The Reverse Spiral: Why Ocean Models Have Been Getting the Bottom Wrong

Wind does not push water in the direction it blows. Anyone who has watched a boat responding to a steady breeze, or sea ice drifting across the Arctic, knows that the surface current runs at an angle to the wind itself. Turn your face toward the North Sea in winter, feel the wind on your cheeks, and know that somewhere beneath that wind, water is moving sideways — not toward you, but perpendicular to your left or right. This peculiar lateral drift is one of the most fundamental facts about the ocean's surface layer, and it shapes everything from the transport of heat around the planet to the behaviour of oil spills and the upwelling of nutrients that sustain fisheries.

But here is a problem that has quietly troubled physical oceanography for decades: the textbooks say the wind-driven transport of water runs at exactly 90 degrees to the wind. Field measurements mostly confirm this. Yet when researchers try to model this process mathematically — to predict how the ocean will respond to changing wind patterns in a warming climate — they keep making the same mistake. They assume that somewhere below the surface, at an arbitrary depth they call the bottom of the Ekman layer, the water simply stops experiencing the wind's influence. At that depth, they set the stress on the water to zero, and move on.

A new paper by Christian Puntini of the University of Vienna and Luigi Roberti of Leibniz University Hannover shows that this assumption is mathematically impossible. Imposing a stress-free boundary at an intermediate depth does not merely simplify the physics — it produces an unphysical result: a current that accelerates below the depth where it should be decaying, spinning in the opposite spiral direction, like watching a whirlpool reversibly unfold rather than dissipate. The water, in this model, gains speed as you go deeper, which is not what happens in the real ocean.

The finding is not merely a mathematical curiosity. It strikes at the foundations of how scientists represent the ocean's response to wind in climate models — models that underpin predictions about everything from the slowdown of the Atlantic Meridional Overturning Circulation to the redistribution of heat by ocean currents in a warming world. If the boundary condition is wrong, the circulation patterns those models predict may be subtly but systematically biased.

"Wind-driven processes at the ocean surface is a classic topic in physical oceanography, with important implications for the ocean circulation and, consequently, Earth's climate," the authors note with characteristic understatement. The paper, published on arXiv in July 2026, is titled with a directness that is rare in the academic literature: Why a mid-depth stress-free boundary condition is incorrect for Ekman flows.

The Science

To understand what Puntini and Roberti discovered, you need to understand what an Ekman layer is — and to do that, you need to go back to a Norwegian ship in the Arctic Circle more than a century ago.

In 1893, the polar explorer Fridtjof Nansen steered his vessel Fram into the Arctic ice and deliberately let it freeze. He wanted to drift across the North Pole with the pack ice, and he wanted to study the physics of the polar ocean. What he observed, and what he described in his 1897 account of the expedition, was peculiar: the sea ice did not drift in the direction of the wind. It drifted at an angle to the right of the wind direction. In the Northern Hemisphere, the deflection was always to the right. In the Southern Hemisphere, it would be to the left.

Nansen shared these observations with a Swedish mathematician named Vagn Walfrid Ekman, who in 1905 published the first theoretical explanation. Ekman's insight was to combine two facts about the ocean's surface layer. First, the wind drags the water at the surface along, transferring momentum downward through the water column via turbulence. Second, the rotating Earth deflects moving objects — this is the Coriolis effect, the same apparent force that makes projectiles curve in the Northern Hemisphere. As the wind's momentum diffuses downward through the water, each successive layer is deflected further by the Coriolis force. The result is a spiral: at the surface, the current runs 45 degrees to the right of the wind; at increasing depths, both the speed and the angle decrease, until at some depth the water effectively stagnates.

This is the Ekman spiral, and the depth-integrated water transport — what oceanographers call the Ekman transport — runs at exactly 90 degrees to the wind (to the right in the Northern Hemisphere, to the left in the Southern). This perpendicular transport is one of the most consequential features of the ocean's dynamics. It drives coastal upwelling, where winds parallel to a coastline push surface water offshore and cold, nutrient-rich water rises from depth to replace it. It shapes the subtropical gyres, the vast clockwise spirals of water that dominate the ocean basins. It influences the formation of sea ice and its drift patterns in polar regions. It is, in the phrase oceanographer Joseph Pedlosky once used to characterise foundational concepts in his field, the kind of thing that once you understand it, you cannot look at the ocean the same way again.

