Abstract
A finite element formulation for the vertical discretization of a global spectral model is presented. Results obtained from a linearized version of the model are compared with both exact analytical solutions and those of a vertically staggered finite-difference scheme. A series of seven-day global integrations using the fully nonlinear model and simple physics is presented and compared with the corresponding series obtained using a vertically staggered finite-difference model. The finite-element version of the model seems to give better performance, particularly at medium range. The new formulation tested here is also shown to be free of a noise problem present in an older version of the model.
