Thank you for your review. Our replies to the individual comments are listed below in bold font.

In their study “Phylochemistry: exploring evolutionary constraints on belemnite rostrum element composition” Alexander Pohle and co-authors propose the use of chemical parameter to assess evolutionary links between extinct fossils. The main message is that of the four studied element/Ca ratios, Mg/Ca appears most variable, but also Sr/Ca shows some distinctions amongst taxa that are of potential taxonomic value. The authors conclude that there may be value in utilising fossil chemistry for the study of evolution in extinct taxa.

The chemical composition of macrofossils has received comparatively scant attention beyond the use for identifying the preservation state of materials taken for geochemical studies. For belemnites, there have been attempts to link element/Ca ratios to palaeotemperature, but also to extract from such data information about changes in ocean chemistry and details on biomineral formation. A dedicated assessment of the taxonomic and phylogenetic value of geochemical indices, however, has so far not been done for rostra, even though the existence of species-specific chemical signatures has received some mention in various contributions.

I find the idea of “phylogeochemistry” quite exciting and think there is merit in pursuing this line of investigation further. If species-specific chemical signatures of belemnite rostra can be determined and linked to inform evolutionary relationships, there would be an added layer of robustness to the designation and identification of taxa, and an improved understanding of how such extinct taxa link together. This stated, I feel that there are a few points that may be worth considering in relation to the current contribution and how it may add to the establishment of such assessments.

Thank you for this detailed review and overall positive assessment of our study. Below, we reply to each of the raised points and suggest how we plan to account for them in the revised manuscript.

The nature of the data:

1) Given that no primary chemical data was obtained for this work, the available base of data is driven by the published literature and studies that were not designed to feed into phylogenetic analysis. However, it is I think important to keep in mind that the data are heavily skewed towards a small number of individuals (c. 1/7 of the dataset come from one specimen). For a number of the two dozen or so genera, therefore, information will be quite limited, and potentially skewed and biased. Can it be shown confidently that the raw data are suitable and sufficient to carry out the analysis both for taxonomic delineation and supposed evolutionary rates of change in element/Ca ratios?

We are fully aware of this and explicitly mentioned the various (intrinsic!) shortcomings of the dataset. This is detailed in the methods and discussion sections. Collecting new primary data would be, of course, optimal. However, this would represent an enormous effort, requiring sampling many different sites around the world, as we cover the entire belemnite clade from the Early Jurassic until the Late Cretaceous. This material then would need to be assessed taxonomically and in terms of diagenesis. We are aware of only one study so far with a comparable design as proposed by the reviewer. Three decades ago, Prof. Veizer here at Bochum received two million Euros in the context of the prestigious Leibnitz Preis, and even then, it took him and his students and collaborators eight years to compile a set of samples (albeit not belemnites, but mainly brachiopods) that reached publishable dimensions. That is clearly (!) beyond our scope and possibilities.

Our study instead serves as a first exploration (hence the title) of the applicability of these tools, thus maximising information input, at the sacrifice of potential accuracy and precision. Nevertheless, we think that we were very open about the potential pitfalls and discussed them extensively. No taxonomic assessments were made on the basis of geochemistry alone and we excluded considerable amounts of data if taxonomic affinity was unclear or other conditions were met (see methods). The conservative approach is documented in L193-194: only 2241 out of 2749 data points were used for our analyses, excluding around 20% of the data (the entire dataset is included in the supplement for convenience).

Prior studies where no taxonomic assessment had been carried out were not considered at all. Furthermore, note that evolutionary rates were independent of the tree topology used (Fig. 10), suggesting that the estimates are robust independently of phylogenetic arrangement of the taxa. We expect that only a substantial number of incorrectly reported element ratios would impact our results, which would imply that a significant number of previous studies were flawed. We also mention the possibility that evolutionary rates are not constant; however, accounting for this would add another level of complexity to the analyses. Obviously, better taxonomic sampling and more detailed taxonomic assessments in geochemical studies would further refine our analyses, and we hope that our study encourages further efforts in this direction.

2) It seems to me that the study only utilises two chemical indices – Mg/Ca and Sr/Ca, given that Mn/Ca and Fe/Ca yield little taxonomic value. I agree that the latter are not meaningful for the understanding of the biomineralisation process, but rather indicate post-depositional processes in belemnite rostra. However, are two geochemical indices enough to allow for an independent assessment of phylogenetic relationships? It feels to me that rather than being interpreted in isolation such chemical features should be used as additional characters alongside morphological and structural features.

