Correcting a systematic bias in an ocean drilling project site 882 alkenone sea surface temperature record

Novak, Joseph B.; Caballero-Gill, Rocío P.; Herbert, Timothy D.; Dowsett, Harry J.; Martínez-García, Alfredo

doi:10.5194/cp-22-1277-2026

Articles | Volume 22, issue 7

https://doi.org/10.5194/cp-22-1277-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/cp-22-1277-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 22, issue 7

Research article

|

08 Jul 2026

Research article |

| 08 Jul 2026

Correcting a systematic bias in an ocean drilling project site 882 alkenone sea surface temperature record

Joseph B. Novak, Rocío P. Caballero-Gill, Timothy D. Herbert, Harry J. Dowsett, and Alfredo Martínez-García

Download

Final revised paper (published on 08 Jul 2026)
Supplement to the final revised paper
Preprint (discussion started on 15 Jul 2025)

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2025-1975', Anonymous Referee #1, 25 Aug 2025

Novak and colleagues present research that corrects for biases in previously published alkenone sea surface temperature records at ODP882. These biases stem from differences between analytical techniques for quantifying alkenone concentrations (GC-FID vs GC-CI-MS). The GC-CI-MS method enables analysis of alkenone-poor sediments that were inaccessible with GC-FID. The authors demonstrate that the original GC-CI-MS ODP882 measurements overstated climate variability, though the overall patterns and trends remain unchanged.
While the manuscript is well-written and the research methodology is sound, the manuscript's critical weakness is the absence of a meaningful discussion section. The authors successfully identify and correct the proxy bias, and highlight that the ODP882 record is very important, but fail to explore what this corrected record reveals about our understanding of Pliocene climate and Northern Hemisphere glaciation—which should be a key contribution from this methodological innovation.
MAJOR ISSUE: MISSING DISCUSSION OF IMPLICATIONS
The research convincingly quantifies and corrects the ODP882 bias, but stops short of addressing why this matters for the current understanding of Pliocene temperatures and Northern Hemisphere glaciation. The authors need to address fundamental questions about their findings' significance:

How does this correction alter our understanding of Pliocene temperatures and the timing/intensity of Northern Hemisphere glaciation?
What are the implications for North Pacific temperature evolution during this critical climate transition?
How might this refined record affect estimates of Earth System Sensitivity (ESS), particularly given that ODP882 is frequently cited in multi-proxy compilations?

These questions are only meant to be illustrative and motivate deeper discussion as the current manuscript provides little context for understanding the broader implications of their correction. The paleoclimate community needs to understand not just that the record was biased, but what new insights emerge from the corrected data.

SECTION 2.1
This section could use some more text, particularly since the methods of this manuscript are intertwined with the key message: that the ODP882 record is biased and this is how you quantified and corrected for that bias. In particular, a clearer explanation of how synthetic UK37 were generated is needed for the broad paleoclimate readership of Climate of the Past.

FIGURE 4
A direct comparison between the original and corrected SST records at ODP882 is conspicuously absent. Figure 4 would be the logical place to show this comparison, allowing readers to visualize both the magnitude of the correction and its impact on key climate transitions.

MINOR COMMENT
Without demonstrating the impact of the corrected record, the manuscript somewhat overstates its importance. For instance, the abstract begins by discussing Earth climate sensitivity but many studies of Earth System Sensitivity rely on multiple records, so the authors should either: (a) demonstrate how this correction specifically affects multi-proxy compilations, or (b) focus on what unique insights about regional climate dynamics this corrected record provides.

RECOMMENDATION
This manuscript makes a valuable methodological contribution by identifying and correcting an important bias in a widely used proxy from an important site. However, it currently reads more as a technical note than a full research article. To maximize its impact, the authors must add a robust discussion section that explores what this corrected record teaches us about Pliocene climate that we didn't know before. I enjoyed reading this manuscript and hope the authors find my comments useful.

Citation: https://doi.org/10.5194/egusphere-2025-1975-RC1
- AC1: 'Reply on RC1', Joseph B. Novak, 19 Jan 2026
  
  Dear Reviewer #1,
  Thank you for taking the time to evaluate our work. We are appreciative of your comments and suggestions, which will result in an improved resubmission of our manuscript.
  Please find our detailed response to your comments attached as a .pdf.
  Warm regards,
  Joseph Novak
  
  Citation: https://doi.org/10.5194/egusphere-2025-1975-AC1
CC1:
'Comment on egusphere-2025-1975', David Naafs, 12 Nov 2025
Summary
In this manuscript Novak and co-authors use published data to propose a correction for the original alkenone-based (Uk37’) sea surface temperature data from ODP Site 882 (North Pacific) that spans the Plio- and Pleistocene and was published as part of an not public PhD thesis and (partly) in (Haug et al., 2005).
The reason for this correction is that the original alkenone data was not obtained using the established GC-FID technique, but with GC-CI-MS, which could introduce a bias. For this purpose, the manuscript presents an approach based on comparing the original GC-CI-MS data with more recently published GC-FID based data from a brief Pliocene interval for Site 882 (Studer et al., 2012) as well as published GC-FID based data from a Pliocene interval from nearby Site 883 (Novak et al., 2024). The main conclusion of this manuscript is that the original SST data from Site 882 is biased, predominantly overestimating the magnitude of SST change at Site 882, but that the main conclusions of the influential (Haug et al., 2005) paper still hold.

Main Conclusion
The fundamental basis for this paper; namely that for the brief Pliocene interval covered by both datasets (~2750-2650 ka) the comparison between the GC-CI-MS-based SST data from Site 882 (Haug et al., 2005) with the GC-FID-based SST data from Site 882 (Studer et al., 2012) the data do not fall on the 1:1 line (e.g. Figure 3), is a valid observation. The other basis that GC-MS based approaches can lead to different UK37’ and hence SSTs is also well-known (in this case I also suggest to include studies like (Hefter, 2008) into this manuscript). So there is clearly a basis that warrants a correction of the data and I appreciate the effort to correct (published) data.
            However, in my opinion the approach presented in this version of the manuscript is too simplistic and needs to be more comprehensive. As such I recommend major revisions for this manuscript.

David Naafs 11^th November 2025

Main Problems
Basis for (linear) correction for whole dataset is not well explained or supported by data

The basis for the specific linear correction applied here is that the GC-CI-MS and GC-FID data for Site 882 do not fall on a 1:1 line AND that the same holds for the GC-CI-MS data from Site 882 and GC-FID data for Site 883 (shown in figure 3). However, the justification for why a linear correlation (eq. 3) is the best option to correct the data is not explained. Other options appear not explored using statistics. This while previous work suggests that the expected bias between GC-MS and GC-FID methods could be non-linear (Hefter, 2008).
Similarly, it is not clear why SSTs are used for this correction and not the raw UK37’ indices. It is the index that is potentially biased, the SST is just a result from that biased index. And with the use of BAYSPINE, using SSTs might introduce an additional (non-linear) bias.
            In addition, the assumption that SSTs at Site 882 should be identical to those at Site 883 during the Pliocene and across periods of major climate change (e.g. iNHG) is not well justified in the current manuscript. Present-day SSTs are the two sites are not given for reference and we know that during past climate states like the Pliocene, sites in the same ocean basin can display differences in absolute as well as SST evolution (Naafs et al., 2020).
Lastly, on several occasions the statistical evidence that is needed to support statements (and importantly the correction) is lacking. For example, in lines 132-134, the manuscript states that the SST data from Sites 883 and 882 do not appear offset and this is used to justify the correction, but no statistical evidence is given. Same for lines 153-155, stating “more closely resemble” and “improved agreement” without statistical evidence to support these claims.
            The revised manuscript needs to take these comments into account, provide a proper justification of the methods used, as well as provide statistical evidence to support the approach.

Correction applied outside calibration range

The entire correction for Site 882 is based on a brief Pliocene interval (~2750-2650 ka) were both a GC-CI-MS-based and GC-FID-based SST data exist. For most of the GC-CI-MS-based data that spans the last 5500 kyr, there is no GC-FID-based SST data available (outside of calibration range). Thus, the entire correction assumes that the offset remained stable across all analyses. The manuscript provides no data to support this fundamental assumption. Details are missing, but I assume that the original GC-CI-MS data were obtained across a period of time, during which MS conditions might have varied. Normally, for each batch of GC-MS runs we would run a calibration curve to correct GC-MS to GC-FID UK37’ values. Hence the assumption that the correction holds across the whole record might be invalid. I wonder whether other temperature records are available, for example for during the (late) Pleistocene for Site 882 to test this hypothesis of a stable offset?
            The revised manuscript needs to at least acknowledge this caveat, but ideally addresses it with other published data and/or add a couple of new GC-FID SST data from across the last 5.5 Myr from Site 882 to confirm that the offset is constant. If not properly validated, I propose to only apply the correction to the Pliocene where GC-FID data is available.

