The remapped particle‐mesh semi‐Lagrangian advection scheme

Colin J. Cotter, J. Frank, Sebastian ReichQuarterly 2007 · DOI

TL;DR

This paper introduces a novel "remapped particle-mesh" (RPM) method for solving the density advection equation in fluid dynamics simulations, particularly for numerical weather prediction. The RPM method is mass-conserving, computationally efficient, and achieves accuracy comparable to or better than existing semi-Lagrangian advection schemes, even on spherical geometries.

What Problem Does This Paper Address?

Accurately simulating fluid flow, especially in atmospheric and oceanic models, requires robust methods for solving the advection equation, which describes how quantities like density are transported by the flow. Semi-Lagrangian (SL) methods are popular because they avoid strict time-step limitations imposed by explicit Eulerian methods. However, standard SL methods often rely on interpolation (e.g., cubic interpolation) to estimate values at departure points, which can lead to mass loss or numerical diffusion. While mass-conserving methods exist, they can be computationally expensive or complex. This paper aims to bridge this gap by developing a new, efficient, and accurate mass-conserving advection scheme suitable for integration with existing SL frameworks.

Methods

The core of the proposed method is the "remapped particle-mesh" (RPM) approach. It builds upon the particle-mesh (PM) discretization, which is inherently mass-conserving. The key innovation lies in how it's integrated into an SL framework:

  • Particle-Mesh Discretization: The density is represented by a set of discrete particles, each carrying a "mass." The density at grid points is then calculated by interpolating these particle masses using basis functions (like cubic B-splines). This PM approach ensures mass conservation by construction.
  • Remapping Step: At the beginning of each time step, the particles are "reset" or "remapped" to Eulerian grid points. This is a crucial step that allows for efficient updates of the velocity field on a fixed grid while retaining the mass conservation property of the PM method.
  • Mass Update: After remapping, new particle masses are calculated based on the grid densities and the basis functions. This step is equivalent to solving a tridiagonal system of equations, which can be done efficiently.
  • Forward Stepping: The particles are then advected forward in time using the updated velocity field. The new density on the Eulerian grid is then computed using the remapped particle masses and the advected particle positions.
  • Extension to Sphere: The method is also adapted for spherical geometries, which is essential for global atmospheric models. This involves using spherical coordinates and appropriate basis functions.

The paper contrasts this with standard SL methods that use backward trajectories and bicubic interpolation, highlighting the RPM's advantages in mass conservation and accuracy.

Datasets

The paper does not introduce new datasets. Instead, it uses standard benchmark test cases for advection schemes, including:

  • 1D Convergence Test: Uniform advection of a sine wave on a periodic domain.
  • 2D Planar Advection:
    • Smoothed slotted-cylinder problem.
  • Idealized cyclogenesis problem.
  • Spherical Advection:
    • Solid body rotation.
  • Smooth deformational flow (polar vortex).
  • Rotating Shallow-Water Equations: A test case for a time-dependent and compressible velocity field.

These are standard problems used to evaluate the accuracy and robustness of fluid advection schemes.

Key Insights and Ideas

  • Bridging PM and SL: The RPM method successfully combines the mass conservation of particle-mesh methods with the computational advantages of semi-Lagrangian schemes, offering a powerful hybrid approach.
  • Efficient Mass Conservation: By remapping particles to the grid and recalculating masses, the method achieves mass conservation without the computational burden of some other conservative schemes.
  • Accuracy on Sphere: The extension to spherical geometry is crucial for global models, and the RPM method demonstrates comparable or superior accuracy to existing methods in these complex scenarios.
  • Spline Interpolation Advantage: The use of spline interpolation (especially cubic B-splines) for mass distribution on the grid, combined with forward trajectories, leads to improved accuracy and conservation compared to traditional bicubic interpolation with backward trajectories.

Key Results

  • 1D Convergence: The 1D version of the RPM method achieves fourth-order accuracy, matching results from established parabolic spline methods.
  • 2D Planar Advection: The RPM method is shown to be more accurate than the standard SL advection scheme using bicubic interpolation for the slotted-cylinder problem, achieving second-order accuracy with respect to spatial resolution.
  • Spherical Advection: For solid body rotation and deformational flow on the sphere, the RPM method shows comparable accuracy to state-of-the-art SL schemes, with errors being similar to or better than those reported in previous studies.
  • Shallow-Water Equations: When integrated into a semi-implicit semi-Lagrangian scheme for rotating shallow-water equations, the RPM method produces results very similar to traditional SL advection, indicating its compatibility and effectiveness in complex atmospheric dynamics.

Limitations and Future Work

  • Computational Cost: While efficient, the method still involves solving tridiagonal linear systems, which can be a computational bottleneck for very large grids. The paper notes that efficient iterative or direct solvers are available.
  • Polar Cap Issues: For spherical geometries, the method might lead to non-uniform particle mass distribution near polar caps when meridional Courant numbers are high, potentially impacting accuracy. The paper suggests a smoothing operator as a potential solution.
  • Future Work: The authors mention that a further generalization to 3D would be straightforward. They also suggest that the proposed advection scheme can be used to advect momenta, which is particularly attractive for newly proposed semi-Lagrangian Störmer-Verlet (SLSV) schemes. The paper also hints at a forthcoming publication on extending the method to general triangular meshes.

Summary updated

Abstract

Abstract We describe the remapped particle‐mesh advection method, a new mass‐conserving method for solving the density equation which is suitable for combining with semi‐Lagrangian methods for compressible flow applied to numerical weather prediction. In addition to the conservation property, the remapped particle‐mesh method is computationally efficient and at least as accurate as current semi‐Lagrangian methods based on cubic interpolation. We provide results of tests of the method in the plane, results from incorporating the advection method into a semi‐Lagrangian method for the rotating shallow‐water equations in planar geometry, and results from extending the method to the surface of a sphere. Copyright © 2007 Royal Meteorological Society
BibTeX
@inproceedings{cotter2007remapped,
  author = {Colin J. Cotter and J. Frank and Sebastian Reich},
  title = {{The remapped particle‐mesh semi‐Lagrangian advection scheme}},
  booktitle = {Quarterly},
  year = {2007},
  doi = {10.1002/qj.11},
}
Figure