The Hidden Order in Coastal Wave Dynamics

The ocean's coastline is not merely a boundary where land meets water. For thousands of kilometers along continental margins, it acts as a waveguide—a kind of underwater highway along which energy travels in the form of coastal trapped waves. These waves, invisible to casual observers but detectable by sensitive instruments, help shape everything from the coastal response to tidal forces to the way sea level anomalies propagate and dissipate. Understanding them matters not just for academic oceanography but for anyone who lives near the coast and cares about accurate predictions of ocean behavior.

Now, a new result reveals something fundamental about how these waves behave: their natural oscillatory patterns—what physicists call "modes"—are orthogonal. In plain terms, each mode contributes to the total energy of the wave field independently, without blending into or interfering with the others. This orthogonality, demonstrated for the first time in full generality by Jörn Callies of Caltech in a paper submitted to the Journal of Physical Oceanography, is a property that mathematicians recognize as the signature of a well-structured system. It means that when you decompose a complex ocean wave field into its constituent modes, each piece tells you something clean and distinct about the system.

But perhaps more practically, this work provides oceanographers with a new mathematical framework—formulated as a generalized Schrödinger equation—that both exposes why this orthogonality exists and offers a robust numerical method for calculating these modes. Callies uses finite-element discretization that preserves the underlying symmetries of the problem, ensuring that computer models inherit the same elegant mathematical properties as the real ocean.

The Science

Coastal trapped waves have been studied since the 1970s, when researchers including Wang, Clarke, and Huthnance developed the theoretical framework for understanding these phenomena. Their work established that coastal trapped waves propagate along continental margins—in the northern hemisphere, they travel with the coast to their right—and that they play a crucial role in how the coastal ocean responds to tidal and wind forcing. These waves can be tracked for thousands of kilometers as they propagate along coastlines after being excited by local wind events. They mediate the ocean's response to sub-inertial tidal forcing and shape how open-ocean sea level anomalies are expressed near the coast.

The dynamics that govern these waves are the linearized hydrostatic Boussinesq equations on an f-plane. The Boussinesq approximation simplifies the full equations of fluid motion by assuming that density differences, while important for buoyancy-driven flows, are small enough to ignore except where they generate buoyancy forces. Hydrostatic balance means that pressure at any point is determined simply by the weight of the fluid above—a reasonable approximation for motions with horizontal scales much larger than vertical scales, which describes most large-scale ocean circulation. The f-plane approximation assumes a constant Coriolis parameter, appropriate for a limited region of the ocean.

The domain considered is a two-dimensional cross-section adjacent to a coastal wall at x = 0, with depth h(x) varying in the cross-shore direction but constant along the coast. After applying a Fourier transform in the along-shore direction—with along-shore wavenumber denoted by l—the governing equations become a system of five coupled partial differential equations. These relate the cross-shore velocity (u), along-shore velocity (v), vertical velocity (w), pressure (p), and buoyancy (b), which here refers to the density deviation from a background stratification characterized by the squared buoyancy frequency N².

The boundary conditions encode the physics of the ocean surface and bottom. At the free surface, a linearized condition relates the pressure to the vertical velocity. At the bottom and any coastal wall, no normal flow is permitted. Together with the interior equations, these conditions define a complete eigenvalue problem: finding the frequencies ωₙ and spatial structures ψₙ of the normal modes that satisfy Hψₙ = ωₙMψₙ.

What Callies does differently from previous researchers is to recast this eigenvalue problem in the form of a Schrödinger equation. In quantum mechanics, the Schrödinger equation describes how quantum states evolve in time. Its key mathematical feature is that the Hamiltonian operator H is Hermitian—a property that guarantees real eigenvalues and orthogonal eigenfunctions. By exposing the same structure in the coastal trapped wave equations, Callies demonstrates that their eigenmodes must be orthogonal in exactly the same mathematical sense.

The key insight is that the operator pair (H, M) governing these waves is not arbitrary. M is real, symmetric, and positive semi-definite; H is Hermitian. When two eigenmodes ψₙ and ψₘ have distinct eigenvalues (ωₙ ≠ ωₘ), they are guaranteed to be M-orthogonal—meaning their inner product weighted by M is zero. This orthogonality is not an assumption or an approximation; it follows necessarily from the mathematical structure of the equations.

