Dynamic Normal Mode Initialization

Abstract

A new dynamic initialization scheme for a primitive equation model is proposed. In the proposed scheme, the balanced initial state is obtained by forward-backward integration of the linear part of the model equations. During the integration the nonlinear part is kept constant. After the integration, the nonlinear terms (including physical processes) are updated. The updating of the nonlinear terms'and the linear integration are repeated several times like the iteration procedure of nonlinear normal mode initialization (NMI). The scheme is proved to be equivalent to NMI with a frequency dependent underrelaxation factor, and therefore, it is called dynamic normal mode initialization (DNI). Since only the linear part is integrated during the dynamic integration, unlike the conventional dynamic initialization, the proposed scheme is efficient. The scheme has been successfully applied to the FSU multilevel global spectral model. The scheme is expected to be applicable to limited area models without difficulty.