General comments

The study by Stevenard et al. presents the first high-resolution synthesis of sea surface temperatures (SST) across the final portion of Marine Isotope Stage 10 (MIS 10), Termination IV and MIS 9, incorporating 98 carefully recalibrated and re-dated marine records. By applying a consistent chronological framework (AICC2023), updated SST calibrations, and a Monte Carlo approach to quantify uncertainties from both dating and SST reconstructions, the authors generate global and regional SST stacks that reveal multi-millennial climate dynamics and discuss their potential drivers.

This work is highly relevant in the context of understanding warm climate intervals, particularly as MIS 9e is associated with the highest atmospheric CO₂ and CH₄ concentrations of the past 800 kyr. The study is of significant value to the paleoclimate community, as these SST stacks provide a basis for evaluating the global climate response to natural forcings (global and regional perspectives) and allow meaningful comparisons with other interglacial periods. Furthermore, the authors are transparent about the limitations of their approach and provide thoughtful perspectives on how future work can build upon these findings.

Importantly, the study offers refined estimates of deglacial mean surface temperature (DGMST) for the target period, based on a larger and higher-resolution dataset than previous compilations.

The manuscript is scientifically sound, methodologically robust, and well-structured. Overall, I find the study to be a valuable contribution and worthy of publication following minor revisions as outlined below.

Specific comments

1. Abstract: homogenize the use of “MIS 9” and “MIS 9e” throughout the abstract for consistency.
2. The introduction could benefit from improved flow and narrative cohesion. Consider reorganizing some paragraphs or refining transitions between concepts to enhance clarity and better guide the reader through the motivation, context, and objectives of the study.
3. Lines 76–79: Please provide references for the mentioned sea surface temperature proxies during the target interval.
4. Table S1: Consider providing more detailed information about the proxies used to estimate SST. For example:

• Specify the foraminifera species used in the Mg/Ca and δ¹⁸Oₚ records to clarify whether the signal reflects seasonal or annual conditions.
• Consider also indicating the ocean basin or region from which each record originates, even though latitude and longitude are already provided.

1. Lines 101-103: a more detailed description of the used proxies is needed in the main text. For example:

• How each proxy works
• The type of signal it reflects (seasonal vs. annual)
• Why proxies with seasonal and annual signals are considered, comparable and suitable for joint inclusion in the compilation

1. Line 525: The title of Section 4.1 could better reflect its content by including a reference to Termination IV and MIS 10. For example:
   
   “Limitations associated with our MIS 10–TIV–MIS 9 spatio-temporal SST synthesis” or simply “Limitations associated with our spatio-temporal SST synthesis.”
   
   This logic could also be applied to the titles of other manuscript sections for improved clarity.
2. Lines 604–608: You describe delayed SST warming in the North Pacific (relative to the North Atlantic and Northern Hemisphere stacks) and compare it to Asian monsoon and insolation records. However, as with the Atlantic, the South Pacific appears to have warmed earlier than the North Pacific. For the Atlantic, you suggest this interhemispheric difference may be linked to intensified Agulhas Leakage during deglaciation. A brief discussion of possible mechanisms explaining the earlier South Pacific warming is currently missing.
3. Line 632: Regarding “(~333.5 ka)”: consider providing an age interval for the hemispheric heat transfer plateau, rather than a single date. Alternatively, clarify that ~333.5 ka marks the onset of the plateau.
4. Line 638: It seems that the most appropriate term in this context is Heinrich Stadial. Since Heinrich Stadial (HS) and Heinrich Event (HE) have distinct definitions, please double-check the use of the term “HE” here and throughout the manuscript to ensure consistency and accuracy.
5. Line 642: The phrase "In contrast, Pacific DSST changes are less abrupt" is vague. Less abrupt than what? The North Atlantic DSST? The Atlantic as a whole? Please specify.
6. Lines 656–657: A reference to Figure 6H would be appropriate here.
7. Lines 688–689: “Indeed, a slowdown of AMOC may have resulted in increased heat storage” — please clarify: increased heat storage where? In the Southern Ocean? In the deep ocean?

Technical corrections

1. When providing time intervals, please list the older date first (i.e., from past to present). For example, use “335 to 300 ka” instead of “300 to 335 ka”.
2. Line 12: Correct to “…from the fact that it is associated…”
3. Line 14: Correct to “…of sea surface temperatures (SST) are available…”
4. Line 41: Correct to “sometimes exhibiting conditions as warm as…”
5. Line 193: Typo — correct to “planktic foraminifera”.
6. Line 200: Typo — correct to “Three scenarios”.
7. Line 447: Typo — “(i.e. when the global temperature starts to rise sharply)”.
8. Figure 5D: The behaviour of the tropical stack during MIS 10 is not clearly visible. Please also check the other figures.
9. Lines 575–576: Typo — “characterized by a ~2 ka lead in the SH compared to the NH”.
10. Lines 598 and 603: Standardize the formatting of “Heinrich event” throughout the manuscript (choose either “Heinrich Event” or “Heinrich event” and apply consistently).
11. Figure 6: Double-check the y-axis title for panel D.
12. Line 615: Update citation to refer to Figure 6D instead of 6E.
13. Standardize the terminology for “greenhouse gases” throughout the manuscript (e.g., use consistently: “GreenHouse gases”, “GHG”, or “greenhouse gas”).

General comments

Overall, this is a well-written and thoughtful paper. Below, I briefly address the criteria outlined by the journal: the manuscript tackles relevant scientific questions within the scope of the journal and presents novel approaches that contribute to the field. The authors reach substantial conclusions, which are generally supported by the results. However, the scientific methods and underlying assumptions require further clarification, particularly regarding experimental procedures and calculations, to ensure reproducibility. The authors appropriately acknowledge related work and clearly delineate their original contributions. The title and abstract accurately reflect the content of the paper. While the language is mostly fluent and precise, the structure of the paper could be improved for clarity. Technical terminology and mathematical expressions are correctly used and defined. Certain sections of the text, figures, and tables would benefit from clarification or streamlining, as detailed below. The references are sufficient and appropriate, and the supplementary materials are well-prepared.

I particularly enjoyed reading the discussion and would like to congratulate the authors on this aspect of the manuscript. I also appreciate the efforts made to account for uncertainties in both the proxies and the age model. However, I found the methods and results sections somewhat difficult to follow. As outlined below, my main concerns relate to the following aspects:

• Age tie-point assignment: How were these determined? Were peaks and troughs used? Inflection points would be preferable, as they are less sensitive to noise and aliasing.
• Age references: Different age references appear to have been used for visual alignment and age model revision. Are there systematic differences between records based on the LR04 stack versus the Greenland temperature record?
• Definition of pre-industrial (PI): The definition is unclear, particularly for seasonally biased proxies. If annual means are used for seasonal proxies, the resulting anomalies could be misleading.
• Results presentation and description: In several instances, I interpreted the figures differently than described. To aid clarity, it would be helpful to outline at the beginning of the Results section the specific aspects being examined, and focus the result description on those. The manuscript is already quite long, and much of the detailed description of individual / regional patterns is not discussed later. I suggest moving the current sections 3.1 to the Supplementary Information and replacing them with a more focused summary of the main results that are subsequently discussed in section 4.

I hope these comments are constructive and helpful in revising the manuscript.

Specific comments

Line 12: “…the fact that is associated…” — It appears that “it” is missing; please revise for clarity.

Line 14–15: The sentence “…it is challenging to assess the regional and global patterns of climate variability…” is quite long and requires several readings to grasp the context. Consider rephrasing for clarity and conciseness.

Line 78: “magnesium and calcium” should be revised to “magnesium-to-calcium.”

Line 79: “Tex” should be capitalized to “TEX.” Please ensure consistent use throughout the manuscript.

Line 79–83: “old SST data” — Suggest rephrasing to “existing SST data” to better reflect that the key issue is differences in calibration methods rather than data age.

Line 95–98: It is unclear whether the 98 records cited are those included after or before applying the selection criteria. Please clarify this point.

Line 98–99: “propose to include SST” — Suggest revising to “decided to include SST,” since this reflects the authors’ actual methodological decision.

Line 103–105: The sentence is somewhat confusing, implying TEX86 reconstructs SST but is associated with subsurface temperatures. Consider rephrasing.  

Additionally, since it is well established that TEX86 can reflect subsurface conditions, please consider citing earlier studies that first demonstrated this, rather than more recent ones.

Line 112–113: The connection between the cited studies and the “efficiency” of Bayesian calibrations is unclear. Consider rephrasing to state that these studies applied the Bayesian calibrations, rather than demonstrated their efficiency.

Line 116–117: Seasonal production is not generally cited as a reason for UK′₃₇ saturation. Please revise or clarify this claim.

Line 121–124: The Mg/Ca–temperature relationship is not linear. Please revise accordingly.

Line 133: “dimensions of about N × 1000” — Some clarification is needed on the variability of this number. Why not use N × 1000 as for other proxies? Is it a small variation (e.g., 999–1001) or more substantial (e.g., 900–1100)?

Line 136: Consider using Breitkreuz et al. (2018, JGR Oceans) instead of LeGrande and Schmidt (2006), as it is more recent and may be more appropriate here.

Line 137–139: More detail is needed about the ice volume correction — specifically, how large is it?

Line 139: “other published SST records” — Does this refer to geographically nearby records? If so, please define “nearby” more precisely.

Section 2.3: This section is somewhat difficult to follow and warrants improvement, especially given its importance for anomaly calculations.

• How were the 1000 proxy values derived? Were HadISST values used as inputs to the Bayesian calibrations, and if so, how were these converted back to SSTs?
• What pre-industrial (PI) references are used for seasonal proxies? If annual mean SST is used for proxies with seasonal production, this may introduce bias to the anomaly calculation.
• Since many Pleistocene records lack modern/PI sediments, an alternative would be to use published core-top proxy values from the same grid cell (e.g., those from global calibration databases).

Line 160: “limit” → “limitation”

Line 160: The phrase “discrepancies of the HadISST database” is unclear. Consider clarifying what discrepancies exist within the dataset or relative to.

Line 176: “propose to use of” — Please rephrase for grammatical correctness.

Line 178: “to account the ‘bipolar seesaw’” — Consider rephrasing for clarity, e.g., “to account for the effects of the bipolar seesaw.”

Line 193: Typo — “planctik” should be corrected.

Line 196–198: This is a critical methodological step and needs more detail:

Please clarify the criteria used to select age tie-points. Based on the supplementary figures, it appears that peaks and troughs were used. However, it is more common in the community to use inflection points, as they are less susceptible to aliasing and more robust against noise, particularly in lower-resolution records.

The authors wrote that uncertainty estimates were visually assigned but it is not immediately clear to the reader how. I suggest indicating these uncertainties in the supplementary figures and including them in a table in the SI.

Line 199: Please clarify how “basin references” are defined / selected.

Line 204–205: See comment above on Line 196–198.

Line 233–234: It’s not necessarily a drawback that high-resolution records contribute more to the stack. One might argue that such weighting improves the robustness of the stack. Consider clarifying this point or adjusting the framing.

Line 237: “inspired from” should be revised to “inspired by.”

Line 240–241: The phrasing “random redrawn” and “different age-SST” is awkward. Please consider rephrasing for clarity.

Line 243–245: This description is difficult to follow. Does “random grid mean” refer to the averaging of values within a grid cell or the shifting of grid boundaries? The rationale for randomly adjusting the latitudinal mean is unclear. Also, the 20° range seems quite large and may blur distinctions between tropical and extratropical settings. Please clarify.

Section 3.1: This section is lengthy and contains descriptive detail that is not followed up in later sections. I recommend summarizing key observations and moving some site- and region-specific details to the Supplementary Information. Additionally, please clarify:

• How is glacial maximum defined? Lowest temperature within MIS 10?
• How is the beginning of deglacial warming defined — change-point analysis, or visual inspection?
• What is the distinction between “peak” and “plateau” in terms of their duration?
• For marginal seas, some description glosses over differences between proxies and records. Consider explicitly stating which proxies/records show which patterns.

