The authors provide a nice western Pacific warm pool planktonic foraminifera oxygen isotope during the past 800 ka with novel stack method. This is important and timely for the further paleoceanography studies in this region. There are just few minor points I would like to ask the authors to address.  

Lack of references in the introduction section

In the introduction section, the authors largely rely on only very few references (Lea et al., 2000 and Tachikawa et al., 2014) to present some basic description of western Pacific warm pool region. Although these 2 references are crucial and important, however, these are definitely not modern physical oceanography papers. If the authors would like to make the reference list be more concise, then perhaps the authors could add “and references therein” to clarify is not just these 2 references for the whole previous physical oceanography studies in the warm pool region. Or the better way, to cite some physical oceanography observational papers in the first 2 paragraphs of introduction section.

Figure 1: what is the software to make this figure? The authors only mentioned the reference for SST dataset.

Line 78-80, should be “30 m” and “20-75 m”; Also, although the original reference from Chuang et al. (2018) used “G. sacculifer”. The genus name has been revised as “Trilobatus sacculifer; T. sacculifer”.

Table 1: I wonder why the authors did not include Medina-Elizalde et al. (2005)’s ODP 806 data?

Also, “180-1115B” data should be “ODP 1115B”, please revise it through the text, figure 1, and Tables.

Lastly, Lo et al. (2017) only report data back to 350 ka, but the dataset of MD05-2925 here is back to ~462(?) ka. Please clarify.

Lines 128-130, please describe why the authors would like to set the reservoir age as zero?

Line 150, should be “29oC”

Lines 186-187, if the authors take the uncertainty into account would the 1.3 permil significantly different from 1.7-1.8 permil glacial/interglacial changes? Please clarify.

Section 6.1, perhaps the authors could also refer to cores used in this study with both benthic and planktonic foraminifera d18O stratigraphy. For example, MD05-2925, Lo et al. (2017) several other reports in this core have reported there is no clear timing differences for the past 5-6 terminations (Liu et al., 2015, Lo et al., 2022).

The title of Figure 7 is not clear, what kind of “contributions” that the authors would like to address in this figure?

Bowman et al. 2003

An 800-kyr planktonic δ¹⁸O stack for the West Pacific Warm Pool

ESSD-2023-335

Review

Manfred Mudelsee

21 September 2023

The concept of the presented manuscript is fine: (A) stack construction for the WPWP, (B) usage of Bayesian age model algorithms, and (C) comparison of stack with other records (e.g., LR04) in terms of variability and spectral properties. However, there are flaws with data analysis and an amount of minor errors that render the current manuscript not publishable. My advice is to give authors enough time to re-submit a manuscript that overcomes data-analytical flaws and minor errors.

Data analysis

(1) Table 1

The MD97-2141 record stands out against the others in terms of temporal resolution. The manuscript should inform readers that indeed the 0.33-kyr resolution (Oppo et al. 2003) is a reasonable value since that record has a high sedimentation rate (5 to 15 cm/kyr) and also a fine sampling (1 cm). It would further be informative to study the effects ex- or inclusion of that high-resolution record on results (variability and spectra). This can be done by repeating stack construction and variability and spectrum estimation without MD97-2141. Of course keep MD97-2141 for calculation of the final stack.

(2) Output resolution of WPWP stack (end of Section 4.1; Bowman et al. 2023)

The stack has 8101 data points covering the interval from 0 to 810 kyr at a constant temporal resolution of 0.1 kyr. While it is fine to present such a stack for visualization purposes, it is not OK to use it for variability or spectrum estimation because the 0.1 kyr resolution is smaller (by a factor ranging from 3.3 for core MD97-2141 up to nearly 40 for core MD97-2140). This boost-up of the sample size may lead to significantly overstated claimed statistical uncertainties (for variability or spectrum estimation). One analysis strategy to assess the effect would be to repeat variability and spectrum estimation for various other, coarser stack resolutions (say, from 0.1 kyr up to 1.0 kyr in steps of 0.1 kyr). Such a sensitivity study could make an interesting appendix for other researchers wishing to study time-resolution effects.

(3) Spectrum estimation (Section 4.3)

Usage of FFT is obsolete since it renders bad (in terms of estimation bias, variance, RMSE, etc.) estimates. This is known for decades (Thomson 1982, Percival and Walden 1993, Mudelsee, 2014). And since the stack is evenly spaced (at 0.1 kyr or up to 1.0 kyr resolution), one needs not invoke the Lomb-Scargle Fourier Transform (Schulz and Mudelsee 2002) but can work with Thomson’s multitaper estimation (MTM), which is the method of choice here. See again the mentioned works (Thomson 1982, Percival and Walden 1993, Mudelsee, 2014) and literature cited therein. Mudelsee (2014) lists also software tools for MTM estimation in case there is need for the authors.

