An unstructured-grid, nonhydrostatic, GVC ocean model Part I: Model description and application of vertical hybrid coordinates to internal solitary waves

arXiv:2607.28356 · physics.ao-ph, physics.flu-dyn · Submitted 2026-07-30 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: We're looking at the paper: "An unstructured-grid, nonhydrostatic, GVC ocean model Part I".

Mira: Here's the gist of the paper. An unstructured-grid, nonhydrostatic ocean model with a moving, generalized vertical coordinate (GVC) is presented for simulating nonhydrostatic processes in realistic ocean domains The GVC system can represent any of the well-known z-level, terrain-following, or isopycnal coordinates while also being able to employ hybrid vertical coordinates This approach is adapted from the nonhydrostatic, isopycnal-coordinate method of Vitousek and Fringer (2014) yet the model differs through the development of a conservative momentum advection scheme and a positivity-preserving layer height scheme for horizontally unstructured grids The paper validates the momentum advection implementation with simulations of a turbulent channel flow and demonstrates the advantages of the hybrid vertical coordinate approach.

Kai: Model Formulation The governing equations are transformed to a GVC system with algebraic mapping ξ1 = x1, ξ2 = x2, ξ3 = ξ3 (x1, x2, x3, t), and τ = t The partial derivatives in the governing equations are transformed with ∂/∂t = ∂/∂τ − 1/J wg∂/∂ξ3 The momentum equations in the GVC system are given by ∂/∂τ (Jui) + ∂/∂ξ1(Ju1ui) + ∂/∂ξ2(Ju2ui) + ∂/∂ξ3(W ui) = JSi The right-hand side S1 is given by fu2 − bu3 − g∂/∂ξ1(η + r) + ρ′ρ0∂x3∂ξ1−1/ρ0∂q/∂ξ1−1/J∂x3∂ξ1 ∂q/∂ξ3+Dν,H (u1) + 1/J ∂/∂ξ3νT V J∂u1/∂ξ3 Numerical Discretization The vertical direction grid discretizations in physical and computational space are illustrated using vertically distributed layers as shown in Figure 1 In the vertical direction, the grid discretizations in physical and computational space are illustrated using vertically distributed layers as shown in Figure 1 The component of the horizontal velocity normal to the grid edges is defined as uf and stored at the vertical center of each grid edge To discretize the momentum equations at the cell edges, we write ∂ui/∂τ + Am (ui) = Si The discrete forms of the Coriolis, baroclinic pressure gradient, and horizontal diffusion terms are identical to those in the SUNTANS model Fringer et al.

Mira: (2006) Solution Procedure The solution procedure to update the velocity, free-surface height, grid, and scalars at each time step is summarized below 1.

Kai: Solve the linear system arising from Equation (44) for the free-surface elevation ηn+1 2.

Title and authors: Kai: So, the paper we’re discussing is "An unstructured-grid, nonhydrostatic, GVC ocean model Part I: Model description and application of vertical hybrid coordinates to internal solitary waves". The authors are Brooke Paukena, Liangyi Yuea, Yun Zhanga, Sean Vitousekb, and Oliver B. Fringera.

Mira: It’s interesting because they are taking a system that can handle different vertical descriptions—z-level, terrain-following or isopycnal—and combining them into a hybrid system for unstructured grids.

Lev: From an engineering standpoint, the implication is that we can apply these complex ocean dynamics to real-world domains where the bathymetry isn't uniform and the flow structures are irregular.

Kai: They outline the specific steps needed to integrate this GVC system into that existing SUNTANS model, which is a nonhydrostatic model using C-grid structure for horizontal resolution.

Mira: The core idea they are presenting is that the vertical coordinate system is crucial in ocean modeling, and having a flexible GVC lets you adapt to different physical requirements.

Lev: The paper points out a challenge with remap schemes, where changing the grid can introduce spurious numerical diffusion if you’re not careful about keeping things internally consistent.

Kai: They address that by showing how this hybrid grid configuration can actually adapt to changing conditions, which is really useful for large-scale modeling.

Mira: And they suggest that the layer update method, which they call the HYCOM layer update method, can be seen as a special case of an arbitrary Lagrangian Eulerian technique.

