## 1. Introduction

At midlatitudes, the atmospheric heat transport is performed by transient disturbances, large-scale cyclones, and anticyclones, as discussed, for example, by Kuo (1956). At the same latitudes, in the ocean, both the time-mean circulation and the transient eddies contribute to the meridional heat transport (Smith et al. 2000). The importance of the time-mean circulation, in the case of the oceanic heat transport, is due to the existence of large-scale currents such as the Gulf Stream, which flows poleward along the western boundary of the North Atlantic. The oceanic heat transport can be inferred from air–sea surface fluxes (Large and Yeager 2009), and these estimates show that despite its relatively small width compared to the Pacific, the Atlantic Ocean performs half the global oceanic heat transport in the 20°–40°N latitude range. This has been recently confirmed at 26.5°N by the in situ measurements of the RAPID array, which gave an Atlantic heat transport of 1.25 ± 0.36 PW over 8 years (McCarthy et al. 2015).

This paper is focused on the eddy heat transport, namely, the time average of the product of velocity and temperature temporal fluctuations. In our definition, “eddy” includes all the time variability at periods larger than a few days; this variability consists of coherent eddies but also waves, meanders, and large-scale modes of variability. With this definition, the eddy heat transport is small at the latitude of RAPID: 0.08 ± 0.03 PW (McCarthy et al. 2015). On the contrary, farther north, downstream of the Gulf Stream separation, satellite observations of surface temperature and geostrophic velocity have revealed a large, localized eddy heat flux (Zhai and Greatbatch 2006). These authors have estimated an equivalent horizontal eddy diffusivity of 1000 to 2000 m^{2} s^{−1} for this eddy flux, but they have not investigated its origin further.

Estimating the eddy contribution to heat transport over the whole North Atlantic requires knowledge of the time-evolving, three-dimensional velocity and temperature fields, information that only high-resolution numerical simulations can provide. Since the pioneering study of Smith et al. (2000), numerical models at resolutions of ⅝° or higher have been used to estimate the meridional heat transport in the Atlantic Ocean. These models all suggest that the eddy heat transport is largest in the 35°–40°N latitude band. There it can exceed 0.3 PW and accounts for one-quarter to one-third of the total meridional heat transport (Smith et al. 2000; Hecht and Smith 2008; Treguier et al. 2012). This contrasts with lower-resolution models, such as the Parallel Ocean Climate Model (POCM) model with ½° resolution at the equator analyzed by Jayne and Marotzke (2002), where the eddy heat transport is lower (0.1 PW, one-tenth of the total, at 40°N). Comparing numerical models at increasing resolution, Hecht and Smith (2008) and Treguier et al. (2012) find that the total meridional heat transport increases with resolution, but they also note that at most latitudes the increase is due to changes in the time-mean circulation more than an increase of the eddy transport. Indeed, an increased contribution of transient eddies at a given latitude does not necessarily cause a larger total transport: the time-mean flow is modified at higher resolution, in such a way that eddy and time-mean flow contributions tend to compensate each other, as demonstrated by Cox (1985) and further discussed by Bryan (1986).

The dynamics of the Gulf Stream and its eddy fluxes are often considered in the framework of a free eastward jet subject to baroclinic instability. In the classical models, such as Phillips or Charney’s (e.g., Pedlosky 1979), instability leads to the development of waves and eddies and thus generates a cross-jet eddy heat flux that tends to reduce the temperature gradient across the front. However, the Gulf Stream is a western boundary current, much more complex than a free zonal jet. In a recent analysis based on a high-resolution model (7-km grid), Kang and Curchitser (2015) have considered separately three regions with different dynamics, upstream and downstream of the Gulf Stream separation from the coast, and confirmed that both barotropic and baroclinic energy conversions are active. However, they have not analyzed the heat balance or eddy heat fluxes.

Here, we use recent satellite observations and ⅞° model simulations to understand the origin of the eddy heat flux in the 35°–40°N latitude band. We argue that the structure of the Gulf Stream, and more precisely the spatial shift between velocity and temperature, plays a key role in the generation of a localized eddy flux, which has a significant contribution to the total meridional heat transport in the Atlantic.

