After reading Francis Albarede’s comments on Samuel Epstein, I note that the authors’ paper does not complement the paleotemperature studies of Epstein and the last 50 years, but instead gives reason to dismiss them. If one accepts the authors’ conclusions, then decades of d18O-paleothermometry studies need to be reassessed.

The major problem lies in the authors' treatment of diagenetic alteration. The paper claims that diagenesis has caused 13C/12C and 18O/16O ratios to increase in their partially calcitised, Carboniferous samples (p.199). This assumption implies an inexplicably high d13C value for secondary calcite in sample B81-40 of almost 20 per mil, and an increase in d18O in the same sample of almost ten per mil, despite there being no known process that can achieve either increase in normal marine sediments.

This assumption of widespread diagenetic alteration is necessary because the authors need to dismiss almost all of their DELTA47 values as the products of alteration; otherwise they are left with an unmanageable spread of temperatures, many of which are unreasonably high. However, standard d18O-paleothermometry tells an entirely different story and reveals an acceptable temperature range for the same Carboniferous samples, while for the Silurian samples no significant range can be discerned. Moreover, d18O-paleotemperature estimates range in the Carboniferous samples in exactly the opposite direction from DELTA47 estimates! How can this be? If the authors are correct, not only has d18O paleotemperature information been obliterated in otherwise pristine carbonates, and even in aragonite-rich samples, but some as-yet unknown diagenetic process has caused 18O/16O to increase substantially, presumably as alteration temperatures increased (according to DELTA47 values).

Clearly, one of these paleothermometry methods must be wrong. The obvious susceptibility of DELTA47 values to alteration, shown by the extreme range in temperature estimates in this study weakens considerably the case for the new method, while d18O-paleothermometry of fossils has an excellent track record correctly recognising independently established cool intervals as far back as the Ordovician Period.

The most parsimonious explanation for this dataset, in my opinion, is that DELTA47 values in fossil carbonate simply do not record seawater temperature, but more study is required before we can be sure. The authors admit that the temperature and seawater d18O estimates in their study are only maximum values, so their conclusions regarding temperature-CO2 coupling based on two maximum temperature estimates over 500 million years still require additional support.

Jan Veizer
Department of Earth Sciences, Ottawa-Carleton Geoscience Center, University of Ottawa, Ottawa, Canada K1N 6N

Commentary on Came et al. “Coupling of surface temperatures and atmospheric CO2 concentrations during the Paleozoic era: Nature 449, 198-202 (2007)”

Came et al. 1 unveiled a new experimental method that has a potential to open the million years of “deep time” to paleoceanography in a way that earlier isotope techniques transformed our understanding of the Quaternary and Tertiary oceans. This nascent study proclaimed that “our results are consistent with the proposal that increased atmospheric carbon dioxide concentrations drive or amplify increased global temperatures”, leading to a widespread community perception of a unique and exclusive interpretation of the initial data. While CO2 is a greenhouse gas and temperature amplifier (only its relative importance at differing time scale is open to debate), the published nascent results do not preclude consistency with alternative scenarios.

As pointed out by Came et al. 1, the alteration trends for the two studied sample populations are at odds with the standard view of post-depositional resetting of isotope signals. In particular, the behavior of oxygen isotopes with elevated “alteration” temperatures, either showing no response (Silurian) or trending opposite to theory (Pennsylvanian), is puzzling. Moreover, the isotopic signature of the secondary calcite, presumably present as a minor component in the shells, would have to be rather exotic to account for the “alteration” trends. Presence of secondary calcite as a major component, apart from being easily detectable by our screening procedures, is even less viable a proposition, since, for example, the Pennsylvanian shells are all aragonites and thus well preserved by definition. Note also that a Tertiary set of samples 2 encounters almost identical problems. The alternative interpretation 3 that the observed spreads of 18O and 13C values in most, not just some, shells in Veizer et al. collection reflect their near-original natural variability cannot be therefore discounted, the more so that the dispersion of data in ancient assemblages is well within the dispersion ranges in modern counterparts 4.

Came et al. 1 also pointed out that the major reason for the discrepancy in the amplitude of temperature variations between the GEOCARB 5 and the 18O-based 6 approaches arises from the “ice volume” correction in the latter. In analogy to Quaternary, Veizer et al. 6 assumed that ice caps were twice the present-day size during the heights of icehouses and absent during greenhouse maxima, imposing almost 2 ‰ variability on the 18O of Phanerozoic ocean water. This alone accounts for almost the entire temperature discrepancy with the GEOCARB. The fact that their temperature trend is a running mean with a 50 Ma window also contributes to the diminution of the amplitudes and to broader peak shapes. In contrast to interpretation in Veizer et al. 6, the results in Came et al. 1 suggest little change in 18O of seawater between the warm Silurian (-1.2‰) and the cold Pennsylvanian (-1.6‰). This is at odds with the Quaternary analogy, and the latter also implies melting of almost two equivalents of present-day ice caps during the intense Late Paleozoic icehouse.

