The consistent in-situ global gridded temperature and salinity dataset, CORA OA 1960&ndash;2024

Szekely, Tanguy; Miadana, Jean-Renaud

doi:10.5194/essd-2026-376

Preprints

https://doi.org/10.5194/essd-2026-376

Preprints

29 Jun 2026

| 29 Jun 2026

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

The consistent in-situ global gridded temperature and salinity dataset, CORA OA 1960–2024

Tanguy Szekely and Jean-Renaud Miadana

Abstract. Gridded ocean products play a pivotal role in oceanographic and climate research, providing a framework for the cross- validation and calibration of satellite observations and in-situ data. This paper describes the latest evolution of the global gridded ocean objective analysis fields for temperature and salinity, distributed by the Copernicus Marine Service. It accounts for the update of the objective analysis first guess using monthly fields derived from the temporal interpolation of World Ocean Atlas 2023 (WOA23) decadal temperature and salinity climatologies. The new product provides monthly objective analysis fields for temperature and salinity with a 0.5° horizontal resolution and 187 vertical levels covering depths from 0 to 5500 m. The time series spans from 1960 to December of the previous year. A full reprocessing is conducted every November, while an interim update covering the first six months of the current year is released each June. This product was compared with other reference 3D temperature and salinity datasets, as well as SST and SSS products. It exhibits strong consistency with all reference data, particularly over the 1985–2024 period. Notably, the product proved robust against the "Argo fast salty drift" frequently detected in other datasets after 2016. The gridded product is freely accessible on the Copernicus Marine data store (https://data.marine.copernicus.eu/) under the ID INSITU_GLO_PHY_TS_OA_MY_013_052.

Received: 18 May 2026 – Discussion started: 29 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

Tanguy Szekely and Jean-Renaud Miadana

Status: final response (author comments only)

RC1: 'Comment on essd-2026-376', Anonymous Referee #1, 10 Aug 2026

Overall comment

This is a timely paper which honestly presents the temperature and salinity analysis product CORA

Great care is given in the reference time evolving mean field over which the objective mapping is done.

The product is very smooth in time, once spatially averaged, which suggests that with the mapping scales and parameters used the data density was insufficient to fully cover the relevant space and time scales globally before the post-200 Argo and other network period (at least, when one gets below the surface).

This dooes not come as a surprise and is correctly taken into account into the error budget I believe. Thus the average for the earlier period (once it is spatially averaged over a basin, and for T below the surface) is mostly the climatology product used (from the decadal NODC climatology) except after 2000.

What is really interesting is the comparison done with other products. What is presented and discussed is mostly visual and qualitative (except for a table with the trends). I suggest that this could be made more quantitative, and in addition to trends, the authos could provide rms differences between the different time series (for example on figure 5 and 6, or even for the OHC), estimates of detrended variability for the different products (at least spatial averages, but why not also examples of spatial maps).

Minor comments

Bottom of page 3: it would be interesting to know to which extent the data set is MEOP-based, thus qualified/adjusted, and to which extent it is not the case. I was aware of turtle data in the southwest Indian Ocean for example, not in MEOP, and presenting large salinity biases. I believe that I see some of their tracks on Fig3 top right panel), but the worse biased ones (in salinity) might have been filtered out. Although it is a small number of bad data, it can locally degrade the product (towards lower S).

Lines 106: a bit strange. Should it be instead of Tclim, Sclim, Tstd and Sstd. (otherwise what are Tstd and Sstd?). As there is a combination of two terms, not obvious to the reader what is the proportion of the data that is filtered out in this step. I believe that it should be reported somewhere (maybe a percentage as a function of the depth?)

114: add ‘to a threshold, noted Thres..’ I guess later on line 17, the numbers reported are non-dimensional, and thus in units of vertical grid steps?

Lines 125-129 duplicated in lines 130-133.

Figure 4: there is a change of data density increasing just after 2000m before the early 1990s (also near 800m). Is it linked to the change of grid spacing with depth and the criterium on maximum vertical data spacing. If the case, I would argue that it is rather arbitrary, and would suggest (for later versions) to revise the selection criteria on line 177. In the figure 4 caption, for the lower panel, it indicates the median of the profiles anomalies, whereas on line 163, this is described as the ‘objective analysis profiles’. Is it the same thing and thus after mapping analysis, or before the mapping.

