The manuscript by Stevenson and co-authors uses XRF and geochemical data in a snow petrel stomach oil deposit to reconstruct summer sea-ice dynamics in the eastern Weddell Sea over the Holocene.  The investigation of this novel type of archive provides complementary information to ice and marine cores and, here, new information from an under-sampled area. The present study is, therefore, timely and of great interest to the paleo-community and beyond.

The manuscript is written in a complicated way, with much diluting information, and the structure is not always sensible. I started listing the comments below, but it was taking too long. I, therefore, present the most important ones only and direct the authors to the annotated manuscript for additional ones.

Sea-ice extent vs sea-ice duration: There is constant confusion between sea ice expansion (winter sea ice edge farther to the north), which you can infer from ice core data or a transect of marine cores, and sea ice conditions over the continental shelf. For example, Lines 342-343 the cooling of surface and freshening of shelf surface waters since the middle-Holocene (Ashley et al., 2021) conducted to increase sea-ice concentration and sea-ice duration on the continental shelf (Crosta et al., 2008; Mezgec et al., 2017; Johson et al., 2021), but did not result in greater sea-ice expansion that conversely recessed in the open ocean (Nielsen et al., 2004; Xiao et al., 2016). This dichotomic behaviour is possibly related to the latitudinal insolation and thermal gradients (Denis et al., 2010) as well as the multi-centennial expression of climate modes (Crosta et al., 2020).

Regional settings: Section 2.1 does not contain a regional context despite the title. I would recommend to detail the regional oceanographic and cryospheric system that will help understand the interpretations thereafter. In this optic, the seasonal sea-ice cycle must be described to know the mean location of the modern sea-ice edge each month as a basis of past changes inferred from the stomach oil deposit analyses. Visualising the seasonal cycle in the northeastern Weddell Sea would also allow to better understanding of the spatio-temporal foraging behaviour of the snow petrel and clarify whether the Maud Rise polynya could have been a foraging zone (section 4.5) despite being very remote (Figure 1).

Age Model: All published Bayesian age models have the mean age (red curve in figure 2) that follows more or less the 14C dates and their envelopes (blue ellipses). Here, it is out of the ellipses by a few hundred years and seemingly follows very thin darkish "lines" scattered throughout the record. Why and what are these thin lines? It is not even clear what represents the blue ellipses. For example, the three at ~5cm are centered at 4200 years BP, while the raw ages are ~5000 years BP, and the calibrated median is ~4500 years BP (Table 1). Finally, the caption of Figure 2 must be detailed for the red lines and its envelope, the blue ellipses, etc.

Statistical analyses: The input data for the PCA must be detailed in section 2.7. From Supplementary Figures 9-10, I understood that the input data are log10 transformed and centred datasets (not said), but the data presented in Figures 3-4 are raw data (cps for XRF data) normalised to accumulation rates. This explains why there is little resemblance between the PCs and the XRF data that make them (PC2 and Cl and S, for example). Overall, this is very confusing and must be better explained. Overall, I question the utility of the PCA as downcore PCs are hardly interpreted.

Structure: The Results present many interpretations, which complicate the reading. For example, lines 217-218 and 220-221 refer to previous publications; lines 227-229 and 310-320 identify the meaning of the proxies. The former type of interpretation can be sprinkled throughout the Discussion, while the former can be gathered in a section dedicated to the use of the proxies.

There are also many typos and non-consistent use of sea ice vs sea-ice.

