Scale-Selective Digital-Filtering Initialization

Abstract

Digital-filtering initialization (DFI) of atmospheric models relies on the fact that the gravity?inertia waves have higher frequencies than the meteorologically relevant rotational modes and assumes that a frequency exists that separates them. This note shows that a Doppler effect of fast-propagating storms may ?shift? the frequencies of the small-scale rotational modes into the frequency categories that are deemed to be the ones of the gravity?inertia waves. A forecast is presented in which the impact of this on DFI manifests itself to a substantial extent (i.e., a reduction the depth of the eye of the storm by about 6?7 hPa). As a cure it is proposed to make the filtering scale selective (i.e., filtering the large spatial scales more than the small ones). It is shown that, not only does this leave the storm almost intact, but it also leads to a more balanced initial state. The implementation of such a filter is straightforward in a spectral limited-area model.