The above considerations demonstrate that we do not have yet an uncontested explanation of the vagaries of Phanerozoic climate or of the composition of its oceans. Momentarily, the “ground truth” geological data 7 argue for a four-fold greenhouse/icehouse climate pattern during the Phanerozoic, as do the 18O-based 6, 8 reconstructions, both consistent more with the alternative scenarios (e.g. celestial 8, 9) than with the two-fold GEOCARB-type 5 causality. Nevertheless, considering all the uncertainties involved, we need new tools to advance our knowledge beyond its present limitations. Fortunately, we may now possess the tool 1, the background database and the samples 3 to tackle these issues of considerable concern to society.

(1) Came, R.E. et al. Coupling of surface temperatures and atmospheric CO2 concentrations during the Paleozoic era. Nature 449, 198-202 (2007).
(2) Snell, K.E., Eiler, J.M., Dettmann, D. & Koch, P.L. Continental temperatures from the Paleocene-Eocene boundary in the Big Horn Basin, WY from carbonate clumped isotope thermometry. Goldschmidt Conf. Abstr., A950 (2007).
(3) Veizer, J. et al. 87Sr/86Sr, 13C and 18O evolution of Phanerozoic seawater. Chem. Geol. 161, 59-88 (1999).
(4) Brand, U., Logan, A., Hiller, N. & Richardson, J. Geochemistry of modern brachiopods: Applications and implications for oceanography and paleoceanography. Chem. Geol. 198, 305-334 (2003).
(5) Berner, R.A. & Kothavala, Z. GEOCARBIII: A revised model of atmospheric CO2 over Phanerozoic time. Am. J. Sci. 301, 182-204 (2001).
(6) Veizer, J., Godderis, Y. & Francois. L.M. Evidence for decoupling of atmospheric CO2 and global climate during the Phanerozoic eon. Nature 408, 698-701 (2000).
(7) Boucot, A.J., Xu. C. & Scotese, C.R. Phanerozoic climatic zones and paleogeography with a consideration of atmospheric CO2 levels. Paleont. J. 38, 3-11 (2004); www.scotese.com/climate.htm.
(8) Scherer, K. et al. Interstellar-terrestrial relations: the variable cosmic environments, the dynamic heliosphere, and their imprints on terrestrial archives. Space Sci. Rev. 127, 327-465 (2006).
(9) Shaviv, N.J. & Veizer, J. Celestial driver of Phanerozoic climate? GSA Today, 13, 4-10 (2003).

I write to respond to comments by Shields and Veizer regarding the role of post-depositional alteration in the d13C, d18O and ∆47 compositions of samples examined by Came et al. (other issues raised by Veizer will be addressed in a separate response). Three issues of this kind are raised:

(1) Shields suggests that the depositional record of ∆47 values in carbonates are exceptionally sensitive to post-depositional modification because a large proportion (roughly half) of the samples examined by Came et al. yielded temperatures in excess of earth-surface conditions. This aspect of Came et al.'s data set is explained by the fact that the study was designed to examine both samples that pass the most stringent criteria for preservation and samples that were known prior to analysis to have evidence of post-depositional modification (such as trace-metal enrichment, XRD evidence for replacement of aragonite by calcite, and microscopic evidence of recrystallization). Our purpose was both to quantify the effects of post-depositional modification (so its magnitude and correlated sample properties would be explicitly known in each suite) and to test Veizer et al.'s criteria for identifying unmodified samples. For this reason, it is unsurprising that half our samples were partially or completely 'reset': They were chosen to be so. This purpose was clearly explained in the text and table, and should have been obvious to anyone who reads the paper in its entirety.
We found Veizer's criteria to be mostly (though not universally) predictive of samples that have temperatures within the range of earth surface conditions, and thus these data support both Veizer's longstanding contention that well-preserved samples can be identified using evidence other than stable isotope data, and the ability of ∆47 thermometry to recover plausible depositional temperatures in the best-preserved samples.