## 2. Methods

*υ*and temperature

*T*, the time-averaged heat transport results from the sum of two terms:

Following Zhai and Greatbatch (2006), we compute the eddy flux ^{−2} K^{−1} in the North Atlantic (Barnier et al. 1995). Further details as well as validations with observations are found in Serazin et al. (2015), who used ORCA12 to estimate the intrinsic variability of the sea level, and in Barrier et al. (2015), who studied the interannual variability of the North Atlantic Subpolar Gyre. The Gulf Stream dynamics in our ⅞° models have been studied in detail by Maze et al. (2013), who analyzed the potential vorticity budget of the 18°C water in an earlier regional version of the Drakkar model. The analysis of eddy fluxes of salt in the same regional configuration by Treguier et al. (2012) has motivated the present investigation of the eddy heat flux.

## 3. Structure of the eddy heat transport near the Gulf Stream separation

The surface meridional eddy flux

*υ*, temperature

*T*, Coriolis parameter

*f*, gravitational acceleration

*g*, and thermal expansion coefficient

*α*:

*z*and taking the derivative in

*x*, one obtains

*z*, the lhs of (3) vanishes at the longitude of maximum velocity

*x*

_{m}(

*z*). If the location of the maximum temperature gradient changes with depth (is tilted), there is no reason for this longitude

*x*

_{m}(

*z*) to correspond to the longitude of the maximum temperature gradient at the same depth, due to the vertical integral on the right-hand side.

Such a spatial shift between the maximum velocity and the maximum temperature gradient is probably present in the observations of Rossby (1987), but it is marginally resolved by the spacing between the profiles (24 km); it seems present in Fig. 2 of Toole et al. (2011). The characteristic shape of the vertical profiles in Fig. 2 was found in the earlier simulations analyzed by Treguier et al. (2012) and Treguier et al. (2014). It is likely that all models with high enough resolution reproduce this observed feature of the Gulf Stream, although this has not been documented in the literature.

## 4. Mechanisms of eddy heat transport

What source of variability can be responsible for the localized

The top two panels of Fig. 3 show the surface temperature, meridional velocity, and their correlation at 36°N in the ORCA12 simulation. The shift in longitude between the maximum temperature gradient and the maximum velocity is 0.58°. Regarding the origin of this shift at the surface, (3) shows that in a geostrophic current it may be due to the reference velocity *υ*(*x*, −*H*) or to the tilting with the depth of the maximum temperature gradient, that is, its origin can be barotropic or baroclinic. For ORCA12, in the region shown in Fig. 3, the barotropic velocity is 10% of the surface velocity, and the shift is mainly baroclinic. In the top-right panel of Fig. 3, the eddy temperature–velocity correlation is positive. Locally, the eddy term appears small compared to the mean, but the mean includes a nondivergent component that would cancel out in a basinwide average. For this reason, the eddy to mean ratio is not significant in this small region, although the eddy flux itself is relevant, as will be shown in section 5. Assuming that the eddy temperature–velocity correlation keeps the same amplitude over the top 300 m of the water column, we can compute an equivalent heat flux. The integral over the longitude band 76°–70°W is 0.25 PW, comparable to the basinwide eddy heat flux in ORCA12 as well as in other numerical models (Smith et al. 2000; Hecht and Smith 2008; Treguier et al. 2012).

*υ*:

*x*is perpendicular to the jet axis,

*υ*

_{0}is the velocity amplitude, and

*L*is the jet width. We assume that the jet axis meanders sinusoidally with typical amplitude

*x*

_{m}, as a function of time

*t*, so that the argument

*x*

_{υ}is a function of

*x*and

*t*:

*T*

_{0}= 16°C,

*δT*= 8°C,

*υ*

_{0}= 1.2 m s

^{−1}, and jet width

*L*= 0.58°. We assume that the meander width

*x*

_{m}is equal to

*L*. The velocity and tracer averaged over one meander period are plotted in Fig. 3 (middle-left row). Note that if the meander width was much larger than the jet width, the time-mean velocity would exhibit two peaks rather than a single one. Our choice of parameters is meant to produce a qualitative agreement with the ORCA12 model, where the mean velocity has a single maximum. The tracer transport

*δx*. Let us define

*x*

_{t}:

*x*

_{t}as argument instead of

*x*

_{υ}. We take

*δx*= 0.58° to fit the ORCA12 model. The resulting time-mean velocity, temperature, and their correlations are shown in Fig. 3 (bottom row). With the shift between the velocity and temperature gradient, the eddy flux is no longer symmetric, and there is a net positive eddy flux across the jet. Its order of magnitude is similar to the one diagnosed in ORCA12; the eddy correlation, integrated over longitude and over a depth of 300 m, would generate a total eddy heat transport of 0.26 PW. Thus, in this case of noncoincident maximum velocity and temperature gradient, when the jet meanders, the eddy/mean decomposition based on the time average gives a significant eddy flux integrated across the jet. This eddy contribution is compensated by a change in the mean tracer flux so that the total remains the same as in the absence of meanders. We think that this simple kinematic model is relevant to explain eddy fluxes of heat and salt in the model Gulf Stream near its separation from the coast. It demonstrates that a meandering jet can produce transient eddy fluxes, provided that a spatial shift exists between the velocity and tracer gradient structures [a case not investigated by Jayne and Marotzke (2002)]. This may also explain why the eddy fluxes are at their maximum near the Gulf Stream separation because although both meander amplitude and eddy kinetic energy increase downstream, the jet becomes less asymmetric where it flows in open water far from the continental slope, and the shift between the maximum velocity and temperature gradient is reduced.

Of course, the Gulf Stream in the ORCA12 model has a more complex behavior than just a regular meandering. Hovmöller diagrams of surface temperature and velocity (not shown) display irregular meanders and variability at multiple time scales. The sea surface temperature also exhibits a strong seasonal cycle (although this seasonal cycle has a small impact, less than 10%, on the peak of maximum *x*, *z*) is considered. The temperature field can be defined as a function of depth, and the velocity can be calculated as the sum of a geostrophic velocity and a barotropic reference velocity. In that case we find that a regularly meandering Gaussian jet can account for 25% of the total ORCA12 heat transport, but we have not explored the parameter space further. Here, the kinematic model is used merely to illustrate a contribution to eddy heat transport. One should not expect it to reproduce the complex dynamics of the Gulf Stream separation.

## 5. Consequences for the Atlantic meridional heat transport

Although localized near the western boundary, the eddy heat flux pattern pictured in Fig. 1 is a significant contribution to the meridional heat transport integrated over the basin in the numerical model. The total meridional heat transport in the North Atlantic, averaged over the years 2003–12, is presented as a function of latitude in Fig. 4 with its decomposition into mean and eddy (top). The total heat transport, reaching a maximum of 1 PW, is quite realistic, in the lower range of uncertainty of the RAPID observations at 26°N (Johns et al. 2011; McCarthy et al. 2015).

As appears from the thick line in Fig. 4, the main effect of the eddy

Here, we focus on the latitude band corresponding to the Gulf Stream separation. There is a sharp increase (divergence) of the eddy heat flux in that latitude range, with the eddy flux increasing from a small negative value at 33°N to reach a maximum of 0.31 PW at 36.6°N. Our new simulation confirms the finding of Treguier et al. (2012); this large eddy heat flux is due to the dynamics at the western boundary. This appears very clearly when the eddy heat transport at that latitude is plotted, cumulated from the west, as a function of longitude (Fig. 4, bottom). The eddy heat transport quickly reaches a large value within 5° from the coast, due to the positive pattern of

## 6. Conclusions

We have confirmed, in observations and a numerical model, the result of Zhai and Greatbatch (2006); there is a large positive eddy heat flux northwest of the Gulf Stream axis, right after its separation from the coast at Cape Hatteras. This localized positive

Our simple kinematic model implies a large compensation between time-mean and transient eddy flux. The analytical meandering jet has the same transport of tracer, whether it meanders or not, but the phase shift between the jet axis and the temperature profile causes a nonzero anomaly in the transport by the time-mean flow that compensates the transient eddy transport. The kinematic model is overly simplified, ignoring dynamics and diabatic mixing, but it is interesting to note that the same behavior occurs in ORCA12. A similar compensation between eddy and mean heat transport appears clearly in the basin-averaged meridional heat flux (Fig. 4). A large eddy mean cancellation in that latitude band is found not only in ORCA12, but also in many other numerical simulations. Consider, for example, the rapid growth of the eddy heat transport from −0.2 to 0.2 PW between 34° and 36°N in the POP 0.1° model (Hecht and Smith 2008): it is not reflected in the total heat transport and must therefore be compensated by a corresponding decrease of the transport by the time-mean flow. This compensation is consistent with our simple kinematic model of the meandering jet. Furthermore, it suggests that this compensation between eddy and mean operates on short time scales (the time scale of a meander). Therefore, if our hypothesis is right, local changes in the characteristics of the Gulf Stream separation in a changing climate would not affect the total heat transport much but would simply modify the repartition between the compensating transient eddy and time-mean transports. More complex mechanisms are certainly at play farther northeast, along the North Atlantic drift at the boundary between the subtropical and the subpolar gyre. There meridional eddy fluxes are likely governed by baroclinic instability, and the relationship between large-scale gradients and eddy fluxes must therefore involve longer time scales (interannual to decadal) with nontrivial consequences for the ocean response to climate change. Finally, the fact that eddy and mean heat fluxes compensate each other locally does not mean that a low-resolution model can represent the Gulf Stream accurately. In noneddying models, the Gulf Stream is laminar and its transport is governed by linear Sverdrup dynamics [with an amplitude of about 30 Sv (1 Sv ≡ 10^{6} m^{3} s^{−1})]. In high-resolution models, as in the real world, the Gulf Stream transport is enhanced by inertial and eddy rectification mechanisms and exceeds 100 Sv. The nonlinearity of the Gulf Stream has a far-reaching influence on the shape of the subtropical gyre; on the air–sea exchanges of heat, freshwater, and carbon; and on the ocean ecosystems.