179: ‘SSS product described in Sammartino et al (2022)’. The choice is interesting, but there is the caveat that the product combines different variables including SST, etc… Why not use the product of N. Kolodziejczik L4 product (combining satellite salinity with in situ data) or the CCI V4.4 S product combining different satellite salinity data, of course only since 2010, and only an update of the last one to the end of 2024 (but notice how little the Nardelli product captures at least globally or by basin for the period 2000-2010).
210: the comment puzzled me. Which data version of the sea mammal data is used. What are the temperature (and pressure) validation steps (if it is MEOP, or if not). There might also be issues of non-randomness of sampling by the sea mammals at time, thus potential biases. Also, some temperature/pressure sensors, for example used for the northern Pacific Ocean sea lion data set, might have had biases and less stable sensors, I believe (but that’s not the case for the Southern Ocean situation commented here?). There certainly are further issues if the real time data are used.

237: I don’t see from Table 3 a larger spread in the first period than in the last one (when spatially averaged!). The average trend is less in first period, but the spread is comparable (at least averaged over 20 years). What is really clear is the very small interannual variability of CORA product when basin-averaged.

On Figure 5, means are not estimated over the same period for the ISAS20 product than for the others (thus shifted negative (and even for the IAPV4 product with the last two years missing). It gives some visual impression of disagreement (for example for ISAS20) that should be avoided. Instead of anomalies, as these are basin-scale averages, why not present instead of anomalies total SST on Figure 5 (as is done on figure 6 for SSS). For the Southern Ocean, I am wondering whether the different products use different defaults for temperature under sea ice, and how is the spatial average done. Is SST in all products, a foundation SST? And how is it estimated if not directly measured?

On Fig. 6, IAP presents a strange large negative peak in SSS in 2001, not seen in all the other products, coming from Pacific and Indian Ocean north of 35°S. Any reason? This suggests some bug (TAO or Rama uncorrected and bised mooring data, perhaps?). Also, in caption, add that deep ocean is areas with bottom depth > 1500 m.

On most figures, when referring to the Southern Ocean, the panel title reads ‘South ocean’, which should be changed. On other panels, ocean is not capitalized. It should be: Atlantic Ocean, etc…

262: ‘… has been partly resolved.’

Some comparison mentioned at the beginning, such as SSS with the Sammartino et al product, are not really described (they appear on one figure).

Citation: https://doi.org/10.5194/essd-2026-376-RC1

AC1: 'Reply on RC1', Tanguy Szekely, 08 Sep 2026

We thank the referee for this thorough and constructive review. Below we respond in detail to the main comment, concerning the quantitative comparison with the other T/S products, which we agree deserves a substantive answer. We will study the RMS differences and the detrended standard deviation for GOHC, SST and SSS. We leave spatial maps of detrended variability for future work / supplementary material. The minor comments (data provenance, notation, Figures 4–6, terminology, etc.) will each be addressed through the corresponding textual and figure changes in the revised manuscript, rather than discussed point by point here; a full list of these changes will accompany the revised version.

OHC detrended std :

GOHC detrended standard deviation (ZJ):

1960-1984 1985-2005 2006-2020

NODC 8.93 5.87 2.54

GCOS 9.04 8.62 5.54

CORA OA 4.18 6.14 5.63

Cheng 2023 9.28 6.90 4.38

This table shows that all reference products except CORA_OA display a decreasing interannual variability, from ~9 ZJ in the early period to 2.5–5.5 ZJ during the Argo era. CORA_OA, by contrast, has the lowest variability of the four products during 1960–1984, peaks during 1985–2005, and plateaus during the Argo era at a level close to GCOS, moderately above Cheng et al. (2023), and more than twice that of NODC. This pattern is consistent with the objective-analysis correlation-scale constraint being weakly satisfied in the sparse-sampling era, damping the analysis increments toward the background climatology and suppressing spurious interannual noise. On the other hand, the other products appear to retain sampling-driven variability that only recedes once Argo provides dense, homogeneous coverage. Once the correlation-scale constraint is well satisfied in the Argo era, CORA interannual variability settles at a level comparable to or above the other products, suggesting it retains more independent interannual signal than the more strongly statistically-smoothed reference products.