(4) Uncertainty presentation of stack

Time-varying standard deviation is certainly interesting, but I think that more about stack uncertainty can be learned from calculation of internal and external errors (and hence use weighting for stack calculation). Internal errors refer to individual records, while external errors measure the spread among the various contributing errors. Individual records with smaller uncertainties should heavier contribute to the stack. Of course the challenge is to do justice to the fact that the number of records available depend on the investigated age. Details about weighting, internal and external errors can be found in the paper by Mudelsee et al. (2014), who constructed a Cenozoic δ¹⁸O stack.

Minor errors

I refer only to Abstract and References since already there appeared quite a number.

Abstract, l. 1

The expression “different ... than” may sound strange to British ears.

Abstract, l. 2

Write “greenhouse gas concentrations”.

Abstract, l. 3

Insert a hyphen: “orbital-scale climate response”.

Abstract, l. 6

Two commas inserted makes it more readable: “... and benthic δ¹⁸O stacks, also constructed using BIGMACS, demonstrate that ...”.

Abstract, l. 7

Insert a bit information: “Sixty-seven radiocarbon dates from the upper parts of four of the WPWP cores ...”.

Abstract, l. 11

The expression “0 - 450 ka” (with a hyphen) looks ugly. Either use an en-dash without spaces or else write “0 to 450 ka”.

References

(1) Do not capitalize (headline style) titles of listed journal articles (e.g. Huybers & Wunsch 2004 “Uncertainty estimates” ... and not “Uncertainty Estimates”).

(2) Do properly use superscripts (e.g., Imbrie et al. 1984 “ ... revised chronology of the marine δ¹⁸O ...”, and not “δ18O”; Lee et al. 2022 wrong “d¹⁸O”, Lisiecki and Raymo 2005 wrong “δ18O”).

(3) Do give editor names for cited chapters from edited books (e.g., Imbrie et al. 1984).

References cited in review

Mudelsee M (2014) Climate Time Series Analysis: Classical Statistical and Bootstrap Methods. Second edition, Springer, Cham, Switzerland, 454 pp.

Mudelsee M, Bickert T, Lear CH, Lohmann G (2014) Cenozoic climate changes: A review based on time series analysis of marine benthic δ¹⁸O records. Reviews of Geophysics 52:333—374.

Percival DB, Walden AT (1993) Spectral Analysis for Physical Applications: Multitaper and Conventional Univariate Techniques. Cambridge University Press, Cambridge, 583 pp.

Schulz M, Mudelsee M (2002) REDFIT: Estimating red-noise spectra directly from unevenly spaced paleoclimatic time series. Computers and Geosciences 28:421—426.

Thomson DJ (1982) Spectrum estimation and harmonic analysis. Proceedings of the IEEE 70:1055—1096.

Bowman et al. 2003

An 800-kyr planktonic δ¹⁸O stack for the West Pacific Warm Pool

ESSD-2023-335

Review

Manfred Mudelsee

21 September 2023

The concept of the presented manuscript is fine: (A) stack construction for the WPWP, (B) usage of Bayesian age model algorithms, and (C) comparison of stack with other records (e.g., LR04) in terms of variability and spectral properties. However, there are flaws with data analysis and an amount of minor errors that render the current manuscript not publishable. My advice is to give authors enough time to re-submit a manuscript that overcomes data-analytical flaws and minor errors.

Data analysis

(1) Table 1

The MD97-2141 record stands out against the others in terms of temporal resolution. The manuscript should inform readers that indeed the 0.33-kyr resolution (Oppo et al. 2003) is a reasonable value since that record has a high sedimentation rate (5 to 15 cm/kyr) and also a fine sampling (1 cm). It would further be informative to study the effects of ex- or inclusion of that high-resolution record on results (variability and spectra). This can be done by repeating stack construction and variability and spectrum estimation without MD97-2141. Of course, keep MD97-2141 for calculation of the final stack.

