The Effects of the Asselin Time Filter on Numerical Solutions to the Linearized Shallow-Water Wave Equations
Year: 1983
Type: article
Abstract: A one-dimensional linear analysis of the shallow-water wave equations without rotation is used to evaluate the effects of a time filter developed by Asselin (1972) on both stability and phase error of the leapfrog advection scheme and each of three diffusion schemes: Dufort-Frankel, “lagged” and Crank-Nicholson. The analysis considers propagation and diffusion both separately and in combination over all resolvable wavelengths on a uniform discrete grid. The main objective is to establish a suitable range of the filtering intensity parameter ν for suppressing the computational wave mode without materially altering the physical wave mode. Without propagation, the “lagged” scheme has the most accurately damped physical mode. Serious overdamping at intermediate wavelengths renders the lagged scheme inferior to the other two when propagation is present, and Dufort-Frankel is preferable to Crank-Nicholson because of much smaller computing demands. With or without filtering, the space truncation error of all three diffusion schemes retards the physical damping rate much more strongly than even a heavy time filtering weight, especially for short waves. With propagating velocities defined to include both mean wind and wave propagation, the main findings are as follows: 1) For diffusion alone, heavy filtering (ν ≈ 1.0) is desirable. When propagation is present (with or without diffusion), the optimal filtering intensity is smaller, with ν ≈ 0.5 for “long” time steps approaching stability limits. 2) In the presence of propagation, filtering has little effect on the dispersive phase lag of the physical mode except to aggravate it, appreciably, for long time steps. Filtering reduces the large positive phase error due to the computational mode. 3) Heavy filtering degrades results for long time steps with propagation and diffusion combined. The physical mode is damped excessively at intermediate wavelengths and is dominated by the computational mode, with a phase error causing waves to accelerate spuriously. From these results, the optimal filtering parameter ν is 0.5–0.6 with diffusion and propagation both present.
Source: Monthly Weather Review
Language: English
Subfield: Computational Mechanics
Field: Engineering
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W2005602558