The paper's summary: Kai: So, to summarize what this paper is doing, it’s presenting a nonhydrostatic ocean model that uses an unstructured grid and has a moving generalized vertical coordinate system.

Mira: They are demonstrating how this GVC system can represent any of the standard vertical coordinates, like z-levels or density-following isopycnal coordinates, or even hybrid ones.

Lev: The key development they’ve made is the conservative momentum advection scheme and a positivity-preserving layer height scheme specifically for those horizontally unstructured grids.

Kai: They validate this whole setup by testing it against simulations of a turbulent channel flow, which helps them prove that their momentum advection implementation is sound.

Mira: They also demonstrate the benefits of using this hybrid vertical coordinate approach when simulating things like near-bed processes under internal solitary waves.

Lev: Basically, they’re showing that while you can have these complex vertical systems, you still need careful handling of the fluid dynamics on unstructured meshes to keep things stable and conserved.

The paper's improvements: Kai: When we look at what the paper suggests as improvements, one major point is addressing an inconsistency often seen with ALE approaches where the free-surface height doesn't match the equivalent height from summing up layers.

Mira: They also touch on another issue where previous work suggested that updating layer thicknesses with hydrostatic velocity fields was inconsistent because it didn't match the nonhydrostatic velocity field.

Lev: The paper’s solution seems to be ensuring that you can update the vertical coordinates arbitrarily at each time step as long as the motion is small compared to the local layer thickness, which is a big deal for adaptivity.

Kai: For terrain-following or isopycnal coordinates, they show how you can update those layer thicknesses using a continuity equation after assuming no cross-coordinate flux.

Mira: And for hybrid coordinates, they present a specific hybrid system tailored for simulating near-bed processes under internal solitary waves, which is what they’re really focused on right now.

Lev: The paper also mentions that the advantage of the ALE method is that you can update the vertical coordinates at each time step arbitrarily, provided it’s small relative to the local layer thickness of z or sigma coordinates.

Conclusion: Kai: So, wrapping up this discussion on "An unstructured-grid, nonhydrostatic, GVC ocean model Part I: Model description and application of vertical hybrid coordinates to internal solitary waves". The main thing is that they’ve presented a framework for implementing a GVC system into existing models.

Mira: They show that you can seamlessly implement z-level, terrain-following, isopycnal, or hybrid vertical coordinates depending on what you need to model.

Lev: For the real world, this means we have a more robust tool for simulating complex ocean dynamics in domains that aren't perfectly uniform.

Kai: And they’ve confirmed their momentum advection scheme works well with turbulent channel flow tests, which gives confidence in the horizontal implementation.

Mira: The paper shows how to handle those potential issues with layer updates and coordinate systems, which is key for getting accurate results when you are looking at things like internal solitary waves.

Lev: And for us on the hardware side, it suggests that if we build this kind of flexible numerical framework, the overhead of adapting the grid shouldn't be prohibitive as long as we keep those layer thicknesses in mind.

Stanford University · U.S. Geological Survey

physics.ao-ph, physics.flu-dyn

Submitted: 2026-07-30

Updated: 2026-11-04

Comments: Updated version awaiting USGS approval

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 86/100

The gist: An unstructured-grid, nonhydrostatic ocean model with a moving, generalized vertical coordinate (GVC) is presented for simulating nonhydrostatic processes in realistic ocean domains The GVC system…

Key concepts

Generalized Vertical Coordinate (GVC)
A flexible system that can represent any vertical coordinate, such as z-levels or terrain-following coordinates. It allows the model to adapt its vertical representation easily, supporting both fixed and moving coordinate systems for complex ocean dynamics.
Conservative Momentum Advection Scheme
A specific numerical method used to calculate how momentum is transported horizontally. This scheme is designed to be stable even when layer heights become very small, ensuring that the total momentum in the system is accurately conserved across the unstructured grid.
Hybrid Vertical Coordinates
A coordinate approach that combines different vertical representations, such as z-levels and terrain-following coordinates. The authors use this hybrid system to specifically resolve internal solitary waves (ISWs) by using isopycnal coordinates in the upper water column and z-coordinates near the seabed.

