Entropy stability and well-balancedness of space-time DG for the shallow water equations with bottom topography

Andreas Hiltebrand, Siddhartha MishraNetworks Heterog. Media 2016 · OpenReview

Abstract

This paper presents a novel space-time Discontinuous Galerkin (DG) method for solving the shallow water equations (SWE) in the presence of variable bottom topography. The proposed scheme achieves two critical numerical properties simultaneously: entropy stability, which ensures physical consistency and energy dissipation, and well-balancedness, which preserves the "lake at rest" steady state to machine precision. These properties are maintained even on unstructured grids, providing a robust framework for simulating complex geophysical flows.

Introduction

Simulating water flow in oceans, rivers, and coastal regions using the Shallow Water Equations (SWE) presents significant numerical challenges. Two of the most prominent hurdles are the maintenance of the "well-balanced" property and the assurance of entropy stability.

In oceanography, many phenomena of interest are small-scale perturbations (such as waves) occurring on top of a steady state. A classic example is the "lake at rest" scenario, where the water surface remains flat despite a complex, uneven bottom. If a numerical method is not "well-balanced," it introduces artificial numerical oscillations—essentially "numerical storms"—that can overwhelm the actual physical waves being studied.

Furthermore, because the SWE are nonlinear, they can develop shocks or hydraulic jumps. Ensuring that a high-order method remains stable without "blowing up" requires rigorous mathematical guarantees. While many existing methods address one of these issues, few provide a unified framework that is both well-balanced and entropy-stable (respecting the laws of thermodynamics regarding energy dissipation).

Methodology

Space-Time Discontinuous Galerkin Framework

The authors employ a space-time DG approach, which differs from traditional semi-discrete methods by treating time as an additional dimension. The solution is solved over a space-time "slab," allowing for a fully implicit discretization that is inherently suited for complex geometries.

Entropy Variables and Stability

To ensure stability, the method utilizes entropy variables as the primary degrees of freedom rather than standard conserved variables (water height and momentum). This choice simplifies the proof of energy stability, ensuring the scheme remains physically consistent. To handle sharp gradients and shocks, the authors incorporate:

  • Streamline Diffusion and Shock-Capturing: Specialized operators adapted for balance laws to suppress numerical oscillations near discontinuities.
  • Rusanov-type Diffusion: An entropy-stable numerical flux that provides sufficient artificial friction to maintain stability without sacrificing the accuracy of the underlying physics.

Achieving the Well-Balanced Property

A significant insight of this work is the demonstration that by performing integration by parts on spatial terms and carefully discretizing the source term (bottom topography), the numerical errors that typically degrade the "lake at rest" state cancel out exactly. This allows the scheme to maintain the steady state regardless of the mesh configuration.

Numerical Experiments

The method was evaluated using several standard benchmarks designed to test both accuracy and robustness:

  1. 1D Lake at Rest: A fundamental test to verify the preservation of a still water surface over a curved bed.
  2. Perturbed 1D Lake: A small water bump is introduced to a quiescent state to evaluate the method's ability to track small-amplitude waves without interference from topography-induced noise.
  3. 2D Benchmarks: The tests were extended to two dimensions using unstructured triangular meshes. These meshes are essential for modeling jagged coastlines but are notoriously difficult to balance compared to simple Cartesian grids.

Key Results and Discussion

Precision and Accuracy

In the "lake at rest" simulations, the relative error for water surface height reached machine precision, ranging from $10^{-15}$ to $10^{-18}$. This confirms that the method is perfectly well-balanced. Furthermore, high-order versions of the scheme ($p=1$ and $p=2$) demonstrated superior resolution of small perturbations, capturing wave dynamics with minimal numerical smearing even on relatively coarse meshes.

Unconditional Stability

As an implicit, entropy-stable method, the scheme is unconditionally stable. This allows for significantly larger time steps than those permitted by standard explicit methods, which is a major advantage for long-term climate or tidal simulations where computational efficiency is paramount.

Flexibility on Unstructured Meshes

Unlike many previous energy-stable schemes restricted to square grids, this method maintains its properties on unstructured triangular meshes. This flexibility is vital for practical engineering applications involving complex river bends and coastal topographies.

Conclusion and Future Work

The space-time DG method presented provides a mathematically rigorous and computationally robust approach to the shallow water equations. While the current scheme is perfectly well-balanced for the "lake at rest" state, the authors identify the extension to moving steady states (e.g., constant river flow over a bump) as a primary area for future research. Additionally, the integration of this method with goal-oriented adaptive algorithms—where the mesh automatically refines in regions of high physical complexity—promises further gains in computational efficiency and resolution.

Summary updated

BibTeX
@inproceedings{hiltebrand2016entropy,
  author = {Andreas Hiltebrand and Siddhartha Mishra},
  title = {{Entropy stability and well-balancedness of space-time DG for the shallow water equations with bottom topography}},
  booktitle = {Networks Heterog. Media},
  year = {2016},
  url = {https://openreview.net/forum?id=MiQWJp7HKEL},
}
Figure