the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Atlantic Meridional Heat Transports from Ocean Reanalyses, Monitored Sections and Heat Budget Constraints
Abstract. Meridional ocean heat transport (MHT) plays a central role in the climate system by redistributing energy across latitudes and influencing global and regional climate variability. While sustained observing systems in the Atlantic, subpolar North Atlantic and Arctic gateways provide valuable observational constraints on Atlantic heat transport, their spatial coverage remains limited and assessing their uncertainties remains challenging. Ocean reanalyses offer a complementary, spatially complete view of the ocean, but their ability to represent Atlantic heat transport across those gateways has so far not been consistently assessed.
Here we evaluate MHT estimates from 17 global ocean reanalyses using in-situ observational transport estimates and complementary heat-budget-derived transport estimates. The assessment covers major observing sections across the Atlantic and Arctic Oceans, including the South Atlantic Meridional Overturning Circulation Basin-wide Array (SAMBA), the RAPID array at 26.5°N, the Overturning in the Subpolar North Atlantic Program (OSNAP), the Greenland-Scotland Ridge (GSR), Davis Strait, Fram Strait and the Barents Sea Opening.
The two sets of estimates generally agree within two sigmas, despite the fact that biases and errors cannot be fully accounted for (whether in mooring-based estimates or in reanalyses). We demonstrate this level of agreement both for the meridional structure and for temporal variability in MHT. Temporal correlations are as high as 0.75–0.8 at RAPID and 0.65–0.75 at the GSR, but significant differences exist in both mean state and variability. Increasing resolution in reanalyses tends to improve agreement with observational estimates when computed from monthly temperature and velocity output. However, for one degree reanalyses it is essential to include the parameterized eddy terms (advective and diffusive) rather than using temperature and velocity output. Accounting for these parameterized contributions removes significant but artificial biases in MHT at this resolution.
The reanalysis ensemble mean shows lower mean heat transport than the mooring based estimates at RAPID and Fram Strait, while showing higher transport through the Barents Sea Opening. Hence there is significant uncertainty in terms of the partitioning of Arctic heat inflow pathways. Heat-budget-derived transport estimates show generally good agreement with the available transport observations across the evaluated sections. The representation of variability depends strongly on region and timescale. At RAPID and the GSR, the considered reanalyses show the highest consistency with observational estimates across seasonal to interannual timescales, whereas OSNAP is the most challenging section. Trend analyses from the reanalysis ensemble indicate a broad reduction of northward Atlantic heat transport across most gateways since the early 1990s, although substantial uncertainties remain prior to the Argo era. At the RAPID latitude (26.5°N), the full reanalysis ensemble indicates 1993–2023 declines of -0.058 ± 0.039 PW decade-1 in MHT.
Competing interests: At least one of the (co-)authors serves as editor for the special issue to which this paper belongs.
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.- Preprint
(12205 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- AC1: 'Author comment: Correction to Fig. 2', Susanna Winkelbauer, 17 Sep 2026
-
RC1: 'Comment on egusphere-2026-5435', Anonymous Referee #1, 05 Nov 2026
Review of the manuscript egusphere-2026-5435.pdf
General comments
The manuscript, entitled “Atlantic Meridional Heat Transport from Ocean Reanalyses, Monitored Sections and Heat Budget Constraints”, is an evaluation of MHT from 17 global ocean reanalyses products. The performance of the reanalyses is compared to MHT observed at several monitoring sections in the Atlantic Ocean from SAMBA in the south to the Arctic Gateways in the north. Inferred mean heat transport estimates are also calculated for the major observational sections and compared to the reanalyses. The manuscript presents comparison of mean MHT, temporal variability and trends.
Generally, the manuscript is within the scope of OS and the language is good. The methods are well described and the results section – which also seems to include discussion – includes several good figures illustrating the results of the analysis. The manuscript ends with a Conclusion section.
Scientific significance: Excellent (a large number of ocean reanalyses products are compared to several of the monitoring sections in the Atlantic)
Scientific quality: Good (Generally good, but some additional suggestions are added below)
Presentation quality: Good (The results are presented in a clear way, but there are some inconsistencies and lack of clarification)
Specific comments
You are comparing the ocean reanalyses to five different monitoring sections (Table 3), but there are (at least) two additional sections, namely the Tracos at 11S and NOAC at 47N (discontinued). Would it be possible to include at least the Tracos section in your analysis?
The trend analysis seems to indicate a general reduction of northward HT along the whole Atlantic Ocean. Here I miss a discussion of the contrasting increasing northward OHT observed at the GSR (e.g. Tsubouchi et al., 2021; Hansen et al., 2023 (https://doi.org/10.5194/os-19-1225-2023); Casanova-Masjoan et al., 2020 (https://doi.org/10.1029/2020JC016283) reporting increased volume transports and temperatures).
Throughout the manuscript the Inferred heat transport estimates are mentioned in several different ways (e.g. “heat-budget-derived transport estimates”, “heat-budget-based inferred estimates”, “inferred mean heat transport estimates”, “inferred ensemble estimates”, etc.). Especially in the beginning, this is quite confusing and I suggest to decide on one name and use this throughout.
