Jonathan,
> The intention of recording the units as 1e-3 (dimensionless) was to
> suggest a canonical unit of PSU i.e. approximately the same as parts
> per thousand. However, this is unclear and therefore
> unsatisfactory. We have discussed this before, in fact, and I believe
> we have decided in principle that the unit should be indicated as
> psu. Since that is not a legal unit in the standard udunits.dat, in
> past discussions we have also decided we would publish our own CF
> version of udunits.dat, including psu as a unit (not convertible to
> other units). The CF udunits.dat should also include sverdrup (=1e6 m3
> s-1) and bel B (hence decibel dB, dimensionless). The latter is
> already given as the canonical unit for some standard names, although
> it is not a legal udunit.
The ability to represent logarithmic units such as decibels has been
added to udunits, although the string representation for such units
includes a reference level. For example, decibels can be represented as
"0.1 lg(re m/(5 s)^2) _at_ 50". The new udunits software also has an XML
database that can be extended with new units, a new API, and support for
the current udunits API:
http://www.unidata.ucar.edu/software/udunits/
We're planning on including the new udunits with the upcoming netCDF 4.1
release, but you can try it now by configuring with --enable-udunits in
the current snapshot release.
> I'd suggest that the CF udunits.dat, with these changes, should be put on our
> website, mentioned in 3.1 section about units in the CF standard, and used by
> the CF checker.
That seems like a good idea, and I think we should also include the XML
version if the new udunits package becomes more widely used.
--Russ
Received on Wed Jun 17 2009 - 07:10:10 BST