Interpretations: (1) It seems to me that most of the interpretations are indicative of environmental conditions in the central Weddell Sea, where summer sea ice is present today but represents a fifth of the foraging area. I am less convinced it is valid for the northeastern Weddell Sea and Lazarev Sea, where the continental shelf is very narrow. (2) I sometimes did not get the reasoning by which foraging in coastal polynya is inferred (lines 353-355), which is not supported by the data (lines 360-363). Overall, it seems to me that authors infer either an open ocean or coastal polynya foraging, while snow petrels may have foraged at the marginal ice zone as it recessed from offshore to the coast over the feeding season. Marine cores show open ocean conditions during summer off northern Ross Sea, Wilkes Land, Prydz Bay, Western Antarctic Peninsula, etc. I do not understand why the authors here suggest the presence of summer sea ice off the northeastern Weddell Sea (Figure 5). (3) Some data are over-interpreted. For example, it is said that the cholesterol is higher in Org-C zone, which is not true in Figure 3 and 6. Authors infer feeding in coastal polynya (lines 425-426), which is contradictory with krill feeding as these organisms need a deep water column for their biological cycle. (4) By comparison to other coastal sites (marine cores), open summer sea-ice conditions may have prevailed off the nesting sites during the late Holocene. Additionally, nesting was possible during the last glacial at the nearby Untersee Oasis when sea-ice conditions were much harsher (McClymont et al., 2021). Do you really think that “increased sea ice extent restricted access to foraging grounds and by ~1700 cal. yr BP resulted in abandonment of the nest site”? (5) Eventually, I do not see the added-value of section 4.6.

I hope these comments and the ones listed in the annotated manuscript will help improve an important study.

Xavier Crosta

General comments

The manuscript is about the use of a fossil stomach oil deposit of snow petrels as an archive for paleoenvironmental conditions in the Weddell Sea region during the Holocene. This is a relatively novel approach as a limited number of studies have been published on this topic so far, making this study a valuable contribution to a better understanding of paleoenvironmental conditions in this sector of the Southern Ocean. The authors analysed bulk isotopic (¹³C and ¹⁵N) parameters, lipid biomarker compounds (n-fatty acids, n-alcohols, Phytol and sterols) and elemental analysis (XRF-Scanning) to infer past changes in the composition of snow petrel diet. These changes are linked to the prevailing sea ice conditions in the foraging range of the birds.

Although the approach is innovative, the discussion presented in the manuscript is vague and gives the impression that the interpretation is not necessarily supported by the data presented.

The use of individual lipid compounds to reconstruct snow petrel diet is a significant simplification, given the complex patterns in the prey organisms and potential post-depositional alteration, but a necessary one to derive qualitative paleo-proxies. Therefore, the derivation of dietary composition from the lipid data should be dealt with in greater detail. Similarly, the link between diet and foraging region, sea ice and marine productivity needs to be outlined in a more concise way. I therefore suggest that the manuscript should be revised in order to better justify and elaborate on the interpretation of the data. Below you will find specific comments on individual paragraphs in the text. I have made only a few comments on the discussion chapters, as these should be streamlined overall.

Specific comments

Line 102 – 108 Description of ¹⁴C sample preparation procedures

The description for the two procedures (bulk sample and acid treatment) differ in the detail given. Please add how the samples were graphitized at BETA in order to provide the same level of detail. Also, please rephrase the sentences in line 106-107, as it is not clear, whether the CO₂ that was released by the acid treatment was graphitized or whether the residue was further processed for ¹⁴C analysis. If the latter is the case- how was the sample transferred into CO₂?

Line 110 ff: Age-depth model

The age-depth model shown in figure 2 does not fit to the calibrated ages of individual ¹⁴C dates. The authors state that this is an effect of the Bayesian approach that does not necessarily lead to an age-depth model that passes through the median of each date.

However, the age-depth model they present seems to systematically overestimate the ages of the sample in units Org B and C. Given that the interpretation of the results is mostly based on the resulting accumulation rates, the author should re-calculate the age-depth model by modifying some of the priors to achieve a better fit.

Line 120ff: Reporting of ¹⁴C ages in Table 1

Mixing the measured ¹⁴C values with the results from the age-depth model is confusing. Please clearly separate these. You could for example add the modelled ages in the data table provided via Pangea, remove the two depths (0.5 and 18.25 cm) without ¹⁴C ages from Table 1 and add calibrated values for each ¹⁴C age. To complete the documentation of ¹⁴C analysis, please add F¹⁴C values and round ¹⁴C ages.

Line 192ff: Statistical analyses

Did you use original values (counts, concentrations) of accumulation rates to conduct the cluster analysis and PCA? Please state in the text.