It might seem that the reviewer missed an important methodological detail: no taxonomic assessment nor phylogenetic inference was performed based on geochemical data, neither in isolation, nor in combination with morphology. All trees were solely reconstructed from the updated morphological character matrix of Stevens et al. (2023), see also our Fig. 1. For the analyses of ancestral states and evolutionary rates, the resulting trees from the first step were held constant for each analysis. In other words, we “plotted” the values on the tree and calculated how quickly they change (i.e., evolutionary rates), accounting for uncertainties at the internal nodes. Thus, our approach is even more conservative than what is suggested by the reviewer. We will state this two-step approach more clearly in the revised manuscript.

3) The placement in time of the data seems to align with the first appearance of the taxon and there does not appear an allowance for the evolution of geochemical data within the taxon as a function of changes in environmental parameters, such as temperature (if it indeed has an effect) or ocean chemistry. A considerable body of research exists that attempts to utilise Mg/Ca and Sr/Ca in belemnites for temperature reconstruction (with varying conclusions), and some studies have made use of belemnite data to attempt to reconstruct seawater Sr/Ca ratios (see Ullmann et al., 2013, Geology). The authors implicitly dismiss the latter as an issue (L350-356) but remain vague about data backing up this interpretation.

We see your point, but this is, unfortunately a necessary limitation of our study since we only looked at the genus level (i.e., only a single value per genus). Recording within-genus changes would require sampling several congeneric species, but we have mentioned the sampling problem already. That said, we think that this comment is rooted perhaps in the above-mentioned misunderstanding of our methodology: We do not infer phylogeny based on the geochemistry, and we do not claim that differences between taxa are only due to vital/evolutionary effects. On the contrary, we discuss several potential reasons for the observed evolutionary geochemical pattern (L365-400) and mention that element ratios are the result of a complex interplay of different factors.

Even if the element data are fully dependent on the environment, they do not detract from the usefulness of our approach because this would allow us to link environmental conditions with the evolution of the clade directly. Ancestral states can be used to trace parameters across a phylogeny. A great example is phylogeography, which is an entire field studying, e.g., dispersal and migratory patterns of related species. Nobody would seriously consider a phylogeny inferred solely based on geography to be reliable. In the revised manuscript, we will highlight more clearly the potential applications of our approach.

4) The authors appear to misunderstand some of the more significant controls on element incorporation into the rostrum, noting a tendency to enrich Mg and Sr during early ontogeny (L383). There is strong evidence to suggest that enrichments of Mg and Sr close to the apical line are not related to ontogenetic trends. The defining feature rather than ontogeny is proximity to the apical line, and also growth increments formed late in ontogeny will show these enrichments where they are meeting the apical line (Ullmann et al., 2015). All other published datasets showing serial measurements from apical line to margin of the rostrum that I am aware of align with this statement, but do not allow to disentangle spatial controls from ontogenetic controls, because only one transect was analysed for a specimen, rather than multiple ones. Overall this point has consequences for the interpretation stated in L385, though: Ultimately, these Sr and Mg enrichments can be avoided by sampling calcite in intermediate growth bands away from apical line and margin of the rostrum, and this substantially reduces observed data variability within individuals (Ullmann and Pogge von Strandmann, 2017). This is something which is likely to continue to skew datasets assembled by researchers that are unaware of the spatial controls on Sr and Mg distribution in the rostrum.

Thank you for this comment. We will adjust our statements accordingly to include spatial effects. We, however, politely disagree with the reviewer that our statement has severe consequences for the future assembly of datasets; after all, this depends on the aims of the study. From an evolutionary perspective, it is necessary to sample homologous areas, thus ideally sampled at the same position in every rostrum. This is obviously not possible for our dataset but needs to be considered in potential future studies aimed at collecting primary data. Depending on the research goals, sampling the apical line and earlier ontogenetic stages might be necessary. We will include a sentence highlighting that the exact sampling position(s) needs careful consideration in the study design.

5) The authors state in L80 that data identified as compromised were excluded from consideration, but plots in figure 2 do not support this statement. For instance, according to published data (Sørensen et al., 2015), Sr/Ca ratios lower than 1.5 mmol/mol are not observed in well-preserved Bellemnellocamax and this is the only cited reference for this genus. However, data as low as 0.5 mmol/mol appear to be included in the analysis for this genus given the representation in panel B of Figure 2. Also, Fe/Ca ratios are not reported for Bellemnellocamax, but rather Mn/Ca ratios, so I assume this was erroneously attributed in this plot.