Abstract
Line 9-11: Since all other sections are spelled out, RAPID should too.
Introduction
Line 48: “Arctic gateway monitoring programs (Tsubouchi et al., 2021)”: Arctic gateways usually refer to the Fram Strait, Barents Sea opening, Davis Strait and also the Bering Strait as in Tsubouchi et al., 2024. Tsubouchi et al., 2021, on the other hand, is looking at the Greenland Scotland Ridge or the “Arctic Mediterranean”. Please update or clarify what you mean here.
Line 61: 2023a instead of 2023b
Line 81: 2023b
Data and Methods
Line 103: in situ. Please also check for consistency through out as I think you also use in-situ.
Line 108: ??
Line 124: Abbreviation of MHT is introduced here, but first usage is at Line 87. Please update
Line 144: This title seems unfinished. Please see my comments above on “Heat Budget Derived estimates” and decide on a common term.
Line 150: Some of the variables/scalars in the equation are not sufficiently described. Please add a description for all of them.
Line 154: Do you mean the squared brackets?
Table 2: Also here, the variables are not sufficiently described in the caption or text.
Line 183-189: This paragraph is quite discussion like. Consider to move to an appropriate place in the Results section.
Line 198-202: There are no references for the uncertainties at the Arctic gateways, OSNAP and GSR. Is it possible to add these?
Line 207: MHT instead of “meridional heat transport”?
Line 224: The section below only refers to RAPID. I therefore suggest to change the title to “AMOC calculations at RAPID”.
Line 232: Similarly, I suggest here to change the title to “Mass flux constraints at OSNAP”
Line 240: StraitFlux was introduced on page five and not mentioned again until here. Please add the reference here also.
Line 241 and 245: Here you use “wet” cross-sectional area and “wet” grid cell. I suppose this is a terminology used by modellers? Although I think I know what you mean, I find that it would be more appropriate with a proper description.
Results
Consider to rename this section to "Results and discussion"
Line 288: Please add a reference for Gent-McWilliams parameterizations.
Line 296: Maybe add a reference to Table 3.
Conclusions
Line 538: Please clarify “UVT approach” as UVT is not mentioned elsewhere in the manuscript.
Citation: https://doi.org/10.5194/egusphere-2026-5435-RC1 -
RC2: 'Comment on egusphere-2026-5435', Anonymous Referee #2, 09 Nov 2026
This manuscript evaluates Atlantic and Arctic heat transports in 17 global ocean reanalyses against monitored sections and heat-budget-derived estimates. The broad coverage of sections and timescales makes it a useful reference for both producers and users of ocean reanalyses. The demonstration that coarse-resolution products require parameterized eddy and diffusive transport terms is also a useful practical result. My main concerns are the robustness of the multidecadal MHT trend, which is a central conclusion of the manuscript, and the interpretation of the systematic biases and resolution effects. I recommend major revision.
Major comments
1. Robustness of the multidecadal MHT trends
The 1993–2023 MHT trend at 26.5°N (−0.058 ± 0.039 PW/decade) is one of the main conclusions of the manuscript, but its uncertainty and robustness require further examination.
As described in Section 2.5 (lines 254–255), the uncertainty of the ensemble-mean trend combines the inter-product spread with the regression uncertainty. However, in Table A3 the individual 1993–2023 trends range from −0.150 to −0.007 PW/decade, yet the reported uncertainty of the ensemble mean (±0.039) is about the same as the regression uncertainty of a single product. For 2004–2023, the reported ±0.052 is even smaller than every individual uncertainty in the table. It is therefore unclear how the inter-product spread enters the combined uncertainty. If the spread is converted into an uncertainty of the ensemble mean, the assumption of independence among reanalysis products needs to be justified, particularly given their shared model configurations. A different treatment of the spread could affect the statistical significance of the 1993–2023 trend. Please provide the exact formula used and check these values.
The trend is also sensitive to how the ensemble is constructed. Differences among products are much larger before Argo, and Table A3 reports trends calculated over each product's available record rather than a common period. Trends over common analysis periods should be provided, together with their sensitivity to the starting year, particularly around the onset of Argo. SODA4 is excluded because of its low-frequency variability (line 446), which is a selection based on the results themselves. Its trend and the effect of including it in the ensemble should therefore be reported. Given the shared model configurations among some products, a sensitivity test excluding one model family at a time would also be useful.
The abstract states a reduction of heat transport across most gateways since the early 1990s, but trends at the individual gateway sections are not shown. Tsubouchi et al. (2021) reported increasing heat transport across the GSR, where observations are available from 1993. The reanalysis trends at the gateway sections should be shown and compared with these observations. A comparison with the longer AX07 record at 30°N (since 1994) and the Dong et al. (2021) estimate at 26.5°N (since 1993) would also help assess the trend before the RAPID period.
The heat-budget-derived trends in Table A4 also use the GSR/Davis boundary for 1993–2023, although the Davis record starts in 2004 and the GSR record ends in 2021 (Table 3). Please explain how the boundary transports are specified outside these periods.