(2) Both Shields and Veizer point out that the directions of d18O changes and relative magnitudes of d13C and d18O changes we associate with post-depositional alteration are contrary to common expectations. A general rule of thumb in this field holds that post-depositional modification of carbonates should involve relatively dramatic decreases in d18O, accompanied by smaller or even negligible changes in d13C, as would occur by reaction with low-d18O waters at elevated temperatures. In contrast, Came et al. suggest that post-depositional modification of Carboniferous samples increased both d18O and d13C, and alteration of Silurian samples decreased d13C and subtly increased d18O.
First, the trends of changing d13C and d18O and their relationships to independent indices of post-depositional modification are not based on ∆47 data. The d13C and d18O data plotted in Figure 1 of Came et al. are closely similar to previous measurements of these samples in other labs, and the designations of altered vs. pristine samples are based on trace metal, XRD and textural analysis all made before the ∆47 measurements. It is simply true (if somewhat enigmatic) that changes in composition and texture in these samples during burial are well correlated with isotopic changes that run contrary to the conventional rule of thumb. The fact that altered samples with unexpected d13C and d18O values also have elevated apparent temperatures by ∆47 thermometry is simply extra, complementary information.
Second, the 'rule of thumb' on which Shields' and Veizer's comments are based is clearly an over simplification. Formation waters in limestone sequences typically have d18O values in excess of 0 per mil, and only must drive decreases in d18O of carbonate when fluid-rock reaction occurs at very high integrated fluid/rock ratios and involves no large temperature gradients along the fluid flow path. It is understandable that our attention has been drawn to limestone sequences that show reductions in d18O during burial: The large changes in d18O associated with vigorous fluid flow are noticeable and straightforward to interpret. However, more subtle isotopic effects that occur at lower fluid/rock ratios are largely unexplored and could lead to increases in d18O under many plausible circumstances. For example, in systems that are not fluid-buffered and where fluid flows down a temperature gradient, 18O will be effectively removed from high temperature rocks and 'deposited' in lower temperature rocks. The magnitude of such effects will vary with the integrated fluid/rock ratio and the sharpness of the temperature gradient. Similar arguments could be made for C isotopes.
My point here is not that a specific interpretation regarding the fluid-flow history of the studied sections can be based on the few measurements reported in Came et al. Rather, it is that that ruberic we have relied on to decide such issues is overly simple and too strongly guided by the extreme cases of fluid-dominated systems.

(3) Shields and Veizer contend that the trends of isotopic change associated with post-depositional alteration in the samples examined by Came et al. require end-members with unacceptably extreme isotopic compositions.
First, as with comment (2), above, we note that this problem, if it is one, is implied by the conventional isotopic data and previously-used alteration indicators, and so cannot simply be explained through some sort of artifact or preservation problem associated with ∆47 data.
Second, both commentors fail to explain exactly how they have calculated the isotopic composition of the putative secondary calcite end member. This makes it difficult to respond in detail. It seems likely that they have assumed that isotopic trends in these suites reflect two-component mixing, but how did they decide on the composition of the secondary component and/or mixing proportions of the two components? Clearly, some extra constraint has been assumed or applied to solve the set of linear equations involved in such mixing models. Shields and Veizer must specify how this was done if their comments are to be understood.
Finally, I reject the implicit premise of these comments, that isotopic trends associated with post-depositional alteration should be understood as two-component mixing problems. What of isotopic exchange? And, by what criteria should we identify secondary components? There are accepted industry standards (e.g., textural analysis of SEM or optical images) really sufficient to recognize carbonate that has been modified in its isotopic composition?
It is true that the data in Came et al., particularly for Carboniferous samples, suggest that post-depositional processes have either modified large fractions of some of these samples (by re-crystallization and/or exchange) or have created domains with extreme compositions. Veizer's and Sheilds' comments seem to permit only the latter possibility. On what grounds? What independent and clear criteria can be called on to identify exchanged or recrystallized domains of the more altered samples? And, most importantly, how do Shields and Veizer propose to come to grips with the conventional evidence (d18O, d13C, trace metals, etc.)? It is insufficient to pass these off as fortuitous trends, as Veizer suggests.

Francis Albarede is correct in drawing analogies between the Came et al. paper and the seminal work of Sam Epstein and other members of the original Chicago group. All the early workers wrote about two major impediments to the application of oxygen isotope paleothermometry, that of the possible loss of the isotope signal during burial, diagenesis and alteration of the fossil and the uncertainty of the delta value of ancient seawater itself. The Came et. al paper resolves the previously contentious question of the delta value of seawater in the Paleozoic, namely Came et al. found that Silurian seawater had a delta 18-O value similar to today’s value of 0 but not the -8 per mille suggested previously by some of the coauthors of the Came et al. paper and others.

Came et alia’s conclusion about the delta 18-O of the ocean, and indirectly the veracity of the new “clumped carbonate method” can be tested by comparing their results to those derived from silicates, namely Paleozoic ophiolites and ore deposits that have been altered by seawater. Indeed we (1) have inferred from a detailed oxygen isotope study of an Ordovician ophiolite, both from its reconstructed isotope mass balance and from separated, coexisting quartz/epidote mineral pairs, that Paleozoic seawater and black smoker fluids were near zero, just as Came et al. concluded from the Silurian carbonates. The stages of diagenesis and alteration of carbonates may need more analyses and discussion, but the debate re Paleozoic ocean delta 18-O values has been closed by Came et al.

