Climate Simulations with a Spectral, semi-Lagrangian Model with Linear Grids

Open Access
Publisher Website
Google Scholar
Abstract
Semi-Lagrangian, semi-implicit, spectral transform atmospheric models do not require conventional quadratic unaliased Gaussian grids because advection is not expressed as an Eulerian quadratic product in their formulation. The conventional T42 quadratically unaliased grid with 128 (64) points in longitude (latitude) supports a T63 spectral truncation, and similarly for other resolutions. We refer to these types of truncation/grid combinations as aliased or linear grids, since they can be considered as being aliased for quadratic products or unaliased for linear terms. In this paper, climate simulations based on linear grids are compared with ones based on quadratic unaliased grids. It is shown that for spectral resolutions above T42, the spectral truncation, rather than the transform grid on which the physical parametrizations are calculated, dominates the accuracy of simulations. Thus the linear grid provides a 50 percent increase in resolution of both horizontal dimensions at a negligible additional cost.

This publication has 14 references indexed in Scilit: