Model-Adaptive Simulation of Hierarchical Shallow Water Moment Equations in One Dimension

arXiv:2510.25351 · physics.flu-dyn, cs.NA, math.NA · Submitted 2025-10-29 · 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: "Model-Adaptive Simulation of Hierarchical Shallow Water Moment Equations in One Dimension".

Mira: Here's the gist of the paper. Model-adaptive simulation of hierarchical shallow water moment equations in one dimension develops adaptive simulations using Shallow Water Moment Equations to model subdomains with varying flow complexity in time and space The proposed strategy involves domain decomposition based on error estimators and coupling different-order models at interfaces using either a padded buffer cell or a conservative interface flux method.

Kai: Model Adaptivity Strategy The paper introduces a space- and time-adaptive procedure for the simulation of shallow free surface flows modeled by the hierarchical SWME, which consists of domain decomposition and spatial coupling.

Mira: The order of the SWME model is treated as space- and time-dependent, denoted as M = M(t, x).

Kai: This adaptivity is achieved by using a high-order SWME model in regions of high complexity and a low-order SWME model in regions of low complexity.

Paper summary: Kai: So, to recap what we just covered, this paper introduces the Model-Adaptive Simulation of Hierarchical Shallow Water Moment Equations in One Dimension. The core idea is that free surface flows have areas that are complex and areas that aren't, and these changes happen over time.

Mira: And the authors propose an adaptive strategy using hierarchical SWME to model these flows by switching between high and low order models depending on the local complexity of the flow at any given space and time point.

Lev: So they are essentially trying to avoid using a single, overly complex model everywhere, which would be computationally wasteful when things are simple.

Kai: Exactly. The mechanism for this switch is based on domain decomposition and error estimators derived from decomposing the higher-order SWME to identify regions that need coarsening or refinement.

Mira: They define specific criteria for model coarsening based on residuals and they also have refinement criteria derived from assumptions about how well lower-order models approximate the higher-order ones.

Lev: The paper mentions analytical formulas for delta AML, MH w MH and delta SML, MH w MH to help set up those error estimation procedures.

Kai: That means they provide the mathematical tools needed to actually implement those criteria—the Jacobian matrices and source term differences are what you need to evaluate those switching conditions.

Mira: So while the abstract tells us *what* they did, the summary clarifies that it's not just applying a fixed decomposition; it's an active procedure where the model order M is treated as dependent on time and space, M equals M of t comma x.

Lev: That dependence on time and space is what makes it truly adaptive rather than just spatially uniform refinement.

Kai: And they also discussed how to couple these different-order models at the interfaces using either the Padded Buffer Cell coupling or the Conservative Interface Flux coupling methods for spatial continuity.

Mira: They are testing both, showing that both approaches work and can yield accurate results when used with their adaptive model switching strategy.

Lev: It's interesting that they tested these two very different interface methods, one being potentially less conservative than the other, just to see which one integrates better with the adaptivity scheme.

Kai: The numerical validation part shows they simulated a dam-break wave hitting a smooth wave and both coupling methods gave accurate results and speedups up to sixty percent compared to a higher order model.

Mira: That comparison is key because it shows that the adaptive approach isn't just theoretical; it performs well against what we already have, which is the standard higher order SWME5.

Lev: And they showed across different friction cases, the adaptive models produced similar results and still managed to capture those specific regions requiring a higher order model.

Kai: The runtime comparisons were quite telling too, showing a speedup of up to sixty percent using either the Padded Buffer Cell or Conservative Interface Flux compared to the standard SWME5.

Mira: It really highlights how this hierarchical approach handles non-uniform flow dynamics in a way that is both accurate and computationally efficient for these kinds of problems.

Conclusion: Kai: So, wrapping up, the Model-Adaptive Simulation of Hierarchical Shallow Water Moment Equations in One Dimension by Verbiest and Koellermeier proposes a space- and time-model-adaptive procedure for simulating free surface flows using hierarchical SWME. They use domain decomposition and spatial coupling methods like PBC or CIF to achieve accurate results with speedups up to sixty percent.

