CMG RESEARCH: An Adaptive Mesh, Spectral Element Formulation of the Well-Posed Primitive Equations for Climate and Weather Modeling

University Corporation For Atmospheric Res, Boulder CO

Investigators

Abstract

DMS Award Abstract Award #: 0222282 PI: Thomas, Stephen Institution: University Corporation for Atmospheric Research Program: Collaborations in Mathematics and the Geosciences (CMG) Program Manager: Catherine Mavriplis Title: CMG RESEARCH: An Adaptive Mesh, Spectral Element Formulation of the Well-Posed Primitive Equations for Climate and Weather Climate simulation is a grand challenge problem requiring multiple, century-long integrations of the equations governing the Earth's atmosphere. Recently, it has been recognized that localized flow structures may play an important role in obtaining the correct climate signal. Higher-resolution climate simulations may be required in the near future at the National Center for Atmospheric Research (NCAR) and other climate modeling centers in convergence studies and to assess model uncertainty. Rather than increase resolution uniformly over the entire sphere, recent developments in adaptive mesh refinement techniques may be applicable to atmospheric general circulation models. High-order polynomial based spectral element and discontinous Galerkin methods offer the geometric flexibility required to implement adaptive methods and also exhibit the exponential convergence of the traditional spectral transform approach. The spectral element dynamical core is ideally suited to high-performance parallel computers based on microprocessors and a team at NCAR was awarded second place in the 2001 IEEE/ACM Gordon Bell competition for achieving 370 Gigaflops, representing a sustained climate simulation rate of over 100 years per day. This work represents a major advance in geophysical fluid flow simulations. By implementing an adaptive non-conforming spectral element dynamical core for atmospheric general circulation models, we propose to study how small scale flow features develop and feedback to larger scales. We also plan to study simple idealized physical forcings in this context and explore, with NCAR and university researchers, how adaptive multi-scale physical parameterizations can be implemented. Weather and climate simulation is an extremely complex problem that requires many long time computer modeling runs for the Earth's atmosphere. Recently, with the advances in computer power and with advances in the understanding of the dynamics, it has been recognized that local flow structures can play an important role in determining specific weather and climate conditions. Computer models should thus advance to fully resolve these structures in an efficient way. This project is concerned with the development of new, efficient computer models that will automatically resolve local features while providing overall accuracy. The impact on the next generation of weather and climate models should be significant. Date: June 28, 2002