Historical Nansen Cast Temperature Profiles and the Ocean Heat Content Change since 19<sup>th</sup> Century

Gouretski, Viktor; Cheng, Lijing; Du, Juan

doi:10.5194/essd-2026-416

Preprints

https://doi.org/10.5194/essd-2026-416

Preprints

23 Jun 2026

| 23 Jun 2026

Status: this preprint is currently under review for the journal ESSD.

Historical Nansen Cast Temperature Profiles and the Ocean Heat Content Change since 19^th Century

Viktor Gouretski, Lijing Cheng, and Juan Du

Abstract. The estimation of the ocean heat content (OHC) is based on the subsurface temperature observations which abundance, quality and accuracy changed over time following the development of oceanographic techniques and methods. The majority of the available global OHC time series begins in 1950s when the International Geophysical Year 1957–1958 (IGY) observational campaign resulted in a rapid improvement of the observational basis. However, a significant amount of subsurface temperature observations was accumulated during the period before the IGY. We perform a three-step quality control (QC) of the Nansen cast temperature profiles with the focus on the time period before 1970. The QC includes automated and manual control of temperature observations along with the quality control of temperature anomalies. The World Ocean Circulation Experiment/Argo Global Hydrographic climatology based on the high-quality temperature observations between 1990 and 2025 is used as the reference. To interpolate temperatures vertically we use a combination of the parabolic Reiniger and Ross (1966) method and the temperature anomaly method by Frankignoul (1981). The method is shown to result in a smaller interpolation error. Correction for depth bias in Nansen cast profiles are also suggested. The oldest temperature profiles from 1870s are treated separately, with the data from “Challenger” expedition used as reference. Finally, we estimate the OHC changes between 1870s and 2020s for six selected multi-year time periods. We find that for the layer 0–2000 m the OHC increase between 1920–1944 and 1945–1969 amounts to 30 % of the total OHC increase between 1920–1944 and 2010–2025.

Received: 01 Jun 2026 – Discussion started: 23 Jun 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 2797 KB)

Supplement (4172 KB)

Download & links

Viktor Gouretski, Lijing Cheng, and Juan Du

Status: open (until 29 Aug 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on essd-2026-416', Anonymous Referee #1, 23 Jul 2026 reply
General Assessment
This manuscript describes the construction, quality control, and application of a global archive of Nansen cast temperature profiles spanning 1873–2025, and uses it to extend estimates of 0–2000 m ocean heat content (OHC) change back to the 1870s. This is a significant and welcome contribution. The pre-IGY hydrographic record is underutilized in the OHC literature, and the authors—who are among the most experienced groups worldwide in historical hydrographic data rescue and bias correction—have produced a dataset that will be of lasting value to the climate community. The methodology is sound, the treatment of uncertainty is generally honest, and the central result (that roughly 30% of the post-1920 OHC increase occurred before 1970) is important and well supported. I recommend acceptance after the authors address the items below. The most substantive concerns are internal numerical inconsistencies and citation/reference mismatches; the remainder are language and presentation issues that require careful copy-editing but no new work.
Strengths of the Manuscript
The dataset itself is a major contribution. Extending quality-controlled, bias-adjusted subsurface temperature coverage back to the Challenger, Gazelle, and Tuscarora expeditions of the 1870s—including four newly digitized station-log datasets—enables the first observation-based assessment of 0–2000 m OHC change over a full century. The 2.8 million retained profiles, with quality flags and interpolated levels in standard formats, will serve many downstream users.

The three-step quality control (automated statistical screening, manual inspection in 507 regional boxes, and anomaly-based screening against the WAGHC climatology) is rigorous and transparently documented, with outlier statistics quantified at each stage (4.2%, 0.17%, and 0.55% of observations, respectively). The temporal trend in outlier rates is itself a useful diagnostic of evolving observational practice.

The hybrid vertical interpolation scheme is well justified. The authors do not simply adopt the Reiniger-Ross method; they quantify its failure mode for large vertical gaps (Figs. 6–9), demonstrate the superiority of the Frankignoul anomaly method below the threshold gap sizes, and provide a clear, reproducible decision rule (Table 1). This level of methodological care is exactly what an ESSD data-description paper should provide.

The depth-bias treatment is clever and appropriately cautious. Using the sample-depth histogram method to derive corrections for the pre-CTD era (when collocated comparisons are impossible), quantifying the impact on OHC (a 12.9% increase in 1930–2015 energy gain), and explicitly declining to apply steel-wire-based corrections to the hemp-rope-era expeditions all reflect sound physical judgment.

The validation against independent estimates (Pan et al. 2026; Zanna et al.; Wu et al. 2025) is convincing, and the authors are commendably candid in labeling the inferred 1870s–1920s warming (67 ZJ) as statistically insignificant rather than overselling it.

