I went back to original concerns, as the discussion is a little bit here and there. From this I conclude the canonical unit for P.S. should be changed to 1.
(For historians, I see there was a brief thread 2009.06.17 titled 'salinity units', which started down the same question but then got sidetracked into another topic. As this was pre-TEOS publication, previous advocacy for 1e-3 over 1 might now be dropped in light of current well-defined oceanographic practice.)
Nan started (5/22) with
> I was recently notified that my salinity data was likely to be overlooked by
> some users, because I'd used '1' as the units, not '.001'. ?
>
> Some members of the OceanSITES project are interested in revising our
> format spec to encourage the use of '1' as an indication that salinity does
> not have units - but, of course, we'd mostly rather remain CF-compliant.
Whether '1' or '.001' is used as the unit, it's still CF-compliant, as Alison pointed out. Indeed the only purpose of canonical units is to specify what _class_ of units applies, right? So CF should feel free to change the canonical units all day long, to match current practice, as long as the unit class is unchanged. (Hypothesis)
To me this implies that using the exact unit as an indication of whether data is a particular type of salinity data is dubious. And, it also suggests that using '.001' (the canonical value), or any non-unitary value, as units for a dimensionless quantity is dubious. (Because it implies you *can* multiply by 1000 and get a meaningful result, and that you *should* multiply by 1000 if you want your data to match the data of someone who uses '1' as the unit. What _else_ would .001 mean?)
All of which implies .001 is a poor choice of canonical unit for the dimensionless salinity standard names.
Nan's more recent (5/27) question goes to the human nature part of the issue:
> The '.001' units for P.S. doesn't mean that stored values of practical
> salinity differs from A.S. by 'a factor of around a 1000', as far as I
> know. If that's the logical inference, then this unit is really a problem,
> and maybe we should do something about it.
I agree with Nan's first sentence here, and while I don't think that it expresses a _logical_ inference, I think it is a _natural_ inference for less expert users who miss the fact P.S. is dimensionless (and '.001' as the canonical unit will help them make that mistake).
So there is potentially a problem already if anyone used .001 as the unit, and reported their 0-40 values as 0 to 0.040, as Rich just said. (I'm assuming using the values 0 to 40 is the only correct range, because P.S. is dimensionless and you can't multiply it meaningfully. Right?) But that problem exists no matter what we do with canonical unit from here forward, and to avoid it happening in the future, I think the canonical unit should be changed.
John
---------------
John Graybeal
Project Lead
Marine Metadata Interoperability Project:
http://marinemetadata.org
MMI Ontology Registry and Repository:
http://mmisw.org/orr
email: jbgraybeal at mindspring.com
skype: graybealski
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <
http://mailman.cgd.ucar.edu/pipermail/cf-metadata/attachments/20150527/27623d70/attachment.html>
Received on Wed May 27 2015 - 12:30:55 BST