To make this concrete, Callies works with the weak form of the equations—essentially multiplying by test functions and integrating by parts. This formulation makes energy conservation transparent and provides the natural framework for numerical approximation. The weak form leads directly to a finite-element discretization: approximating the solution using piecewise linear basis functions on a triangular mesh of the domain. This discretization preserves the essential symmetries of the continuous problem, so that the discrete matrices ᵋ and ᵅ6 retain the same Hermitian and positive-definite properties as their continuous counterparts.

Callies demonstrates this approach with several numerical examples. The first uses a randomly perturbed exponential bathymetry—a more realistic scenario than the idealized slopes typically considered in theoretical work. The depth follows h(x) = h₀(1 − e^(−x/d)) with a decay scale of 50 kilometers, perturbed by random variations drawn from a Gaussian distribution with Matérn covariance. Using uniform stratification (N = 3 × 10⁻³ s⁻¹), an inertial frequency f = 10⁻⁴ s⁻¹, and an offshore depth h₀ = 3 kilometers, he computes the first five modes for an along-shore wavenumber l = 2π/200 km. The resulting mesh contains 41,907 nodes, and the Arnoldi iteration converges in just a few iterations, producing results in under 10 seconds on a laptop.

A second example follows Huthnance's classic 1978 setup with a linear bathymetric slope transitioning to a flat bottom offshore, using a slope Burger number of 1/2. A third example, following Kelly (2022), uses a tanh-shaped bathymetry representing an idealized shelf geometry with a diurnal tidal period of 24 hours.

What They Found

The central result is mathematical in nature but carries profound practical implications: coastal trapped wave modes are orthogonal in the sense that they make independent contributions to the energy of the wave field. This orthogonality holds whether or not the wave evolution is much slower than the inertial period—a point that previous researchers had considered only in the slow limit.

The proof proceeds by demonstrating that the operator pair (H, M) has the structure of a generalized eigenvalue problem with a Hermitian operator H. If ψₙ and ψₘ are two eigenmodes with distinct eigenvalues ωₙ ≠ ωₘ, then their M-inner product vanishes: ⟨ψₘ, ψₙ⟩ₘ = 0. Moreover, when the eigenmodes are normalized so that ⟨ψₙ, ψₙ⟩ₘ = 1, the total energy of any wave field can be written as:

$$E=\frac{1}{2}⟨\psi ,\psi {⟩}_M=\frac{1}{2}\sum _n|a_n{|}^2$$

where the aₙ are the modal amplitudes. The energy decomposes into a clean sum of independent contributions, one from each mode. There are no cross terms—no mixing or interference between modes in the energy budget.

This orthogonality property was known previously only in the slow limit where ω² ≪ f², using a reduced set of dynamics based on the geostrophic momentum approximation. Callies shows that the full hydrostatic Boussinesq dynamics retain this property without any approximation. The Schrödinger formulation exposes the underlying mathematical structure that guarantees orthogonality in the general case.

The numerical examples confirm that the finite-element approach correctly captures this orthogonality. For the randomly perturbed exponential bathymetry, the computed modes exhibit the expected properties: they are bottom-trapped (energy concentrated near the seafloor), and the cross-shore structure decreases in scale with increasing mode number. The pressure modes have a number of nodes in the vertical direction equal to the mode number.

Normalized frequencies of coastal trapped wave modes for random bathymetry

Label	Value
Mode 1	-0.52
Mode 2	-0.31
Mode 3	-0.25
Mode 4	-0.19
Mode 5	-0.18

The modal frequencies span a range from sub-inertial (below the Coriolis frequency f) to values approaching f, consistent with trapped wave dynamics. Mode 1 has ω₁/f = −0.52, mode 2 has ω₂/f = −0.31, mode 3 has ω₃/f = −0.25, mode 4 has ω₄/f = −0.19, and mode 5 has ω₅/f = −0.18. The negative sign indicates the direction of phase propagation.

For the linear bathymetry case, the frequencies follow a similar pattern but with slightly different values reflecting the different geometry: mode 1 at ω₁/f = −0.35, mode 2 at ω₂/f = −0.18, mode 3 at ω₃/f = −0.12, mode 4 at ω₄/f = −0.085, and mode 5 at ω₅/f = −0.067.

Normalized frequencies of coastal trapped wave modes for linear bathymetry

Label	Value
Mode 1	-0.35
Mode 2	-0.18
Mode 3	-0.12
Mode 4	-0.085
Mode 5	-0.067

The dispersion curves—the relationships between along-shore wavenumber and frequency—show how these modes behave across a range of scales. At small along-shore wavenumbers (long waves), all modes converge toward ω ≈ 0, the geostrophic limit. As the along-shore wavenumber increases, the modes diverge, with mode 1 consistently having the highest (most negative) frequency and higher modes progressively lower frequencies.