Thank you for pointing this out; the Fe/Ca and Mn/Ca ratios were indeed mixed up for Belemnellocamax, which we will correct in the revised manuscript. The reason that these data were not excluded is that Sørensen et al. (2015) considered them as poorly preserved, mostly based on the Mn/Ca ratio with a more or less arbitrarily set threshold. Excluding these data would be problematic for the comparability to the data from other publications, where different thresholds were used (see also discussion). Because well-preserved samples are by far the majority in Sørensen et al. (2015), we think that the effect is negligible. In the paper, we also discuss the possibility of bias through diagenetic alteration. Nevertheless, we will add a statement in the methods section that specifies our approach to the exclusion of poorly preserved data.

Accessibility and focus of the text

The text, where it relates to phylogenetic work, is overall quite rich in technical terminology and dense methodological detail that for a non-specialist in these techniques makes it relatively inaccessible. This stands in opposition to the description of geochemical data, which appears somewhat superficial. It is understandable that technical detail is required and terminology, where established, should not be compromised by oversimplification, but I think more could be done to write the text in a way that can be understood by non-specialist readers.

We are fully aware of this challenge and believe a paper must be stand-alone. There are limits, however. It is for this reason that we provided a table explaining the most important terms and concepts. The length and detail of the phylogenetic part are unfortunately necessary for reproducibility reasons and comply with the standards in the research field. We will attempt to make the text even more approachable, paying particular attention to the parts that caused, unfortunately, misunderstandings. Note, however, that a full introduction to the phylogenetic methodology is out of the scope of our paper. There are several excellent review papers out there for interested readers, and they have already been cited. After all, the principle of scientific writing is that the author must provide the fundamental information to a research field but should also refer to published work. That said, we would appreciate very specific comments on which portion of the text requires more clarity for the tangential reader. Ideally with line numbers.

In my opinion, part of this challenge also relates to understanding exactly what aspects of the chemical data led to the construction of the phylogenetic trees. From my point of view, a good starting point would be an attempt to use Figure 2, and plot ranges of Mg/Ca and Sr/Ca for taxa (amended by number of analyses per box plot) against time such as in Figure 7, but without any inferred phylogenetic links. Personally, I am not convinced how much added value for the analysed dataset derives from sophisticated probability-driven modelling. An appraisal of chemical data in the framework of existing phylogenetic schemes at this stage may be equally if not more powerful, until a broader and more reliable database (in terms of number of chemical proxies and number of analyses for all taxa) for belemnite rostra becomes available.

See the previous comments; we did not use any geochemical data to reconstruct our phylogenies. In fact, our approach is very similar to what is suggested by the reviewer: “an appraisal of chemical data in the framework of existing phylogenetic schemes”. The twist is that there is only a single “existing phylogenetic scheme” (Stevens et al., 2023), which is missing several taxa for which geochemical data is available. For this reason, we had first to update the latter phylogeny (with morphological data only!). Ancestral state estimation then accounts for uncertainties at the nodes and along branches, while evolutionary rates provide an estimate for the tempo of these changes. As already mentioned, we will ensure in the revised version that this misunderstanding will not happen.

Detailed comments:

Title: Is there a reason why one should call this line of study “phylogeochemistry” rather than “phylochemistry”?

“Phylochemistry” would be a lot broader and could, in principle, apply to any tissue (in principle, even to DNA data). Our approach specifically relates to the use of comparative phylogenetics to geochemical proxy data, which is why we think this is an appropriate name. Our approach is also of more interest for application in palaeontology because, for modern organisms, genetic data can be acquired, which will, of course, yield phylogenetically informative data far superior to the chemical composition of biominerals.

L1: It may be questionable whether biogenic carbonate hard parts are the most important geochemical archives. Certainly they are very important, but also organic geochemistry has yielded important insights regarding carbon cycle and climate dynamics, and significant work has been done also on detrital silicates and other substrates.

Thank you for pointing this out. We will modify this to “among the most important”.

L6: Might it be better to speak about element/Ca ratios in general here? Sr may be a minor or trace element in belemnite rostra and Mg is also not usually a major element in these materials depending on exact definitions.

We actually had a similar discussion among the authors during the writing phase of the manuscript, so your suggestion is important to us. We will change this throughout the manuscript.

L10: Personally I strongly doubt this inference of evolutionary rates and the reliability of this estimate.

See text for details. Note that there are uncertainty ranges associated with these estimates, which we will add to the abstract in our revised manuscript. We also would like to highlight the misunderstanding mentioned above: the phylogenetic inference is not based on the geochemical data but on morphological data only.

L44: I think this statement needs a bit more quantitative detail. What is meant by significant geochemical variation? How is it determined that it is significant? What kind of proxies and/or fossils are compared here?

Thank you. We will add more details in our revision.

L47: There seems to be a wrongly formatted reference here. From a chemical point of view quite a bit has been published on brachiopods in this relation and there is a reasonable amount of information compiled also for bivalves and barnacles – maybe it would be worth mentioning some of this research given the focus of the paper?

Thank you for pointing out the references that were wrongly formatted. We will add a couple of more references.

L52: I fully support this, but as stated above – why do geochemical data need to be seen in isolation from other characters valuable for taxonomic work? If one accepts the identification of the rostra taken for chemical work to be correct, why cannot morphological features such as grooves, striations, cross-sectional shape, placement of the apical line, bluntness of the apex etc. be used in conjunction with this data?

See replies above, only morphological data were used for the phylogenetic placement of taxa.

L62: See above for definitions of main and trace elements.

Thank you.

L63: Why must isotope data be the focus of upcoming work? These data could already now be included, but it might not be a surprise that objectively disentangling taxon-specific effects from environmental effects would be very challenging? This issue is for instance at the core of the debate about the utility of Mg/Ca in rostra as a thermometer.

The manuscript is already relatively long and isotope data would require a different discussion. The suggestion of the reviewer would be more suitable for an Earth Science Reviews paper. As noted by the reviewer, disentangling evolutionary and environmental effects would be even more challenging for isotope data, which is why we focused on element/Ca first, as they are expected to be more strongly biologically controlled.

L92: Equations 1-3 seem quite cumbersome to me and are also not correct in their current form in the case of equation 2. I assume the authors meant for the concentration of Sr to be in µg/mol rather than g/mol? A simpler way to get to this result would be to calculate Sr/Ca in µg/wt% and multiply by m(Ca)/(10*m(Sr)).

We will adapt the formulas and the text and hope this will be better comprehensible. Note that we double checked the formulas by also calculating known values, i.e., where both raw data and element ratios were given. Thus, our calculations are correct.

L116-87: As opposed to the assessment and debate about chemical data – which is comparatively simplistic and avoids dealing with some of the complexity of some of these data – the phylogenetic analysis is extremely densely written, with little explanation of the rationale of parametrisation and adopted approach. I am not sure why this bi-modal approach to complexity was adopted, but feel that there is scope to make the study much more accessible to the audience by laying out the rationale and consequences of the chosen methodology in some more detail.

Thank you. We will add further detail on the rationale of the chosen methodologies.

L124: I assume what is meant here are “100,000” generations rather than “100.000” generations? See also L179

Yes, we will change the format accordingly.

L166: What is the rationale to adopt the mean of these ratios as the most meaningful? My feeling is that the median will be more reliable.

This was an arbitrary decision. However, as the distributions of Mg/Ca and Sr/Ca ratios are relatively symmetric (see Fig. 2), we do not think that there would be a big difference between the two.

L258: In the figure caption, should it read “within a genus”?

Apologies. This is actually the incorrect caption, as it is identical to the caption of Fig. 3. We will add the correct caption in the revised version.

L310: I am not sure how these values are meaningful? As far as I can tell, the observed rates of change in long-term records are smaller than postulated (Ullmann et al., 2013, Geology). Over short periods of time, a well-documented case of a succession of taxa is at the Toarcian OAE, where Passaloteuthis is followed by Acrocoelites. Acrocoelites already at its first appearance has a high Mg/Ca ratio which may reduce somewhat towards the middle Toarcian (Ullmann et al., 2014, PNAS) as opposed to a hypothetical ramping up to a stable value that one may read off Figure 8.

