[CF-metadata] fixed sensors, depth, datum

From: Jonathan Gregory <j.m.gregory>
Date: Fri, 12 Sep 2008 08:44:58 +0100

Dear Dale et al.

You're quite right, altitude is height above geoid. Hence the names for the
tidal datums should be altitude_of_DATUM, which is neater. We should spell
them out, I think e.g. altitude_of_mean_low_water, as we generally avoid
abbreviations in standard names for the sake of intelligibility by people who
aren't experts in the relevant area.

The way you propose to use attribute ancillary_variables isn't quite what it
was intended for, I think. Do you envisage there could be several data
variables in the file giving alternative values for a particular offset? If
so, I can appreciate there may be a need for a link. But if there is only one,
is a link required? You could find the variable you want by examining the
standard names.

While I think that it's better not to have different standard names for
altitude wrt different geoids as I would consider them to be varying
definitions of the same geophysical quantity, I think that height above geoid
(altitude) and height above reference ellipsoid are different geophysical
quantities and do need different standard names. Hence I would not be in
favour of a generic "height above vertical datum". I don't think that is
sufficiently well-defined to have a standard name. I also think that "height
above reference ellipsoid" is a different quantity from "height above mean low
water" etc. The fact the Dale wants to distinguish the various tidal datums
shows that they define distinct quantities.

Ethan argues for putting the definition of vertical coordinate system with
vertical coordinate variables, not in the grid mapping, which is so far for
horizontal coordinate systems. I would note that

* we have already defined the ellipsoid in the grid mapping. To define it
separately somewhere else, even if only by name, could be confusing or
inconsistent.

* it's not needed only for vertical coordinate variables. If you have a *data*
variable with the standard name of altitude, for instance, you might want to
identify the geoid.

However I do feel that it's a bit cumbersome to introduce a grid_mapping
variable with a "dummy" horizontal coordinate system just to name the geoid.
What if we allowed the grid_mapping_name to be optional if the grid mapping
was included only in order to describe the geoid or reference ellipsoid?

float altitude; // data variable
  altitude:standard_name="altitude";
  altitude:grid_mapping="geoid";
char geoid; // grid mapping variable
  geoid:geoid_name="WGS84";

Cheers

Jonathan
Received on Fri Sep 12 2008 - 01:44:58 BST