A crucial validation comes from the convergence analysis. By computing modes on progressively finer meshes and comparing with analytical solutions for the internal Kelvin wave, Callies demonstrates that the numerical error decreases with the square of the mesh size—a hallmark of a well-behaved finite-element method. For the full dynamics (not just the slow approximation), the error for mode 1 on the coarsest mesh (characteristic size 0.05) is about 0.5%, dropping to around 0.1% on the finest mesh. Higher modes show larger errors but still converge systematically.

Convergence analysis: numerical error vs mesh size

Label	Value
0.05	0.005
0.025	0.0012
0.0125	0.0003
0.00625	0.00008

Solid lines: full dynamics. Light lines: slow approximation.

Why This Changes Things

Orthogonality is not just a mathematical curiosity. It is the property that allows us to decompose complex signals into simple parts and understand each part independently. When a musical instrument produces a complex tone, the sound can be decomposed into a sum of pure harmonics, each with its own frequency and amplitude. Because these harmonics are orthogonal, the energy of the complex tone is simply the sum of the energies of the individual harmonics—no complicated interactions to track. The same principle applies to coastal trapped waves.

In practice, this means that the response of the coastal ocean to forcing—whether from tides, winds, or remote disturbances—can be understood mode by mode. Each mode acts as a nearly independent channel through which energy propagates. The response of each channel depends on two things: how strongly the forcing projects onto that mode's spatial structure, and how close the forcing frequency is to the mode's natural resonant frequency. When forcing is near resonance, the response amplifies dramatically; away from resonance, it diminishes according to a simple inverse relationship.

This decomposition matters because coastal trapped waves mediate critical processes. They transmit the ocean's response to tidal forcing along thousands of kilometers of coastline. They help determine how sea level anomalies from the open ocean express themselves near the coast—a question of direct relevance to coastal communities concerned about sea level rise. They shape the coastal response to wind events, influencing upwelling, nutrient fluxes, and ultimately biological productivity.

Before Callies' work, the orthogonality of these modes was established only in the slow limit, where the wave frequency is much less than the inertial frequency. In this limit, the dynamics reduce to the geostrophic momentum approximation, and the modes become edge waves governed by boundary potential-vorticity anomalies. But the slow approximation is just that—an approximation. Real coastal forcing often involves frequencies comparable to the inertial period, where the full dynamics matter.

Callies shows that orthogonality holds more generally, without requiring the slow approximation. The Schrödinger formulation exposes the underlying mathematical structure that guarantees this property: the Hermiticity of the operator H. This Hermiticity emerges from the same integration-by-parts argument that yields energy conservation—the boundary terms cancel because of the no-normal-flow boundary conditions at the bottom and any coastal wall.

The practical implication is that oceanographers can now compute coastal trapped wave modes with confidence that their decomposition is physically meaningful. The finite-element method preserves the essential symmetries of the problem, so numerical solutions inherit the same orthogonality as the continuous equations. No spurious modes contaminate the spectrum; no artificial coupling between modes arises from numerical approximation.

There is also a more fundamental point about the nature of these waves. Callies emphasizes that low-frequency coastal trapped waves are edge waves—waves trapped at a boundary by the contrast between the ocean interior and the coast. Their dynamics are governed by boundary potential-vorticity anomalies. The classical Bretherton boundary condition, familiar from quasi-geostrophic theory, applies to gently sloping bottoms. But for steep slopes and coastal walls, additional lateral contributions to the boundary potential vorticity become important. This extended understanding clarifies the physical interpretation of the modes and connects to a broader framework for thinking about wave-boundary interactions.

The Schrödinger formulation also generalizes to more complex geometries that are not uniform in the along-shore direction. In such cases, the along-shore wavenumber l is no longer a good quantum number, and the standard separation of variables approach fails. But the eigenvalue problem with frequency as the eigenvalue remains well-defined, and its eigenmodes are guaranteed to be orthogonal. The response to tidal or wind forcing can then be understood mode by mode, with no interaction between modes in the linear equations, even in a geometry with along-shore variations.