Note that we use a “simple” model of Brownian Motion that assumes that the evolutionary rate is constant across the tree. This is probably not realistic in many cases, but allowing for variable rates (various versions of “relaxed Brownian Motion” or “Ornstein-Uhlenbeck” models, the latter accounting for evolutionary “optima”) would add further levels of complexity to the analyses, which we think is currently not warranted due to the biases in the primary data and the phylogenetic uncertainties. That said, constant rates are a useful approximation as potential rate variation is “averaged out”. Furthermore, the model provides explicit uncertainty intervals, which are shown in Figs. 9 and 10, and also mentioned in the text.

We noticed that we only included the 95% confidence intervals for σ², but not for σ, which we will add in the revised version. As we restricted our analysis to the genus level, the model can obviously not detect changes within a genus, and a considerable number of genera were not included. The important point is that our analyses show that evolutionary rates are independent of the uncertainty in the phylogeny (Fig. 10), suggesting that they are robust unless the primary data (i.e., the element ratios) are very heavily biased, which we do not think is the case. Our results should serve as rough first-order approximations, which can be refined in the future by dedicated studies with more uniform sampling and testing of further models. We will extend the discussion on this topic and the reliability of the estimates in our revised version.

L342: I am surprised about the way that this is presented. Most authors use Mn and Fe as the basis for their quality control, and it is well established that enrichments of Mn and Fe (qualitatively) link to diagenetic modification of calcite in rostra. The levels of Fe and Mn that are observed in the samples that have not been excluded from further interpretation are therefore expected to be erratic and too small to correlate strongly with other proxies than other redox-sensitive metals. However, they are still likely to be dominated by post-depositional enrichment rather than original incorporation, only that they otherwise signify (hopefully) negligible degrees of chemical overprint on other proxies.

We are unsure what the suggested change is here. We report on the variability of our data, noting that there is no strong taxon-specific effect for these ratios. The observed pattern that variability is higher within the same taxon than between taxa is exactly what would be expected if the values are dominated by post-depositional enrichment. We realised that the latter was missing from the following interpretation sentence, which we assume is what the reviewer is referring to. We will modify the statement accordingly. However, because the primary Fe/Ca and Mn/Ca ratios are unknown, biomineralization effects and kinetic factors cannot be excluded. This is well exemplified by modern calcifying organisms, which show variation in Mn and Fe concentrations within the same environmental setting, despite being unaffected by diagenesis. Since we did not use Mn or Fe data in our analyses, differences in interpretation do not affect our study.

L349: The authors are quite vague here in what “seawater properties” may be. Is this to mean that some physical properties of seawater influence the way in which belemnites would have incorporated elements into their rostra, or is this rather envisioned to be the chemical signature of the water itself? It would be useful to explain in some more detail the envisioned mechanisms of this, especially as both physical and chemical parameters of seawater have been proposed to have had an influence on chemical signatures of belemnite rostra.

This comment comes as a surprise as we specifically mention several examples of seawater properties. This essentially represents environmental effects, which would influence the concentration of elements in the body fluid of belemnites, in turn providing control on the element ratios in the rostrum (which is further altered by biogenic effects). The exact mechanism is not relevant to our study, and there are certainly better ways to study this directly. Nevertheless, we will incorporate a few more references for possible mechanisms that have been .

L351: I am not sure if I understand fully what the authors want to express here (what exactly is meant by non-systematic in terms of duration/spatial context), but it appears to me that this is the scenario that has been previously put forward regarding the Toarcian OAE.

Non-systematic here refers to shifts in element ratios that are not parallel, i.e., some lineages may experience an increase, others a decrease and yet others stasis within the same temporal or spatial context. Major ecological disturbances such as oceanic events could, of course, be an exception to this general pattern, which we will add in the revised version; thank you for this suggestion! Nevertheless, as seen for the Toarcian OAE, during a similar time interval, Mg/Ca decreases in the lineage leading toward Holcobelus and Belemnopsis, so the shift seen between Passaloteuthis and Acrocoelites does not necessarily affect all taxa (i.e., it is non-systematic).

L354: However, one finds in sedimentary successions without major changes in palaeoenvironment different taxa of belemnites through time with quite variable Mg/Ca and Sr/Ca?

Yes, see previous reply to previous comment. Taxon-specific shifts are not necessarily parallel, and they therefore may result in an increase in variability. This could be clarified through taxon-specific plots and diagenesis screening in the respective studies.

L376: Should “Hibolithes” here be in italics? I think this element of research is the most promising regarding the chemical signatures and should be explored in some more detail.

