In the early days of computing, geophysical fluid dynamics (GFD), predominantly numerical weather prediction (NWP), was a dominant factor in the design of computer architecture and algorithms. This early work focused initially on finite difference algorithms on rectangular computational grids and later on spectral methods. After the initial work of von Neumann, Charney, and Arakawa however, the focus shifted from the basic algorithms to improvements in the physical models. Further work on fundamental numerical algorithms shifted to other disciplines—predominately the then emerging aerospace community. As a result, for 40 years, the GFD community has been using numerical techniques that are virtually unchanged. The two primary numerical methodologies that have been used for modeling the atmosphere and the ocean have been spectral methods for global modeling and structured rectilinear grids for regional modeling. In 1992, work began on the Operational Multiscale Environment model with Grid Adaptivity (OMEGA), a new atmospheric simulation tool based upon an adaptive unstructured grid. Over the past 14 years, this model has found application to atmospheric dispersion, point weather forecasting, and, the topic of this paper, hurricane track forecasting. The unstructured grid paradigm employed in OMEGA has the advantage of flexibility in providing high resolution where required by either static physical properties (terrain elevation, coastlines, land use) or the changing dynamical situation. This paper provides a description of this new paradigm and presents its use in atmospheric simulation. Copyright Springer Science+Business Media, Inc. 2007
