🔢 mathematics

Barotropic-Baroclinic Splitting for Multilayer Shallow Water Models with Exchanges

This paper presents and numerically analyzes an exact operator splitting method for nonlinear multilayer shallow water models with inter-layer exchanges in terrain-following coordinates, which preserves total energy, satisfies discrete maximum and entropy principles, maintains geostrophic equilibrium through a well-balanced barotropic step, and significantly reduces computational cost in low Froude simulations without sacrificing accuracy.

Original authors: Nina Aguillon, Sophie Hörnschemeyer, Jacques Sainte-Marie

Published 2026-01-26

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Nina Aguillon, Sophie Hörnschemeyer, Jacques Sainte-Marie

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the ocean as a giant, multi-layered cake. The top layer is the surface, which ripples and waves quickly when the wind blows. The layers deep down move much slower, shifting and swirling in complex patterns.

For a long time, computer models trying to simulate this ocean cake faced a frustrating problem: they had to take tiny, tiny steps to track the fast surface waves. Even though the deep layers were moving slowly, the computer had to wait for the fast surface waves to finish their dance before it could take a single step for the whole cake. This made simulations incredibly slow and expensive.

This paper introduces a clever new way to simulate these ocean layers, called Barotropic-Baroclinic Splitting. Here is how it works, using simple analogies:

The Problem: The "Fast Runner" vs. The "Slow Walker"

Think of the ocean simulation as a relay race.

The Fast Runner (Barotropic): This represents the surface waves. They move at the speed of a sprinter (about 200 meters per second).
The Slow Walker (Baroclinic): This represents the deep currents and the mixing between layers. They move at the speed of a leisurely walker (about 3 meters per second).

In the old method, the computer had to make the Slow Walker take steps as small as the Fast Runner. If the Fast Runner took a step every millisecond, the Slow Walker had to wait and take a step every millisecond too, even though it only needed to move a tiny bit. This wasted a massive amount of computing power.

The Solution: Splitting the Team

The authors propose splitting the team into two groups that run on different schedules:

The Barotropic Step (The Fast Lane):
The computer focuses only on the surface waves and the average speed of the water. It takes many, many tiny, rapid steps to keep the surface waves accurate. Think of this as the Fast Runner sprinting back and forth many times while the rest of the team stands still.
The Baroclinic Step (The Slow Lane):
Once the Fast Runner has finished its sprint for a "big" chunk of time, the computer switches gears. It looks at the deep layers and how they exchange water and heat with each other. Because these changes are slow, the computer only needs to take one large step for this part.

The Magic Trick: The paper proves that you can run these two steps separately and then stitch them back together perfectly. The result is that the computer can take huge steps for the slow, deep parts of the ocean while still keeping the fast surface waves accurate.

Why This Matters

Speed: The paper shows that for simulations where the waves aren't super fast (low "Froude number," which is common in real oceans), this method is much faster. It's like the Slow Walker finally getting to take a long stride instead of shuffling in place.
Accuracy: Even though they are taking shortcuts, the math proves the energy is conserved. The "cake" doesn't magically gain or lose height or speed. The model stays stable and doesn't blow up with errors.
Stability: The authors added a special "balance" feature. Imagine a lake that is perfectly still. Old models sometimes made the water ripple or drain away due to tiny computer errors. This new method keeps the lake perfectly still, just like in real life, which is crucial for long-term climate predictions.

The Bottom Line

The authors have built a smarter way to calculate ocean movements. By separating the "fast surface stuff" from the "slow deep stuff" and letting them run on their own time schedules, they can simulate the ocean much faster without losing accuracy. This is a big win for scientists trying to understand how the ocean moves and how it affects our climate over long periods.

Technical Summary: Barotropic-Baroclinic Splitting for Multilayer Shallow Water Models with Exchanges

Problem Statement
Global Ocean Circulation Models (OGCMs) face a severe computational constraint: the time step is limited by the fastest wave speed, specifically surface gravity waves (approx. 200 m/s), even though the primary phenomena of interest (stratification and currents) evolve on much slower time scales (approx. 3 m/s). While previous approaches like the rigid lid approximation or implicit surface height methods have been used, they have limitations. The barotropic-baroclinic decomposition offers a solution by separating fast surface gravity waves (barotropic) from slower vertical adjustments (baroclinic). However, existing splitting methods often struggle with maintaining consistency in multilayer frameworks with mass exchanges, preserving energy properties, and ensuring stability without losing accuracy, particularly in terrain-following coordinates.

