Hi Jon,
I agree with your reading of the spec (that there is currently no way to
relate a vertical coordinate variable with a vertical datum) and Rich's
suggestion that we need a new ticket for vertical datum.
It seems there are (at least?) two ways to capture the vertical datum
information:
1) Extend the existing grid mapping/projection/CRS framework to include
a vertical datum id/name.
2) Extend the existing vertical coordinate set of attributes to include
a vertical datum id/name.
I prefer the second option because it keeps all the vertical information
in one place and it also more easily allows two or more variables to be
on the same horizontal CRS but different vertical CRS.
However, the first option seems appealing in that the entire CRS would
be in one location and if the vertical datum is an ellipsoid the
existing attributes can be used (as Seth suggested).
[Has anyone thought about how time might fit into all this CRS
discussion? I haven't looked at ISO 19108 in awhile but it seems
relevant if we want to consider temporal CRS. Should we consider that at
the same time as vertical? Or should we keep that separate in the
interest of not slowing down vertical CRS?]
Ethan
Jon Blower wrote:
> Hi all,
>
> Looking through the CF conventions document (1.3) there seems to be no
> way to distinguish some of the possible vertical coordinate systems.
> CF distinguishes dimensioned vertical coordinates (e.g. height or
> depth in metres) from dimensionless coordinates (e.g. sigma) but does
> not seem to distinguish between the different possible heights/depths.
> Here are a few possibilities that I think can't be distinguished in
> CF:
>
> - Depth below the geoid (also, which geoid?)
> - Depth below an ellipsoidal approximation of the earth (e.g. WGS84)
> - Depth below instantaneous sea surface (e.g. where depth is inferred
> from pressure)
> - Depth below a spherical approximation of the earth (used in many
> models, I believe)
>
> (the same goes for heights of course) Am I right or have I
> misunderstood something? To provide a case study, a colleague of mine
> is combining ocean model results with other GIS data sources in order
> to drive a high-resolution (50m in the horizontal) model of an
> estuary. The ocean model data files (which I think are CF-compliant)
> don't give a vertical datum, and this leads to an uncertainty in
> *horizontal* positioning of around 100m, which is two whole grid
> cells!
>
> This problem is most marked, of course, in very high-resolution
> studies - it's not likely to be too important for larger-scale work,
> although it does seem important to be able to describe vertical
> positioning as accurately as we can describe horizontal and temporal
> positioning.
>
> Any thoughts?
> Jon
>
>
--
Ethan R. Davis Telephone: (303) 497-8155
Software Engineer Fax: (303) 497-8690
UCAR Unidata Program Center E-mail: edavis at ucar.edu
P.O. Box 3000
Boulder, CO 80307-3000 http://www.unidata.ucar.edu/
---------------------------------------------------------------------------
Received on Wed Aug 06 2008 - 23:31:08 BST
