Review of manuscript submitted to Climate of the Past by Philip Meister and colleagues: A global compilation of diatom silica oxygen isotope records from lake sediment – trends, and implications for climate reconstruction

The temperature dependence of oxygen isotope fractionation makes them a widely applied (paleo)climatology tool. Oxygen isotope ratios (expressed as δ¹⁸O) in diatom biogenic silica represent a valuable archive, but their interpretation and model-archive comparison can be complicated, particularly in lakes. Here, Meister and colleagues compile published lake diatom δ¹⁸O records to address the extent to which a common signal can be observed. They find 54 lakes in total, from 71 publications, that have available data. Compiling, binning and filtering the data, they observe commonalities in lake diatom δ¹⁸O records over the common era and a consistent, decreasing trend over the Holocene, which is broadly consistent with existing palaeoclimate (particularly temperature) records and understanding of the δ¹⁸O proxy.

In general, this is a useful compilation that seems comprehensive and thorough and achieved in a sensible way, and is being made available via an appropriate repository (Pangaea.de). The introduction largely sets out the state-of-the-art (and Fig. 1 is particularly useful, I feel). The compilation and standardisation seems like a large effort the authors should be commended for. The topic of the manuscript falls within the scope of Climate of the Past, and the conclusions are largely supported by the data. The manuscript is generally well written (although with scope for tightening the language and a few typological/grammatical errors that could be caught with a through proofread) and the figures are clear. Overall, I think this is a manuscript and a data compilation that deserves to be published, and will hopefully stimulate more work in the field (especially given their Fig. A1B which shows declining production of lake diatom δ¹⁸O data).

My major criticism is that the discussion of the compilation is rather descriptive and qualitative. There is very little in the way of statistical analysis, which might be useful in parts of the discussion related to the magnitude of δ¹⁸O trends, regional differences in the timing of minima/maxima, and the presence or not of periods of stasis. To what extent can they be demonstrated to reflect ‘real’ underlying phenomena, vs. just being an artefact of low and noisy data availability? The comparison of the new NH compiled record to existing proxy data and insolation curves is also rather qualitative. Beyond this, the discussion focuses almost entirely on sites from >45 deg N, which is motivated only on L249 (unless I miss it elsewhere). Given the general paucity of data, it seems a shame not to exploit the compilation as much as possible. While other regions might be too data scarce to do the binning/filtering steps in e.g. Figs 5 and 6, do they contain useful information on the spatial pattens which could complement the discussion in e.g. section 4.2 and Fig. 7? Fig 5G and 6G each contain only two sites that meet the quality control criteria – is it really the case that there are not even two sites from e.g. South America or tropical Africa sites that are useful?

Minor comments

Fig 1/main text L105: Could diagenesis also be included in this figure? And/or more detail given in the main text? Presumably older samples are more susceptible to diagenetic overprinting? How can we be sure this is not a major driver of the observations?

L210: I would suggest a brief summary of the hydroLakes database is warranted, since L219 refers to a ‘geostatistical approach’, implying that values are not specific to a given lake, and many readers (including myself) would not be familiar with it. Would it not make sense to preferentially use the parameters as given in the original publications, supplementing them with hydroLakes only when necessary?

Fig 4: excludes 33 (of the 49 extant) lakes because the HydroLakes dataset doesn’t include their catchment area. I would suggest this is something that is relatively easy to define in a consistent way given a digital elevation model, with readily available topographic analysis tools, e.g. the ‘TopoToolbox’ for Matlab, or similar capabilities within ArcGIS or qGIS.

L233 “presumably” instead of “supposedly”?

L261: Does this approach of subtracting the mean of a record only work when every record covers the full timespan under consideration? Otherwise a record that covers only e.g. the mid to late Holocene would bias that period (because the mean subtracted from that record would be smaller than a record with an identical gradient/trend but longer coverage).

L270: It’s a bit unclear how many sites/records are actually used. L271 states 64 sites; L283 is states both 54 and 56 (i.e. 7 + 49); Table A1 has 53 lakes. Can this be clarified?

L275: Table A2 is mentioned before Table A1. Is A2 complete? There doesn’t seem to be enough lakes here to match the numbers given in the main text. Also, it might be helpful to include the number #X in both tables A1 and A2 somehow to allow cross referencing.

L285: “extant”? (rather than ‘still existing’)

L310: A reference to Downing and Duarte (2009) (their Fig 5) or similar might be useful here.

L351: ‘may correspond to more than one record’

L379: There can be some buffering even when t_res is less than sampling frequency, meaning the "full" amplitude is not necessarily displayed See e.g. Richter and Turekian (1993). 

Fig 5: Presumably the axis labels shouldn’t read ‘kyr’ but ‘yr’?

