Sensitivity of a spectrally filtered and nudged limited-area model to outer model options

Abstract

Numerical filters required to control spatial computational modes in a limited-area model (LAM) that uses the unstaggered A grid are developed and tested over the complex topography of the Great Basin of the western United States. The filters are founded upon Fourier expansions of forecast deviation fields and function equally effectively for both periodic and aperiodic local structures. Unlike other spatial filters, the approach used here avoids any direct contamination of larger scales. Provided that the shortest resolved wavelength of two grid intervals is removed, the results do not depend strongly on the range of filtered short waves or on the type and order of horizontal space difference approximations. This approach leads naturally to methods in which the large scales predicted by an ambient outer model can be directly incorporated within the complete domain of the inner LAM, rather than just through conditions applied at the lateral boundaries of the LAM. This technique has some similarities to methods used both in operational regional models in Japan and in recent regional research models at the National Centers for Environmental Prediction (formerly National Meteorological Center) of the United States. Several methods to incorporate the large scales into the LAM are evaluated in a winter storm case study and in an ensemble of seven forecasts.