But there is a difficulty. The classical Ekman solution assumes a very simple ocean: constant density, constant viscosity, and an infinite depth. Real oceans are more complex. Density increases with depth as cold, salty water sinks below warm surface water. Viscosity — or more precisely, the vertical eddy viscosity, a measure of how efficiently momentum is mixed downward by turbulence — varies with depth in ways that are still not fully understood. And the ocean has a bottom, at some finite depth, beyond which there is only seabed.

When researchers try to model wind-driven ocean currents more realistically, they must decide what boundary condition to impose at the ocean floor. There are two natural choices. The first is the no-slip condition: water at the bottom of the ocean sticks to the seabed, so its velocity is zero. This is physically sensible — there is, after all, friction at the seafloor, and water touching rock is not moving relative to it. The second is the stress-free condition: there is no shear stress at the bottom, meaning the gradient of the velocity vanishes — the velocity is not necessarily zero, but it is not changing with depth either. In the infinite-depth approximation, the stress-free condition at the ocean floor was often replaced by the requirement that velocity itself decays to zero as depth becomes infinite.

For a full-depth ocean, there is a third approach that has become common in the literature: choose a depth -D somewhere between the surface and the ocean floor, call that the bottom of the Ekman layer, and impose a stress-free condition there. The idea is to separate the problem into two regions. In the upper layer, extending from the surface down to depth -D, you solve the Ekman equations with the stress-free condition at -D. Below -D, you assume the wind-driven flow is negligible. It is an a priori assumption — you decide in advance where the Ekman layer ends — and it is convenient mathematically. But Puntini and Roberti's central claim is that it is also wrong.

The mathematics of this is not trivial, but the core idea can be stated clearly. The Ekman equations describe how velocity changes with depth under the combined influence of wind stress and the Coriolis force. If at some depth -D, the velocity gradient is zero — if the stress-free condition holds — then the equations force the velocity itself to be at a minimum at that depth. This is not a choice or an assumption; it is a mathematical consequence of the governing equations. Below -D, the velocity must increase again. The current cannot simply stop at the bottom of the chosen layer; it must reverse direction of descent and accelerate toward the seabed. The result is what Puntini and Roberti call a "reverse Ekman spiral" — the velocity increases with depth below the depth where it should be decaying.

To prove this rigorously, Puntini and Roberti work with the governing equations in a form that combines the two horizontal velocity components into a single complex variable:

$$U(z)=u(z)+iv(z)$$

where u and v are the eastward and northward velocity components, and z is the vertical coordinate measured downward from the surface. The equations for wind-driven flow in the f-plane approximation — which treats a small patch of the ocean as a flat plane rotating at constant angular velocity — take the form:

$$\begin{array}{rl}(m(z)U^{\prime }(z){)}^{\prime }& =if\rho (z)U(z),z\in (-H,0),\\ m(0)U^{\prime }(0)& =\tau ,\\ U^{\prime }(-D)& =0.\end{array}$$

Here m(z) is the depth-dependent eddy viscosity, f is the Coriolis parameter, ρ(z) is the water density, and τ is the wind stress vector. The condition U'(-D) = 0 is the stress-free boundary condition imposed at depth -D. The final equation is the statement that the derivative — the velocity gradient — vanishes at that depth.

The authors' proof proceeds by multiplying the governing differential equation by the complex conjugate of the velocity and integrating from the stress-free depth -D to an arbitrary depth z below it. This integration by parts yields two relations: one connecting the rate of change of the speed |U(z)| to the integral of the velocity gradient squared, and another connecting the rate of change of the flow direction φ(z) to the integral of the density-weighted velocity squared. The details are technical, but the conclusion is not:

• For depths z between -D and the surface, the speed |U(z)| increases as you go upward (toward the surface), meaning it decreases as you go deeper — exactly what the Ekman spiral predicts.
• At depth z = -D, the speed reaches a minimum. The derivative |U'| = 0.
• For depths z below -D (i.e., z < -D), the speed |U(z)| must increase again as depth increases — the current accelerates downward toward the ocean floor.

The same logic applies to the direction of the current (in the Northern Hemisphere, with f > 0): the angle between the current and the wind increases with depth above -D, reaches a maximum at -D, and then decreases below it. The spiral reverses direction.

