The Application of a Multi-Level Spectral Model to Data Assimilation

Abstract

The techniques developed and tested by Simmonds (1976) for inserting grid-point data into a barotropic spectral model have been extended to the multi-level case. Theoretical and numerical analyses suggest that such a model can provide a good data assimilation vehicle if inertia gravity waves are adequately suppressed. Efficient mechanisms in the present model involve the application of divergence diffusion, a time filter and a modest amount of vorticity diffusion. Two high-resolution versions of this model were constructed, the first containing all the physical processes usually included in general circulation models, the second being the same but for the omission of moist and radiative processes. Data assimilations were undertaken with these two models to determine the necessity of including the physical processes. In the upper levels and at the surface the latter model was found to produce better analyses in the one 6-day assimilation carried out. Forecasts initialized from these analyses supported the view that the assimilation without physics produced a more accurate representation of the atmosphere. The experiments show that, at the very least, the tested spectral model is an adequate data assimilator.