[CF-metadata] Vertical Coordinate

From: Jonathan Gregory <jonathan.gregory>
Date: Tue, 07 Jan 2003 10:06:19 +0000

Dear Burkhardt

Is there any way to compute the P of the model levels from vertical coordinates
which are a function of lev only (like a and b) and some other fields which
are a function of (lon,lat,lev)? If so, we could define a new kind of vertical
coordinate, to add to those already in Appendix C.

If not, I agree that you *could* regard P(lon,lat,lev) as a kind of coordinate
and in fact we have had previous discussions about supporting this concept. But
in the end we decided that this is going too far away from pure metadata, into
the domain of scientific things that you know about the data. You know that if
T and P are functions of (lon,lat,lev), then you can associate T with P, but
mapping T onto P in this way is really a kind of scientific transformation of
the data, I think. To take another example, I suppose that specific humidity
q is also q(lon,lat,lev) in the model, so in principle you could regard q as
a coordinate for T. Formally there is no difference between q and P.

Does that help?

Best wishes

Jonathan


> Dear member of CF community,
> we are using the non-hydrostatic weather forecast model of the German
> Weather Service. The vertical coordinate used in hydrostatic models is
> often defined as described in the CF-conventions (e.g. dimensionless
> sigma or hybrid pressure-sigma coordinates). With given height
> dependent coefficients (e.g. "a" and "b" ) and surface pressure field
> the pressure at the vertical levels of the model can be computed. For
> the non-hydrostatic model we are using its a bit more complicated
> since the pressure at the model levels is P=Pref+Pdev, where Pref is
> the pressure of a reference atmosphere and Pdev is the deviation from
> the reference profile.
> My Questions:
> Is there a CF conformal way to define a three-dimensional variable
> (in my case P(lon,lat,lev)) as vertical axis?
> I guess simply giving P the attribute "axis=Z" would be not
> sufficient?
> Should a three dimensional variable like air temperature
> (T(lon,lat,lev)) have the attribute "coordinates="P" in this case?
> Thank you, I am looking forward to your answer.
> Burkhardt
> -----------
> Dr. Burkhardt Rockel
> GKSS Forschungszentrum
> Max-Planck-Strasse
> D-21502 Geesthacht
> Germany
> Phone: +49 4152 87 2008
> Fax: +49 4152 87 2020
> Email: rockel at gkss.de
> ----------
Received on Tue Jan 07 2003 - 03:06:19 GMT