A Global Multilevel Atmospheric Model Using a Vector Semi-Lagrangian Finite-Difference Scheme. Part I: Adiabatic Formulation

J. R. Bates National Aeronautics and Space Administration/Goddard Laboratory for Atmospheres, Greenbelt, Maryland

Search for other papers by J. R. Bates in
Current site
Google Scholar
PubMed
Close
,
S. Moorthi National Aeronautics and Space Administration/Goddard Laboratory for Atmospheres, Greenbelt, Maryland

Search for other papers by S. Moorthi in
Current site
Google Scholar
PubMed
Close
, and
R. W. Higgins National Aeronautics and Space Administration/Goddard Laboratory for Atmospheres, Greenbelt, Maryland

Search for other papers by R. W. Higgins in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

An adiabatic global multilevel primitive equation model using a two time-level, semi-Lagrangian semi-implicit finite-difference integration scheme is presented. A Lorenz grid is used for the vertical discretization and a C grid for the horizontal discretization. The momentum equation is discretized in victor form, thus avoiding problems near the poles. The 3D model equations are reduced by a linear transformation to a set of 2D elliptic equations, whose solution is found by means of an efficient direct solver.

The model (with minimal physics) is integrated for 10 days starting from an initialized state derived from real data. A resolution of 16 levels in the vertical is used, with various horizontal resolution. The model is found to be stable and efficient, and to give realistic output fields. Integrations with time steps of 10 min, 30 min, and 1 h are compared, and the differences are found to be acceptable.

Abstract

An adiabatic global multilevel primitive equation model using a two time-level, semi-Lagrangian semi-implicit finite-difference integration scheme is presented. A Lorenz grid is used for the vertical discretization and a C grid for the horizontal discretization. The momentum equation is discretized in victor form, thus avoiding problems near the poles. The 3D model equations are reduced by a linear transformation to a set of 2D elliptic equations, whose solution is found by means of an efficient direct solver.

The model (with minimal physics) is integrated for 10 days starting from an initialized state derived from real data. A resolution of 16 levels in the vertical is used, with various horizontal resolution. The model is found to be stable and efficient, and to give realistic output fields. Integrations with time steps of 10 min, 30 min, and 1 h are compared, and the differences are found to be acceptable.

Save