[CF-metadata] Rotated-pole grids

From: Seth McGinnis <mcginnis>
Date: Thu, 18 Jun 2009 12:45:19 -0600

One anecdotal point about lat-lon grids and GIS: ArcGIS 9.3 now can read NetCDF
data natively -- *IF* the NetCDF file is fully CF-compliant and the projection
is properly specified. That can be harder to do than you might think; there
are many fiddly little details that are irrelevant from a modeling perspective
but matter greatly for importing to GIS, and the implementation itself is
imperfect (parameters specified as floats instead of doubles don't get read in
properly, not every projection scheme has been implemented yet, etc.).

There's no argument that we want the map projection to be specified properly,
but providing the lat-lons *in addition* is valuable because it provides a
fallback in case of error: if, for some reason, a GIS client can't understand
the projection information to import the data as raster, it's still possible to
import it as a set of point features. It may be clunky to work with that way,
but it's a lot better than nothing.

Likewise, there are a lot of homebrew systems out there that do GIS-like things
without having a full suite of sophisticated projection tools behind them.
 Allowing the users of those systems to figure out where the grid-cells are
located without needing to research and implement the relevant coordinate
transformations by hand is a big win for usability.

Cheers,

--Seth McGinnis
Associate Scientist
NARCCAP Data & User Community Manager
ISSE / NCAR


> Just a word about GIS clients, picking up on one of Steve's
> comments - many GIS clients would prefer to have the
> coordinate system properly specified (as a CRS code or
> Well-Known Text) rather than by listing the lat-lons
> exhaustively. (Another conversation could be started to
> discuss the value of including these as optional attributes,
> since most of the CF projections are commonly used in GIS.)
Received on Thu Jun 18 2009 - 12:45:19 BST