The general concept used to consider optimization of the computational system of atmospheric prediction equations is to find, if possible, eigensolutions of the most so-phisticated linearized form of the nonlinear equations. If these solutions henceforth to be denoted as 'structures' can be used in a spectral expansion, then the numerical inte-grations will be carried out using the best representation available. If a finite grid of points is needed, then a grid is sought on which the structures to the linear system are also eigensolutions. Since these structures represent variability in the space domain, they are scale dependent. It is not always practical or even possible to use structures for the most complex system, thus we search for structures which satisfy at least some linear form of the equations. The structures can also be selected from observations pro-vided they have significant amplitude in the largest scales or they might be selected from EOF analysis of data.
展开▼