The role of CO2 in regulating cli- CO as a primary driver of mate over Phanerozoic timescales has 2 recently been questioned using δ18O records of shallow marine carbonate Phanerozoic climate (Veizer et al., 2000) and modeled pat- terns of cosmic ray fluxes (Shaviv and Dana L. Royer, Department of Geosciences and Institutes of the Environment, Veizer, 2003). The low-latitude δ18O Pennsylvania State University, University Park, Pennsylvania 16802, USA, compilation (Veizer et al., 1999, 2000), [email protected] taken to reflect surface water tempera- Robert A. Berner, Department of Geology and Geophysics, Yale University, New tures, is decoupled from the CO2 record Haven, Connecticut 06520, USA and instead more closely correlates with the cosmic ray flux data. If correct, Isabel P. Montañez, Department of Geology, University of California, Davis, cosmic rays, ostensibly acting through California 95616, USA variations in cloud albedo, may be Neil J. Tabor, Department of Geological Sciences, Southern Methodist University, more important than CO2 in regulating Dallas, Texas 75275, USA Phanerozoic climate. Here we scrutinize the pre-Quaternary David J. Beerling, Department of Animal and Plant Sciences, University of Sheffield, records of CO , temperature, and cos- Sheffield S10 2TN, UK 2 mic ray flux in an attempt to resolve current discrepancies. We first compare proxy reconstructions and model pre- ABSTRACT INTRODUCTION dictions of CO2 to gauge how securely Recent studies have purported to Atmospheric CO2 is an important we understand the major patterns of show a closer correspondence between greenhouse gas, and because of its short Phanerozoic CO2. Using this record of reconstructed Phanerozoic records of residence time (~4 yr) and numerous CO2 and Ca concentrations in cosmic ray flux and temperature than sources and sinks, it has the potential Phanerozoic seawater, we then modify between CO2 and temperature. The role to regulate climate over a vast range the δ18O record of Veizer et al. (1999, of the greenhouse gas CO2 in control- of timescales, from years to millions of 2000) to account for the effects of sea- ling global temperatures has therefore years. For example, the 30% rise in at- water pH. This modified δ18O record is been questioned. Here we review the mospheric CO2 concentrations over the then compared to records of continental geologic records of CO2 and glacia- past 100 years has been accompanied glaciations and cosmic ray fluxes. tions and find that CO2 was low (<500 by significant global warming (Mann et ppm) during periods of long-lived and al., 1999, 2003). Most studies incorporat- COMPARISON OF PROXY CO2 widespread continental glaciations and ing all known climate forcings implicate RECORDS TO GEOCARB MODEL high (>1000 ppm) during other, warmer CO2 as the primary driver for this most RESULTS periods. The CO2 record is likely robust recent rise in global temperatures (Mann Multiple geochemical models of atmo- because independent proxy records et al., 1998; Crowley, 2000; Mitchell spheric CO2 evolution have been devel- are highly correlated with CO2 predic- et al., 2001). At the longer timescale oped in recent years; the most complete tions from geochemical models. The of glacial-interglacial cycles (105 yr), a models track the exchange of carbon Phanerozoic sea surface temperature tight correlation between CO2 and polar between buried organic and inorganic record as inferred from shallow marine temperatures has long been established sedimentary carbon and the atmosphere carbonate δ18O values has been used (Barnola et al., 1987; Petit et al., 1999). plus oceans (Berner, 1991; Tajika, 1998; to quantitatively test the importance of Although debated for many years, it Berner and Kothavala, 2001; Wallmann, potential climate forcings, but it fails is clear that CO2 acted as either a cli- 2001; Kashiwagi and Shikazono, 2003). several first-order tests relative to more mate driver or an important amplifier The CO2 predictions from these models well-established paleoclimatic indi- (Shackleton, 2000). For pre-Quaternary are highly convergent; for the purposes cators: both the early Paleozoic and climates, ice core records do not ex- of this study, we will use GEOCARB III Mesozoic are calculated to have been ist, but a multitude of CO2 proxies and (Berner and Kothavala, 2001), which too cold for too long. We explore the models have been developed. As with predicts CO2 over the whole Phanerozoic possible influence of seawater pH on the Recent (101 yr) and Quaternary (105 in 10 m.y. time-steps. The GEOCARB the δ18O record and find that a pH-cor- yr) records, a close correspondence model (Berner, 2004) is based on quan- rected record matches the glacial record between CO2 and temperature has gen- tifying over time the uptake of CO2 dur- much better. Periodic fluctuations in the erally been found for the Phanerozoic ing weathering of Ca and Mg silicates cosmic ray flux may be of some climatic (e.g., Crowley and Berner, 2001). Taken and its release during the weathering of significance, but are likely of second- together, CO2 appears to be an impor- sedimentary organic matter. Also consid- order importance on a multimillion- tant driver of climate at all timescales. ered is the burial of carbonates and or- year timescale. ganic matter in sediments and the fluxes ————— GSA Today; v. 14; no. 3, doi: 10.1130/1052-5173(2004)014<4:CAAPDO>2.0.CO;2. 