The authors present this as a theorem — Theorem 1 in the paper — and the proof is clean and mathematically airtight. There is no ambiguity about the conclusion. If you impose a stress-free boundary condition at an intermediate depth -D, the mathematics demands that the velocity reach a minimum there and increase below. This is not an approximation or a simplification. It is what the equations say.

The second part of the paper addresses a more subtle question. If the no-slip condition at the ocean floor is physically more appropriate — if the water genuinely sticks to the seabed — then does the Ekman transport remain perpendicular to the wind? The classical result of a 90-degree angle between transport and wind stress depends on the stress-free condition at the boundary. With no-slip, the bottom stress contributes an extra term to the transport integral, and in principle the orthogonality could be lost.

Puntini and Roberti tackle this using a method called the WKB approximation — after its inventors, Wentzel, Kramers, and Brillouin — which is a technique for solving differential equations where the coefficients vary slowly. The key observation is that at great depths, where wind effects are negligible, water density varies only slowly and gently. The authors treat the region below the Ekman depth -D separately, assuming constant (molecular) viscosity there, and solve the governing equation approximately using the WKB ansatz. They show that the velocity gradient at the ocean floor, U'(-H), decays exponentially with the depth ratio H/D. For a realistic ocean of 1000 metres depth with an Ekman depth of 100 metres, their numerical estimate gives:

$$|U^{\prime }(-H)|\lesssim 35|U_D|10^{-29710}\approx 0$$

The number is effectively zero. Even allowing for density variation rather than assuming constant density, the result holds: for a sufficiently deep ocean, the no-slip condition at the seabed effectively implies zero velocity gradient there as well — and therefore the Ekman transport is perpendicular to the wind, as observations confirm.

This is the paper's second significant result: the no-slip condition is not only physically correct but also effectively produces the same orthogonality condition that the stress-free assumption was invented to enforce. You get the right answer for the right reason.

What They Found

The results can be distilled into three concrete findings.

Finding 1: The stress-free intermediate-depth boundary condition produces a reverse spiral. Puntini and Roberti's central theorem proves mathematically that imposing U'(-D) = 0 — stress vanishes at an intermediate depth — forces the current speed |U(z)| to be at a minimum at that depth and to increase below it. There is no way around this conclusion. The reverse spiral is not a numerical artefact or an artefact of a particular choice of parameters; it is a mathematical necessity.

Finding 2: This result holds regardless of the depth-dependent structure of viscosity or density. The authors note that the classical Ekman solution assumed constant viscosity and constant density, and even with these simplifications, the reverse spiral emerges. When viscosity and density are allowed to vary with depth — which is more realistic — the same conclusion follows, provided the eddy viscosity is positive (which it physically must be). Different viscosity profiles can change the quantitative details, but they cannot eliminate the minimum at depth -D and the subsequent acceleration below it.

Finding 3: A no-slip condition at the ocean floor is consistent with the observed perpendicularity of the Ekman transport. Using the WKB approximation to analyse the deep ocean below the Ekman layer, the authors show that the velocity gradient at the seabed is exponentially small when the ocean is much deeper than the Ekman layer. For a 1000-metre ocean with a 100-metre Ekman layer, the correction to the 90-degree orthogonality is astronomically small — essentially unmeasurable. The no-slip condition gives the right result: it produces the observed perpendicular transport between the wind and the ocean current, but through a physically correct mechanism rather than an assumption that turns out to be mathematically inconsistent.

These findings carry significant implications for how ocean dynamics is modelled and understood. The stress-free intermediate-depth condition has been used extensively in the literature — Puntini and Roberti cite papers by McWilliams, Huckle, and Shchepetkin (2009), Price and Sundermeyer (1999), Lewis and Belcher (2004), among others, where this approach appears. The new paper does not merely criticise these choices; it shows that the assumption is mathematically self-contradictory in a specific and identifiable way.

The classical Ekman spiral: the surface current is deflected 45° from the wind, and progressively deeper layers show decreasing velocity and reduced deflection. The vertically integrated transport runs at right angles to the wind. This image is correct. The problem is what happens below the layer where the textbook model stops.

Why This Changes Things

To appreciate why this matters, you need to understand how deeply the Ekman layer sits inside the rest of oceanography.

