Abstract
Results are presented from the application of the semi-Lagrangian method to a barotropic model with high-order differencing on a nonstaggered grid. The nonstaggered grid, or A-grid in the standard nomenclature, is not widely used in numerical weather prediction models mainly because of the relatively poor accuracy of second-order finite-difference schemes formulated on the A-grid. However, a nonstaggered grid can accommodate high-order differencing with comparative ease and with essentially no reduction in time step size with the semi-implicit, semi-Lagrangian formulation. When the resulting gains in accuracy over the range of scales resolved well by the grid are taken into account, the nonstaggered grid now emerges as an attractive framework for numerical weather prediction.
The formulation of a semi-implicit, semi-Lagrngian barotropic model is described. Extensive tests on a large number of cases with staggered and nonstaggered versions of the model confirm the superiority of the high-order methods over both a second-order, staggered version of the model, and low-order versions on the A-grid at the same grid resolution.
