A Controlled Comparison of the Spectral Element and Discontinuous Galerkin Methods for Two-dimensional Linear Advection
Franklin Okorodudu
Department of Computer Science, Faculty of Computing, Delta State University, Site III, Abraka, Nigeria.
Nkonyeasua Cleon Nkonye *
Department of Computer Science, Faculty of Computing, Delta State University, Site III, Abraka, Nigeria.
*Author to whom correspondence should be addressed.
Abstract
This study compares reported single-element implementations of the spectral element method (SEM) and the discontinuous Galerkin (DG) method for a controlled two-dimensional linear temperature-advection problem with a prescribed velocity field and an analytical solution. Both implementations employ Gauss–Lobatto–Legendre nodes and fourth-order Runge–Kutta time integration. Their numerical accuracy is assessed using the reported maximum absolute and discrete L2 errors for polynomial orders N = 2, 4, 6 and 8. The tabulated errors decrease with increasing polynomial order for both implementations. At N = 8, the SEM maximum absolute error is 1.007243 × 10⁻⁵, whereas the corresponding DG error is 1.224984 × 10⁻⁷. The reported L2 errors at the same order are 2.382468 × 10⁻⁶ for SEM and 3.764281 × 10⁻⁸ for DG. Although the DG results are smaller at all investigated orders, the calculations were performed on a single computational element, so internal numerical fluxes between neighbouring DG elements were not evaluated. Moreover, the available computational record does not fully specify the boundary treatment, numerical operators or error-norm conventions. Temporal-refinement results and reproducible execution-time measurements are also absent. Accordingly, the findings establish only the behaviour of the documented single-element calculations under polynomial-order refinement. They do not support a general ranking of SEM and DG for multi-element advection problems, computational efficiency or complete climate-modelling applications.
Keywords: Linear advection, spectral element method, discontinuous Galerkin method, GLL nodes, error norms, polynomial-order refinement