The National Center for Atmospheric Research | UCAR | UOP
photo Home Our Organization Our Research News Center Education Libraries Community Tools

Quarterly logo

Modeling roundup Beyond latitude and longitude

Peter Laurizen

Peter Lauritzen was the principal organizer of a 2008 NCAR Advanced Study Program colloquium on numerical techniques for global atmospheric models, including alternative gridding schemes like those illustrated below. (Photo by Carlye Calvin.)

Scientists at NCAR, DOE, and universities are joining forces to explore how the standard 3-D ­latticework of model grid points might be revamped to better serve future global models. The usual grid of rectangles oriented by latitude and longitude works less well as resolutions tighten, mainly because of the narrowing of longitude lines near the poles. As NCAR’s Joseph Tribbia puts it, “Latitude and longitude will not scale.” Two of the possibilities being explored:

  • Cubed sphere (bottom left). This design largely equalizes the size of grid boxes by centrally projecting the faces of an inscribed cube onto the spherical surface, as if each side of a cube had been stretched outward from its center to form a sphere. Mark Taylor (Sandia National Laboratories) and NCAR colleagues have tested a version of the Community Atmospheric Model on an ocean-only “aquaplanet,” using 86,200 processors at ¼ degree resolution, and found that it scales well. They’re also exploring how variables can be remapped between latitude-longitude and cubed-sphere grids.

“The exchange of information between spherical grids is nontrivial if higher-order accuracy is desired,” says NCAR’s Peter Lauritzen. Graduate student Paul Ullrich worked with Lauritzen last summer to develop a new remapping algorithm through NCAR’s Summer Internships in Parallel Computational Science program.

  • Hexagonal grid (bottom right). Much like a soccer ball, this grid covers the globe with hexagonal boxes. This eliminates the poles entirely and eases the boundary problems that often emerge when a regional model is nested inside a global one. Experiments carried out by NCAR’s William Skamarock and Joseph Klemp with Todd Ringler (Los Alamos National Laboratory) and John Thuburn (University of Exeter) show that the “hex grid” can support a dynamical core as accurate as other state-of-the-art systems, and it could be more efficient as well.

alternate gridding systems: the cubed sphere (left) and the hexagonal grid (right).

Alternative gridding schemes: the cubed sphere (left) and hexagonal grid (right). (Illustrations by Peter Lauritzen)

Also in this issue...


Untitled Document