Methodology
The authors propose a numerical analysis of a barotropic-baroclinic splitting within a nonlinear multilayer framework with mass exchanges between layers, formulated as an exact operator splitting. The model is based on inviscid Euler equations with a passive tracer in terrain-following ( $\sigma$ ) coordinates, assuming constant density.

The core methodology involves decomposing the layer-averaged Euler equations into two distinct operators:

Barotropic Step: This step handles the evolution of the free surface height ( $h$ ) and the depth-averaged velocity ( $\bar{u}$ ). It is formulated as a well-balanced one-layer shallow water model that includes pressure terms and topography source terms but excludes baroclinic forcing. Crucially, this step treats the transport of individual layer velocities ( $u_\alpha$ ) and tracers ( $T_\alpha$ ) at the mean velocity $\bar{u}$ .
Baroclinic Step: This step manages the vertical exchanges between layers and the adjustment of velocities relative to the mean. It solves for the deviation from the mean velocity ( $u_\alpha - \bar{u}$ ) and enforces the constraint that layer heights remain proportional to the total height ( $h_\alpha = l_\alpha h$ ) via Lagrange multipliers (exchange terms $G_{\alpha \pm 1/2}$ ). This system is fully 3D but lacks pressure and bathymetry terms, allowing for a larger time step.

The numerical scheme employs a Lie splitting with a subcycling strategy:

The baroclinic step is solved on a large time step $\Delta t$ .
The barotropic step is solved on the same interval using $K$ small sub-steps $\delta t$ constrained by the gravity wave speed.
A correction mechanism ensures that the mean velocity predicted by the barotropic step coincides with the mean of the individual velocities updated in the baroclinic step.

Key Contributions

Exact Operator Splitting Formulation: The paper reformulates the splitting as an exact operator sum ( $M_L = BT + BC$ ), clarifying the mathematical structure of the multilayer system with exchanges.
Preservation of Physical Properties: The authors prove that the splitting preserves total energy conservation (or dissipation) and satisfies a discrete entropy inequality. Specifically, the barotropic step satisfies an energy equality, while the baroclinic step satisfies an energy inequality due to the dissipative nature of the exchange terms.
Discrete Maximum Principles: The scheme is shown to satisfy discrete maximum principles for velocity and tracer concentrations, ensuring that values remain within physical bounds.
Well-Balancing: By utilizing a well-balanced solver (hydrostatic reconstruction) for the barotropic step, the overall scheme preserves the "lake-at-rest" equilibrium and, when extended with recent linear well-balancing strategies, maintains geostrophic equilibrium on coarse meshes.
Hyperbolicity Analysis: The paper investigates the hyperbolicity of the split systems, proving that the barotropic system is hyperbolic and analyzing the eigenvalues of the baroclinic system, noting that it remains strictly hyperbolic under specific conditions regarding velocity deviations.

Results
Numerical experiments demonstrate the efficacy of the proposed scheme:

Computational Efficiency: In low Froude number simulations (where the ratio of flow velocity to gravity wave speed is small), the splitting method drastically reduces computational cost compared to unsplit schemes. The gain is proportional to the Froude number.
Accuracy: Despite the splitting and subcycling, the method maintains accuracy comparable to unsplit schemes. Convergence studies against analytical solutions of the hydrostatic Euler equations show first-order convergence.
Diffusion Characteristics: In certain test cases (e.g., wind-driven cavity), the splitting method was found to be less diffusive than the unsplit scheme, though in others (stationary solutions), it exhibited slightly higher diffusion.
Geostrophic Equilibrium: The application of linear well-balancing strategies in the barotropic step significantly improves the resolution of western boundary currents in long-time integrations, capturing narrower and faster currents compared to standard schemes on coarse meshes.

Significance and Claims
The paper claims that this work provides a rigorous numerical analysis of a barotropic-baroclinic splitting that is both computationally efficient and physically consistent. The primary significance lies in the ability to simulate long-term ocean circulation with large time steps (constrained by slow baroclinic waves rather than fast gravity waves) without sacrificing the conservation of energy or the preservation of critical equilibria (lake-at-rest and geostrophic). The authors emphasize that the well-balancing strategy used in the barotropic step is inherited by the overall scheme, allowing for accurate simulations of geostrophic flows on coarse grids. The method is presented as a robust framework for both global and regional applications where free surface waves are not the primary focus.

The Problem: The "Fast Runner" vs. The "Slow Walker"

The Solution: Splitting the Team

Why This Matters

The Bottom Line

Technical Summary: Barotropic-Baroclinic Splitting for Multilayer Shallow Water Models with Exchanges

More like this