What's Next

Several questions remain open. The most technical concerns the well-posedness of the eigenvalue problem. The governing equation changes from elliptic (well-behaved) for sub-inertial frequencies (ω² < f²) to hyperbolic (potentially ill-behaved) for super-inertial frequencies (ω² > f²). The sub-inertial case, which includes the trapped modes that are most relevant for coastal dynamics, is well-posed and converges nicely under mesh refinement. The super-inertial case is more delicate; in semi-infinite domains, a continuum of modes exists that propagates in from infinity and back out, but the generic case is ill-posed and requires careful treatment.

There is also the question of what happens in three dimensions, where the simplifications that make the two-dimensional analysis tractable no longer apply. Callies notes that the phase relations between variables that hold in 2D—specifically, that cross-shore and vertical velocities are in phase with each other and in quadrature with along-shore velocity and pressure—do not generalize to three dimensions. The physical interpretation of 3D modes, and whether they retain the orthogonality property, remains to be explored.

The connection to energy flux is more subtle than it might appear. While modes are orthogonal in the sense of energy content, the energy flux loses its simple interpretation in complex geometries. For each mode, the time-averaged flux is non-divergent, but when multiple modes are present, the flux emerges from the phase relation between modes. This is how initially localized energy can spread across the domain—through mode coupling mediated by phase differences, not through direct energy transfer between orthogonal modes.

From a practical standpoint, the finite-element framework opens possibilities for computing coastal trapped wave modes in realistic geometries with irregular coastlines, varying stratification, and complicated bathymetry. The method preserves the essential physics while handling arbitrary domain shapes naturally. This could improve the accuracy of coastal ocean models and ultimately lead to better predictions of coastal sea level, tidal dynamics, and wind-driven response.

There are also connections to make with the growing body of work on sub-inertial tides and their energetics. Musgrave (2019) showed that coastal trapped wave modes make independent contributions to along-shore energy flux only in the slow limit. Kelly (2022) showed that independence beyond the slow limit requires a redefinition of modes that treats along-shore wavenumber as the eigenvalue rather than frequency. Callies' work clarifies the status of the standard frequency-based modes: they are orthogonal in energy but not in energy flux. This distinction matters for understanding how energy propagates along coastlines and how it eventually dissipates.

The potential applications extend beyond pure oceanography. Coastal trapped waves interact with other coastal processes—circulation, mixing, biological productivity—in ways that are not fully understood. Having a clean decomposition of the wave field into orthogonal modes provides a conceptual framework for thinking about these interactions. If each mode acts as an independent channel, then changes in one channel (say, due to altered stratification from climate change) propagate independently and can be tracked and predicted without worrying about complex cross-mode interactions.

There is work to be done in validating these theoretical results against observations. The orthogonality of computed modes is a mathematical property, but does the real ocean respect it? Do observed coastal wave fields decompose cleanly into independent modal contributions? Testing this against the growing database of coastal measurements—from moored instruments, satellite altimetry, and coastal tide gauges—would be a natural next step.

The theoretical framework might also be extended to other wave systems. The Schrödinger formulation exposes a mathematical structure that has proven fruitful for understanding orthogonality and developing numerical methods. Similar structures may exist in other coastal and boundary-trapped wave phenomena—in atmospheric waves trapped by mountain ranges, in waves trapped by sea ice edges, in internal waves trapped by density interfaces. Recognizing the common mathematical structure could illuminate these related systems and suggest analogous numerical approaches.

Finally, there is the question of what this means for coastal communities. Coastal trapped waves influence sea level variations on timescales from days to years, affecting storm surge, tidal flooding, and long-term sea level change. Accurate prediction of these variations requires accurate models of the underlying dynamics. A better theoretical understanding of the modal structure—the fact that each mode contributes independently to the energy—provides a foundation for more accurate and interpretable models. The mathematics may seem abstract, but its practical implications reach all the way to decisions about coastal infrastructure, flood protection, and adaptation to a changing climate.

The ocean's coastline is a dynamic, complex boundary where many processes interact. Callies' work reveals that at least one aspect of this complexity—coastal trapped waves—is more orderly than it might appear. These waves decompose into modes that are mathematically orthogonal, each contributing its own independent share of energy to the total field. This orthogonality emerges from the fundamental structure of the governing equations, preserved in numerical approximations, and holds regardless of whether we invoke the slow approximation or consider the full dynamics. Understanding this order is both a scientific advance and a practical tool—an example of how deep mathematical insight can illuminate even the most empirically complex systems.