Global Atmospheric Solvers for Next-Generation Weather and Climate Models

global cores

Workshop held on 23-24 September 2008 in Boulder Colorado

Report from the workshop

Global Atmospheric Dynamical Core Assessment

28 October 2008

This is an assessment of the state-of-the-art in global solvers for atmospheric fluid flow by the participants of the Global Atmospheric Core Workshop held in Boulder on 23-24 September 2008. The workshop was hosted by the National Center for Atmospheric Research and formally was a meeting of the Atmospheric Dynamical Core Working Group that was formed as part of the effort to develop a new combined climate and weather modeling system.

Summary

We are evaluating state-of-the-art approaches for dynamical cores in a new combined weather and climate modeling system. A consensus exists for the necessary attributes of this solver - it must be nonhydrostatic, global, solve the deep atmosphere equations, conserve mass and scalar mass, have local refinement and limited-area capabilities, be capable of relatively uniform resolution over the globe, have positive-definite and monotonic transport options, and must be efficient on planned petascale MPP machines. Exact energy conservation, while potentially desirable, is not viewed as a strict requirement. Existing operational models use latitude-longitude grids. They have some problems with the poles and critically will not scale well to MPP architectures. The development efforts that are moving forward most quickly are converging on two classes of grids for the sphere - cubed-sphere grids and icosahedral grids using finite-volume and finite difference solvers. Efforts have not completely converged - there are problems with all of the approaches but solutions are actively being explored. There are also development efforts centered for hydrostatic solvers based on Galerkin techniques using cubed-sphere grids. Viable nonhydrostatic solvers using these methods are far off. Test cases for nonhydrostatic (and hydrostatic) solvers need further development. There is a relative dearth of discriminating test results for models and, equally important, there are almost no computational costs reported for these potential cores that would allow for a determination of solver efficiency that is necessary for evaluating candidate architectures for applications.

Complete Global Core Assessment Report [pdf available]


Agenda from the 23-24 September Global Atmospheric Core Workshop and links to the presentations.

Tuesday 23 September

9:00-9:20 am: Welcome - Guy Brasseur (ESSL Director), Introductions.

9:20-9:30: Introduction and Overview, Bill S. [pdf available]

9:30-10:45: Core Requirements. Bill S. (Priority applications, core requirements).

John Thuburn (conservation). [pdf available]

11:00-12:00 Methods 1 - Implicit solvers/techniques on MPP machines - Ross Heikes [pdf available]

1:00-2:00: Methods 2 - Galerkin methods - Mark Taylor [pdf available]

2:00-3:00: Methods 3 - Finite Difference/Volume methods. George Bryan [pdf available]

(energy conservation in nonhydrostatic solvers).

3:15-4:15: Discretizing the Sphere 1 - Icosahedral grids (triangles). Bob Walko [pdf available]

4:15-5:15: Discretizing the Sphere 2 - Icosahedral grids (hexagons). Jin Lee [pdf available]

5:15 - Closing Comments. Bill S.

Wednesday 24 September

9:00-9:30 am: High order finite volume methods. Phil Collela. [pdf available]

9:30-10:30: Discretizing the Sphere 3 - Cubed-sphere grids, FV approaches. Bill Putnam. [pdf available]

10:45-11:30: Discretizing the Sphere 3a - Cubed-sphere grids, Galerkin approaches. Ram Nair. [pdf available]

11:30-12:00: Discretizing the Sphere 4 - Yin-Yang grids. Bill S. [pdf available]

1:00-2:00: Vertical coordinates, Dave Randall. [pdf available]

2:00-3:00: Test suites. Christiane Jablonowski. [pdf available]

3:15-4:00: Summary and recap.

4:00: What's next?

End of Workshop.