Interactive comment on “ Reconstruction of secular variation in seawater sulfate concentrations ” by T

. Long-term secular variation in seawater sulfate concentrations ([SO 2 − 4 ] SW ) is of interest owing to its relationship to the oxygenation history of Earth’s surface environment. In this study, we develop two complementary approaches for quantiﬁcation of sulfate concentrations in ancient seawater and test their application to late Neoproterozoic (635 Ma) to Recent marine units. The “rate method” is based on two measurable parameters of paleomarine systems: (1) the S-isotope fractionation associated with microbial sulfate reduction (MSR), as proxied by (cid:49) 34 S CAS-PY , and (2) the maximum rate of change in seawater sulfate, as proxied by ∂δ 34 S CAS / ∂t (max). The “MSR-trend method” is based on the empirical relationship of (cid:49) 34 S CAS-PY to aqueous sulfate concentrations in 81 modern depositional systems. For a given paleomarine system, the rate method yields an estimate of maximum possible [SO 2 − 4 ] SW (although results are dependent on assumptions regarding the pyrite burial ﬂux, F PY ), and the MSR-trend method yields an estimate of mean [SO 2 − 4 ] SW . An analysis of seawater sulfate concentrations since 635 Ma suggests that [SO 2 − 4 ] SW was low during the late Neoproterozoic ( < 5 mM), rose sharply across the Ediacaran–Cambrian boundary ( ∼ 5– 10 mM), and rose again during the Permian ( ∼ 10–30 mM) to levels that have varied only slightly since 250 Ma. How-ever, much short Early and as a consequence of anoxia, intense and The procedures in this offer potential for fu-ture high-resolution quantitative analyses of paleo-seawater sulfate


General Comments
Rate method.The application of modern S fluxes and associated d34S values to ancient systems is likely an over-extension and probably produces some of the uncertainty (and some of the unrealistic values) in reconstructed sulfate concentrations.Whereas there are ways to get at output d34S (through d34Spyr, for example), it is quite difficult to accurately predict the source d34S.Indeed, previous authors infer that the sulfur isotope composition of the source flux has differed from modern values quite significantly (e.g., Fike and Grotzinger, 2008).To a first order, it is hard to envision the source d34S value as invariant over long timescales.Changes in the fractional burial of S as pyrite and sulfate minerals through time (thought to drive much of the marine sulfate d34S variability) almost requires a change in the source as rocks of differing ages are later weathered on land in different proportions.Ultimately, it would be useful if the authors included model sensitivity analyses to changing source d34S.
Response: We agree that source flux d34S has probably varied through time, and that such variation may have influenced the d34S of seawater sulfate.We also agree that sensitivity analysis might be applied to test the potential influence of the source flux on seawater sulfate d34S.However, this is beyond the scope of the present study.Our rate method (Equations 2-4) does not depend on source flux d34S, so there is no need to engage in this exercise.
MSR method.The linear relationship between D34Ssulfate-sulfide from modern aqueous systems is striking and suggests that there is hope in reconstructing ancient sea-C8305 water sulfate concentrations with this approach.It would be useful if the authors distinguished which data points in Fig. 2 are derived from water column S phases, pore water S phases, solid S phase, etc.It seems somewhat coincidental that aqueous sulfate concentrations near the modern seawater sulfate concentration happen to yield the maximum d34S, above which fractionations are essentially constant.Might the hypersaline environments explored be unrepresentative due to high ionic strength or some other dissolved constituent that limits isotopic discrimination?In other words, can we be certain based on the current data set that seawater with higher sulfate contents (>29 mM) would not exhibit higher fractionations?Response: In Figure 2, all sulfate d34S values used in calculation of D34Ssulfatesulfide are based on measurements of aqueous sulfate, as stated in the text.For sulfide d34S, we used four different sulfur phases: pyrite, sediment acid-volatile sulfur (AVS), sediment total reduced sulfur (TRS), and aqueous H2S (note: this information has been added to the sulfide d34S column of Table A1).At the reviewer's request, we have constructed a version of Figure 2 that shows the different sulfide phases, and we calculated separate regressions for each phase (Fig. B3â ȂŤ see Supplementary document).The following points should be noted about this figure.First, each of the four phases yields a statistically significant regression (r = 0.81-0.92;p(a) <0.05; see Table B1 below).Second, the four phases have similar regression slopes although slightly variable y-intercepts.For this reason, TRS and AVS yield D34SCAS-PY values that are, on average, slightly larger for a given [SO42-]SW value than pyrite and aqueous H2S.Third, the four regression lines generally converge at higher [SO42-]SW, and the largest differences occur at low [SO42-]SW, where data is sparser.Whether there are real differences in the regression relationships among these four sulfide phases is an issue that will require further inquiryâ ȂŤ the regression lines in Figure B3 are not statistically different.One could argue in favor of using the pyrite d34S data alone, which would result in a small change in the regression relationship used to calculate paleoseawater [SO42-]SW values.We opted to use a larger sulfide d34S dataset, especially one containing more data at low [SO42-]SW, in order to generate a stable C8306 relationship over a wider range of [SO42-]SW values.The second part of the reviewer's comment concerns the reasons why the hypersaline environments in our dataset (Table A1) do not conform to the 'MSR trend', i.e., the regression relationship for environments with salinities of <40 psu (= practical salinity units) (Fig. 2).Whether MSR fractionations reach a maximum at the salinity of modern seawater (35 psu) and then remain essentially unchanged at higher salinities is uncertain.Our dataset certainly suggests that this might be the case, but the number of examples of hypersaline environments (n = 6) is too small to reach firm conclusions.Because we are not even certain that the MSR fractionation trend changes above 35 psu, it would not be useful to speculate on what factors might make this small set of hypersaline environments "unrepresentative".We simply raise the possibility of a change in the MSR fractionation trend at salinities >40 psu with the intention of encouraging further research into this issue.
that foraminifera CAS records agree with the Neogene barite record, but they analyzed pelagic planktonic foraminifera more closely associated with open ocean environments and not margin platforms.Lyons et al. (2004) show that very recent carbonate platform muds conform to the modern marine d34Ssulfate record, but these do not extend very far back in time.The authors do a good job critically choosing specific sulfur phases (e.g., shallow pyrite) to construct the MSR method equations.Whereas, modern environments provide the opportunity to be picky, ancient environments can only be probed through rock-bound proxies.Pyrite records are particularly sensitive in this regard, how can we be confident that the rock-bound pyrite is in fact shallow and therefore that D34S(CAS-pyr) accurately reflects cogenetic D34Ssulfate-sulfide?
Response: First, fractionation of S isotopes during precipitation of sulfate evaporites and incorporation of CAS in carbonates has been shown to be small (<1‰ (Schidlowski et al., 1977;Burdett et al., 1989;Kampschulte et al., 2001).The Phanerozoic records of CAS d34S and evaporite d34S were compared by Kampschulte and Strauss (2004), who found considerable overlap and no systematic bias toward higher values in one or the other dataset.
Second, we agree that the type of pyrite present in ancient sediments needs to be evaluated in order to assess whether it is syngenetic/early diagenetic and, thus, useful for calculating paleoseawater sulfate concentrations.There are well-established petrographic and geochemical techniques for this type of evaluation (e.g., Wilkin et al., 1996;Lyons and Severmann, 2006).This is an issue that each researcher making use of the methods developed in this study for estimation of paleoseawater sulfate concentrations will need to consider in regard to his/her specific study units.
Action: We have added a brief synthesis of these points to the manuscript.
Heterogenous marine d34S records.Unfortunately, d34S records of most time intervals have only been developed from one or two locations.The multiple records from the Neoproterozoic indicate both lateral (horizontal; Loyd et al., 2012;2013) and strat-C8308 ified type (vertical; Li et al., 2010) variability probably stemming from overall low, but likewise variable, marine sulfate concentrations (as the authors mention, P13209-10; lns 34-30, 1-7).Similar heterogeneity may occur during other time intervals as well.In the face of potentially large heterogeneity, how reflective is a single succession of the global ocean?Furthermore, how can we be confident that intervals with data from only one or two successions can be used to accurately constrain a global signal?
Response: We agree that spatial heterogeneity in seawater sulfate concentrations may become pronounced at low average concentrations, as during the Neoproterozoic.This does not invalidate an estimate of seawater sulfate concentrations for a particular time and locale.It does mean that a single estimate will not suffice to characterize seawater globally, and that a number of estimates from widely separated locales would be desirable to characterize the range of variation in seawater sulfate concentrations at a given time.These considerations in no way invalidate our methodology for estimating seawater sulfate concentrations.
Action: We have added a brief synthesis of these points to the manuscript.

