On Wednesday 06 April 2005 20:19, Jonathan Gregory wrote:
> The data provider probably doesn't care very much. This information is not
> seen as critical for interpretation or even formulation of the model. Hence
> you can probably legitimately serve it using any reasonable assumption.
I would go even further than this. While it is possible that the coordinate
transformation error becomes more significant at very high resolution global
models and mesoscale model scales, the actual significance of the model
results at those scales is poor (typically, for a grid point model, one
should be looking at an effective resolution about four times the grid
resolution). What that means is the error in the grid projection
transformation (if there is one) is vastly smaller than the effective grid
registration (ie. how accurately the model coordinates reflect the real
world). This is probably true even in a data assimilation model, where the
model knows a bit more about the real world ...
Nowcasting models at the very finest scales may be a different story, but in
that case the grid formulation should probably be rather different than a
global spherical formulation ... ("should" I say, don't know that they
are :-)
I don't think that CF can cope with everything ...
Cheers
Bryan
--
Bryan Lawrence
Head, NCAS/British Atmospheric Data Centre
Director, CCLRC/Environmental Data Archival and Associated Research
badc.nerc.ac.uk, home.badc.rl.ac.uk/lawrence, +44 1235 445012
(CCLRC, Rutherford Appleton Laboratory, Chilton, Didcot, OX11 0QX)
Received on Thu Apr 07 2005 - 01:20:49 BST
