ESMValTool and observational (OBS) data uncertainties

[valeriupredoi]

This is a broad issue (and not even an issue per se) - and pretty skinny at the moment, since it contains just a shout out - we should start thinking about implementing the use of OBS data uncertainties in dedicated diagnostics, but before that we should probably ask ourselves if there are already any functions that deal with this. We will have a discussion about this at the UKESM retreat (November 11-12, 2019) and I will post whatever useful conclusions we get to. In the meantime, can people add anything to this if relevant functionality is planned/already exists pls? Another aspect would be to start thinking what sort of preprocessor functions we'd need to write to support various stats for OBS's :beer:
@ESMValGroup/esmvaltool-developmentteam

[rjp23]

I also raised this at the ESA CCI CMUG integrration meeting. A lot of obs data doesn't include uncertainties and how these uncertainties are propagated when averaging or regridding needs a lot of care and expertise from the data providers.

As a starting point, if Obs4MIP data contains an estimate of uncertainties then ESMValTool should at least be able to utilise them but this is a very complex issue and probably not just something that can be easily integrated generically into the code.

It'd be useful to collect whatever thoughts/plans that ESA and CMUG might have in this area for ESMValTool.

[zklaus]

We should also be aware of the place this kind of data has in the current and upcoming standards.
For example, in CF there is ancillary data for this purpose (see also CF appendix C).

There was a discussion of some ESMValTool implications about that also on the iris mailing list recently.

As for the state of support for ancillary variables in Iris, I give SciTools/iris#3483 as an entry point.
It is an area of rather active development and it seems Iris 3.0.0 will offer good support.

[valeriupredoi]

@rjp23 would you be ok to upload your presentation from yesterday's retreat session? That'd be great, it was very informative to me :beer:

[valeriupredoi]

Rob @rjp23 did you have a chance to plop them slides here, pls :beer: