the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 Aug 2025
Subpolar Atlantic meridional heat transports from OSNAP and ocean reanalyses – a comparison
Abstract. Ocean reanalyses are potentially useful tools to study ocean heat transport (OHT) and its role in climate variability, but their ability to accurately reproduce observed transports remains uncertain, particularly in dynamically complex regions like the subpolar North Atlantic. Here, we evaluate currents, temperatures, and resulting OHT at the OSNAP (Overturning in the Subpolar North Atlantic Program) section by comparing OSNAP observations with outputs from a suite of global ocean reanalyses. While the reanalyses broadly reproduce the spatial structure of currents and heat transport across OSNAP West and East, systematic regional biases persist, especially in the representation of key boundary currents and inflow pathways.
Temporal variability is well captured at OSNAP West, but none of the reanalyses reproduce the observed OHT variability at OSNAP East, especially a pronounced peak in 2015. This discrepancy in 2015 is traced to the glider region over the eastern Iceland Basin and Hatton Bank, where OSNAP data show a strong, localized inflow anomaly associated with the North Atlantic Current (NAC). This signal is absent from all reanalyses as well as from independent, indirect heat transport estimates based on surface heat fluxes and heat content. Investigation of sea level anomalies and implied geostrophic currents further confirm that this mismatch is mainly driven by differences in flow structure rather than temperature anomalies alone.
Our results highlight both the value and limitations of reanalyses in capturing subpolar heat transport variability. While higher-resolution products such as GLORYS12V1 better represent circulation features, significant mismatches remain, especially in regions with sparse observational coverage. The findings underscore the need for improved observational networks and higher-resolution modeling to more accurately constrain subpolar OHT.
- Preprint
(7171 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 03 Jan 2026)
- RC1: 'Comment on egusphere-2025-4093', Anonymous Referee #1, 19 Oct 2025 reply
-
RC2: 'Comment on egusphere-2025-4093', Anonymous Referee #2, 16 Dec 2025
reply
Review of “Subpolar Atlantic meridional heat transports from OSNAP and ocean reanalyses - a comparison” by Winkelbauer et al.
The manuscript presents an interesting comparison of the Ocean Heat Transport (OHT) over the subpolar gyre between reanalyses and the OSNAP observations over the OSNAP period. To my view, one of the main conclusions of the manuscript is that none of the reanalyses evaluated in this study reproduce the OHT variability at OSNAP East, which is explained by biases in both the overturning (biased high) and gyre (flattens the curves) terms. They further evaluate the lack of peak in OHT in 2015 from reanalyses, but were inconclusive as to what can explain this disagreement, observations being too sparse over the interior of the OSNAP array.
The study provides a valuable comparison of the OHT between reanalysis and observations, it is well-written and the results are timely, but I recommend clarifying some of the conclusions, as indicated in my major comments, before publication of the manuscript.
Major Comments
- As I understand it, the authors use the OSNAP data as gold standard for the evaluation of OHT in different reanalyses. However, they repeatedly conclude that most of the differences could be attributed to a lack of observations at specific depths and areas along the section, as OSNAP uses constant fields in the interior and relies on end-point dynamic height moorings to capture the total integrated transport and its variability (e.g. of conclusion at l.213, l.229, l.233, l.289, l.295, l.316, l.394, l.411). In this context, can the authors comment on the choice of OSNAP data as a gold standard to evaluate these reanalyses while there are not enough OSNAP observations in the interior to conclude anything on the possible causes for the total OHT differences? In a way, the reanalyses might assimilate more observations in the interior than the OSNAP product, meaning that their OHT distribution could be considered as closer to the truth than the one from the OSNAP product. Hence, how can we reliably assess inconsistencies between reanalyses and OSNAP?
- A specific example of my main comment #1 is the discussion of an anticyclonic eddy in the NAC region at lines 309–317. Can the authors clarify their conclusion: is the inconsistency coming from the resolution of the reanalyses at the boundaries or from a lack of observations in the available OSNAP dataset? If these two points are valid, how a validation of one dataset as compared to the other can be convincing? Related to this specific point, I am not sure to understand why the anticyclonic eddy can be observed in the glider data (as discussed in Lozier et al., 2017) but not in the final OSNAP product?
- To add some clarity in the differences between the data used in this study, I suggest changing the structure of section 2 by adding a ‘section 2.1 Data’ that would introduce the data used in this study: OSNAP (including lines 101–106 currently in the following section), the reanalyses and the altimetry data. Sections 2.1 would become section 2.2 etc..
- In the new section 2.1, I strongly recommend the authors discussing the differences and similarities between the reanalyses in terms of observations assimilated in these reanalyses. For example, can the authors clarify if the reanalyses assimilate OSNAP observations? Are they assimilating the same observations otherwise (e.g., Argo, altimetry, hydrographic sections…) meaning that their differences in OHT (or for example the results discussed at l.239-241) can be interpreted as a result of different horizontal resolutions and dynamical models only?
- Finally, I recommend the authors discussing another OHT dataset produced by combining Argo, altimetry and gravimetry data from Calafat et al., 2025. https://doi.org/10.5194/os-21-2743-2025
Minor Comments
L. 62: use AMOC instead of MOC, as it was the authors’ choice for the rest of the manuscript
L. 108: consider using the term ‘derived’ instead of ‘calculated’
L. 114-115: Consider clarifying the grid that is used for the bilinear interpolation.
L. 128-130: The sentences don’t read properly. Maybe: ‘Additionally, the potential temperature and a reference temperature are needed for estimating the heat transport. An unambiguous heat transports require closed volume transports, which is […].’
L. 133: Can the authors describe in few sentences what is the StraitFlux’s line integration method?
L. 198: Typo ‘Iceland basin’
L. 203-205: EN4 includes a large number of Argo profiles and has probably a better spatial coverage over the subpolar North Atlantic than OSNAP that is missing observations in the basin interiors. However, there are possible issues of data QC in EN4. Consider also discussing in more details what structural and methodological uncertainties in OSNAP can explain these differences. From my understanding, OSNAP uses EN4 in the interior?
L. 270: Related to my main comment #4, I recommend the authors to clarify in the new section 2.1 if they use independent data sources in the reanalysis.
L. 290-292: I am confused, didn’t the authors say that OSNAP cannot represent the circulation over this portion of the interior array because there aren’t enough observations there?
L. 295: Consider clarifying if the potential uncertainties in these areas relates to uncertainties in the reanalyses or OSNAP observations?
L.320: Is the barotropic compensation applied at OSNAP uniform in time or in space (horizontally and vertically) or both? How can it impact the correlation of OSNAP with SLA?
L. 338-346: Related to my main comment #4, it would be easier to interpret this result with more details on the dataset assimilated by the different reanalyses. Do these two reanalyses (GLORYS2v4 and GLORYS12V1) assimilate the same SLA data? Why not showing the results for the other reanalyses?
L. 347-355: Can the authors clarify why smooth all fields over 1deg resolution while the coarser resolution from reanalysis or altimetry is 1/4deg? In my view, only GLORYS12V1 should be smoothed at 1/4deg.
Citation: https://doi.org/10.5194/egusphere-2025-4093-RC2 -
EC1: 'Comment on egusphere-2025-4093', Meric Srokosz, 18 Dec 2025
reply
Given the overall positive comments by the reviewers I would encourage the authors to submit a revised version taking account of and responding to those comments.
Citation: https://doi.org/10.5194/egusphere-2025-4093-EC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,302 | 130 | 22 | 1,454 | 47 | 40 |
- HTML: 1,302
- PDF: 130
- XML: 22
- Total: 1,454
- BibTeX: 47
- EndNote: 40
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of “Subpolar Atlantic meridional heat transports from OSNAP and ocean reanalyses – a comparison” by Winkelbauer et al.
The authors compare the meridional heat transport in various ocean reanalysis products to the Overturning in the Subpolar North Atlantic Program (OSNAP). The authors find that the reanalyses capture many of the relevant components of the subpolar property field and circulation patterns, and thus express a confidence in using reanalyses. There are notable limitations in the ocean reanalyses that the authors explain well.
I found the paper to be interesting addition to the literature. The manuscript is well-organized and well-written, and the figures clearly show the key results of the work. I recommend the paper be accepted pending minor revisions, which I detail below.
Revisions to be addressed prior to publication:
Overarching comment:
Heat transport cannot be calculated for cross-sections that do not conserve mass because the results are sensitive to the choice or reference temperature. Either the authors need to show that the lines are mass-conserving (or at least volume-conserving), or use the term “temperature transport” rather than “heat transport” (as in Johns et al. 2011). To this end, it would be good to show the net volume transport for each of the reanalyses. This is shown to some degree in Fig. A3, but a time series of volume transport each month for each product would be enlightening.
Specific comments:
l. 54-55: “As reanalyses generally do not assimilate direct observations of ocean currents, their transport estimates depend largely on model dynamics and parameterizations rather than observational constraints” – this is not entirely true. Ocean reanalyses assimilate SSH and T/S, which together constrain the geostrophic circulation. Most of the AMOC (and resulting MHT) is in geostrophic balance, thus the components of the velocity field that are important to this paper are indeed assimilated. The one exception to this would be the boundary currents, where direct velocity measurements from ADCPs and current meters are indeed not assimilated by the reanalyses. This sentence should be rewritten to convey this information.
l. 83-85: are the vertical cross-sections from GLORYS12V1 re-mapped onto a ¼° grid to be comparable to the other reanalyses? If not, the mean RMSE shown in Figs. 3 and 5 could be aliased by the different spatial resolution.
l. 86: “…they differ in their data assimilation methods…” it would be good to clarify what these differences are. A table would be a good way to organize this information.
l. 90: “they can be considered independent of OSNAP in that regard”. As mentioned above, though the velocities are not assimilated, much of the OSNAP velocity field is determined from SLA and geostrophy so the only place there is any independence is in the boundary currents. This should be specified.
Fig. 1: what is the mooring in the center of the Labrador Sea?
l. 133: the reanalyses used in this paper are not volume (or mass) conserving so to which ‘conservation properties’ are the authors referring?
l. 136: when the heat transports are calculated at monthly time scales, is it calculated from the monthly mean of the heat transport or calculated from the monthly mean velocity and temperature fields? The former accounts for the v’T’ term, while the other does not.
l. 136: the authors refer to a 0.5% error… is this a percentage of PW? Heat transport has very small variability compared to its mean value. So it would be more clear if the authors just reported a value of heat transport in PW rather than a %.
l. 157: What is meant by “Mass-consistent heat transport estimates”?
Table 1: this is an impressive list of data sets. Why was JRA-55 used rather than the updated version (JRA-3Q)?
Fig. 3: Consider using a different colorbar to depict RMSE – at first look, this appears as a consistent high bias in the reanalyses compared to OSNAP.
l. 315 and 404: it is unclear to me whether this 2015 event was captured by OSNAP because there was a glider in that year (and not afterwards), or if this was truly an anomalous event. It would be interesting to analyze an OSNAP gridded section that does not include the glider in 2014-2016. Does the event appear if the glider is not included? Determining whether the event is real or an artifact of changing observational structure would go a long way toward understanding the authors’ thoughts in the conclusions about the importance of a consistent set of observations.
Fig. 11: the units on the x-axis are a bit strange… 2.5-3.3 x 106 m… I suggest using km and specifying that this refers to the along-section distance from the OSNAP western boundary.
l. 412-418: In this paragraph, the authors express more confidence in reanalyses products than is justified from the results of this paper. While it is true that the discrepancy in OHT between OSNAP and the reanalyses in 2015 is interesting and raises questions about the coverage and consistency of OSNAP, the authors have not presented any independent evidence that reanalyses can provide error estimates for the observing system (OSNAP in this case). I agree that this is a possible use of ocean reanalyses once they are validated, but the authors would need to present independent data that justify this usage. Given how much the reanalyses disagree with one another (in this paper and in others, e.g. Jackson et al. 2019), I would proceed down this path with a lot of caution – and much more caution than interpreting the direct observations from OSNAP.
Conclusions: the authors could also mention the use of reanalyses to replace the use of moorings in regions where lower frequency variability is dominant. This would save costs and is currently being pursued by the RAPID team (Petit et al., (in review)).
References:
Johns et al. (2011): https://doi.org/10.1175/2010JCLI3997.1
Jackson et al. (2019): https://doi.org/10.1029/2019JC015210