Which elements/compounds did you include in the statistical analyses? Please name them here. XRF data: Is it really useful to include “all” parameters in the analysis? I’d recommend to use those that are relevant for answering the research question, as e.g. Te, Se, Rb and Pb are not further discussed in the manuscript.

Line 210 ff: Sampling and validity of accumulation rates

The interpretation of environmental changes is strongly relying on the accumulation rate of specific lipids. However, I see some issues with this procedure and don’t think that accumulation rates can be used here.

The first concern is a geometrical effect of sampling: In figure 2 the age-depth model and the stomach oil deposit with the profile and sample positions for ¹⁴C ages (and biomarkers) are shown. The sampled profile is not perpendicular to the layering in the deposit, and hence the “depth” distance between two samples in the profile does not correspond to the thickness of material that was deposited between two points in time. Therefore, the interpolation of ages between ¹⁴C-dated samples may be valid procedure to estimate the age of a sample, but the depth scale is not reflecting the actual built-up of the deposit through time (see also figure uploaded in separate file).

The second concern is that the depth-scale is only valid along the one sampled profile and not representative for the whole deposit. In SI figure 1 the complex depositional history/internal structure of the deposit is shown. The lithological units can be correlated through the deposits but they vary in thickness and the same unit can have different thickness depending on where in the deposit you measure. Assuming that the boundaries of a unit correspond to specific time horizons, this means that different thicknesses were deposited during the same period, depending on where you look in the deposit and hence accumulation rates are variable within one lithological unit.

Line 224: Origin of fatty acids in stomach oil deposits

The authors suggest that “samples likely comprised predominantly wax esters” – I don’t think that this conclusion is supported by the data presented by the authors. The concentrations of fatty acids are > 10 times higher than those of n-alcohols (line 232 and line 233). Given the 1:1 ratio of fatty acids and n-alcohols in wax esters, this suggests additional sources of fatty acids. Please discuss further, what other sources of fatty acids may occur in the deposit and what the implications are for the interpretation of fatty acids with regard to snow petrel diet.

Line 230: Identification of C18:1 homologue

In avian dietary studies and also in lipid studies of krill, fish and other marine organisms it has been shown that more than one C18:1 compound is abundant and the position of the double bond is diagnostic for specific sources (e.g., Connan et al. 2008, Yang et al. 2016). Which C18:1 compound are you referring to here or did you sum up all C18:1 compounds? Please clarify.

Line 235: Cholesterol as marker for krill

Here it is stated that cholesterol is a marker for krill, because it accounts ”for more than 76% of total sterols in krill (Ju and Harvey, 2004) and is typically lower in concentration in fish.” This conclusion should be better substantiated, as cholesterol is probably the most common sterol in both groups, so it cannot be said that an increase in cholesterol is associated with more krill. Eating more fish also leads to more cholesterol. My suggestion would be to refer to compound-ratios when discussing relative changes in the composition of snow petrel paleo diet (Also for fatty acids). Please provide refences for the concentration and abundance of cholesterol in fish.

Line 246 ff: PCA results

To what extent can the assumptions about the sources of biogenic components be confirmed by PCA? Are there any indications of a correlation between cholesterol content and C14 fatty acids? Please add this information to this chapter or insert a paragraph on this topic elsewhere in the discussion.

Line 255 ff: Biomarker zones in the deposit

How do the units defined by the cluster analysis of organic compounds correlate to the visible lithology and to the units defined by XRF?

The deposit shows some changes in colour and was divided into lithological units (supplement). How do these relate to the cluster units identified by the cluster analysis? Is the change in colour/texture related to changes in inorganic and organic composition? And in the cluster analysis? Please point that out in the results or discussion sections.

Line 294ff: Interpretation of inorganic composition

The discussion on the sources of inorganic elements in the deposits needs to be more concise.

The sources of the elements in the deposit are not “local erosion” or “wind-blown” as this ascribes processes and not distinct sources. E.g., some of the local erosional products are likely also windblown to the snow petrel nest. And if Cl, S, Br and P are “windblown” they still have to come from somewhere. I’d suggest to distinguish between minerogenic material derived from bedrock, reflected by the elements Fe, Al, Mg and Ca and other elements such as Si, Ti and Zr. In the lithological description (SI Fig. 1) the rock fragments were assigned as “granite”. What rock type was the sample analysed for elemental composition?