(2) Output resolution of WPWP stack (end of Section 4.1; Bowman et al. 2023)

The stack has 8101 data points covering the interval from 0 to 810 kyr at a constant temporal resolution of 0.1 kyr. While it is fine to present such a stack for visualization purposes, it is not OK to use it for variability or spectrum estimation because the 0.1 kyr resolution is smaller than the individual resolutions (by a factor ranging from 3.3 for core MD97-2141 up to nearly 40 for core MD97-2140). This boost-up of the sample size may lead to significantly overstated claimed statistical uncertainties (for variability or spectrum estimation). One analysis strategy to assess the effect would be to repeat variability and spectrum estimation for various other, coarser stack resolutions (say, from 0.1 kyr up to 1.0 kyr in steps of 0.1 kyr). Such a sensitivity study could make an interesting appendix for other researchers wishing to study time-resolution effects.

(3) Spectrum estimation (Section 4.3)

Usage of FFT is obsolete since it renders bad (in terms of estimation bias, variance, RMSE, etc.) estimates. This is known for decades (Thomson 1982, Percival and Walden 1993, Mudelsee, 2014). And since the stack is evenly spaced (at 0.1 kyr or up to 1.0 kyr resolution), one needs not invoke the Lomb-Scargle Fourier Transform (Schulz and Mudelsee 2002) but can work with Thomson’s multitaper estimation (MTM), which is the method of choice here. See again the mentioned works (Thomson 1982, Percival and Walden 1993, Mudelsee, 2014) and literature cited therein. Mudelsee (2014) lists also software tools for MTM estimation in case there is need for the authors.

(4) Uncertainty presentation of stack

Time-varying standard deviation is certainly interesting, but I think that more about stack uncertainty can be learned from calculation of internal and external errors (and hence use weighting for stack calculation). Internal errors refer to individual records, while external errors measure the spread among the various contributing records. Individual records with smaller uncertainties should, hence, stronger contribute to the stack. Of course the challenge is to do justice to the fact that the number of records available depend on the investigated age. Details about weighting, internal and external errors can be found in the paper by Mudelsee et al. (2014), who constructed a Cenozoic δ¹⁸O stack.

Minor errors

I refer only to Abstract and References since already there appeared quite a number.

Abstract, l. 1

The expression “different ... than” may sound strange to British ears.

Abstract, l. 2

Write “greenhouse gas concentrations”.

Abstract, l. 3

Insert a hyphen: “orbital-scale climate response”.

Abstract, l. 6

Two commas inserted makes it more readable: “... and benthic δ¹⁸O stacks, also constructed using BIGMACS, demonstrate that ...”.

Abstract, l. 7

Insert a bit information: “Sixty-seven radiocarbon dates from the upper parts of four of the WPWP cores ...”.

Abstract, l. 11

The expression “0 - 450 ka” (with a hyphen) looks ugly. Either use an en-dash without spaces or else write “0 to 450 ka”.

References

(1) Do not capitalize (headline style) titles of listed journal articles (e.g. Huybers & Wunsch 2004 “Uncertainty estimates” ... and not “Uncertainty Estimates”).

(2) Do properly use superscripts (e.g., Imbrie et al. 1984 “ ... revised chronology of the marine δ¹⁸O ...”, and not “δ18O”; Lee et al. 2022 wrong “d¹⁸O”, Lisiecki and Raymo 2005 wrong “δ18O”).

(3) Do give editor names for cited chapters from edited books (e.g., Imbrie et al. 1984).

References cited in review

Mudelsee M (2014) Climate Time Series Analysis: Classical Statistical and Bootstrap Methods. Second edition, Springer, Cham, Switzerland, 454 pp.

Mudelsee M, Bickert T, Lear CH, Lohmann G (2014) Cenozoic climate changes: A review based on time series analysis of marine benthic δ¹⁸O records. Reviews of Geophysics 52:333—374.

Percival DB, Walden AT (1993) Spectral Analysis for Physical Applications: Multitaper and Conventional Univariate Techniques. Cambridge University Press, Cambridge, 583 pp.

Schulz M, Mudelsee M (2002) REDFIT: Estimating red-noise spectra directly from unevenly spaced paleoclimatic time series. Computers and Geosciences 28:421—426.

Thomson DJ (1982) Spectrum estimation and harmonic analysis. Proceedings of the IEEE 70:1055—1096.

Dear Authors, Dear Editor,

As stated in the introduction of the manuscript recent studies have advocated the development of regional 𝛿¹⁸O stacks to distinguish spatial differences in the timing and amplitude of 𝛿¹⁸O signals. Hence the contribution is timely and has the potential to provide a useful benchmark record in the Quaternary research. The manuscript is principally well-written, however there are two points in the methodology and some other specific ones which need revision before the study can be accepted for publication.

General comments:

1. It is not clear how was the MD97-2141 data resampled (lines 90-91): I note at this point that binning rather than smoothing and resampling would be a more adequate data processing to reduce the resolution of this record. In addition, the chosen 0.33 kyr mean sample spacing is still much finer compared to the cores from the WPWP (according to Table 1 the mean sample spacing of those cores ranges from 0.76 to 3.9 kyr). Binning to ~1ka might be more suitable to get close to the median resolution of the records representing the core area of the WPWP.
2. According to my understanding the study applied a reservoir age offset (R) as 0+/-0.2 kyr (lines 129-130). The appropriateness of this R is debatable. I suggest checking Sarnthein et al., 2015 (DOI: https://doi.org/10.2458/azu_rc.57.17916 ). In particular, MD01-2378 was scrutinized in the study. Based on the inferred planktic reservoir ages typically >200yrs and >1kyrs during LGM (and probably in glacial conditions in general).

Specific comments:

line 3: perhaps “greenhouse forcing” instead of “greenhouse gas”

line 4: perhaps “covering the…” instead of “of the…”

lines 24 to 26: Despite these are almost common knowledge some references can be needed. e.g. Wefer and Berger 1991 (https://doi.org/10.1016/0025-3227(91)90234-U) could be a pertinent reference.

line 46, 49, 226, 234, and 329: Please correct and update the citation: “(Lee and Rand et al., accepted)”

line 50: “between 0-43 kyr BP” sounds strange

line 76: the sentence sounds strange. I suggest rephrasing as follows: “Six of the cores span the last 300 to 500 kyrs, and five extend back to 750 ka.”

line 84: Please change to “from 450 to 800 ka”

line 88: Ditto. Please change to “from 0.33 to 3.9 kyr”

lines 149-150: The sentence is somehow repetitive. Please rephrase it.

line 185: The sentence needs grammar checking.

line 212: Please change to “between 36 and 38 ka”

line 214: Ditto. Please change to “between 30 and 40 ka”

lines 235-236: I suggest replacing “our WPWP…” with “the new WPWP…”

Within their manuscript: “An 800-kyr planktonic δ¹⁸O stack for the West Pacific Warm Pool” Christen Bowman et al. present a regional planktonic δ¹⁸O stack record that is created upon the basis of previously published δ¹⁸O records from the area by application of a novel dating and stacking software tool. The dataset might be useful for paleoceanographers in the future. Hence, I generally support its publication in ESSD, but find that the article has some flaws that should be remedied prior to publication. My concerns are outlined in more details below. The authors should also pay attention to a careful and precise wording/phrasing throughout the manuscript.

General comments:

• Choice of records: I wonder, if the records chosen to be included in the stack are representative for the WPWP. There are apparently many more regional planktonic (G. ruber) δ¹⁸O records from the WPWP available, which are not considered in the stack record. The authors do however include two records from the Timor Sea. Strictly speaking, this is not part of the WPWP. What are the selection criteria to include / exclude records within / from the stack? The criteria should be stated clearly in the text.
• Related to my previous point, I have another comment: Throughout the manuscript the authors refer to the “WPWP”. I am wondering if it is reasonable here, because the authors include two core sites from the Timor Sea (SO18480-3, MD01-2378). Wouldn’t it be more precise to refer to the Indo-Pacific Warm Pool (IPWP)? I however note that additional records from the tropical eastern Indian Ocean might be needed to cover the entire IPWP. If the authors use WPWP, shouldn’t it be “Western Pacific Warm Pool” instead of “West Pacific Warm Pool” throughout the manuscript?
• Section 3 – Data: There are radiocarbon dates of cores KX21-2 (Dang et al., 2020) and MD97-2141 (Oppo et al., 2003), which should be included in this study. The KX21-2 data are presented within the original publication, the MD97-2141 data can be found here: https://www.ncei.noaa.gov/pub/data/paleo/contributions_by_author/oppo2003b/oppo2003b.txt
• Please reference the original datasets, not only the original publications. If I regard it correctly, the original records are mostly deposited online and subsequent users should be able to cite the original datasets directly.
• Comparison of the planktonic WPWP stack to the LR04 / LS16 benthic stacks: Why don’t the authors compare planktonic and benthic δ¹⁸O records of the cores they are using to create the stack to get more direct assessments of the offsets between planktonic and benthic δ¹⁸O records? Benthic δ¹⁸O records are available for at least some of the cores.
• The authors present a stack record with a temporal resolution / time steps of 0.1 kyr, although the resolution of the individual records that go into the stack ranges between 0.33 and 2.3 kyr. The stack thus feigns a higher resolution than given, which should be avoided. The question is how this affects the statistical analyses presented in the article.
• The authors align their planktonic δ¹⁸O stack to the LR04 stack. They argue that there is almost no time shift between the planktonic and benthic records. Considering this, and by looking at the two stack records in comparison (Figure 4), the question arises, why scientists should use the planktonic δ¹⁸O stack instead of the LR04 or more recent regional LS16 benthic stacks in the future. I think that should be pointed out more clearly within the article.
• If the δ¹⁸O records are shifted and scaled “to better match the target stack” (line 114), how useful is it to compare amplitudes or spectral power of the WPWP and LR04 stacks?
• The authors introduce that they “seek to characterize WPWP climate on orbital timescales and its differences from high-latitude climate, which can help test hypotheses about the sensitivity of the WPWP to orbital forcing, ice volume, and greenhouse gas concentration” (lines 20-22). Later on, they compare their regional planktonic stack to a regional benthic stack, to the LR04 stack and to a regional WPWP SST stack record. They however miss to draw conclusions from their results. What can be inferred from the comparisons? What is, for instance, the value of comparing spectral power and variability of the regional planktonic stack to the global benthic stack?

