Development of an Effective Non-Linear Normal-Mode Initialization Method for a High-Resolution Global Model

Abstract

An efficient non-linear normal-mode initialization (NNMI) scheme for a super high-resolution atmospheric general circulation model (AGCM) has been developed. The conventional NNMI method, based on Machenhauer (1977), is difficult to apply to such an AGCM, with a grid spacing of 20 km, since it demands an extremely large computational memory. The new scheme combines two techniques; one is the venical-mode initialization method proposed by Bourke and McGregor (1983), and the other is the incremental NNMI method proposed by Ballish et al. (1992). Comparisons with the conventional Machenhauer scheme, using a 60 km-resolution AGCM, show that the new scheme favorably suppresses undesirable gravity-wave oscillations, while maintaining meaningful tidal waves in the initial states. A medium-range forecast experiment with data assimilation shows that the new scheme is also competitive with the conventional scheme in terms of forecast skill. It is demonstrated that the new scheme is efficient and effective for a 20 km-resolution AGCM.