For the second group of elements, please discuss, where the windblown particles come from.

Line 304: Is it possible that Ca is derived from carbonate in that specific layer (e.g. incorporated fragments of egg shell)?

Line 335ff: Role of ice shelf retreat on accessibility of foraging areas

This section is not linked to the findings of this study. Only in line 340-341 is stated that “Maintenance of the ice fronts in a retreat scenario from the start of the record is consistent with our evidence of increased availability of productive foraging habitat”. This statement needs to be clarified: Why is the retreat of Brunt Ice shelf consistent with productive foraging habitats, and how do you infer the productivity of the foraging habitat from your data?

Line 429: Nitrogen isotopes

Please clarify what the potential end-members for your isotopic composition are. The range of 9 to 19 permill in ¹⁵N values is quite large. Discuss if a shift in the order you find is reasonably explained by nutrients in the ocean (glacial/interglacial shift are c. 4 permill only, Horn et al. 2011). What could be other effects? E.g., Alteration by weathering and degradation, different fractionation depending on the sample composition (in the nitrogen-containing compounds of the sample, what are these?).

Line 516: Interpretation of abandonment of the nesting site at 2000 due to sea ice: I don’t think that the ¹⁴C age from the top of the deposit can be inferred to mark the timing of abandonment because 1) the deposit may have been degraded/eroded from the top when the nest was no longer occupied and therefore the age can only indicate a maximum age of abandonment. 2) The abandonment of the nest could as well have other reasons, such as physical properties of the nest cavity (See Einoder et al. 2014).

Technical corrections

Line 21 and 516: 2000 cal yr BP instead of 6700?

Line 68: You state here, that the deposit is well preserved - please explain how you come to this assessment.

Line 70: “radiocarbon-based age-depth model” instead of “radiocarbon dated age-depth model”

Line 150ff: Biomarker analysis- Which fractions did you separate/in which fractions did you recover n-alcohols, phytol and sterol? Please add.

Line 193-196: The sentence seems to be incomplete, please revise

Line 193-194: What is Org 1-3 and XRF 1-3? Is it the “Units” that are defined by the cluster analysis? In Figure 2, 3 and 4 and in the subsequent text, the authors refer to Org A-C and XRF A-C. Please clarify

Line 199: Please list the elements (XRF) and compounds (organic) that were included into the PCA.

Lines 204-207: Incomplete sentence, please revise.

Line 220-221: Please check the values given for ¹⁵N: lowest value is equal to mean value

Line 238: Please add reference for the cholesterol concentration in fish to support that cholesterol is indicative for krill in stomach oil deposit. As stated in line 236 cholesterol is ubiquitous.

Line 233: What are “key n-alcohols”? Better name the compounds. Up to here the results of n-alcohol analysis have to been mentioned in the text.

Line 320: “Bedrock contamination is unlikely given local gneiss bedrock” This sentence is out of context, please revise.

Line 322: typo “lotted”

Rounding of ¹⁴C ages: Please check through the manuscript and use the general rounding convention for ¹⁴C ages (see table below)

Figure S4: Which parameters are shown here? What are blue data points and what are the black data points?

References:

CONNAN, M., et al. 2008. Interannual dietary changes and demographic consequences in breeding blue petrels from Kerguelen Islands. Marine Ecology Progress Series 373. 123-135

EINODER, D., et al. 2014. Cavity characteristics and ice accumulation affect nest selection and breeding in snow petrels Pagodroma nivea. Marine Ornithology 42, 175-182.

HORN, M.., et al. 2011. Southern ocean nitrogen and silicon dynamics during the last deglaciation. Earth and Planetary Science Letters 310, 334-339.

YANG, G. et al 2016. Fatty acid composition of Euphausia superba, Thysanoessa macrura and Euphausia crystallorophias collected from Prydz Bay, Antarctica. Journal of Ocean University of China 15, 297-302.