Review of manuscript egusphere-2024-3686 submitted for publication in EGUsphere by Qiang Zhang and colleagues: Faithful transfer of radiolarian silicon isotope signatures from water column to sediments in the South China Sea

There is growing interest in how the silicon isotope composition (expressed as δ30Si) of radiolarians can be used to supplement/complement those of the more established proxy archives in diatoms and sponges. As with these two groups, there is a need to understand and account for any post-mortem alteration of the initial isotope signal. Here, Zhang et al compare bulk-assemblage radiolarian δ30Si values from water-column plankton tows and underlying coretop sediments. They demonstrate the two are statistically indistinguishable, lending more confidence to the use of radiolarian δ30Si as a window on to past Si cycling.

In general, this manuscript is very well written and the figures are clear (with possible exception of Fig.3). The referencing is generally appropriate – with some notable absences (see below). The data are generated by appropriate techniques (though some more details may be warranted). Overall, there is little to criticise in terms of the central conclusion – that the radiolarian δ30Si signal is not altered during water column sinking – which is supported by the data (though I have a series of minor comment/suggestions that I detail below). Nevertheless, this is a relatively small dataset and I have the impression that with a slightly expanded dataset much more could be done. I note in the supplement, Fig. S1 contains 22 ‘unpublished’ radiolarian datapoints. By integrating these, and dissolved Si δ30Si data, a much more impactful paper would result. Some suggestions are below, but this is ultimately an editorial decision.

Minor comments and suggestions

L16-27 – the prominence of the SALH in the introduction is a bit strange to me, considering it’s not the focus of the manuscript (and isn’t returned to)

L48: Two papers that deserve citation/discussion here and elsewhere are Closset et al. 2015 (doi: 10.1002/2015GB005180) and Grasse et al. 2021 (doi: 10.3389/fmars.2021.697400) – both present a comparison of plankton tow diatoms and sediment trap (Closset) or core-top (Grasse) material, concluding that the transfer of biogenic silica from surface ocean to depth isn’t associated with a resolvable change in δ30Si. See also Varela et al. 2004 (doi: 10.1029/2003GB002140; sediment trap data) and Fripiat et al. 2012 (doi: 10.5194/bg-9-2443-2012; water column biogenic silica data).

L85 and introduction: In general, there is a growing awareness that ‘bulk’ assemblage δ30Si data have disadvantages as a paleo-archive, and that where possible single-species records are much stronger. Therefore it would be good to see some justification for why this was not attempted here.

Section 2.3: I would suggest more detail is needed here. Specific suggestions include:

• Define what is ‘sufficient’ radiolarian tests (L80; what is the typical mass of Si processed)
• Give a brief overview of Zhang and Swann (L81). Is there potential for larger diatoms or sponge spicules to ‘contaminate’ the sample?
• Confirm that Na (used in dissolution) was successfully removed by the ion exchange chromatography
• Give approximate mass resolution (m/Dm, L93)
• Give details of how Mg measured/which ratio(s) (presumably in ‘dynamic’ mode), and a reference to Cardinal et al. 2004 (doi: 10.1039/b210109b) is probably appropriate
• Give approximate sample introduction rate, concentration, and instrument sensitivity
• Confirm what one analytical replicate represents (just one standard-sample-standard bracket, or (as is usual) three or four?)

L114: It’s not clear what volume the 28025-102443 individuals refer to – in a 1m2 water column? It is also not clear how these numbers are derived – presumably because the volume of water passing through the nets (L66) is known? This could be clarified.

L126: Can an approximate detection limit be given for these elements?

Discussion section: In general, there is no discussion of any spatial pattern in radiolarian assemblage. But I feel there is probably useful insight here. For example, Station 28 is located away from the cluster of other stations, and visually in Fig. 3 looks different. What physicochemical parameters influence the community assemblages? As an aside, I note that the assemblage data is not made available. Could the species in Fig 3 be amalgamated at a higher taxonomic level in order to make the similarities and differences clearer? And/or condensed, via an appropriate multivariate statistical approach, to 2 axes?