## Acknowledgments

We thank the reviewers and Paola Cessi, the editor, for their remarks, which helped improve the manuscript. This work is a contribution of the Drakkar project, which is funded by the Centre National de la Recherche Scientifique (CNRS), the Institut National des Sciences de l’Univers (LEFE-INSU), the Groupe Mission Mercator Coriolis (GMMC), and Ifremer. A. M. Treguier acknowledges support from LabexMER (Grant ANR-10-LABX-19-01). The numerical simulation ORCA12.L46-MJM88 presented in this study has been performed at the Centre Informatique National de l’Enseignement Superieur (CINES) computing center, operated by GENCI.

## REFERENCES

Barnier, B., L. Siefridt, and P. Marchesiello, 1995: Thermal forcing for a global ocean circulation model using a three-year climatology of ECMWF analyses.

,*J. Mar. Syst.***6**, 363–380, doi:10.1016/0924-7963(94)00034-9.Barnier, B., and Coauthors, 2014: DRAKKAR: Developing high resolution ocean components for European Earth system models.

*CLIVAR Exchanges*, Vol. 65, International CLIVAR Project Office, Southampton, United Kingdom, 18–21.Barrier, N., J. Deshayes, A. Treguier, and C. Cassou, 2015: Heat budget in the North Atlantic Subpolar Gyre: Impacts of atmospheric weather regimes on the 1995 warming event.

,*Prog. Oceanogr.***130**, 75–90, doi:10.1016/j.pocean.2014.10.001.Brodeau, L., B. Barnier, A. Treguier, T. Penduff, and S. Gulev, 2010: An ERA40-based atmospheric forcing for global ocean circulation models.

,*Ocean Modell.***31**, 88–104, doi:10.1016/j.ocemod.2009.10.005.Bryan, K., 1986: Poleward buoyancy transport in the ocean and mesoscale eddies.

,*J. Phys. Oceanogr.***16**, 927–933, doi:10.1175/1520-0485(1986)016<0927:PBTITO>2.0.CO;2.Cox, M., 1985: An eddy resolving model of the ventilated thermocline.

,*J. Phys. Oceanogr.***15**, 1312–1324, doi:10.1175/1520-0485(1985)015<1312:AERNMO>2.0.CO;2.Gula, J., M. J. Molemaker, and J. C. McWilliams, 2015: Topographic vorticity generation, submesoscale instability, and vortex street formation in the Gulf Stream.

,*Geophys. Res. Lett.***42**, 4054–4062, doi:10.1002/2015GL063731.Hall, N., B. Barnier, T. Penduff, and J. Molines, 2004: Interannual variation of Gulf Stream heat transport in a numerical model forced by reanalysis data.

,*Climate Dyn.***23**, 341–351, doi:10.1007/s00382-004-0449-2.Hausmann, U., and A. Czaja, 2012: The observed signature of mesoscale eddies in sea surface temperature and the associated heat transport.

,*Deep-Sea Res. I***70**, 60–72, doi:10.1016/j.dsr.2012.08.005.Hecht, M., and R. Smith, 2008: Towards a physical understanding of the North Atlantic: A review of model studies.

*Ocean Modeling in an Eddying Regime*,*Geophys. Monogr.*, Vol. 177, Amer. Geophys. Union, 213–240.Jayne, S., and J. Marotzke, 2002: The oceanic eddy heat transport.

,*J. Phys. Oceanogr.***32**, 3328–3345, doi:10.1175/1520-0485(2002)032<3328:TOEHT>2.0.CO;2.Johns, W. E., and Coauthors, 2011: Continuous, array-based estimates of Atlantic Ocean heat transport at 26.5°N.

