Some numerical aspects of developing an operational primitive-equation model for the tropics.

Abstract

A barotropic primitive equation model on a global tropical grid using operational real data is used to test various boundary conditions and methods of initialization including numerical variational analysis. Comparisons of forecast accuracy are made between staggered and non-staggered grids. All prediction models produce better verification statistics than persistence, with the staggered grid verifying better than the full grid. This appears to be a result of two-gridlength noise being of greater magnitude in the full grid model. Less altering of individual synoptic systems occurs in the variational analysis initialization compared to two forms of the nonlinear balance equation. As a result, nondivergent variational analysis input fields to the prediction models result in the best verification values.