L435: it seems like more could be done here with the records from other regions of the world.

L465: This seems like repetition/overlap with paragraphs starting L486 below. Is a ref here to fig 7 appropriate?

L473-5: An example of where more statistical rigor might help: how robust is this particular interpretations relative to a simpler view of consistently decreasing δ¹⁸O? (a straight line could be drawn through the ±1sd shading).

L489: ‘show a tendency’ (also L503)

Fig 7: Even if not discussed (see above) could the minima/maxima of records outside the northern hemisphere high latitudes be displayed here?

L505: Again, without more statistical rigor, from e.g. Fig. 6E/G I would be cautious about over interpreting these differences.

L520: presumably also because most biogenic production is in summer (particularly relevant for short residence time systems).

Fig 8 caption: subscripts and superscripts not displayed correctly.

L558: Fig. 8D is labelled June, here it says July.

L564: How far behind?

Section 4.4: given some of the discussion previously I would have expected some stronger/more explicit recommendations in this section, for how diatom δ¹⁸O can become a more useful proxy (beyond the rather generic ‘further research is needed…’).

L595: ‘would be consistent with’ or similar?

References

Downing, J.A., Duarte, C.M., 2009. Abundance and size distribution of lakes, ponds and impoundments, in: Likens, G.E. (Ed.), Encyclopedia of Inland Waters. Elsevier, Oxford, UK, pp. 469-478.

Richter, F.M., Turekian, K.K., 1993. Simple models for the geochemical response of the ocean to climatic and tectonic forcing. Earth and Planetary Science Letters 119, 121-131, doi:

This study compiles 71 δ18OBSi records from lake sediments covering various time periods and locations, predominantly >45° N during Holocene. Despite originating from different geographic locations, the records feature common patterns that correspond to known climate events of the Holocene and the Common Era. The collection of δ18OBSi records, binned to a common temporal resolution and complemented with metadata on hydrological parameters, is a valuable contribution that will facilitate climate reconstructions and proxy–model comparisons.

This manuscript is very well written and the records compiled for this study are well presented. Although the statistical analysis presented here is rather simple, this is probably the correct approach considering the large differences in temporal coverage and sampling frequency between the records, as well as the sparse spatial coverage. This work is valuable for the paleoclimate community and clearly deserves to be published in Climate of the Past. However, I found several issues that should be addressed before publication.

My only major criticism has to do with the analysis in Section 4.4, which, in my opinion, is unconvincing and requires revision. Please see my comment below.

Information provided in the two tables is useful and well presented. However, these tables and the associated PANGAEA dataset would benefit from including additional information discussed in the text, such as the dating method, temproal coverage, and temporal resolution. These details could be incorporated into Table 1A or shown in a separate table.

The Introduction is comprehensive and well-written; Fig. 1 provides a helpful overview of the processes shaping the δ18OBSi signal. However, it does not become clear until Section 1.4 why this study focuses on diatoms but not other sources of d18O data. It would help the reader understand the novelty and purpose of this study if this was mentioned in the Abstract and/or earlier in the Introduction.

L54: This sentence could be made more clear. Does the phrase "common δ18OBSi patterns" describe a comparison between the different lake records or a comparison between the lake records and previously known climate events?

L58: Typo "extratopic"

L182: "Such compilations, however, generally do not include δ18OBSi–records." Why not?

L232-233: "the effect of different 14C–calibrations and different age model approaches is supposedly minor." Could you please elaborate on that?

Section 3: The authors have grouped the records according to Marine Isotope Stages. It would be helpful to define the temporal boundaries of the MISs within the text.

L276-280: Can these dating methods be included in Table A1?

L288-297: Could the temporal coverage information be included in Table A1?

Fig. 2B: It appears that there is missing information for some records. Please provide an explanation.

Section 3.4: Section title is "Temporal coverage and resolution of combined records" but is "temporal coverage" discussed here?

Section 3.4: Can the temporal resolution be included in Table A1?

Section 3.4: Is "sampling resolution" the same as "temporal resolution"? Additionally, it should be noted that temporal resolution is nonuniform across some (all?) of the records, with generally higher resolution for more recent time intervals (e.g. see Fig. 5A). This impacts the comparison shown in Fig. 4B.

L488: The authors state that Eastern Eurasian sites feature a Holocene maximum at 12 kyr BP, but it seems that this is because the start of the Holocene has been defined as 12 kyr BP. As can be seen in Fig. 9A, the actual maximum occurs earlier in some records, around 13-14 kyr BP. Therefore, it might be more appropriate to show 13-14 kyr BP as the Holocene max. in Fig. 7A for these records.

