@gleckler1 you mean the name of climatology file used to compute the metrics?

Yes, the name of the climatology used to compute the metrics!

so NOAA-OISST-v2 (tos) or GPCP (pr) etc etc

@gleckler1 if you can clearly define what needs to be done here, I can implement the changes and close this up - it's not dependent on @doutriaux1 or UV-CDAT changes..

The md5sums in an updated observation json dictionary should also be done

@durack1 , look at src/python/pcmdi/scripts there is a script in there that @gleckler1 wrote that generates the obs dict. This is where it needs to sit eventually.

Ok so https://github.com/PCMDI/pcmdi_metrics/blob/master/src/python/pcmdi/scripts/build_obs_meta_dictionary.py - and will need calls to md5sum

@gleckler1 is this done - is the issue close-able?

Following the 24th Sept 2014 email to GFDL the following improvements were listed:
Fine tune the metadata in json files and file versions (json example attached). We may seek feedback from the WIP but would welcome your input.

Can we make this one a high priority? I've looked at it, but right now this one is too hard for me... the hope is that each metrics json file, for each model will include the following:

1. The filename.nc of the model climatology
2. md5sum of the model climatology
3. creation_date

@gleckler1 so to clarify, what you're asking for is the files that are being interrogated to create the results in the CMIP_results/CMIP5/historical *.json also contain the 3 attributes above? Which file that you run (and is in the repo) creates these jsons?

Don't quite follow you but I think yes... the metrics json files should include of for each model/climatology these 3 attributes... e.g.
{
"ACCESS1-0": {
"SimulationDescription": {
"Center": "N/A",
"Experiment": "historical",
!!!! "ClimatologyFilename': filename.nc,
!!!! "Md5sum of climatology file": MD5results,
!!!! "creation_date": creation_data
"Login": "N/A",
"MIPTable": "Amon",
"Model": "ACCESS1-0",
"ModelActivity": "CMIP5",
"ModelFreeSpace": "N/A",
"ModellingGroup": "CSIRO-BOM",
"SimName": "r1i1p1",
"SimTrackingDate": "2012-01-15T10:45:49Z"
},
Actually SimTrackingDate I think == creation_date, so there are only 2 new items here but both involve operating on the file object on the fly and I don't fully understand the object in the driver to do this myself

@gleckler1 are you talking about the files that we provide (so those in CMIP_results), or the output files that are created by the GFDL, NCAR or whoever folks?

This only makes sense to me for us to do this, as we have the CF-compliant CMIP5 data that we're using to create the CMIP3/5 metrics benchmarks stored in CMIP_results..

If the first one (files that @gleckler1 creates), where is that script? Is it in the repo?

creation date is simtrackingdate there's an alias for that, we can take the alias out, it was originally designed after Jerome's recommendation as it was deemed ok for a first pass.

@gleckler1 the md5sum of the climatology file is kinda arbitrary, as we are creating this climatology from the CMIP5 data that we have stored locally - would it be more useful to include more specific information which is obtained from the source files from which the climatology has been generated?

We could include both pieces of information, source files (filename, creation_date, tracking_id) and also info about the derived climatology file (filenameandpath, clim_creation_date - which will be dependent on whom created the file and whether they wrote suitable attributes)

UPDATE
all that is needed now is attributes for model climatology filename.nc and md5sum of that climatology:
climatology_md5sum
climatology_filename

I am taking care of all other attributes (CMIP5_creation_date and CMIP5_tracking_id)