Items to Be Addressed
Reconcile internal numerical inconsistencies in the OHC results. The OHC increase between the 1920–1944 and 1945–1969 periods is given as 149 ZJ in Section 8.5 (line 424) but as 139 ZJ in the Conclusions (line 484). Similarly, the total warming since 1932 is 456 ZJ at line 424 but 457 ZJ at line 444. These are presumably the corrected vs. uncorrected values or simple typographical slips, but as written the paper contradicts itself on its headline numbers. Please reconcile these figures throughout the Abstract, Section 8.5, and Section 9, and state explicitly whether quoted values are for original or depth-corrected profiles.

Correct the standard-depth list in Section 8.5 (lines 436–438). The list of 67 standard levels contains apparent duplications and omissions: “30” appears twice (“15, 30, 25, 30, 35…”), “2000” appears twice (“2100, 2000, 2300”—presumably 2200 was intended), and “3200” appears twice (presumably 3100). As listed, the levels do not sum to 67 unique depths, and the claim that they represent 1.2% of all 1-m levels cannot be verified. Please correct the list and re-verify the 49.5% statistic if affected.

Resolve citation-year and author-name mismatches between text and reference list. (a) Reiniger and Ross is cited as 1966 throughout the text (lines 20, 163, 471) but dated 1968 in the reference list; (b) Chapman is cited as 1955 (line 100) but listed as 1959; (c) Zanna et al. is cited as 2018 throughout but listed as 2019; (d) “Spillhouse, 1938” (line 50) should be Spilhaus, consistent with the reference list; (e) “Gourteski” at lines 30 and 85 should read Gouretski. These are easily fixed but essential for a data paper that readers will mine for sources.

Make the Challenger expedition dates consistent. The expedition is variously dated 1872–1876 (lines 35, 74), 1873–1875 (line 320), and 1873–1876 (lines 413, 578). The voyage departed in December 1872 and returned in May 1876; please adopt one convention (or clarify that the dates at line 413 refer to the analysis period rather than the voyage) and apply it throughout.

Check Figure 15 legend against the stated analysis periods. The legends in Fig. 15 appear to read “2010–2017” for the most recent period, whereas the text and Section 8.1 define the period as 2010–2025. If the figure was carried over from an earlier analysis, please regenerate the legend; if the plotted period genuinely differs, the discrepancy must be explained. Relatedly, in Section 8.4 the error-estimation procedure is clearly described, but the resulting uncertainty magnitudes are never stated numerically in the text—adding one or two sentences quantifying the basin- and global-scale standard errors (or pointing to the error bars in Figs. 15–16 with representative values) would materially strengthen the section at negligible cost.

Clarify data access in Section 10. The dataset DOI requires email registration and login. ESSD policy expects data to be publicly and freely accessible without barriers wherever possible. Please either remove the registration requirement, or briefly justify it and confirm that access is granted automatically and without restriction, so the editor can verify compliance with the journal’s data policy.

The manuscript requires a careful language edit. The science is clear, but grammatical errors and typographical slips occur throughout and occasionally rise to the level of technical error. The most important instance: “Jung’s modulus” at line 459 should be Young’s modulus (and note the same sentence refers to “hemp wire” where “hemp rope” is meant, and “four time higher” should read “four times higher”). A non-exhaustive list of additional corrections follows; I would encourage the authors to have the full manuscript professionally proofread rather than relying solely on this list.

Line 11–12: “observations which abundance, quality and accuracy changed” → “observations whose abundance, quality, and accuracy have changed.”

Line 35 and line 483: “marks the begin of” / “the begin of the ocean warming” → “beginning.”

Line 45: “Fritjoff Nansen” → “Fridtjof Nansen.”

Line 49: “hydrographic hydrographic profiles” — duplicated word.

Line 62: “we access for the first time the change” → “we assess.”

Line 126: “An disproportionately large number” → “A disproportionately large number.”

Lines 137, 421, 472: “undergone” used as a simple past (“the data … undergone three steps”; “the global ocean undergone”; “profiles undergone”) → “underwent” (or “have undergone”).

Line 197: “performed for a the resampled profile” — delete “a.”

Line 260: “Here we imply this method” → “apply.”

Line 292: “In a deviation to the widely used World Ocean Atlas climatologies” → “In contrast to.”

Line 309: “mimimum/maximum” → “minimum/maximum.”

Line 415: “with wights equal to” → “weights.”

Line 417: “There several significant differences” → “There are several…”

Reference list: “littele ice age” (Gebbie and Huybers), “Warming of heworld ocean” (Levitus et al.), “Pysics and Chemistry” (Thomson and Murray 1889), “Washigton D.C.” (U.S. Naval Oceanographic Office), “Manual fort hydrological works” (State Oceanographic Institute), and “Her Majesty Stationary Office” → “Stationery Office.” The Frankignoul (1981) entry should also be reformatted to match the journal’s reference style (year at the end).

This is a valuable data paper on an important and under-examined segment of the ocean observing record, and I congratulate the authors on the substantial effort behind it. None of the items above challenges the methodology or the conclusions; they concern internal consistency, referencing, figure/text agreement, and language. Once these are addressed, I fully support publication in Earth System Science Data.

Reply
Citation: https://doi.org/10.5194/essd-2026-416-RC1
RC2: 'Comment on essd-2026-416', Anonymous Referee #2, 12 Aug 2026 reply

Review of "Historical Nansen Cast Temperature Profiles and the Ocean Heat Content Change since 19th Century" by Gouretski et al.
* SUMMARY *
In this work, the authors present a reconstruction of ocean heat content change since the 19th century using historical Nansen Cast temperature profiles. The key steps include a three-stage quality-control procedure, a hybrid vertical interpolation scheme and a depth-bias correction. The result helps address the lack of in-situ ocean heat content estimate prior to 1960 in previous studies.
Overall, I think this work represents an important contribution to the ocean heat content community. Nonetheless, I find that the current version of the manuscript does not provide sufficient technical detail, making it difficult for readers to fully understand the underlying data processing and uncertainty quantification. For instance, three measures of interpolation error are used in Section 4, but none is clearly defined in the text. Similarly, instrumental errors and the mesoscale variability are considered in Section 8.4, but the manuscript does not explain how these quantities are derived. Therefore, I recommend major revision and have listed specific comments and questions below for the authors to consider.
In addition, I was unable to access the link in the Data availability section (http://dx.doi.org/10.12157/IOCAS.20260520.001). Please check if the link is working.

* SPECIFIC COMMENTS *
* Sec 3.1: This section corresponds to Figs 5a-c. However, the text only discusses Fig 5a, not the other two. Are Figs 5b and c not relevant here? If so, consider removing them to simplify the figure.
* Sec 3.2: It is unclear how the manual quality control is done. What are the criteria for determining if a profile is an outlier? Please explain the procedure and cite relevant references.
* Sec 4.1
Please add a short description of the RR and FR methods in the preamble of section 4. In addition, please explain why the two methods are chosen here instead of a simple linear interpolation. The authors show that the polynomial interpolation (the RR method) causes unrealistic excursions in temperature profiles in Fig 6 (lines 182-186). For these cases, it seems that a simple linear interpolation would yield a similar result, but without the unrealistic excursions. Is this correct?
* Sec 4.2
Lines 202-203 "centre of the gap": But in Fig 7, the x-axis label is "depth of the upper gap boundary", not the depth of the centre of the gap.
Is the interpolation error a function of geographical locations? Ocean temperature profiles have different characteristics in different regions, e.g. weak stratification at high latitudes vs strong stratification in the tropics. Hence, I expect that you may get different interpolation errors from the WAGHC profiles at different locations.
* Sec 4.3 and Fig 9
Several key details are missing from this section.
What is the definition of the mean error and the mean absolute error? What is the dimension on which the "mean" operation is applied?
Why is the interpolation error always positive in Fig 7, but has both positive and negative values in Fig 9?
In Fig 9, which panels are the mean error and which are the mean absolute error? (I can't find this information in the caption)
In Fig 9, does the interpolation error includes errors for both climatology and anomaly? If so, it is no surprise that the FR method has a smaller error. This is because: a) the climatology signal is much larger than the anomaly signal and b) the FR method already knows what the climatology signal is. However, we are more interested in temperature change than the climatology. Given this, it may be better to focus on the anomaly signal when evaluating the interpolation error.
It would be more consistent to use the same type of diagram for sections 4.2 and 4.3, since they both examine the interpolation error. Fig 9 shows the error in different depth ranges using different panels. Could those panels of different depth ranges be put into a single panel like in Fig 7?
Line 225: Is the histogram for the interpolated data also normalised by the number of observations? It seems a more sensible choice is to use the number of interpolated data points as the normalisation factor.
Line 226: Do you mean "the most data abundant *temperature* bin for each each depth" here?
* Sec 5
Line 269: If the standard levels are removed, why the frequencies for the standard levels are not zero in Fig 11? Perhaps this is caused by the Gaussian smoothing? What is the width of the Gaussian kernel and how is it determined?

* Sec 6
Lines 311-312: Do you mean "σ_MEAN(z) is the global mean of σ(x,y,z)"?

* Sec 7
Fig 13: The depth is denoted using positive values in (d), but negative values in (f).
Fig 13 caption: Lines 344-345 are inconsistent with the text at line 334. Also, "same as f)" should be "g) same as f)".