2. Interpretation of MHT biases and resolution effects
The ensemble mean at RAPID is approximately 0.30 PW lower than the observational estimate. The manuscript suggests that methodological differences may contribute to this discrepancy (lines 313–316), but their magnitude is not quantified. It would be useful to apply a RAPID-like diagnostic to at least one representative reanalysis and compare the result with the directly integrated transport. The close agreement between the heat-budget-derived estimate (1.18 PW), which does not rely on geostrophic assumptions, and the RAPID observations should also be considered when discussing whether methodological differences can explain the systematic reanalysis bias.
The interpretation of resolution effects also needs qualification. The importance of bolus and diffusive terms is demonstrated for global MHT in ECCOv4r4 and ESTOCv7a (Figure 3), but their contribution to Atlantic MHT at the monitored sections is not shown. Table A1 suggests that the RAPID transport increases mainly from the 1° to the 1/4° products, with no clear further increase at higher resolution. However, the low values of the 1° products are strongly affected by CIGAR (0.50 PW, without parameterized terms) and the two ESTOC products. Resolution also varies alongside model formulation, forcing and assimilation methods, making it difficult to isolate its effect.
The statement in the abstract that increasing resolution improves agreement with observations (lines 15–16) should therefore be qualified. Similarly, the conclusion that the highest-resolution products do not outperform the 1/4° systems is based on only two products, one of which (SODA4) is excluded from the trend analysis because of its unrealistic variability.
Minor comments
Lines 71–87. The Introduction could be shortened, particularly the description of MER-EP and ORA-IP (lines 71–79) and the heat-budget method (lines 80–87), which is discussed again in Section 2.2. More space could instead be given to recent changes in Atlantic heat transport and the AMOC, since the multidecadal trend is a central result but receives little attention in the Introduction.
Equation 1. Please specify the seawater density and specific heat capacity used in the calculations. It would also help to clarify whether the observational and reanalysis transports use consistent section positions and reference temperatures, particularly at the Arctic gateways, where the net volume transport across individual sections is nonzero and the heat transport therefore depends on the reference temperature.
Section 2.2. Please specify the boundary configuration, terms and datasets used for each inferred estimate shown in the figures, and how the adjustment is applied. For example, the caption of Figure 11 lists the atmospheric reanalyses but not the boundary configuration or adjustment method. Some inputs are not independent of the evaluation, as the Davis and GSR boundary transports are also evaluation sections and ORAS5 is one of the evaluated reanalyses (lines 178–181). For the Davis + GSR configuration, the prescribed boundary transport (0.31 PW) constitutes a substantial part of the inferred OSNAP heat transport, which should be considered when interpreting its agreement with the OSNAP observations. The terms “heat-budget-derived”, “inferred” and “energy-budget” estimates should also be used consistently.
Section 2.5. The reanalysis uncertainty is stated to be ±2σ, but at RAPID the reported ±0.16 PW (Figure 4) appears small given that the values in Table A1 range from 0.50 to 1.13 PW. Please check that the uncertainties are consistently defined across sections, and state whether the observational uncertainties represent 1σ or 2σ.
Figure 3 and Table 1. Figure 3 uses ECCOv4r4 (1992–2017), whereas Table 1 and the rest of the analysis use ECCO4r6. Please clarify why a different release is used here. The ECCO Consortium references in Table 1 also describe v4r4 rather than r6.
Figure 4. Please state in the caption the period over which the means are calculated.
Section 3.2. Correlations based on 12-month running means have far fewer independent samples than the number of monthly values, especially for the short OSNAP and SAMBA records. Please provide significance estimates that account for temporal autocorrelation.
Lines 355–358. At SAMBA, the mean-state comparison and the correlation analysis use the altimetry-based estimate of Dong et al. (2021), which the authors note is not independent of the reanalyses. The comparison with the mooring-based SAMBA record should also be given for its period (2013–2017), so that the reader can see how the conclusions at this section depend on the choice of benchmark.
Lines 510–511. The inter-product spread is reported to be about 65% larger in 1993–2004 than in 2005–2023. If this is calculated from anomalies referenced to the 2005–2023 mean (Figure 11), the spread within the reference period is reduced by construction. Please check whether this result holds with a different reference period.
Table A3. Please explain why the coarse-resolution CIGAR product is absent from the trend table, in addition to the exclusions already discussed in the text (lines 445–448).
Citation: https://doi.org/10.5194/egusphere-2026-5435-RC2
Model code and software
StraitFlux Winkelbauer Susanna https://github.com/susannawinkelbauer/StraitFlux/tree/main
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,091 | 343 | 76 | 1,510 | 75 | 70 |
- HTML: 1,091
- PDF: 343
- XML: 76
- Total: 1,510
- BibTeX: 75
- EndNote: 70
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
We noticed a plotting error in Fig. 2b of the preprint affecting the observational estimate for the combined GSR + Davis Strait heat transport. Due to an erroneous unit conversion when generating the figure, this observational marker was plotted close to zero instead of at approximately 0.30 PW.
The corrected version of Fig. 2 is provided below. This error is restricted to the position of this observational marker, the underlying data and all other results and conclusions of the manuscript are unaffected. Fig. 2 will be corrected accordingly in the revised manuscript.