Implications of corrections not clearly explained

Assuming that following my comments in the revised manuscript the correction still holds, the authors need to expand on the implications of this correction for Plio/Pleistocene climate. Site 882 is quite an important site and the current correction leads to lower maximum SSTs and higher minimum SSTs (e.g. lines 132-134). For example, given the corrected record shown in figure 4b, the original warming across the iNHG (~2.7 Myr) that formed the foundation of the (Haug et al., 2005) paper, appears to be largely reduced (if not removed), especially when the data from around 2850 ka is taken into account.

Minor comments:
Line 32: both marine and terrestrial temperatures can be used for this purpose

Line 50-53: also reference (Hefter, 2008) that introduces a method to use GC-MS to quantify UK37’-based SSTs (including a discussion on correcting for offset with GC-FID).

Line 73: justification for non-linear BAYSPINE calibration is needed

Line 85-90: The discussion of “instrument A/B” is not clear, revise and expand to clearly explain what this represents.

Figure 2: the x-axis (time) stops at 2800 ka, but there is younger data shown. Make sure axis covers whole record

Figure 5: give r2 values for both panels

Line 200: Why is Prof Gerald Haug not co-author of this manuscript? It looks like the other scientists involved in creating the published Site 882 and 883 data are co-author and he was involved in discussions (line 208-209), looks weird to me. It would be a strong signal if the original author of the data is part of this correction.

References
Haug, G.H., A. Ganopolski, D.M. Sigman, A. Rosell-Melé, G.E.A. Swann, R. Tiedemann, S.L. Jaccard, J. Bollmann, M.A. Maslin, M.J. Leng, and G. Eglinton, 2005. North Pacific seasonality and the glaciation of North America 2.7 million years ago. Nature 433, 821-825. doi: 10.1038/nature03332
Hefter, J., 2008. Analysis of alkenone unsaturation indices with fast gas chromatography/time-of-flight mass spectrometry. Analytical Chemistry 80, 2161-2170. doi: 10.1021/ac702194m
Naafs, B.D.A., A.H.L. Voelker, C. Karas, N. Andersen, and F.J. Sierro, 2020. Repeated Near-Collapse of the Pliocene Sea Surface Temperature Gradient in the North Atlantic. Paleoceanography and Paleoclimatology 35, e2020PA003905. doi: 10.1029/2020pa003905
Novak, J.B., R.P. Caballero-Gill, R.M. Rose, T.D. Herbert, and H.J. Dowsett, 2024. Isotopic evidence against North Pacific Deep Water formation during late Pliocene warmth. Nature Geoscience 17, 795-802. doi: 10.1038/s41561-024-01500-7
Studer, A.S., A. Martínez-Garcia, S.L. Jaccard, F.E. Girault, D.M. Sigman, and G.H. Haug, 2012. Enhanced stratification and seasonality in the Subarctic Pacific upon Northern Hemisphere Glaciation–New evidence from diatom-bound nitrogen isotopes, alkenones and archaeal tetraethers. Earth and Planetary Science Letters 351-352, 84-94. doi: 10.1016/j.epsl.2012.07.029
Citation: https://doi.org/10.5194/egusphere-2025-1975-CC1
- AC3: 'Reply on CC1', Joseph B. Novak, 19 Jan 2026
  
  Dear Professor Naafs,
  Thank you for taking the time to comment on our work. We are appreciative of the criticisms raised, which will result in an improved manuscript upon resubmission.
  Please find our responses to your specific comments in the attached .pdf file.
  Warm regards,
  Joseph Novak
  