* Sec 8
I suggest swapping "The earliest time period ..." with "The following time periods ..." at lines 351-353. When I read it, I thought "The following time periods" means that the time periods after the earliest time period, which is not what the authors meant here.
Line 365 "the northern part of the Atlantic ocean": The northern part of the Atlantic ocean shows a cooling in Fig 14l. Do you mean the northern part of the Pacific ocean?
Fig 14: This is a busy figure. To improve readability, I suggest labelling the time periods to the left of the figure and the depths at the top of the figure using bold font.

* Sec 8.4
This section would benefit from additional detail on several key points.
The text mentioned the instrumental errors and the mesoscale variability. How are they derived and what are their respective magnitudes?
Fig S8 shows how to estimate spatial covariances. There are several dots for a given lag in Fig S8. What are they?
How are the within-box error and the correlation radius combined to give the error for spatial means?
For the earliest periods, basin means are estimated from observations that cover only a small fraction of the basin (see Fig 14 top row). This will probably introduce another error (a mapping error?). Is this considered in Figs 15 and 16?
* Sec 8.5
Lines 423-426 "149 ZJ" "456 ZJ" "30 %" "67 ZJ": Please include uncertainty ranges when discussing these results. The same comment applies to the "30 %" reported in the abstract at line 25.
Lines 442-443: Intuitively, why does the depth corrections tend to increase the warming rate estimate?
Line 444: The depth correction factor is close to 0.99 at most depths in Table A1. That is, the correction would only change the depth of a profile by about 1%. However, the impact of the correction on the ocean heat content change estimate is about 10%, as reported at line 444. This is interesting. Could you please comment on why a small depth correction could lead to a large increase in the estimate of ocean warming?
Line 446: How is the heating rate estimated in Fig 16b? Is it simply the difference between two adjacent time-mean anomalies in Fig 16a divided by the time span between them?
It would be useful to also show the ocean heating rate from Pan et al., Zanna et al and Wu et al in Fig 16b. In addition, please consider showing heating rate in W per m^2, so that it is easy to compare it with the TOA energy imbalance.
What is the uncertainty range in Fig 16?
Lines 460-461: The text says that applying the depth correction would result in a positive heating rate between 1870s and 1920s. However, Fig 16b shows that doing so reduces the 1874-1932 heating rate from a positive value to near zero, which seems to contradict the text.
* Other items
Line 21 "a smaller ... error": The comparison is unclear. Do you mean a smaller error compared with linear interpolation?
The y-axis label is "percent profiles" in Figs 3a-b and "Frequency %" in Fig 4. They seems to denote the same quantity? If so, please make them consistent.
Fig 5: Some of the panel labels and text are difficult to read. Please consider moving them outside the panels to improve readability.
Line 149: quality controlled -> quality control
Lines 164-166: Please check if there is a missing comma between:

1) "observed temperature" and "and the ..."

2) "(FR method)" and "which interpolates ..."
Line 197 "a the resampled profile" -> "a resampled profile"

Reply

Citation: https://doi.org/10.5194/essd-2026-416-RC2

Viktor Gouretski, Lijing Cheng, and Juan Du

Supplement

https://doi.org/10.5194/essd-2026-416-supplement

Viktor Gouretski, Lijing Cheng, and Juan Du

Viewed

Total article views: 92 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
67	21	4	92	19	13	9

HTML: 67
PDF: 21
XML: 4
Total: 92
Supplement: 19
BibTeX: 13
EndNote: 9

Views and downloads (calculated since 23 Jun 2026)

Month	HTML	PDF	XML	Total
Jun 2026	1	0	1
Jul 2026	43	14	3	60
Aug 2026	23	7	1	31

Cumulative views and downloads (calculated since 23 Jun 2026)

Month	HTML	PDF	XML	Total
Jun 2026	1	0	1
Jul 2026	43	14	3	60
Aug 2026	23	7	1	31

Viewed (geographical distribution)

Total article views: 65 (including HTML, PDF, and XML) Thereof 65 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 25 Aug 2026

Download

Preprint (2797 KB)
Metadata XML

Short summary

This study depicts a three-step quality control procedure applied to the Nansen cast temperature profiles (1873–2025, especially before 1970), including automatic/manual quality control and temperature anomaly quality control. Corrections for depth bias are also suggested with improved vertical interpolation procedure. Based on this dataset we access for the first time the change of the ocean heat content of the upper 2000 m layer over a 100 year time period between 1870s and 2020s.


Total:	0
HTML:	0
PDF:	0
XML:	0

Historical Nansen Cast Temperature Profiles and the Ocean Heat Content Change since 19th Century

Supplement

Viewed

Viewed (geographical distribution)

Historical Nansen Cast Temperature Profiles and the Ocean Heat Content Change since 19^th Century