Section 4.4: The comparison of MIS 1 and MIS 2 means, in my opinion, is very misleading. As shown in Fig. 9A, some records only cover the very end of MIS 2 when d18O is near maximum. Thus, these records have higher mean MIS 2 d18O values compared to records that span the entire MIS 2 period. Fig. 9B shows large differences between the records, which can be wrongly interpreted as differences in the climate signal, when they are likely the result of the records covering different time intervals. In fact, the records in Fig. 9A appear to be in a good agreement, except for one outlier. The authors acknowledge their concern in L582 but unfortunately I feel that there is no benefit in the anaysis presented in Fig. 9B. I recommend either removing the lower panel of Fig. 9 or replacing this analysis with a different one. Section 4.4 should be revised accordingly.

Fig. 9A shows NH records, but Fig. 9B shows all individual records. It would be more consistent to present the same set of records in both panels.

Fig. 9B: It is unclear whether what is shown is the difference MIS 1 - MIS 2 or MIS 2 - MIS 1.

Review of the paper by Meister et al. entitled “A global compilation of diatom silica oxygen isotope records from lake sediment – trends, and implications for climate reconstruction”

General comments

In this study, the oxygen isotopes in biogenic silica (δ¹⁸O_BSi) of 71 down-core records published to date were analysed and interpreted with respect to climate change for different regions and time periods (Common Era, Holocene, MIS 2). The focus is mainly on the correlation of δ¹⁸O_BSi values to temperature. As a specialist in climate dynamics and meteorology, I am not able to assess the geochemical sections of the publication.

The strength of the paper is that the existing data and publications dealing with δ¹⁸O in diatom silica are collected, archived, documented and interpreted together. This must be seen as an important step, because the dynamics of δ¹⁸O are a substantial component of the dynamical climate system. Therefore, δ¹⁸O also forms an important component of modern climate models. For these reasons, the publication of this contribution is desirable and highly recommended.

The paper has several weaknesses. On the one hand, the relevant processes, as excellently illustrated in Figure 1, should be given more consideration in the interpretation of the data. Processes such as evaporation and rainfall, transport distance and season of precipitation, continentality, freezing and melting, etc. play a crucial role if they are related to temperature or precipitation. On the other hand, both the resolution, the number and the spatial representativeness of the data are relatively limited. Overall, the statements made primarily concern the Northern Hemisphere.

For the reasons mentioned, I recommend acceptance of the publication after major revision.

Specific comments

Figure 5: The variability of the data is extremely high. Perhaps instead of plotting maxima and minima (Fig. 7), individual curves should be assessed in relation to the processes shown in Figure 1. In general, it must be acknowledged that in Figure 5 F the two main cooling periods (Dark Ages Cooling or LALIA and Little Ice Age) are quite clearly indicated. Büntgen et al. (2016) have recently characterized the Dark Ages Cooling and Wanner et al. (2022) diagnosed the Little Ice Age cooling (Wanner et al., 2022).

The Holocene subsets clearly indicate a negative temperature trend. This should be commented on in more depth because it is still not clear whether temperatures increased or decreased in the late Holocene (Liu et al., 2014; Wanner, 2021). The most recent reference on Holocene climate should be commented: Kaufman and Broadman, 2023.

Figure 8: I am asking me whether the selection of the time series shown is significant. In contrast to the data of the present study, the time series of Marcott et al. (2013) is primarily based on marine data. I recommend that the latest reconstruction by Kaufman et al. (2020) be used here. Herzschuh et al. have just presented a new reconstruction in Climate of the Past. It would be exciting to further consider why the Eurasia and North America curves in Figure 8 F diverge in the early Holocene. How far are the detected time series temperature-sensitive or humidity/evaporation/precipitation-sensitive?

MIS 2 (Fig. 9 ): As requested above, I think it is right that the course of the individual curves in Fig. 9 A is interpreted in terms of dynamical processes. This should be attempted even more strongly; in particular, it would make sense if the origin of the individual figures were indicated in this figure.

References:

Büntgen, U. et al., 2016. Cooling and societal change during the Late Antique Little Ice Age from 536 to around 660 AD. Nature Geoscience 9(3). DOI:10.1038/ngeo2652.

Kaufman, D., McKay, N., Routson, C. et al. (2020b) Holocene global mean surface temperature, a multi-method reconstruction approach. Scientific Data 7: 201.

Kaufman, D.S., Broadman, E., 2023. Revisiting the Holocene global temperature conundrum. Nature 614, 425-435. doi: 10.1038/s41586-022-05536-w.

Liu, Z. et al., 2014. The Holocene temperature conundrum. Proceedings of the National Academy of Sciences of the U.S. 111, E3501–E3505.

Wanner, H., 2021. Late-Holocene – cooler or warmer? The Holocene, 31(9), 1501-1506.

Wanner, H. et al., 2022. The variable European Little Ice Age. Quaternary Science Reviews 287, 107531.