. the paleo-atmosphere. Journal of the Geological Society Ž.London 149, 675–676.