  Citation: https://doi.org/10.5194/egusphere-2025-1975-AC3
RC2:
'Comment on egusphere-2025-1975', Ronnakrit Rattanasriampaipong, 23 Dec 2025
Novak et al. present a method to correct systematic biases in UK'₃₇-based SST reconstructions from ODP Site 882. The original published SST estimates (Haug, 1995; Haug et al., 2005; Martínez-Garcia et al., 2010) were derived from alkenone measurements using GC-CI-MS, a method known to introduce non-linear, concentration-dependent biases. The original SST record shows systematic offsets relative to subsequent GC-FID-based records from overlapping samples at ODP Sites 882 and 883. The authors propose a linear correction approach based on the relationship between GC-CI-MS (Haug, 1995) and GC-FID (Studer et al., 2012) SST estimates where data overlap at ODP Site 882.
Major comments:
Insufficient discussion of the implications and broader significance of the correction: I echo the other reviewers' concern that the manuscript lacks adequate discussion of why this SST correction matters and how it changes our understanding of North Pacific climate evolution. The following additions would significantly strengthen the manuscript:

Data-model comparison: The authors mention that the original SST record is unsuitable for evaluating Earth System Models but provide no actual model-data comparison. Including this comparison (either with existing model outputs or literature values) would demonstrate the practical impact of the correction and justify the effort.

SST gradients and ocean circulation dynamics: The corrected SST record will substantially alter the temporal evolution of meridional temperature gradients in the northwestern Pacific. Given ODP Site 882's position within the Kuroshio Extension region, the authors should discuss implications for our understanding of western boundary current dynamics and their role in North Pacific climate variability over the past 6 Ma.

Prevalence of GC-CI-MS measurements: The authors state that GC-CI-MS is used when alkenone concentrations are very low but cite only a handful of studies employing this method. A more systematic assessment of how commonly GC-CI-MS has been applied in the published literature would help readers evaluate the broader applicability of this correction approach. If this method has been widely used (particularly in low-productivity regions or deep-time studies), the implications extend well beyond ODP Site 882.

Uncertainties in extrapolating the correction beyond the calibration interval: The linear correction is derived from a relatively brief temporal interval where the Haug (1995) and Studer et al. (2012) datasets overlap. Applying this correction to the entire 6 Ma record likely introduces additional uncertainties that must be properly quantified and propagated. Specifically:
The magnitude of the GC-CI-MS bias may vary systematically with alkenone concentration, sample matrix effects, or instrumental drift over time.

The bias may differ between glacial and interglacial periods due to changes in alkenone preservation, sediment composition, or productivity regimes.

The authors should provide uncertainty estimates for the corrected record and discuss how confidence in the correction degrades outside the calibration interval.

General comments:
Ensure consistency in citing Haug (1995) versus Haug et al. (1995) throughout the manuscript. For example, Figure 3a shows "Haug et al. (1995)..." which appears incorrect if the reference is to a single-author 1995 publication.

Figure comments:
Figure 1 – Consider adding modern SST contours or climatology to emphasize that these sites are located in the subarctic North Pacific, where accurate SST reconstructions are critical for quantifying meridional temperature gradients and evaluating climate model performance in this sensitive region.
Figure 2 – Please extend the x-axis to cover the entire range of the data
Figure 4 – Include a comparison showing the original versus corrected SST estimates for the entire ODP 882 record, not just a subset. This would clearly demonstrate the magnitude of the correction during different climate states (e.g., glacial-interglacial extremes) and help readers assess whether the bias is constant or varies systematically with SST.

Line-by-line comments:
L17-18: Should "Haug 1995" be cited here along with “Studer et al., 2012”?
L18: Either remove “long” or change it to “long-chain.” “Long alkenone SST” sounds odd.
L65: Please provide lat/lon and water depth coordinates of ODP sites 882 and 883.
Citation: https://doi.org/10.5194/egusphere-2025-1975-RC2
- AC2: 'Reply on RC2', Joseph B. Novak, 19 Jan 2026
  
  Dear Dr. Rattanasriampaipong,
  Thank you for taking the time to review our work. We are appreciative of your comments, which will result in a substantially improved resubmitted manuscript.
  Please see our specific responses to your comments in the attached .pdf file.
  Warm regards,
  Joseph Novak
  
  Citation: https://doi.org/10.5194/egusphere-2025-1975-AC2

Peer review completion

AR – Author's response | RR – Referee report | ED – Editor decision | EF – Editorial file upload

ED: Reconsider after major revisions (23 Jan 2026) by Erin McClymont

Dear Joseph Novak and co-authors
Thank you for your detailed replies to the suggestions from the two reviewers and community comment. As you note in your replies, the reviews give constructive comments which should help to clarify the approach and also the implications of the work. At a time when data-model comparisons and data assimilation are important parts of palaeoclimate research, having a clear understanding of data generation and analysis (and its uncertainties) is critical.