Wind-driven ocean circulation is the mechanism by which the atmosphere communicates with the ocean. The atmosphere heats the tropics and cools the poles, and the wind is the primary agent that stirs the ocean's surface layer, transferring heat, salt, and momentum across vast distances. The Ekman transport is the crucial intermediate step: it takes the wind's push and turns it into a sideways drift of water, which then feeds into the larger-scale circulation patterns — the gyres, the boundary currents, the overturning cells — that distribute heat around the planet.

This matters enormously for climate. The Atlantic Meridional Overturning Circulation (AMOC), for instance — the conveyor belt of warm water flowing northward at the surface and cold water returning southward at depth — depends in part on wind-driven processes in the upper ocean. Models of AMOC behaviour under climate change, which have suggested the circulation may weaken or even collapse under sufficient freshwater input, rely on correct representations of how wind momentum is transferred to the ocean. If the boundary conditions used in those models contain a systematic error — one that produces an unphysical current profile in the lower part of the Ekman layer — the implications for the larger circulation might be subtle but cumulative.

The same applies to coastal upwelling. When winds blow parallel to a coastline — the coast of Peru, the coast of California, the Benguela coast of southwest Africa — the Ekman transport runs offshore, pushing surface water away from the coast. This forces cold, nutrient-rich water to rise from depth, fuelling the biological productivity of these regions. Get the boundary conditions wrong, and the model may misrepresent how efficiently this upwelling occurs, how much nutrient is supplied to the sunlit surface layer, and ultimately how the marine ecosystem responds to changing wind patterns.

The finding also matters because the stress-free intermediate-depth condition has been used not only in theoretical work but in numerical ocean models. While the most sophisticated general circulation models resolve the full water column and impose no-slip conditions at the seafloor, the parameterisations they use — the ways they represent the unresolved physics of turbulence and mixing — often incorporate the same conceptual framework. The KPP (K-Profile Parametrisation) scheme, which Puntini and Roberti discuss and which is widely used in ocean modelling, represents the vertical eddy viscosity as a non-monotonic function of depth: it increases from the surface to some maximum, then decreases back toward the molecular value at depth, with the depth of the minimum identified as the effective bottom of the wind-driven layer. This is, in effect, a physically motivated resolution of the dilemma Puntini and Roberti identify: instead of assuming the wind-driven layer ends at a fixed depth with a stress-free condition, the KPP scheme allows the wind effects to taper off gradually within a viscosity structure that reflects the physics.

The authors' suggestion is that the no-slip condition at the true ocean floor, with the Ekman depth emerging naturally from the parametrisation of the eddy viscosity, is the correct framework. The KPP approach, by allowing the viscosity to approach a small but finite molecular value at depth rather than abruptly vanishing, is essentially doing the right thing — it is avoiding the mathematical pitfall of imposing stress-free at an intermediate depth while still capturing the essential physics.

There is a broader lesson here about the relationship between mathematical idealisation and physical reality in oceanography. The Ekman layer was discovered through a combination of observation and theory more than a century ago, and the mathematics has always been clean. But when you try to apply that clean mathematics to the messy, variable, bounded ocean of the real world, every simplification carries a cost. The stress-free intermediate-depth condition was introduced to make the mathematics tractable, and it gave the right answer — the perpendicular transport — by construction. What Puntini and Roberti show is that it gives the right answer for the wrong reason, and the wrong reason turns out to have consequences that are physically unacceptable.

The WKB analysis in the paper's second half is particularly striking in this regard. The authors estimate the velocity gradient at the ocean floor under a no-slip condition and find it is essentially zero for any reasonably deep ocean. In other words, the no-slip condition at the true floor of the ocean already delivers what the stress-free assumption was invented to deliver: a velocity gradient that is negligibly small, and therefore an Ekman transport that is perpendicular to the wind stress. The no-slip condition works not in spite of the additional term it introduces in the transport formula, but because that additional term turns out to be vanishingly small. The correction is not just small in the sense of being a small percentage of the dominant term; it is small in the sense of being essentially unmeasurable — on the order of 10⁻²⁹¹⁰ in their specific numerical example with a 1000-metre ocean.

This suggests a powerful idea: that the observed orthogonality of wind and Ekman transport in the real ocean is not merely a coincidence that the stress-free boundary condition was invented to reproduce. It is a consequence of the ocean being deep relative to the depth at which wind effects become negligible, combined with the no-slip condition at the seabed. The physics produces the orthogonality naturally, without any need for the artificial stress-free intermediate-depth boundary.

