[cf-pointobsconvention] Should Z Coordinates be required?

Dear John

I completely sympathise with encouraging people to provide useful metadata.
I would argue against a mandatory Z (or any) coordinate, though, as the
data really might not have a useful Z coordinate. I'm guessing, but
what about the sea state (its roughness due to waves)? That's a property of
the sea surface one might record, but it doesn't have a height, and the ship
observing it is obviously at sea level. Or the total cloud amount, say? This
also doesn't relate to a particular altitude in the atmosphere. You could
record the altitude of the station that observed it, but I would not say that
was really a coordinate of the cloudiness itself, and again it would not be
useful for marine obs. Requiring instead an "unknown" indicator or a missing
data indicator strikes me as excessive, and I'm not convinced data-writers
would always do it, or that it would be useful if they did. There are lots
of features of CF metadata which are very valuable but optional, such as
bounds, cell_methods and standard_names.

Cheers

Jonathan