L134-140: In general, it’s the fractionation between dissolved Si and biogenic silica that is relevant, not the absolute δ30Si value. So it’s a shame not to see any dissolved Si δ30Si data from the water samples (perhaps this is coming in a later publication?). An existing dataset of δ30Si exists for the SCS (Cao et al. 2012, doi: 10.1016/j.gca.2012.08.039). The overlap isn’t perfect in terms of location or seasonality but its surprising that it’s not mentioned here, given the importance of water δ30Si in setting radiolarian δ30Si, and that it would allow the authors to place constraints on the fractionation of Si isotopes by radiolarians.

L165: Can an indication of the timespan covered by the upper 1cm be given? I presume there are some constraints on sedimentation rates and bioturbation in this well studied region.

Section 4.2: The radiolarian assemblages are similar between water column and core-top, so the inference is that ‘dissolution is expected to have limited impacts on these radiolarian shells’ (L209) – but is this necessarily true? Is is possible weakly silicified parts of the tests are dissolving? This would be interesting to know, as it has different implications for *how* the δ30Si is preserved: if no dissolution occurs, then there’s no real potential for altering the isotopic signature (which therefore means the conclusions here are not transferable to other settings where more dissolution does occur). But if dissolution does occur and the δ30Si remains the same, then either a) different parts of the tests have the same δ30Si or b) a coincidental balance of heavier and lighter parts dissolved. To begin to address this, it would be good to see an independent constraint on the (radiolarian) biogenic silica preservation effeciency, either from the literature, from a comparison of export vs. sediment-trap/burial fluxes, or even a theoretical predicted efficiency based on sinking speeds, water column depth, and dissolution kinetics. Finally, It would be good to see an attempt to engage with what may happen with progressive dissolution in the upper centimeters of the sediment – if there is preferential dissolution of some species here, might that introduce a bias (an apparent fractionation) to the bulk-assemblage δ30Si data?

Review of “Faithful transfer of radiolarian silicon isotope signatures from water column to sediments in the South China Sea” by Zhang et al.

In their manuscript Faithful transfer of radiolarian silicon isotope signatures from water column to sediments in the South China Sea” the authors present the first combined water column and surface sediment dataset of silicon isotope compositions of radiolaria. This data is further supplemented with radiolarian assemblage countings, which allows to investigate the influence of transfer from the water column to the sediment, especially via dissolution. The data presented is of great interest as d30Sirad has a high potential as a proxy to allow surface to intermediate water column reconstructions of the silica cycle, but is limited by few investigations on specific processes.

While I find the dataset of great importance and the presented findings are interesting and presented in a concise format, the authors should re-structure the discussion a bit and not solely discuss the influence of cleaning method and dissolution. This dataset has much more potential, especially as there is a dataset for d30Si-DSi values from the same area that could be used to investigate fractionation factors. I have summarized my main questions and comments below, followed by detailed comments per line.

Main questions:

1. The major question in terms of silicon isotope compositions of radiolaria in the community at the moment is the question about their fractionation factor. As your study provides d30Sirad from the water column and the sediment I am wondering why there is no discussion at all about the fractionation of Si and a calculation of D30Si (fractionation factor) even if you just measured the bulk radiolarian composition, as there are silicon isotope compositions of DSi from the water column available (Cao et al 2012, https://doi.org/10.1016/j.gca.2012.08.039).
2. The sample preparation and evaluation should be described in a bit more detail as you are referring to a newly published method it would be good to at least mention the main steps involved in the preparation (see detailed comments below). Furthermore, I would be interested to know if sample amounts were generally too low to try to pick single species or orders of radiolaria, as this would help better understand how different radiolarians incorporate Si and if there is a difference in Si fractionation.
3. The discussion feels a bit shallow. The highlight of your manuscript is the investigation of any dissolution effect on the sedimentary radiolarian signal. However, as you point out in the discussion, the prominent radiolarian species in your samples turn out to be species that are considered to be quite resistant to resolution. I suggest including some of the references you refer to in the last part of the discussion in the introduction instead and cutting the actual discussion about dissolution shorter, as you basically show that there is no evidence on dissolution based on the abundance data and the silicon isotope data. While you show this with your statistical analyses, you can also highlight that your d30Sirad values of water column and surface sediments are the same within analytical precision as indicated in Figure 5.

Comments per line:

L6: You should remove the comma after “water column”.

L8: Add a comma before “as evidenced”.

L18: Add a comma behind “Based on this”.

L25: Remove “actually”.

L26-27: Remove the commas behind “mechanism” and “impact” and “in both glacials”.

L29: I suggest removing “wider” or replacing it with “broader”.

L42: Remove “so”.

L46: Change to “others”.

L49: Remove “also”.

L 54: The authors claim that purifying radiolarian test from the natural environment have difficulties. Can the authors add a reference here and specify what difficulties they mean?

L70: Remove the comma behind “box corer” and add an article (either “the” or “a”) before “cold storage”

L73-78: Under section 2.2 the authors refer to the preparation of radiolarian slides but only refer to a previous paper concerning the actual method. Please add some specifics of which steps this methodology includes concisely.

L78: Changes to “individual counts”.

L 80-81: Similarly to my comment above the others only refer to a previous paper concerning the preparation of surface sediment, please add a short description what this includes, freezing, freeze-drying, wet-sieving as stated in the section above?

Further, can you specify what “sufficient” radiolarian tests are?

L85: rephrase to “ using NaOH fusion”.

L86: Remove “between”.

L95: You are referring to analytical replicates here. Was the same sample prepared by NaOH fusion measured on a different day, or are you referring to replicates within the standard sample bracketing procedure?

L96: “sampling of the standard (diatomite)” What does sampling mean here? Was the diatomite measured on several days? Were the columns prepared again? Was it prepared again?

L98: What do you mean by 2s absolute?

L98-100: I don’t see the value of adding the error of the diatomite to your analytical uncertainty. The diatomite and other standards are mainly measured to see if you get the correct values. If at all you can use the diatomite value to normalize your values. But why would you add the uncertainty?

L100: Reproducibility and instrument accuracy are generally indicated based on the NBS28, I don’t see any values for the NBS28 at all.

L107: Remove “for” after “To assess”.

Figure 3: The numbers in the figure and species names in the legend are hard to read. I suggest increasing the size of the figure for publication.

L125: Change to “constituents”.

L126: Remove “all”.

Figure 4: I am wondering if samples have been checked at higher magnification as well. If radiolaria are not completely clean, contaminants are found within the radiolarian structure and are not visible at this resolution with EDS. These would rather be seen by looking closer at individual specimens. Also, what contaminations would the authors have expected to see with the help of the EDS and SEM here that would affect the silicon isotope composition? If the main goal was to check for the effects of dissolution, I would have expected a much higher magnification of the SEM to be able to see if there were traces of dissolution or remineralization. which is hard with these pictures. Contamination by other Si phases, such as diatoms and sponges, is already visible with the light microscope. The SEM and EDS with this kind of use only show that there seems to be no clay contamination, which would also be better seen at higher magnification.

Figure 5: D30Si is usually referred to as the fractionation factor, so its use here is a bit confusing. Also, (A) already indicates that the data are the same within error, so I see little use in plotting the difference here as they are within your analytical uncertainty.

L134-140: The discussion starts with a comparison of the absolute d30Sirad values measured in this study with previous studies. While it is important to note that the range in silicon isotopes measured here agrees with previous studies, I don’t see a value in solely comparing absolute values between regions here. While this can serve as an introduction to the discussion, a comparison of the impact of source signatures of d30Si-DSi and potential fractionation factors of radiolaria is missing from the discussion so far.

L152: Correct to “all the radiolarian shells”.

L 153: Why did you expect the d30Sirad values to differ between depths? (100-300m versus 0-100m).

L170: Remove “(vital effects)” here.

L169-172: I don’t think this discussion is correct here. As far as I can see, there is no significant difference in species compositions or d30Sirad between your water column and surface sediment data (line 153-159). So even if there is a difference in d30Sirad and/or fractionation between radiolarian species/order, there is no way to see any effect of this in your data.

L180-215: Section 4.2 Transfer of radiolarian d30Si signatures into the sediment record need some re-structuring and rewriting. The first paragraph is fine, picking up on the fact that there is little difference between the water column and the sediment d30Sirad. The following paragraphs, however, are mainly a summary of the literature and should be shortened, and the information from the radiolarian abundances investigated here should be more highlighted. Additionally, references to the SEM and EDS analysis conducted are not referred to at all here.

L181: Change to “resemble”.

L201: Add a comma after “in the SCS”.

L206: Change to “in this study,…”.

L209: Correct to “expected to have..”.