What's Next

The paper leaves several questions open, and these are where future work will be most valuable.

The first concerns the transition region between the bottom of the Ekman layer and the deep ocean. Puntini and Roberti's WKB analysis applies below the depth -D where wind effects become negligible. But the region around -D itself — the depth at which the eddy viscosity reaches its minimum and transitions from wind-driven turbulent values to the much smaller molecular value — is complex and not fully resolved by the analysis. The KPP parametrisation represents this transition, but there is still debate about the precise depth-dependence of eddy viscosity in the real ocean. Direct measurements of velocity and turbulence in the ocean's interior are difficult to obtain, and our understanding of this region relies heavily on indirect inference from larger-scale observations and on numerical models whose fidelity in this depth range is not well constrained by data.

The second question concerns shallow water. The authors acknowledge that their analysis of the no-slip condition producing negligible bottom stress applies to deep oceans — where H ≫ D, in their notation. In shallow coastal waters, where the depth may be only a few tens of metres, the correction does not vanish. In these regions, the angle between the Ekman transport and the wind can deviate significantly from 90 degrees. Lentz and Fewings (2012) noted this in their review of inner-shelf circulation, where wind stress is balanced by bottom stress and along-shelf pressure gradients rather than by the depth-integrated transport alone. For these environments, the no-slip condition becomes important in its own right, and the orthogonality that holds in the deep ocean does not necessarily apply. Understanding the transitions between these regimes — deep ocean, shallow shelf, coastal boundary layer — is an active area of research, and Puntini and Roberti's analysis provides a useful theoretical grounding for thinking about where and why the classical Ekman results break down.

The third question concerns the spherical geometry of the real Earth. The paper, like much of the theoretical literature on Ekman dynamics, works in the f-plane approximation, which treats a small patch of the ocean as a flat plane. This is appropriate for the horizontal scales at which Ekman dynamics operates — tens to hundreds of kilometres — but larger-scale ocean circulation involves the curvature of the Earth in essential ways. The authors cite their own recent work (Puntini, Roberti, and Stefanescu, 2026) on wind-driven ocean currents in spherical coordinates, and it will be interesting to see whether the boundary condition issues identified here propagate into the spherical formulation in ways that affect the large-scale circulation patterns.

There is also the question of density stratification. The WKB analysis in the paper's second half explicitly treats the density variation at depth as small and slow — a necessary condition for the WKB approximation to be valid. In the deep ocean, where temperature and salinity change only gradually with depth, this is a good approximation. But in regions where there are sharp density interfaces — pycnoclines, where the density gradient changes rapidly over a short vertical distance — the slow-variation assumption breaks down, and the question of what boundary condition is appropriate becomes more subtle. The interaction between density stratification and the bottom boundary condition is an area where more theoretical work is needed.

Finally, there is the question of what this means for climate modelling in practice. The paper is careful not to make grandiose claims about its implications for operational forecasting or climate projection. But the fact that the stress-free intermediate-depth boundary condition is mathematically inconsistent should give modellers pause when they encounter parameterisations or theoretical frameworks that rely on it. The correct physics — no-slip at the seabed, with the Ekman depth emerging from the structure of the eddy viscosity — is already implicit in the KPP scheme and in the more sophisticated ocean general circulation models. What Puntini and Roberti have done is to show, with mathematical rigour, why this is not just a matter of numerical convenience but a matter of physical correctness.

The ocean does not have a stress-free layer buried in its interior. The wind drives the surface, the Coriolis force twists it, the turbulence diffuses the momentum downward, and the current decays. At the seafloor, it stops. That is what the equations say, and that is what the observations confirm. The stress-free intermediate-depth condition was a mathematical simplification that got the right answer, but it did so by creating an artefact — a reverse spiral of unphysical acceleration below the assumed bottom of the wind-driven layer. Correcting this is not merely a matter of tidying up the theory. It is a step toward ensuring that our representations of the ocean's response to wind — in all their complexity, from theoretical frameworks to the parameterisations embedded in climate models — rest on foundations that are not merely approximately right, but exactly right, for exactly the right reasons.

The wind still blows. The ice still drifts. The current still turns. But now we understand a little better why it behaves as it does — and why it cannot, contrary to what a hundred years of simplified modelling suggested, behave otherwise.