More specific comments:

• Line 9: It should be “a smaller glacial-interglacial amplitude”.
• Lines 19-20: “Thus, climate records of the WPWP region are expected to have features which differ from many other locations on Earth.” This statement is rather general and should be more precise.
• Line 25: It might appear a bit old-fashioned, but don’t foraminifers have tests instead of shells? Please correct.
• Lines 24-35: Here, I clearly miss some references.
• Line 78: There are calcification depth estimates from the study area (e.g. Hollstein et al., 2017). Why don’t the authors consider these estimates? For instance, for G. ruber, the study indicates a calcification depth of 0-75 m, rather than the 30 m indicated by Wang et al. (2000).
• Line 79: Shouldn’t it be T. sacculifer?
• Table 1: The reference of Chuang et al. (2018) is missing in the Reference list.
• Line 119: The sentence is not complete.
• Lines 119-121: A short explanation for the shift and scale parameters might be helpful.
• Lines 129-130: The authors do not apply local reservoir age corrections and assume a reservoir age standard deviation of 0.2 kyr. What is the rationale behind these assumptions?
• Lines 135-139: If I understand it correctly, the authors choose a standard deviation of 1 kyr for the alignment of the records to MIS 3 and 4, but a standard deviation of 4 kyr for the first and last δ¹⁸O measurements of each core. I think, this needs some explanation.
• Line 150: “Present day”: Here the authors could be more precise, and indicate the year they are referring to.
• Section 4.3: Please indicate, which software / tool was used to perform spectral analysis.
• Line 185: “weaker the glacial-interglacial amplitude”. Delete “the”.
• Line 254: “most negative”. Please rephrase.
• Line 261: “where glacial surface water δ¹⁸Osw was more positive at the LGM”. Please rephrase. I assume that the authors want to express that glacial δ¹⁸Osw was higher.
• Line 262: Which sites?
• Line 333: It should be “which has a higher resolution”.
• Tables 1 and 2: Could the authors maybe sort the entries of these tables?

References

Dang, H., Wu, J., Xiong, Z., Qiao, P., & Li, T. (2020). Orbital and sea-level changes regulate the iron-associated sediment supplies from Papua New Guinea to the equatorial Pacific. Quaternary Science Reviews, 239. https://doi.org/10.1016/j.quascirev.2020.106361.

Hollstein, M., Mohtadi, M., Rosenthal, Y., Moffa Sanchez, P., Oppo, D., Martínez-Méndez, G., . . . Hebbeln, D. (2017). Stable oxygen isotopes and Mg/Ca in planktic foraminifera from modern surface sediments of the Western Pacific Warm Pool: Implications for thermocline reconstructions. Paleoceanography. https://doi.org/10.1002/2017PA003122.

Oppo, D. W., Linsley, B. K., Rosenthal, Y., Dannenmann, S., & Beaufort, L. (2003). Orbital and suborbital climate variability in the Sulu Sea, western tropical Pacific. Geochemistry, Geophysics, Geosystems, 4(1), 1-20. https://doi.org/10.1029/2001gc000260.