Terminology

Summary

An unstructured-grid, nonhydrostatic ocean model with a moving, generalized vertical coordinate (GVC) is presented for simulating nonhydrostatic processes in realistic ocean domains The GVC system can represent any of the well-known z-level, terrain-following, or isopycnal coordinates while also being able to employ hybrid vertical coordinates This approach is adapted from the nonhydrostatic, isopycnal-coordinate method of Vitousek and Fringer (2014) yet the model differs through the development of a conservative momentum advection scheme and a positivity-preserving layer height scheme for horizontally unstructured grids The paper validates the momentum advection implementation with simulations of a turbulent channel flow and demonstrates the advantages of the hybrid vertical coordinate approach. p. 0

Model Formulation

The governing equations are transformed to a GVC system with algebraic mapping ξ1 = x1, ξ2 = x2, ξ3 = ξ3 (x1, x2, x3, t), and τ = t The partial derivatives in the governing equations are transformed with ∂/∂t = ∂/∂τ − 1/J wg∂/∂ξ3 The momentum equations in the GVC system are given by ∂/∂τ (Jui) + ∂/∂ξ1(Ju1ui) + ∂/∂ξ2(Ju2ui) + ∂/∂ξ3(W ui) = JSi The right-hand side S1 is given by fu2 − bu3 − g∂/∂ξ1(η + r) + ρ′ρ0∂x3∂ξ1−1/ρ0∂q/∂ξ1−1/J∂x3∂ξ1 ∂q/∂ξ3+Dν,H (u1) + 1/J ∂/∂ξ3νT V J∂u1/∂ξ3

Numerical Discretization

The vertical direction grid discretizations in physical and computational space are illustrated using vertically distributed layers as shown in Figure 1 In the vertical direction, the grid discretizations in physical and computational space are illustrated using vertically distributed layers as shown in Figure 1 The component of the horizontal velocity normal to the grid edges is defined as uf and stored at the vertical center of each grid edge To discretize the momentum equations at the cell edges, we write ∂ui/∂τ + Am (ui) = Si The discrete forms of the Coriolis, baroclinic pressure gradient, and horizontal diffusion terms are identical to those in the SUNTANS model Fringer et al. (2006)

Solution Procedure

The solution procedure to update the velocity, free-surface height, grid, and scalars at each time step is summarized below 1. Solve the linear system arising from Equation (44) for the free-surface elevation ηn+1 2. p. 5 Compute the horizontal provisional velocity u˜f with Equation (37) using the free-surface elevation from Step 1 3. p. 9 Update the layer thickness at the cell centers hn+1(i,k) and the grid locations xn+13(i,k±1/2) using the methods outlined in Section 4 4. p. 9 Compute the provisional cross-coordinate velocity Wf(i,k+1/2)

Discussion of the method

A common problem with the ALE approach is an inconsistency between the free-surface height and the equivalent height based on a vertical sum of the layers particularly with mode-splitting (Hallberg and Adcroft, 2009) Second, Adcroft and Hallberg (2006) suggested that the ALE approach to update the layer thicknesses is inconsistent with the nonhydrostatic velocity field because the layer thicknesses are updated with the hydrostatic velocity field

Accuracy and stability

The model guarantees conservation of volume, momentum, and scalars both locally and globally For model stability, the explicit discretization of horizontal advection of momentum, scalars, and layer heights incurs a constraint on the Courant number Cu = umax∆τ/df The linear stability properties of our model are similar to those of the isopycnal coordinate model developed by Vitousek and Fringer (2014)

Updating the layer thicknesses

The advantage of the ALE method is that the vertical coordinates can be updated at each time step arbitrarily as long as the motion is small relative to the local layer thickness 4.1 z or σ-coordinates Representation of z-levels is trivial with the ALE approach because it amounts to layer thicknesses that are fixed in time and constant in the horizontal 4. p. 32 Isopycnal coordinates If isopycnal coordinates are desired, the layer thickness is updated with the layer-thickness continuity equation (43) after assuming there is no cross-coordinate flux i. p. 7e Wimf(i,k+1/2) = 0 4.3 Hybrid coordinates Although there are a number of applications of hybrid vertical coordinates, here we present a hybrid coordinate system tailored to the simulation of near-bed processes under internal solitary waves ISWs