Yes, thank you for pointing this out; this was an oversight. We think that one should be very careful in using chemical signatures to identify taxa and a lot more work in this direction is required. To stimulate further research, we will suggest in the revised version that geochemical data (if available) should be included in future taxonomical studies. If the same ontogenetic and spatial areas are targeted, it may be possible to define “chemospecies” in an assemblage.

L383: This is an incorrect reflection on the published literature (see above in main comments).

We will change this to make it clear that higher Mg/Ca is observed close to the apical line. However, this notion is correct as this topological variation is part of ontogenetic variation and early ontogenetic stages close to the alveolus do show higher Mg/Ca. Also, the bulk material of bulk sampled rostra will still have been derived from sections further away from the apical line, which, due to diagenetic alteration, has usually been avoided during bulk sampling.

L391: One may also take a inorganic geochemical approach to this problem and stipulate that a belemnite taxon has a distinct element partition coefficient for the studied elements, and therefore a change in the element/Ca ratio may simply be driven by a change in this ratio in seawater through time, which, given the typical residence time of Ca, Mg and Sr in seawater in the 10^6 year range, is at least conceptually feasible.

We will add this point. This might be a conceptually feasible “thought experiment” but is largely at odds with the widely held understanding of the biomineralization of molluscs, whose biominerals are not direct precipitates from seawater and differ in all aspects from inorganic precipitates.

L396: EDS analysis is quite problematic for Mg and Sr content determination – if it is SEM-based then it would be nearly impossible to get reasonable data from this; if it is EPMA-based then still, error bars are likely to be significant. Most of the published data, however are very likely from wet chemical analysis, for which the described factors are irrelevant.

We will check whether EDS has been used in any of the cited studies and modify the statement accordingly. Nevertheless, there will still be methodological differences between studies, even if their effect may be negligible. The point is that if all the data were obtained for a single study, one would analyse them all in the same facility with the same technical protocol. This condition is obviously not given for our dataset and we think that it is worth pointing this out, even if the impact is likely very minor.

L399: This also would have an effect if an electron beam technique was employed. This topic has been published on a fair amount and studies could be referenced here that illustrate this point.

We will add further references.

L410: I find this statement to be a bit vague. Mg/Ca ratios are taxon-specific, but one should be careful using them as taxon-specific markers?

Yes, because they are very heterogeneous and rapidly changing, they do not show strong separation and significant overlap between different clades. Furthermore, they may be affected by environmental factors. All in all, this makes them a poorly phylogenetically informative character. We will add some clarification.

L416: If Mg/Ca ratios correlate with Sr/Ca ratios, and Mg/Ca has a link to phylogeny, why then should Sr/Ca ratios not?

Given that the link to phylogeny is rather weak, such a strong correlation would not be expected. Note that there is a single evolutionary rate per tree (i.e., generation of the analysis). This means that if this rate was inferred as low for Mg/Ca, it was also inferred as low for Sr/Ca. The correlation between Mg/Ca and Sr/Ca ratios is between individual samples.

L420: Again, I find this statement quite vague – how are these ionic radii linked to element incorporation into belemnite rostra? Why should one wonder about aragonite if the rostra are calcite (which I think is well established)? How is Sr then linked with Mg? Why would Mg and Sr change if it is always calcite that is being formed?

The Ionic radius of Ca, Mg, Sr etc. is one of the key elements that determine the elemental geochemistry of carbonates. Elemental radius and crystallographic properties are closely linked even when only referring to biogenic and abiogenic calcite (organomineralic versus mineralic ultrastructures). We are not sure if we understood the reviewer properly, but we are happy to provide fundamental literature references to the reviewer if required.

L479: This may be true for some studies, but it is not entirely fair as a generalised statement. What is correct is that thresholds vary, but they also ought to do so for good reason: The levels of Fe and Mn present in pore fluids that are agents of chemical alteration are specific to local context, and therefore spatially (and temporally) quite variable, very likely even within individual sedimentary successions. It thus has to first be evaluated if Fe and Mn are powerful in delineating altered data at all, and – if so – at which level of these metals other tracers are seen to be substantially different from original values. The consequence is that no uniform limit is given in the literature.

We will modify the text to account for these factors. The problem with the thresholds remains that the primary element concentrations cannot be known and thus, the thresholds remain somewhat arbitrary. Thus, a combination of different diagenesis screening tools is important rather than relying on a single technique.

L484: It might be correct that some studies were too optimistic with their threshold values, but it is also worth pointing out that a threshold is a compromise that assumes the fraction of altered material in a sample to be uniform in chemical character, which is most certainly not the case. Diagenetic fluids will evolve and this is reflected in the nature of chemical overprint, and hence will cause knock-on effects for the validity of any given preservation threshold. It is therefore important not to be too dogmatic about such values, but to also use common sense in the interpretation of other chemical and structural observations to identify materials that may inadvertently have slipped through a yes/no exclusion routine based only on set element concentration limits. I do disagree with the concluding statement that chemical assessments can only ever be supplementary to optical and CL screening. Researchers have their preferred modes of preservation screening, and all of these techniques have their benefits and drawbacks. There is – in my opinion – no reason to treat chemical analyses as second-class diagenetic screening; they relate to the exact material that is tested for target geochemical proxies, so in this sense much more appropriate than looking at correlative fragments or sections on an SEM or optical microscope, and they are quantitative in nature rather than qualitative. It is also worth keeping in mind that a CL image is typically akin to a 2D Mn map in a calcite fossil, so in the end does tell the same thing as a Mn/Ca determination, but trades quantitative character for spatial resolution. SEM and microstructural imagery can be very powerful to illuminate what kinds of structural overprint may occur in a fossil, but they do not show exactly where it occurs in the entire specimen, and what the geochemical consequences of this alteration are. So, altogether optical and chemical techniques all have their place, and I do not think that there is reason to use value statements such as the ones in L491.

We will modify our statement, but we maintain the position that a combination of the available optical, microanalytical and geochemical diagenesis screening tools is the most appropriate approach. The reviewer’s statement that CL is akin to a 2D Mn map is not quite accurate, as CL is more sensitive and could show variation in Mn concentration where the EDS map suggests a homogeneous distribution or the lack of Mn (because of the detection limit). We thus agree that all these techniques all have their place, though none of them should be used in isolation.

L496: I am not sure how this approach allows to differentiate between environmental and evolutionary effects? Going through the text, this particular element seemed to still have scope to be looked at in more detail.

We agree that our current study is somewhat limited in this regard (hence, “potential”). Nevertheless, if the results were more strongly aligned with the phylogeny, it would be a clear indication of a strong evolutionary effect. It might be possible to differentiate between the two with more complex models. However, this would likely require improved sampling and a better-constrained phylogeny first. We will modify this statement accordingly.

L504: C and O isotope data were available for incorporation into the study. What was the rationale for not including them but to here point out that they should be? My feeling is that environmental controls are even more important on these isotopic systems than on element/Ca ratios, and that would be a challenge that would need to be addressed.

As explained above, we deliberately restricted our study to element ratios. Including isotopes would lengthen the study. Furthermore, due to the likely greater environmental controls, a quite different discussion would be required. Nevertheless, this is certainly something that should be done in the future.

Thank you for your review. Our replies to the individual comments are listed below in bold font.

This study uses a novel approach, bringing in a powerful palaeontological method to examine geochemical data from belemnites. It looks like there are potentially some very interesting applications of this approach, however, I do not think this comes across well in the discussion and conclusions of the manuscript. It is not clear what the significance of the phylogenetics really is in terms of using the geochemical data to reconstruct paleoenvironmental (oceanographic/temperature) trends. From the current way the manuscript is presented, it appears that there are significant differences between belemnite genera in terms of Mg and Sr incorporation, but that it is not necessary to apply phylogenetics to look at them; identifying the samples to genus level and not comparing different families/distantly related groups will be sufficient to avoid bias introduced by variable minor element incorporation. I think this manuscript would benefit from presenting more of a case for what the phylogenetic part of the methods can reveal; perhaps it is necessary to look beyond the potential use of using minor/trace elements to reconstruct environmental conditions. What else might the phylogeochemistry tell us? What questions can it answer?

Thank you for your suggestion. Phylogeochemistry has many potential applications beyond our case study. One major benefit is that it allows the direct link of geochemical proxies with species in a phylogeny. This means that rather than comparing a phylogenetic tree with a single (e.g., global) curve of proxy data and assessing its impact on the evolution of the group, each species has a distinct value for the proxy data, and the values at the nodes at the trees can be estimated. This can then be used to investigate, e.g., the influence of local (!) temperature on extinction rates or how frequent evolutionary transitions between cold-water and warm-water clades were. We will modify the manuscript to mention examples like these and the broader applications of phylogeochemistry.

Please consider all minor technical and language edits done.