Mira: The implications are that we can simulate these complex fluid dynamics more efficiently by automatically adjusting the model order based on where the flow is complicated, which is a big deal for computational cost.

Lev: It’s about finding a way to use the hierarchy of models—high order for hard areas, low order for easy areas—without bogging down every single calculation with that high complexity everywhere.

Mira: They’re suggesting that future research should look into combining this adaptive procedure with data-driven methods for error estimation to make the model switching even more automated.

Lev: That would be a step toward a fully autonomous system where it doesn't just rely on predefined thresholds, but actually learns the optimal way to adapt in real time based on what it observes.

Rik Verbiest, Julian Koellermeier

Bernoulli Institute, University of Groningen · Department of Mathematics, Computer Science and Statistics, Ghent University

physics.flu-dyn, cs.NA, math.NA

Submitted: 2025-10-29

Updated: 2026-11-04

Comments: 31 pages, 7 Figures

Journal ref: Verbiest, R., Koellermeier, J. Model-Adaptive Simulation of Hierarchical Shallow Water Moment Equations in One Dimension, Comput. Math. Appl., 221, 135-158, 2026

DOI: 10.1016/j.camwa.2026.07.036

License: http://creativecommons.org/licenses/by-nc-nd/4.0/

Importance score: 92/100

The gist: Model-adaptive simulation of hierarchical shallow water moment equations in one dimension develops adaptive simulations using Shallow Water Moment Equations to model subdomains with varying flow…

Key concepts

Model Adaptivity
This strategy means the simulation switches between different levels of mathematical detail depending on the local conditions. If a region has complex flow, it uses a high-order model; if it is simpler, it uses a low-order model. This ensures accuracy where needed without wasting computational power elsewhere.
Domain Decomposition
The simulation space is divided into smaller sections or domains. These sections are analyzed using error estimators to decide whether to make them finer (refinement) or coarser (coarsening). This process helps pinpoint exactly where the flow behavior changes significantly.
Spatial Coupling Methods
These are techniques used at the boundaries between different simulation domains. Two methods were tested: a Padded Buffer Cell, which uses a buffer area with dynamic values, and a Conservative Interface Flux method, which modifies only the flow rates at the edges to maintain conservation.

Terminology

Summary

Model-adaptive simulation of hierarchical shallow water moment equations in one dimension develops adaptive simulations using Shallow Water Moment Equations to model subdomains with varying flow complexity in time and space The proposed strategy involves domain decomposition based on error estimators and coupling different-order models at interfaces using either a padded buffer cell or a conservative interface flux method. p. 0

Where the paper says it

Model-Adaptive Simulation of Hierarchical Shallow Water Moment Equations in One Dimension Rik Verbiest∗1 and Julian Koellermeier1, 2 1Bernoulli Institute, University of Groningen 2Department of Mathematics, Computer Science and Statistics, Ghent University November 4, 2025 Abstract Shallow free surface flows are often characterized by both subdomains that require high modeling complexity and subdomains that can be sufficiently accurately modeled with low modeling complexity. Moreover, these subdomains may change in time as the water flows through the domain.

Model Adaptivity Strategy

The paper introduces a space- and time-adaptive procedure for the simulation of shallow free surface flows modeled by the hierarchical SWME, which consists of domain decomposition and spatial coupling. The order of the SWME model is treated as space- and time-dependent, denoted as M = M(t, x). p. 6

Where the paper says it

Spatially- and time-adaptive order. So far, the order M of the SWME model has been treated as fixed and constant in time and space throughout the literature.

This adaptivity is achieved by using a high-order SWME model in regions of high complexity and a low-order SWME model in regions of low complexity. p. 2
Where the paper says it

Verbiest complexity with sufficient accuracy, while in regions of low complexity, a low-order SWME model with order ML with small computational effort is used with sufficient accuracy. The switching procedure between high-order SWME and low-order SWME should be based on a suitable set of error estimators that estimate the model error made using a SWME model of a certain order.

The switching procedure between models is based on a set of error estimators derived from the decomposition of the higher-order models. p. 0
Where the paper says it

The first approach dynamically updates padded state variables but cannot be written in conservative form, while the second approach uses fixed padded state variable values of zero and reduces to conservative form for conservative moment equations. The switching procedure between high-order models and low-order models is based on a new set of model error estimators, originating from a decomposition of the high-order models.

Domain Decomposition Criteria

The domain decomposition is based on error estimators constructed from a decomposition of a higher-order SWMEMH to identify regions for model coarsening or refinement. Model-coarsening criteria are defined by the set of conditions where the residuals R:ML(t, x) and RML:MH(t, x) exceed chosen thresholds. p. 6

Where the paper says it

The error estimators are then used to decompose the domain into subdomains, each modeled by a SWME model of an appropriate order. This is done by evaluating whether the error estimators exceed chosen thresholds, which results in a set of domain decomposition criteria for model-coarsening, i.

Model-refinement criteria are derived from assumptions about equivalence between lower and higher-order models to identify regions where order should be increased. p. 9
Where the paper says it

Assume now that the flow is modeled by the lower-order SWMEML model. To identify where the order of the SWME model should be increased from lower order ML to higher order MH to capture the complex fluid flow dynamics, the following rationale will be used.

Spatial Coupling Methods

Two methods are proposed for coupling different-order SWME models at boundary interfaces: (1) the Padded Buffer Cell (PBC) coupling, which uses a padded buffer cell with dynamically updated padded values that can be non-conservative. (2) the Conservative Interface Flux (CIF) coupling, which sets the padded values to zero and results in a coupling method that can be written in conservative form by only modifying the fluxes at the boundary interfaces. p. 0

Where the paper says it

To enable adaptive simulations, we propose two approaches for the coupling of the varying-order shallow water moment equations at their boundary interfaces. The first approach dynamically updates padded state variables but cannot be written in conservative form, while the second approach uses fixed padded state variable values of zero and reduces to conservative form for conservative moment equations.

Numerical Validation and Results

Numerical simulations of a dam-break wave colliding with a smooth wave were performed using both A-SWME-PBC and A-SWME-CIF, showing accurate results and speedups up to 60 percent compared to a higher-order model. The comparison across different friction cases demonstrated that the adaptive models yield similar results and accurate captures of regions requiring higher order. p. 0

Where the paper says it

The switching procedure between high-order models and low-order models is based on a new set of model error estimators, originating from a decomposition of the high-order models. Numerical results of the collision of a dam-break wave with a smooth wave yield accurate results, while achieving speedups up to 60 percent compared to a non-adaptive model with fixed modeling complexity.

Runtime comparisons showed a significant decrease in runtime of up to 60 percent using the A-SWME-PBC or A-SWME-CIF compared to the higher-order model SWME5. p. 2
Where the paper says it

In [15], model reduction of a hyperbolic regularization of the SWME using POD-Galerkin and dynamical lowrank approximation is studied and shown to result in accurate numerical simulations and considerable runtime improvements. Previous works have studied the coupling of the fully resolved Reynolds-Averaged Navier-Stokes equations with the SWE [23] and with SWME models of fixed order M [26].

Analytical Error Estimation

Analytical formulas for deltaAML,MH(wMH) and deltaSML,MH(wMH) are provided in Appendix A to facilitate the error estimation process. These formulas allow for the construction of Jacobian matrices and source term differences needed to evaluate the model-coarsening and model-refinement criteria. p. 8

Where the paper says it

Verbiest where δAML,MH ( ⃗wMH ) ∈ R (ML+2)×(ML+2) represents the difference between the system matrices of the SWMEML and the SWMEMH for the first ML + 2 variables and where δS⃗ML,MH ( ⃗wMH ) ∈ RML+2 represents the difference between the source terms of the SWMEML and the SWMEMH for the first ML+2 variables. The analytical formulas for δAML,MH ( ⃗wMH ) and δS⃗ML,MH ( ⃗wMH ) are given in Appendix A.

Conclusion

The paper proposes a space- and time-model-adaptive procedure for simulating free surface flows modeled by hierarchical SWME, incorporating domain decomposition and spatial coupling methods like PBC or CIF to achieve accurate results with significant speedups. The work suggests future research into combining this procedure with data-driven methods for error estimation. p. 2

Where the paper says it

A final approach that is worth mentioning is the micro-macro decomposition [18], which decomposes the time integration of hierarchical kinetic moment equations into potentially several small micro steps with a higher-order model and a single large macro step with a lower-order model. Another approach is a data-driven reduction of the number of variables in the model.

CRediT authorship contribution statement

R. Verbiest: Conceptualization, Methodology, Software, Formal Analysis, Investigation, Data Curation, Writing - Original Draft, Writing - Review & Editing, Visualization.

J. Koellermeier: Conceptualization, Methodology, Writing - Review & Editing, Supervision.

Acknowledgments

This work is part of the project HiWAVE with file number VI.Vidi.233.066 of the ENW Vidi research programme, funded by the Dutch Research Council (NWO).

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. Model-adaptive simulation of hierarchical shallow water moment equations allows AI systems to dynamically adjust their modeling complexity based on local flow features by switching between high-order and low-order models in different spatial regions of the domain. This enables accurate simulation with sufficient physical complexity where needed and computational efficiency where flow is uniform, achieving speedups up to 60 percent compared to non-adaptive models.

  2. The system can perform space- and time-adaptive order selection by decomposing the domain into subdomains based on error estimators that evaluate whether the effect of adding higher-order moments can be directly estimated. This allows the simulation to locally choose an order M(t, x) based on the local flow complexity.

  3. The system can implement two distinct spatial coupling methods: Padded Buffer Cell (PBC) coupling, which uses padded values that are dynamically updated, and Conservative Interface Flux (CIF) coupling, which sets the padded values to zero. This provides flexibility in handling subdomains with different model orders at their boundaries.

  4. The system can perform model-coarsening by reducing the order from higher order MH to lower order ML if the norm of the residual R:ML(t∗, x∗) is less than epsilonR:ML and other criteria are met. This identifies regions where a lower complexity model is sufficient, thereby reducing computational cost.

  5. The system can perform model-refinement by increasing the order from lower order ML to higher order MH if one of the model-refinement criteria in (3.19) is satisfied, such as "SMi:Mi+1(w n,(0) i) > κS." This automatically increases complexity in regions of high flow dynamics, ensuring accuracy where required.

  6. The system can utilize a numerical scheme that reduces to a conservative finite volume scheme for conservative moment equations by setting the padded moments to zero in the CIF coupling method. This ensures that the resulting flux computation is written in conservative form for conservative moment systems.

Abstract

Shallow free surface flows are often characterized by both subdomains that require high modeling complexity and subdomains that can be sufficiently accurately modeled with low modeling complexity. Moreover, these subdomains may change in time as the water flows through the domain. This motivates the need for space and time adaptivity in the simulation of shallow free surface flows. In this paper, we develop the first adaptive simulations using the recently developed Shallow Water Moment Equations, which are an extension of the standard Shallow Water Equations that allow for vertically changing velocity profiles by including additional variables and equations. The model-specific modeling complexity of a shallow water moment model is determined by its order. The higher the order of the model, the more variables and equations are included in the model. Shallow water moment models are ideally suited for adaptivity because they are hierarchical such that low-order models and high-order models share the same structure. To enable adaptive simulations, we propose two approaches for the coupling of the varying-order shallow water moment equations at their boundary interfaces. The first approach dynamically updates padded state variables but cannot be written in conservative form, while the second approach uses fixed padded state variable values of zero and reduces to conservative form for conservative moment equations. The switching procedure between high-order models and low-order models is based on a new set of model error estimators, originating from a decomposition of the high-order models. Numerical results of a dam break and of the collision of a dam-break wave with a smooth wave for different friction parameter values yield accurate results, while achieving speedups up to 40 percent compared to a non-adaptive model with fixed modeling complexity.

Related papers