Any convention on what flag_meaning to use to QC pass?

[David-Rayner-GVC]

Hi! Is there any convention for what flag_meaning code should be used for " qc pass"? Examples:

cf-convention/cf-conventions#205
has flag_values 1 and flag_meanings "PASS",

whereas
http://gyre.umeoce.maine.edu/data/gomoos/buoy/php/view_ncfile_metadata.php?ncfile=/data/gomoos/buoy/archive/I0104/realtime/I0104.accelerometer.realtime.nc
has 0b for "quality_good"
mirroring
http://cfconventions.org/cf-conventions/cf-conventions.html#flags

Feels like it would be useful to standardize this?
Thanks!

[larsbarring]

Hej David

As you maybe saw issue cf-convention/cf-conventions/issues/205 was rather outdated (and now closed) and the discussion continued in cf-convention/cf-conventions/issues/216. And after some further discussion this resulted in the text that you are referring to. So I think there is in fact already agreement.

Lars

[David-Rayner-GVC]

Thanks Lars!

The text in flag_meanings is also completely unrestricted

So, do I have this right: it is fairly simple to automate detecting whether an aggregate quality flag is provided for a variable, but it is impossible to automate extracting only values that pass QC because there is no agreed flag_meaning for "pass"?
dave

[larsbarring]

I am not an expert on these matters, but I imagine that the problem is in defining what is actually meant by QC (=quality control) as such and teh different quality categories. What aspects are assessed, and what does "pass" actually mean for all those diverse variables, data producer and data user communities, and usage situations? What is regarded as "pass" in a specific context may be quite insufficient in another situation. Quality is in the eye of the beholder. To put this on a more unified footing a common framework all these aspects have to be described and agreed upon. Note also that flags can be used for many purposes so you have to first establish that a flag variable in fact contain QC information.

[David-Rayner-GVC]

what does "pass" actually mean for all those diverse variables, data producer and data user communities, and usage situations?
Ok, I agree "pass" probably couldn't be standardiazed as it is too subjective. But maybe "no known quality issues" could be more easily assigned/interpreted?