Regional modeling: A theoretical discussion

Abstract

The goal of regional modeling is to make a detailed forecast for a given limited area of interest by focusing resolution over it and the immediate vicinity. As a consequence, the period of validity is necessarily more restricted than would otherwise be the case, and this is the price that must be paid for locally-enhanced resolution. The principal attributes of the non-interactive and interactive strategies for regional modeling are described. For the non-interactive strategy, particular emphasis is placed on the importance, difficulty, and impact, of well-posedness for open-domain problems. A methodology is given for estimating the size of numerical buffer zones required to obtain a forecast uncontaminated by the inward propagation of inaccurately-specified lateral boundary conditions. The interactive strategy addresses the well-posedness issue of (non-interactive) limited-area models. A computational overhead is incurred but this can be reduced through the use of variable resolution. It is argued that regardless of the preferred regional modeling strategy, experiments should be undertaken to today's regional models under carefully-controlled conditions, to reflect the significant reduction over the past two decades of other sources of error. © Springer-Verlag 1997.