,*J. Climate***24**, 2429–2449, doi:10.1175/2010JCLI3997.1.Kang, D., and E. N. Curchitser, 2015: Energetics of eddy–mean flow interactions in the Gulf Stream region.

,*J. Phys. Oceanogr.***45**, 1103–1120, doi:10.1175/JPO-D-14-0200.1.Kuo, H., 1956: Forced and free meridional circulations in the atmosphere.

,*J. Meteor.***13**, 561–568, doi:10.1175/1520-0469(1956)013<0561:FAFMCI>2.0.CO;2.Large, W. G., and S. G. Yeager, 2009: The global climatology of an interannually varying air–sea flux data set.

,*Climate Dyn.***33**, 341–364, doi:10.1007/s00382-008-0441-3.Madec, G., 2008: NEMO ocean engine. Institut Pierre-Simon Laplace Note du Pole de Modélisation 27, 300 pp.

Maze, G., J. Deshayes, J. Marshall, A. Treguier, A. Chronis, and L. Vollner, 2013: Surface vertical PV fluxes and subtropical mode water formation in an eddy-resolving numerical simulation.

,*Deep-Sea Res. II***91**, 128–138, doi:10.1016/j.dsr2.2013.02.026.McCarthy, G., and Coauthors, 2015: Measuring the Atlantic meridional overturning circulation at 26°N.

,*Prog. Oceanogr.***130**, 91–111, doi:10.1016/j.pocean.2014.10.006.Pedlosky, J., 1979:

*Geophysical Fluid Dynamics*. Springer-Verlag, 624 pp.Ratsimandresy, A., and J. Pelegri, 2005: Vertical alignment of the Gulf Stream.

,*Tellus***57A**, 691–700, doi:10.1111/j.1600-0870.2005.00115.x.Reynolds, R. W., T. Smith, C. Liu, D. Chelton, K. Casey, and M. Schlax, 2007: Daily high-resolution-blended analyses for sea surface temperature.

,*J. Climate***20**, 5473–5496, doi:10.1175/2007JCLI1824.1.Rossby, T., 1987: On the energetics of the Gulf Stream at 73W.

,*J. Mar. Res.***45**, 59–82, doi:10.1357/002224087788400918.Serazin, G., T. Penduff, S. Gregorio, B. Barnier, J. Molines, and L. Terray, 2015: Intrinsic variability of sea level from global 1/12° ocean simulations: Spatiotemporal scales.

,*J. Climate***28**, 4279–4292, doi:10.1175/JCLI-D-14-00554.1.Smith, R. D., M. E. Maltrud, F. O. Bryan, and M. Hecht, 2000: Simulation of the North Atlantic Ocean at 1/10°.

,*J. Phys. Oceanogr.***30**, 1532–1561, doi:10.1175/1520-0485(2000)030<1532:NSOTNA>2.0.CO;2.Stammer, D., 1998: On eddy characteristics, eddy transports, and mean flow properties.

,*J. Phys. Oceanogr.***28**, 727–739, doi:10.1175/1520-0485(1998)028<0727:OECETA>2.0.CO;2.Toole, J. M., R. G. Curry, T. M. Joyce, M. McCartney, and B. Peña-Molino, 2011: Transport of the North Atlantic deep western boundary current about 39°N, 70°W: 2004–2008.

,*Deep-Sea Res. II***58**, 1768–1780, doi:10.1016/j.dsr2.2010.10.058.Treguier, A. M., J. Deshayes, C. Lique, R. Dussin, and J. M. Molines, 2012: Eddy contributions to the meridional transport of salt in the North Atlantic.

,*J. Geophys. Res.***117**, C05010, doi:10.1029/2012JC007927.Treguier, A. M., and Coauthors, 2014: Meridional transport of salt in the global ocean from an eddy-resolving model.

,*Ocean Sci.***10**, 243–255, doi:10.5194/os-10-243-2014.Volkov, D., T. Lee, and L. Fu, 2008: Eddy-induced meridional heat transport in the ocean.

,*Geophys. Res. Lett.***35**, L20601, doi:10.1029/2008GL035490.Watts, D., and W. E. Johns, 1982: Gulf Stream meanders: Observations on propagation and growth.

,*J. Geophys. Res.***87**, 9467–9476, doi:10.1029/JC087iC12p09467.Zhai, X., and R. Greatbatch, 2006: Inferring the eddy-induced diffusivity for heat in the surface mixed layer using satellite data.

,*Geophys. Res. Lett.***33**, L24607, doi:10.1029/2006GL027875.