OHC RMS differences :

RMS differences vs CORA_OA by sub-period :

product 1960-1984 1985-2005 2006-2020

NODC 15.95 13.13 8.39

GCOS 8.81 15.00 3.10

Cheng2023 10.10 13.17 7.22

This table gives the RMS differences between CORA and each reference product by sub-period, offering insight into how well interannual anomalies from each product track those of CORA. The three products show markedly different behaviours. The CORA-NODC RMS difference decreases monotonically from 15.95 ZJ (1960-1984) to 8.39 ZJ (2006-2020). GCOS and Cheng et al. (2023), by contrast, follow a non-monotonic, hump-shaped pattern: an intermediate RMS in the early period (8.81 and 10.10 ZJ respectively), a peak in 1985-2005 (15.00 and 13.17 ZJ), and a marked drop during the Argo era (3.10 and 7.22 ZJ), the lowest values of the three periods for both products. The elevated RMS differences before 1985 are consistent with the objective-analysis correlation-scale constraint being weakly satisfied during the sparse-sampling era, which damps CORA_OA's interannual signal relative to the reference products; however, the contrast between NODC's monotonic decline and GCOS/Cheng's hump shape suggests that phase alignment of the interannual anomalies, not only their amplitude, also plays a role and differs by product. The persistence of a 3-8 ZJ residual RMS difference during the Argo era, when sampling is no longer limiting, points to genuine structural/methodological differences between the products (climatology reference, mapping method, QC) rather than a data-coverage effect.

Detrended standard deviation of the SST (°C)

		1960-1984	1985-2005	2006-2020
CORA	Global	0.018	0.018	0.026
Pacific	0.020	0.024	0.023
Atlantic	0.022	0.033	0.025
Indian	0.022	0.018	0.031
South	0.005	0.004	0.02
EN4	Global	0.067	0.065	0.058
Pacific	0.076	0.073	0.060
Atlantic	0.065	0.077	0.060
Indian	0.11	0.10	0.077
South	0.05	0.039	0.051
ERSST	Global	0.069	0.061	0.057
Pacific	0.076	0.077	0.057
Atlantic	0.077	0.081	0.051
Indian	0.095	0.095	0.084
South	0.041	0.041	0.039
IAP v4	Global	0.072	0.059	0.059
Pacific	0.072	0.079	0.063
Atlantic	0.069	0.084	0.061
Indian	0.10	0.095	0.078
South	0.062	0.038	0.04
ISAS 20	Global	-	-	0.045
Pacific	-	-	0.044
Atlantic	-	-	0.042
Indian	-	-	0.057
South	-	-	0.032
SMOS	Global	-	-	-
Pacific	-	-	-
Atlantic	-	-	-
Indian	-	-	-
South	-	-	-

Detrended standard deviation of the SSS (PSU)

		1960-1984	1985-2005	2006-2020
CORA	Global	0.0017	0.0029	0.0035
Pacific	0.0029	0.0018	0.0065
Atlantic	0.0038	0.0035	0.0074
Indian	0.0031	0.0010	0.0089
South	0.0007	0.0009	0.0017
EN4	Global	0.0060	0.0086	0.0073
Pacific	0.010	0.013	0.012
Atlantic	0.015	0.016	0.011
Indian	0.0154	0.016	0.017
South	0.0048	0.0062	0.0061
ERSST	Global	-	-	-
Pacific	-	-	-
Atlantic	-	-	-
Indian	-	-	-
South	-	-	-
IAP v4	Global	0.0075	0.010	0.0053
Pacific	0.016	0.013	0.010
Atlantic	0.019	0.015	0.0094
Indian	0.017	0.027	0.016
South	0.013	0.0048	0.0051
ISAS 20	Global	-	-	0.0062
Pacific	-	-	0.0095
Atlantic	-	-	0.0086
Indian	-	-	0.015
South	-	-	0.0030
SMOS	Global	-	-	0.0067
Pacific	-	-	0.013
Atlantic	-	-	0.011
Indian	-	-	0.017
South	-	-	0.0043

Table 4 gives the detrended standard deviation of the SST and SSS time series for all tested products and for three periods. It shows that CORA almost always gives a lower estimate of the detrended SSS or SST variability, by a factor of 2 to 10. The most severe underestimation occurs before the Argo era in the Southern Ocean, and the gap narrows markedly during the Argo era. It is consistent with a damping of the objective-analysis signal in regions and periods where data coverage is sparse. The comparison also shows that ISAS20 underestimates the detrended standard deviation relative to the other products, though to a lesser extent than CORA. This is likely related to the linear vertical interpolation scheme shared by CORA and ISAS20 (see Section 3.1), which does not extrapolate to the surface for profiles whose shallowest observation lies below 1 m depth, reducing the number of valid surface data points relative to the other products. The residual gap between CORA and ISAS20 likely reflects a difference in data sources: ISAS20 uses the full Argo profiles from the Global Data Assembly Center, whereas CORA discards Argo sub-surface measurements taken while the CTD pump was not activated, further reducing the number of valid near-surface observations available to CORA relative to ISAS20.

SSS RMS differences (PSU)

	1960-1984	1985-2005	2006-2020
ERSST	-	-	-
IAP	0.0081	0.011	0.0071
EN4	0.0063	0.015	0.0061
ISAS 20	-	-	0.0086
SMOS	-	-	0.0050

SST RMS differences (°C)

	1960-1984	1985-2005	2006-2020
ERSST	0.069	0.060	0.046
IAP	0.081	0.060	0.045
EN4	0.074	0.082	0.046
ISAS 20	-	-	0.026
SMOS	-	-	-

This table gives the RMS differences between CORA and the reference products. The SST RMS difference decreases with time for ERSST and IAP, and slightly increases then decreases for EN4. It is worth noting that ERSST, IAP and EN4 SST RMS differences are almost identical during the Argo era, while ISAS20 is smaller by a factor of ~1.8. This suggests that the objective-analysis method shared by ISAS20 and CORA tends to produce more consistent interannual variability between the two than either does with the statistically-based products. The SSS RMS difference (for IAP and EN4, the only products with data across all three periods) is low during both the early period and the Argo era, and peaks during 1985-2005 — the same hump-shaped pattern observed for the OHC RMS differences against GCOS and Cheng et al. (2023), suggesting a common cause (e.g. a transition in the underlying observing network) rather than a variable-specific artifact.

Citation: https://doi.org/10.5194/essd-2026-376-AC1

RC2: 'Reply on AC1', Anonymous Referee #1, 17 Sep 2026

Thank-you. This quantified material is very informative,
and helps better to understand some of the differences between the products.
Good luck with revising your manuscript.

Citation: https://doi.org/10.5194/essd-2026-376-RC2

RC3: 'Comment on essd-2026-376', Zhankun Wang, 18 Sep 2026

Review comments on “The consistent in-situ global gridded temperature and salinity dataset, CORA OA 1960-2024” by Szekely and Miadana.
This manuscript provides a robust, well-structured foundation for introducing the latest CORA OA dataset. The transparency regarding the product's limitations and its comparative performance makes it a strong candidate for publication, though a few structural and grammatical refinements will elevate the final submission.
Vertical interpolation relies on a linear scheme, which could prove problematic across specific regions and depth levels, particularly when dealing with low-resolution profiles. Significant uncertainties may stem from this approach. I suggest comparing this technique against interpolation schemes employed by comparable datasets to better highlight its limitations and associated errors. Addressing this aspect is vital for clarifying some discrepancies observed between CORA and alternative products.
More details should be provided regarding the quality control criteria and processing procedures. Additionally, further elaboration is needed on the bias corrections applied to the dataset. This information is vital for readers to fully evaluate the data utilized in the objective analysis.
Ln 8, "The time series spans from 1960 to December of the previous year." Mentioning the previous year relative to 2026 implies 2025, contradicting the 1960–2024 interval indicated in your title. Revising this phrase for consistency would prevent potential reader confusion.
Ln 15, The sentence reads, "According to IPCC (2023), the global ocean has warmed and that the salinity changes...". Remove the word "that" for proper grammatical flow.
The citation formatting throughout the paper is somewhat disorganized and integrated awkwardly into the main text. While I expect this will be corrected prior to final publication, addressing it before formal submission would greatly enhance readability. The current layout makes certain passages difficult to parse (for instance, lines 18–19: "International initiatives were subsequently established to coordinate basin-scale measurement campaigns Chapman (1998); Bourlès et al. (2008); Teng et al. (2008) and to implement sustained observation systems such as the Global Drifter Program Niiler (2001)").
Ln 30, periodically updated. Please be more specific about the update frequency.
Ln 40-43 , INSITU_GLO_PHY_TS_DISCRETE_MY_013_001 and INSITU_GLO_PHYBGCWAV_DISCRETE_MYNRT_013_030. More information is needed for readers who are unfamiliar with these products
The caption of Figure 1 is too short. At least what in the figure should be described.
Ln 87, The text states, "historical data where first only integrated...". Change "where" to "were".
The text font in Figure 2 is too small to read.
Ln 101, Equation 6 is cited before other Equations. Equations should be cited in order.
The entire paragraph explaining the discontinuities introduced by decadal climatologies (starting with "Climatologies covering decadal periods...") appears twice verbatim between lines 124-129 and lines 130-133. The authors should proof-read the paper before submitting.
Ln 134. What does “yearly climatology" mean? Please explain.
Table 1 uses "Celsius degree". Standardizing this to "Degrees Celsius" or simply "°C" is standard practice.
It is not clear to me how the temperature and salinity anomalies are determined and which climatology is used as the baseline. This is important information. Please clearly describe this in the methodology section.
Equation (5) alongside its subsequent explanation remains confusing. Could you please clarify what Rii represents?
Ln 177, spell out “ LOPS/UMR”
The manuscript oscillates between "Argo" and "ARGO". The official program name is "Argo" (not an acronym), so this should be standardized throughout the text.
Figure 5, Are the mean SST from different products used the same baseline and comparable? If different baselines are used. please explain the differences are induced by different baselines or convert the data to the same baseline to make them comparable.
Ln 233, NODC was merged into NCEI more than 10 years ago. NCEI should be refereed, not NODC. Remove "one".
In both Figure 5 and Figure 6, the subplot titles and the captions refer to the "South ocean". This should be corrected to its formal name, the "Southern Ocean",
Ln 293, The transition "To the contrary, the products based on a near real time processing..." is slightly awkward. Replace it with standard phrasing such as "In contrast, products based on..." or "Conversely, products based on...".

Citation: https://doi.org/10.5194/essd-2026-376-RC3

Tanguy Szekely and Jean-Renaud Miadana

Data sets

Global Ocean in situ - Delayed Mode temperature and salinity CORA -objective analysis Tanguy Szekely https://data.marine.copernicus.eu/product/INSITU_GLO_PHY_TS_OA_MY_013_052/description

Tanguy Szekely and Jean-Renaud Miadana

Viewed

Total article views: 302 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
162	62	78	302	102	78

HTML: 162
PDF: 62
XML: 78
Total: 302
BibTeX: 102
EndNote: 78

Views and downloads (calculated since 29 Jun 2026)

Month	HTML	PDF	XML	Total
Jul 2026	50	21	4	75
Aug 2026	40	17	8	65
Sep 2026	72	24	66	162

Cumulative views and downloads (calculated since 29 Jun 2026)

Month	HTML	PDF	XML	Total
Jul 2026	50	21	4	75
Aug 2026	40	17	8	65
Sep 2026	72	24	66	162

Viewed (geographical distribution)

Total article views: 262 (including HTML, PDF, and XML) Thereof 262 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 29 Sep 2026

Short summary

The ocean stores vast amounts of heat and plays a central role in regulating Earth's climate. Monitoring its temperature and salinity is essential to understand climate change. We present an updated global ocean dataset covering the past six decades, built by combining historical measurements from ships, floats, and buoys. The dataset is freely available and regularly updated, providing scientists with a reliable reference for climate studies, ocean modeling, and satellite data verification.


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