4 MARCH 2004, GSA TODAY ���� A of CO2 to the atmosphere and oceans from the thermal de- ��������� ��������� composition of carbonates and organic matter at depth. ������������� ��������� Weathering fluxes are modified over time as changes occur in ����������������� ������ ���� global temperature, continental size, position and relief, and � �������������� land plant colonization. This includes incorporating solar radi- ����������� ation, due to the slow stellar evolution of the sun, and the ���� CO2 greenhouse effect in general circulation model (GCM) calculations of global mean surface temperature and river run- off. Volcanic degassing is guided by the abundance of volca- nics, seafloor spreading rates, and the carbonate content of ���� subducting oceanic crust. �������������� The paleo-CO2 results of Rothman (2002) and U. Berner and Streif (2001), presented by Shaviv and Veizer (2003) as � additional models, are in fact not based on carbon cycle B modeling, but constitute an extension of the δ13C plankton 13 ���� CO2 proxy (see below). These authors apply Δ C, the differ- ������ ence between the δ13C of bulk organic matter and carbonates � (Hayes et al., 1999), to directly calculate paleo-CO2. However, bulk organic matter can include non-photosynthetic com- ���� pounds as well as terrestrial material derived from rivers, and the original method was based strictly on marine photosyn- thetic compounds (Freeman and Hayes, 1992; cf. Royer et al., ���� 2001a). In addition, changes in Δ13C over time can be due �������������� to changes in seawater temperature (Rau et al., 1989) or O2 concentrations (Beerling et al., 2002), and not only atmo- � spheric CO2. �� C Four proxies for pre-Quaternary CO2 levels have been de- veloped over the past 15 years (consult Royer et al. [2001a] for further details). 1. The δ13C of pedogenic minerals (calcium carbonate �� [Cerling, 1991] or goethite [Yapp and Poths, 1992]). The car- bonate in certain pedogenic minerals is formed from biologi- cally and atmospherically derived soil CO2. Because these two components differ in their carbon isotopic compositions, ������������������ � ��� ��� ��� ��� ��� ��� � the concentration of atmospheric CO2 (pCO2) can be esti- mated assuming some knowledge of the biologically derived ��������� 13 pCO2 in the soil, and the δ C of the atmospheric and biologic Figure 1. Details of CO2 proxy data set used in this study. A: constituents. Reliable pedogenic minerals are available back Five-point running averages of individual proxies (see footnote to the Devonian, and the range of errors is comparably small 1). Range in error of GEOCARB III model also shown for at high CO2. Some disadvantages of this proxy include com- comparison. B: Combined atmospheric CO2 concentration record parably high errors at low CO2, and the difficulty of as determined from multiple proxies in (A). Black curve represents extracting organic carbon from the paleosols that contain average values in 10 m.y. time-steps. Gray boxes are ��������standard these minerals. deviations (± 1σ) for each time-step. C: Frequency distribution������������ 2. The δ13C of phytoplankton (Freeman and Hayes, 1992; of CO2 data set, expressed in 10 m.y. time-steps. All data are calibrated to the timescale of Harland et al. (1990). Pagani et al., 1999). Most phytoplankton exert little or no ac- tive control on the CO2 entering their cells. Because of this, 13 the Δ C between seawater CO2 and phytoplankton photosyn- tracted using fossil plants. High resolution, high precision thate is affected by seawater pCO2 and can thus be used as a CO2 records are possible, but because the stomatal response CO2 proxy. High resolution CO2 records are obtainable from to CO2 is species-specific, care must be exercised in pre- appropriate marine sediment cores, but factors such as cell Cretaceous material. geometry and growth rate, which also influence the δ13C of 4. The δ11B of planktonic foraminifera (Pearson and Palmer, phytoplankton, must be carefully considered. 2000). The relative proportions of the two dominant species 3. The stomatal distribution in the leaves of C3 plants (Van of boron