The reviewers have given clear suggestions for edits or additions to be made, and your responses suggest that these changes can be realised. A couple of comments are provided here about some of the assumptions and tests, which would be good to clarify or weave into your responses and edits:
- As with the other reviewers, clarity on the adoption of linear vs non-linear corrections is needed. Given the second-order polynomial relationship between GC-FID and GC-CI-MS UK’37 values identified by Chaler et al. (2003), the rationale for the assumption that a correction for ODP 822 should be linear is unclear.

- Line 100 indicates that a simple correction was made between “adjacent samples” from Haug 1995 and Studer et al. (2012), so not measurements made on the same sediment sample? If this is the case, is there not additional uncertainty introduced to the comparison of the two approaches, because comparing samples of different ages will also be expected to have different SSTs at ODP 882?

- I’m still unclear why the constant values from “Instrument A” were used rather than “Instrument B” (Line 86-87). The text says that the “masses better corresponded to the UK’37 values in the chosen sites”, but Figure 3 in Chaler et al. (2003) doesn’t show UK’37 values, only the masses of the individual alkenones. Can you clarify the reasoning here? The Yamamoto and Kobayashi (2016) paper seems to indicate that at least for some samples, the masses of alkenone present could have been equivalent to Instrument B according to Figure 3 in Chaler et al (2003).

- Results on Figure 3: If the corrections are mass dependent as well as UK’37 dependent, should there not be uncertainty around the solid black line of the theoretical GC-CI-MS / GC-FID relationships given that masses in the 5-30 ng range were used? Given that the Yamamoto and Kobayashi (2016) paper also highlights concentrations exceeding 200 ng g-1 (e.g. Figure 4), should the impact of a wider range of concentrations have been tested?

- The equation used by Chaler et al. (2003) requires the mass of the analyte to be provided as an input to the constant, but you acknowledge that such data is not available from Haug (1995). However, there is alkenone mass data available from the Studer et al. (2012) paper and from Yamamoto and Kobayashi (2016): have you tried to predict the GC-CI-MS values using measured masses and UK’37 data from GC-FID, to test whether the results align with the Haug (1995) paper?

- As noted by the Chaler papers, the GC-CI-MS technique has been used to isolate alkenones for quantification where concentrations are very low and/or where the samples are complex and co-elutant removal is challenging. Under those circumstances, the GC-FID data would not be representative of the actual UK’37 value either, because the uncertainty on quantifying one or more low abundance alkenones would be large, or there may be over-estimation of the alkenone signal due to co-elution. Whilst acknowledging the caveats of the GC-CI-MS approach there could be a note here to the benefit that this technique has offered (perhaps with a recommendation in the Conclusions that undertaking the instrument-specific calibration of Chaler and co-authors ensures a translation of the data to “GC-FID equivalent”).

Hide

AR by Joseph B. Novak on behalf of the Authors (20 Apr 2026) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (05 May 2026) by Erin McClymont

RR by David Naafs (07 May 2026)

RR by Ronnakrit Rattanasriampaipong (02 Jun 2026)

ED: Publish subject to minor revisions (review by editor) (03 Jun 2026) by Erin McClymont

AR by Joseph B. Novak on behalf of the Authors (09 Jun 2026) Author's response Author's tracked changes Manuscript

ED: Publish as is (11 Jun 2026) by Erin McClymont

AR by Joseph B. Novak on behalf of the Authors (16 Jun 2026) Manuscript

Short summary

Accurate information about the surface ocean temperatures are important for understanding Earth's climate history. Here, we correct a sea surface temperature record from the critical subarctic North Pacific region for systematic biases introduced into the algal-biomarker-based geologic temperature estimates by the analytical method used by the original study.

Correcting a systematic bias in an ocean drilling project site 882 alkenone sea surface temperature record

Download

Interactive discussion

Peer review completion

Suggestions for revision or reasons for rejection

Suggestions for revision or reasons for rejection