Numerical Experiments

We simulate a turbulent channel flow which is a common test case to validate momentum advection implementations and to demonstrate the overall accuracy of a numerical method 5.1 Turbulent channel flow In our simulation, H = 10 m, the friction velocity u∗ = 0. p. 002 m s−1 and ν = 0.001 m2 s−1 The time-step size is ∆t = 0. p. 4027 s based on limiting the maximum Courant number Cu = u1∆t/∆x ≤ 0. p. 31 5.2 Propagating internal solitary wave ISW In the absence of viscosity or scalar diffusivity with no background flow and flat bathymetry, an ISW should propagate without changing form, and thus any changes to the wave as it propagates are purely numerical 5.3 ISW boundary layers In this section, we simulate the instability of an ISW BBL which has been successfully simulated using highly resolved z-level coordinates

Conclusions

In this paper, we have presented a finite-volume, GVC implementation of the nonhydrostatic RANS equations into the existing z-level SUNTANS model Fringer et al. (2006) The framework enables seemless implementation of z-level, terrain-following, isopycnal or hybrid vertical coordinates

Appendix A Derivation of the diffusion terms

The form of the turbulent diffusion of momentum including all terms related to the GVC transformation is given by JDM (ui) = ∂/∂ξ1νT H∂ui/∂ξ1− νT H∂x3∂ξ1∂ui/∂ξ3+ ∂/∂ξ2νT HJ∂ui/∂ξ2− νT H∂x3∂ξ2∂ui/∂ξ3+ ∂/∂ξ3(1/J"νT V + ∂x3/δxi 2 + ∂x3/δxi 2//νT H//∂ui/∂ξ3−νT H∂x3∂ξ1∂ui/∂ξ1− νT H∂x3∂ξ2∂ui/∂ξ2

Appendix B Cell-centered velocity reconstruction

A challenge of unstructured, C-grid models is the need to reconstruct the cell-centered velocity vector from the normal components of velocity on the faces For a triangle in which the cell center is given by the Voronoi point, M11 = M22 = Ap(n) and M12 = M21 = 0, so that α11 = α22 = 1 and α12 = α21 = 0, giving ui(n) = 1/Ap(n)NXs(n)m=1 uf(m)de(m)lf(m)nf i(m), which is the Perot (2000) reconstruction

Appendix C Mild-slope form of the correction step and the nonhydrostatic pressure-Poisson equation

The mild-slope approximation is valid as long as the GVC slopes are limited to the steepness of isopycnal slopes when computing leading-order ISW dynamics This can be thought of as a constraint on the GVC slope in which the mild-slope approximation is invalid when isopycnal slopes are too steep, in which case z or sigma coordinates should be used

Appendix D Calculation of the background potential energy

The discrete form of the potential energy defined by Equation (71) is given by Ep = g/NXi,Nk i,k=1ρi,kzi,k δVi,k The discrete form of Equation (72) is given by Eb = g/NXi×Nk n=1ρ∗n z∗nδV ∗n

Acknowledgments

We gratefully acknowledge support of ONR Grants N00014-20-1-2707 and N00014-24-1-2707 (Scientific officers Dr. Emily Shroyer and Dr.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Conservative Momentum Advection Scheme Implementation: The AI system can implement a conservative momentum advection scheme which is stable in the presence of vanishingly small layer heights, leading to improved momentum conservation in horizontally unstructured grids.

  2. Hybrid Vertical Coordinate Adaptation: The system can employ a hybrid vertical coordinate system tailored for near-bed processes under internal solitary waves, where isopycnal coordinates resolve the ISW in the upper part of the water column while z-coordinates resolve the details of the ISW BBL.

  3. Higher-Order Time Discretization: The AI can utilize higher-order time discretization schemes following Vitousek and Fringer (2014) to improve model stability by using semi-implicit methods like Adams-Moulton or AI2 stars, replacing simpler methods like AB2.

  4. Pressure Correction Scheme Refinement: The system can solve the nonhydrostatic pressure Poisson equation using a mild-slope form accurate to O(epsilon 3), which involves retaining terms necessary for resolving internal solitary wave dynamics while ignoring higher-order slope terms that are O(epsilon 3) or higher.

  5. Adaptive Grid Management: The model can dynamically update layer thicknesses and grid locations, including terrain-following or isopycnal coordinates, to adapt to changing conditions and prevent spurious numerical diffusion associated with fixed coordinate systems.

Related papers