[cf-pointobsconvention] Draft 2

[John Caron]

John Graybeal wrote:
> I think 'requiring a Z' and 'requiring a Z with every point' are 
> two different things.  I don't think anyone meant to argue for the 
> latter, or they'll correct me if I'm wrong.

> 
> To take an easy case first, for surface-based measurements like ship 
> point observations, specifying 'surface' in one place for all Z 
> coordinates (e.g., via an attribute) seems like a reasonable approach. 

You can use a scalar, "nominal" coordinate variable, like:

variables;
 char z(z_strlen);

 float lon(obs);
 float lat(obs);
 double time(obs);

 float dataVar(obs);
  dataVar:coordinates = “lon lat z time”;

data:
  z = "tropopause";


> Hurricane track (great example) may be a little more complex -- I don't 
> know if the science would say that is a 'surface' value, or something less 
> concrete.  But it would be OK with me to allow the attribute to take on a 
> value representing an abstraction, if that's necessary.
> 
> John
> 
> At 11:31 AM -0600 9/18/07, Don Murray wrote:
>> John Caron wrote:
>>>> John Graybeal wrote:
>>>> Re the Z axis, it would be nice to know for sure whether the Z axis
>>>> should be interpreted as 'surface'.  (Other possible surfaces
>>>> exist, and I suppose an abstract form of Z is also possible.  So if
>>>> we want determinism, we could either say "lack of characterization
>>>> always means surface" or "Z must be described, if only to say it is
>>>> 'surface'.")
>>> Yes, after thinking more about it, Im not so sure we should allow
>>> data that doesnt have a z coordinate. A standard I think should be
>>> allowed to insist on things like this.
>> Then what do you do with a hurricane track that does not have
>> an z associated with it?  Or a trajectory from ship point obs
>> that only have a latitude/longitude?  It seems like wasted
>> space to have to write out 0 for z for every point if you don't
>> have it and there are plenty of datasets out there without a
>> Z value.
>