Line 265–267: These records are based on different proxies, which could explain the discrepancies.

Line 281: MD97-2141 is located in the western Pacific, not in the South China Sea.

Line 290: The definition of “glacial maximum temperature” is unclear; this made it difficult to verify the interpretation based on the figure.

Line 333: Consider marking the start date in the figure to assist the reader.

Line 342–343: This sentence is unclear; please revise.

Line 348: Site MD02-2575 does not show large glacial–interglacial oscillations.

Line 351: The δ¹⁸Op record does not appear to show an abrupt deglacial warming. Please revise accordingly.

Line 358–360: Same as above — the patterns described are not apparent in the δ¹⁸Op record.

Line 361–365: This highlights the need for a clearly stated method/criteria for visual assessment. For PRGL-1, δ¹⁸Op shows decreasing temperatures during deglaciation, in contrast with the UK′₃₇ record. If these are considered consistent, why are similar discrepancies in the Caribbean interpreted as strong disagreements?

Line 372: The MAT record at ODP-1089 is incomplete and does not clearly show deglacial warming.

Line 373 & 395: While I agree that ODP-1084 shows a plateau, to my eyes MD96-2094 appears to show a peak, more similar to MD97-2120.

Please consider defining “plateau” and “peak” at the beginning of the section.

Line 449: Do the stated uncertainties correspond to the shaded envelope in Figure 5c? They appear slightly different.

Line 450–455: It was difficult to identify the patterns described in Figure 5. Consider marking key transitions in the figure.

Line 453: Is the resolution sufficient to detect such a small change over 2 kyr intervals?

Line 487–489: This depends on how “deglacial warming” is defined. If it’s defined as a significant shift, there would not be multiple discrete steps.

Line 500: It would be helpful to include in the SI a figure showing all records used to construct the basin/hemisphere stacks.

Line 509–510: Please indicate where the σ uncertainty values can be found and how they were calculated.

Line 533–535: Have the authors considered using the Breitkreuz et al. (2018) δ¹⁸Osw dataset? The described unexpected temporal evolution raises the question of whether including δ¹⁸Op records truly adds value to the compilation beyond increasing record count.

As a test, the authors could create stacks with and without δ¹⁸Op to assess their impact on the results.

Line 535–538: Consider sensitivity tests using the Gray and Evans (2018) multivariate calibration to explore the influence of non-thermal factors.

Line 543–544: Please show the effect of including these records in the SI. Including problematic records may weaken the integrity of the compilation — this tradeoff merits discussion.

Line 545–546: The PI reference remains unclear. An alternative would be to compare with climatological products like the World Ocean Atlas.

Line 568–569: This sentence is confusing. Do the authors mean basin-scale temperature trends should be considered independently from global averages? Please clarify.

Line 582–583: Please elaborate — what is the benchmark, and what makes this anomaly stand out?

Line 585: The reasoning is unclear. Is the same mechanism present in other basins? Is the SH stack dominated by Atlantic records?

Line 602: Please specify the latitude and month used for “solstice insolation.”

Line 615: Should this refer to Figure 6D? Please check.

Line 639: Consider rephrasing “redistributing surface temperature” to “redistributing heat.”

Line 677: Typo — “global temperature similar” should be revised.

Line 682–683: Are Holocene anomalies directly comparable to MIS 9e results? Were all anomalies calculated using the same reference period?

Line 741: Why? Is it due to differences in the spatial coverage of these compilations? Please clarify.

Line 769–771: This sentence is unclear. Does the “last 60 years” refer to the Osman stack or HadCRUT5? Please clarify.

Figures

Figures 3 & 4: Please indicate whether sites are located in the east, south, or marginal zones of each basin.

Figure 5: Please mark MIS 10.

Figure 6D: Typo in y-axis label — “transfert” should be “transfer.”