Review of “Global and regional Pleistocene benthic δ18O stacks with a comparison of different age modeling strategies” by Zhou et al.

General comments

The authors propose an update to the LR04 benthic d18O stack (Lisiecki & Raymo 2005) that is widely used to construct age models of marine sediments in paleoceanographic studies. This update, which relies on an extended number of sites (224 cores), includes one global stack presented on 3 different chronologies and two regional stacks (Atlantic, Pacific). The manuscript describes the construction of these stacks and their chronologies. It also discusses the age uncertainties associated with them and the applications for which each stack is best suited.

There is absolutely no doubt that the new stacks proposed by the authors are of high scientific importance and will be widely adopted by the paleoceanographic community. Overall, the manuscript is well written and illustrated. It would have been helpful to include more information on the sites selected for each chronological iteration, as well as a more detailed discussion on specific aspects, which are detailed below.

Specific comments

1. The Methods section starts directly with the description of stacking steps. I think a new subsection should be added first to better describe the marine cores selected for the construction of each stack and the related criteria: number of cores, proportions of new sites vs. cores already included in previous stacks, name of cores, their location (including water-depth), their references, mean temporal resolution of records, which proxies are available for each core apart from benthic d18O (magnetic data, SST, others?). Right now, this information is partly disseminated throughout the text, partly present in the supplementary data or missing. I think presenting an overview of selected cores and criteria in the main text and explicitly referring to the supplementary data for more information would be helpful.
2. I am surprised that the list of geomagnetic reversal and excursions selected for BIGSTACK_magrev (Table 1) does not include recent well-dated magnetic events such as the Mono-Lake, Laschamps and Blake events. Reading the discussion in lines 411-416, I guess it is because only geomagnetic reversal/excursions dates derived from astronomical tuning are included. First, I would place this statement earlier in the text, in the Methods section. Second, I think one advantage of using magnetic reversal/excursions is also to benefit from absolute and relatively accurate dating from Ar/Ar. I find it really unfortunate that the attempt to use Ar/Ar dating did not succeed (lines 414-417). I would encourage the authors to further detail and illustrate this attempt. And if difficulties arise in the interval around 1.9-2.1 Ma only, why not reduce the coverage of the stack to the most recent 1.9 Ma or simply omit the period 1.9-2.1 Ma in the produced stack? More discussion on this issue would be helpful.
3. Given the high scientific importance of the produced stacks, I think that their relative comparisons are too brief and insufficiently discussed. Because stacks are plotted on different Y-axes in available figures (Figures 6, 7, etc), I find it very difficult to compare the amplitude and the timing of the stacks. I think it is particularly important for the comparison of the global BIGSTACKmixed stack to the regional Atlantic and Pacific stacks. I would invite the authors to plot these stacks on the same Y-axes and develop the discussion on their similarities and differences. Also, I would find it interesting to reinforce the discussion by adding a comparison, in addition to LR04, of the new stacks to ProbStack and HN23 stack, which is currently absent in the manuscript. Finally, more discussion on the reasons and the impacts of a higher smoothing effect in the BIGSTACKs compared to LR04 (lines 402-404) would be helpful. Indeed, I have the impression that the five BIGSTACKs are highly smoothed compared to LR04 (see https://yz3062.github.io/htmls/BIGSTACK_bokeh.html). I already found LR04 quite smoothed when using it to derive age models of sediment cores for the last glacial cycles. I anticipate the high smoothing of BIGSTACKs may be a major drawback when using them as references to produce marine chronologies. I am wondering to which extent this smoothing could be minimized during the stacking approach and would expect this smoothing to be better discussed in the text, in particular highlighting this potential limitation in scientific applications.
4. I am not convinced by the temporal lag of 1 ka that is applied to derive the Pacific stack compared to the Atlantic one. Even though this issue is already discussed in lines 466-477, I am wondering whether this approach is the most robust. As already mentioned in the text (line 468), I think the Pacific-Atlantic lag is larger than 1 ka during deglaciations. Have the authors considered using a larger Pacific-Atlantic lag during deglaciations, instead of or in addition to a constant throughout the investigated time period?

Technical comments

Line 17: The reference Ahn et al. 2017 is missing in the reference list.

Lines 68-70: This section describes the “strengths and weaknesses” of chronological approaches. For radiocarbon dating, I would mention in these lines the difficulty to correctly estimate past marine age reservoir values.

Line 153: please clarify what is meant with “multiproxy age models”. Do the authors mean age models that use several proxy-records from the same core?

Lines 227-230: I think this information should arrive earlier in the text.

Line 305: I think the tie-points that have been defined to manually align BIGMACSmagrev to the different targets to produce the mixed BIGSTACK should be provided in the supplementary data files. It may already be the case, but I did not find them to check.

Figure 5: please add the colour code in the figure’s caption so that the reader knows which colored curve corresponds to which stacks.

Figures 6, 7, 9, 10: please add the name of the stacks directly onto the figure (e.g. along the Y-axes or next the record).

Figure 8: similarly please indicate directly onto the figure which age difference is plotted in panels a, b and c.

Supplementary Data file 10: it would be helpful to specify the content of this supplementary data. To which stacking step or chronology do ages given in this excel file correspond for all cores?

This manuscript presents an ambitious and valuable update to the venerable LR04 stack by compiling a much larger dataset and introducing new regional (Atlantic and Pacific) stacks. The study’s goals are highly relevant for Quaternary paleoclimate and stratigraphy. The paper is generally well written and organized, and the figures are informative. In its current form, however, several aspects need clarification or further development to maximize the manuscript’s clarity, transparency, and utility. In particular, the methods and rationale for certain chronological choices should be explained in more detail, comparisons to previous work should be expanded, and some figures and data descriptions require improvements for clarity. I recommend major revisions to address the issues raised above before the manuscript is accepted for publication in EGUsphere/Geochronology. 



Comments:

1. The manuscript would benefit from a clearer description of the cores and data that went into each stack. Currently, information on the 224 cores (sources, locations, water depths, new vs. previously published records, resolution, etc.) is scattered or only in the supplement. I recommend adding a summary table or a dedicated subsection in Methods describing the core selection criteria and dataset characteristics.

2. Given the prominence of LR04 and ProbStack, the manuscript should more fully compare the new stacks to these earlier benchmarks. The text does note that the new global stack includes ~50% more data than ProbStack, but there is little quantitative or visual comparison in the paper. I recommend adding some discussion (and possibly a figure or table) comparing the new global stack in terms of mean trends, variability, and age offsets.

3. Can anything be done to retain more signal variance? While a smooth stack yields a high signal-to-noise ratio for orbital-scale trends, the manuscript should caution that some climatic signals (e.g. abrupt events, smaller excursions) might be blunted. I encourage the authors to highlight this limitation so that users of BIGSTACK understand the potential need to cross-check against individual records for fine-scale events.

4. Figs. 5–7: Please add clear labels or legends identifying each colored curve.