(1) Muehlenbachs, K., Furnes, H., Fonneland, H.C. and Hellevang, B. (2003). Ophiolites as faithful records of the oxygen isotope ratio of ancient seawater: The Solund-Stavfjord Ophiolite Complex as a Late Ordovician example. In Ophiolites in Earth History, Editors Y Dilek and P.T. Robinson, Geological Society, London, Special Publication 218, 401-414.

It is nice to see Sam Epstein’s contributions mentioned again. Although not referenced by Came et al. (2007), it is worth noting that he was co-author of a paper about 30 years ago (Knauth and Epstein, 1976) that gave combined H and O isotopic data for early diagenetic cherts indicating Silurian climatic temperatures of 35 plus or minus about 5 degrees. This is essentially indistinguishable from the result presented in the Came et al. (2007) paper. Also, a paper published as part of the GCA Epstein Volume (Knauth and Roberts, 1991) showed that marine evaporite fluids preserved as fluid inclusions in Silurian, Devonian, and Permian halite all reach a maximum evaporatively enriched delta 18O of approximately +6 per mill, the value expected in coastal evaporite environments (Lloyd, 1966) where the oceans have delta 18O near zero. The fluid inclusion data showed that the oceans could not have been 5 per lower in the Silurian, as had been argued by some from O isotope data in carbonates.

Came, R.E. et al. Coupling of surface temperatures and atmospheric CO2 concentrations during the Paleozoic era. Nature 449, 198-202 (2007)

Knauth, L.P. and S. Epstein, 1976, Oxygen and Hydrogen isotope ratios in nodular and bedded cherts: Geochimica et Cosmochimica Acta 140, p. 1095-1108.

Knauth, L.P. and S.K. Roberts, 1991, The Hydrogen and Oxygen Isotopic history of the Silurian-Permian hydrosphere as determined by direct measurement of Fossil Water. Geochemical Society Special Publication No. 3, p. 91-104.

Lloyd, R.M. 1966, Oxygen isotope enrichment of sea water by evaporation. Geochimica et Cosmochmica Acta, 30, 801-814.

Thanks to Drs Veizer, Eiler, Muehlenbachs and Knauth for their rapid replies. I will attempt to answer the two most important specific points here.

1. Eiler writes:

"And, most importantly, how do Shields and Veizer propose to come to grips with the conventional evidence (d18O, d13C, trace metals, etc.)? It is insufficient to pass these off as fortuitous trends, as Veizer suggests."

This final and "most important" statement reiterates the point that the d18O and d13C values are altered according to conventional evidence. However, no evidence is presented to suggest that the stable isotope values from this study are in any way altered. Indeed, the reported isotope values seem to reflect established trends in d13C and d18O, suggesting that they are not altered. No trace metal evidence is given for the Silurian samples, while trace metal data for the Carboniferous samples show alteration related to conversion to calcite (as marine aragonite would seldom contain such high concentrations of Fe and Mn). If the proposed stable isotope shifts are due to alteration to calcite, then the geochemical composition of the secondary calcite can be calculated from XRD data (suppl. table 2). In the most altered sample, only 31% has reverted to calcite, enough to push Fe conc. up to 500ppm. This means that the calcite portion of this sample must have an Fe conc. of about 1500ppm or so, fairly typical for secondary calcite. Because the authors argue that d18O has also been raised in this sample, one can calculate the d18O value of this secondary component. Clearly, from the authors' own arguments, both d13C and d18O needs to be much higher in this calcite component - something which can merely be tested. Alternatively, if both the aragonite and calcite in these samples have the same isotope values then this would indicate that the aragonite itself has undergone isotopic alteration. Not only would this conclusion be extraordinary, but it would imply that the trace metal evidence for alteration is anyway irrelevant.

2. The later comments of Muehlenbachs and Knauth are dealt with elsewhere in the literature and I do not need to repeat them here. However, one comment of Muehlenbachs requires immediate correction:

"The Came et. al paper resolves the previously contentious question of the delta value of seawater in the Paleozoic, namely Came et al. found that Silurian seawater had a delta 18-O value similar to today’s value of 0 but not the -8 per mille suggested previously by some of the coauthors of the Came et al. paper and others.......The stages of diagenesis and alteration of carbonates may need more analyses and discussion, but the debate re Paleozoic ocean delta 18-O values has been closed by Came et al."

By the authors' own calculations, Carboniferous seawater d18O was lower than -3 per mil during a period of ice build-up possibly equivalent to today. This minimum deviation from the modern value already represents an extraordinary depletion with respect to the modern ocean.