On 20-02-15 15:40, Jim Biard wrote:
> Maarten,
>
> This is life with CF. Just to be sure I understand the details of the
> problem, the vertical set of pressure values at one lat and lon is
> different from the vertical set at a different lat and lon, correct? Is
> there no metric (as in, not just an index) vertical coordinate that you
> can construct that doesn't vary across the lat/lon grid?
Yes, that is the case.
> As far as the pressure bounds being a variable no user will ever need,
> I've got to differ with you. Having an unambiguous, consistent, and
> automation-friendly mechanism for declaring bounds is quite useful.
No: it is the _bounds_ the user is interested in, there just isn't a
meaningful use for the pressures themselves. The ozone content is given
for the sub-columns, not at a particular level.
> It removes uncertainty and avoids the problems that can develop when a
> product is developed by one community of people who say, for example,
> "Everyone knows that the grid boundaries are halfway between the grid
> points" and then gets used by a community of people who say, "Everyone
> knows that the grid boundaries are at the grid points." (I've dealt with
> this sort of problem more than once in my career!) If CF had a standard
> that stated that an extra element at the end of a dimension was
> understood to be an upper bound, then your original proposal would be
> fine. Without a standard, your original proposal is non-standard (by
> definition), and thus open to conflicting interpretations.
>
> I've felt the same frustration about the inefficiencies of coordinate
> grids, but until we define a standard for a more compact representation,
> it's what we've got.
Coordinate grids are very efficient compared to the present problem,
where a 3D dummy variable is needed.
And perhaps CF 2.0 is an opportunity to address this.
Best,
Maarten Sneep
--
KNMI
T: 030 2206747
E: maarten.sneep at knmi.nl
R: A2.14
Received on Fri Feb 20 2015 - 08:25:13 GMT
