News & Views | Published:

Climate science

Unexpected fix for ocean models

Nature volume 535, pages 497498 (28 July 2016) | Download Citation

Computational models persistently underestimate strong currents that redistribute ocean heat. This problem is solved in models in which ocean eddies are damped by coupling of the atmosphere with the sea. See Letter p.533

In all ocean basins, strong currents are trapped at the western boundary as a consequence of Earth's rotation. The currents in the Northern Hemisphere carry warm subtropical water north along the coast to latitudes of about 35–40° N, then turn east to form boundary-current extensions. Much of the heat carried by these currents is released to the atmosphere1, and the remainder provides heat to the ocean at higher latitudes. The eastward currents simulated by ocean models are persistently weaker than those observed. On page 533, Ma et al.2 show how coupling the ocean to the atmosphere in models can strengthen the eastward currents, which should in turn improve climate predictions.

In simulations, a weak eastward current causes ocean-temperature biases, so that water near the coast is warmer than in the real world and water in the ocean interior is too cold. A weak current will also have a weak front (the temperature difference across the current is small). This effect has implications for the atmosphere, because observational analysis3 suggests that strong fronts anchor mid-latitude storm tracks, whereas weak fronts do not. Modelling4 that couples the ocean to the atmosphere at high atmospheric resolution (approximately 50 kilometres) has demonstrated the crucial role of ocean fronts in driving the atmosphere both in the boundary layer — the region in which flow is directly affected by Earth's surface — and deep into the troposphere, the lowest part of the atmosphere, where most weather phenomena occur. Ma et al. focus instead on how those atmospheric modifications feed back to the ocean.

Weak currents in simulations can be partly fixed by increasing the spatial resolution of ocean models from approximately 100 km to about 10 km. The higher resolution increases the ocean's kinetic energy in the regions near the currents to values comparable to estimates based on observations3. However, the distribution of energy in the ocean-only models differs from that observed: in observations, more energy is contained in the core of the current extensions than in the eddies that it sheds, whereas the reverse is true for the models5.

Coupling an ocean model to an atmospheric model would seem likely to further weaken current extensions, rather than to strengthen them, by cooling the warm region south of the current and warming the cold region north of the current. Surprisingly, Ma et al. show that such coupling can instead strengthen the current in high-resolution models.

Focusing on the Kuroshio Extension, a boundary-current extension in the North Pacific Ocean, the authors compare coupled models that resolve ocean-eddy-scale (about 10 km) air–sea heat exchange with models that do not (Fig. 1). They show that ocean eddies drain energy from the current extensions, and that coupling of an eddy-resolving atmosphere model weakens the eddies by removing heat from warm ocean eddies and injecting heat into cold ones. An implication of this study is that the stronger current simulated by coupled models will deliver heat away from the western boundary into the ocean interior, as observed, and that this will improve climate prediction.

Figure 1: Improved modelling of a strong ocean current.
Figure 1

Ma et al.2 modelled the Kuroshio Extension, a strong, eastward-flowing current in the North Pacific Ocean, using two different versions of a computational model. a, In the first version, small-scale (10 km) interactions between the ocean and the atmosphere were suppressed by having the atmosphere interact with only a 'smoothed' version of the ocean. A plot of the simulated kinetic energy of the Kuroshio Extension (shown in units of square metres per seconds squared) suggests a relatively weak, meandering current. Land is shown in grey. b, When the authors used a version that allows small-scale ocean–atmosphere interactions, the Kuroshio Extension is stronger and meanders less than in a, and has shifted southward, more closely resembling the observed current. These findings help to explain why ocean-only models persistently underestimate the strength of such currents.

There are some limitations and further important implications of Ma and co-workers' work. The effect of coupling to a high-resolution atmospheric model on western boundary currents was demonstrated only for the Kuroshio Extension region. Although western boundary currents all have much in common, such coupling will not necessarily improve simulations in all basins. Moreover, the improvement comes at a steep price in terms of computer resources and time. The successful result also implies that ocean-only models will continue to simulate weak currents, because they lack the atmosphere–ocean feedback that lowers the energy of ocean eddies6,7.

Accurate simulation of the transport and uptake of ocean heat and of atmosphere–ocean heat exchange is crucial for climate prediction. Most models for century-scale and decadal prediction in the next assessment of the Intergovernmental Panel on Climate Change are being run at coarse resolution (100 km in the ocean); none will be at sufficiently high resolution to resolve eddies. Ma and colleagues have isolated a specific process that is missing in the lower-resolution climate models, and which leads to poor representation of ocean currents and therefore unrealistic air–sea heat exchange.

Boundary currents make a large contribution to heat transport by the ocean from the tropics to the high latitudes, which contributes to mid-latitude climate variations, storm-track steering and Arctic ice melt8. Accurate representation of these strong currents, such as the Gulf Stream, is thought to be necessary for climate prediction both in the near term (1–20 years)9 and on the century scale. Ma and co-workers' study provides a strong and specific argument for increasing the spatial resolution of climate models to improve these predictions.



  1. 1.

    & J. Climate 14, 3433–3443 (2001).

  2. 2.

    et al. Nature 535, 533–537 (2016).

  3. 3.

    , , & In Earth's Climate (eds Wang, C., Xie, S. P. & Carton, J. A.) (Am. Geophys. Union, 2004).

  4. 4.

    et al. J. Climate 23, 6277–6291 (2010).

  5. 5.

    & Ocean Modelling 8, 31–54 (2005).

  6. 6.

    et al. J. Climate 19, 1970–1989 (2006).

  7. 7.

    , & Ocean Modelling 16, 141–159 (2007).

  8. 8.

    , & Geophys. Res. Lett., 33, L23503 (2006).

  9. 9.

    et al. Bull. Am. Meteorol. Soc. 95, 243–267 (2014).

Download references

Author information


  1. Kathryn A. Kelly is in the Applied Physics Laboratory, Air-Sea Interaction and Remote Sensing Department, University of Washington, Seattle, Washington 98105-6698, USA.

    • Kathryn A. Kelly
  2. LuAnne Thompson is at the School of Oceanography, University of Washington, Seattle, Washington 98195-5351, USA.

    • LuAnne Thompson


  1. Search for Kathryn A. Kelly in:

  2. Search for LuAnne Thompson in:

Corresponding authors

Correspondence to Kathryn A. Kelly or LuAnne Thompson.

About this article

Publication history




By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing