Ferdinand Baer, University of Maryland
Joseph J. Tribbia, National Center for Atmospheric Research
Mark Taylor, Los Alamos National Laboratory

Research Objectives

Our goal is to use the parallel efficiency of SEAM (Spectral Element Atmospheric Model) to perform simulations of three-dimensional decaying stratified turbulence on a rotating sphere. Decaying geophysical turbulence simulations have so far been carried out in spherical geometry only with two-dimensional models. In three dimensions, it has previously only been possible to carry out these types of experiments with filtered equations in periodic Cartesian domains, such as the well known quasi-geostrophic experiment carried out by McWilliams, Weiss, and Yavneh. With the SEAM model on 128 nodes of the T3E, we will be able to perform this simulation of rotating, stratified turbulence for the first time using the full primitive equations at very high resolution and in full spherical geometry.

Computational Approach

	The potential vorticity on Jupiter after integrating the dynamical equations on that planet using the SEAM model for 276 days. Two different truncation experiments are shown. Note how much more detail can be seen in the jet streaks using T1033, one of the highest-resolution experiments yet attempted.

SEAM is a spectral element atmospheric global circulation model which is ideal for MPPs such as the T3E. SEAM achieves almost perfect parallel scalability up to 256 processors. SEAM has proven to be spectrally accurate, producing results of comparable accuracy to the more conventional spherical harmonic based climate models.

Accomplishments

In the past year we made many preliminary 3D decaying turbulence runs and discovered that the conventional initial conditions used for these types of runs (with simplified equations) are ill-posed for the full 3D primitive equations used in atmospheric modeling. We have addressed this issue by developing a 3D nonlinear balance procedure which creates initial data with random vorticity but with physically correct correlations between the horizontal and vertical scales. The nonlinear balance equations need to be solved only once for each resolution, and this is done efficiently with several spherical harmonic expansions. Despite this difficulty, intriguing results were obtained regarding the columnization of vortices in the primitive equations. This was heretofore only observed in quasigeostrophic systems. Similarly, interesting coherent vortex formation was noted in simulations with planetary rotation and radius similar to Jovian values. These had previously been noted in shallow water computations.

Significance

With the strong emphasis on global modeling of the climate system, substantial interest is developing on the evolution of regional climate. This requires an interface among many scales in a model and an efficient way of doing this. SEAM is ideal for this purpose because it not only allows for small-scale prediction at arbitrary locales over the globe-those regions for which the modeler has a particular interest relating to a specific climate or event-but also is optimized to do such integrations on an MPP system. Thus SEAM should be a desirable alternative to other climate models for predicting both regional and global climate events simultaneously and in a highly efficient, real-time environment.

Publications

M. Taylor, R. Loft and J. Tribbia, "Performance of a spectral element atmospheric model (SEAM) on the HP Exemplar SPP2000," Parallel Computing (submitted).

M. Taylor, J. Tribbia, and M. Iskandarani, "The spectral ele-ment method for the shallow water equations on the sphere," J. Comput. Phys. 130, 92 (1997).

D. Haidvogel, E. Curchitser, M. Iskandarani, R. Hughes, and M. Taylor, "Global modeling of the ocean and atmosphere using the spectral element method," Atmosphere-Ocean 35, 505 (1997).