Specific Comments
P 13191, lns 5-7: It seems difficult to rationalize such a broad statement.Local source d34S values and fluxes will be particularly influential, especially if low oceanic [SO42-] lends to short residence times.
Response: Whether such a statement is overly broad or not depends on one's outlookâ ȂŤ there is no inherently correct view on such a matter.We agree with the reviewer that local variations in sulfate concentration and isotopic composition will become more pronounced at low average concentrations.The significance of this point was considered in the preceding response.
Response: This relationship is valid for mixed fluids that contain a seawater component >5% (where the second fluid is low-sulfate freshwater).It would not be valid for a purely terrestrially sourced fluid.Sulfate concentrations for all freshwater systems in our dataset (Table A1, records 1-18) were measured, not calculated from salinity.Of the 36 brackish systems in our dataset (Table A1, records 19-54), we estimated sulfate concentrations for 9 of them from salinity data.By definition, our brackish systems had total salinities of 10 to 30 psu and thus consisted of 28-86% seawater.The calculated sulfate concentrations are therefore reliableâ ȂŤ there are no problems with the sulfate concentrations in our dataset (Table A1).
P 13196, lns 3-5: The Habicht et al. (2002) data show a clear step function, not a linear relationship as seen in the natural samples.
Response: The statement in question is: "Our results are similar to, although more linear and more statistically robust than, those reported by Habicht et al. (2002) on the basis of culture experiments."Our results are similar to those of Habicht et al. in terms of the broad relationship between MSR fractionation and aqueous sulfate concentration, although more linear (as noted by the reviewer).We stand by our statement.
General Note: It would be nice to see how water column sulfide compares to shallow pyrite in modern systems where both are measureable or have been measured.This