twobyfour,

I’m not sure that I completely buy it. Each ring of gaussian grid elements, stepping poleward, would have to be attributed with special rules for insolation, albedo, thermal conductivity, etc. It IS doable, but damned difficult to avoid artifacts and error propagation.

Of course, my perspective is biased. I’m accustomed to multi-layered model spaces where the earth’s curvature is not an important consideration. I can almost accept the possibility of modeling with a gaussian grid as long as it is only one layer thick – treating the atmosphere of the earth as a single surface. Spatial models by necessity rely upon a moving window of neighborhood functions – a grid element is evaluated with its neighbors, the window moves on and evaluates the next element with its neighbors and then feeds back to the previous element iteratively. Pretty straightforward on a 2d surface, the gaussian grid would require special rules for how the window’s algorithms treats each element, but, it is doable. However, when the model has more than one layer, the window must evaluate 3d neighborhoods adding incredibly complex considerations of how to treat the different kinds of boundaries between elements.

How to calibrate all of this complexity?