Improve profile QC output #54
Conversation
* OBSERVATION_QC only rejects if all data rejected at a given location * fix whole profile QC variables for profiles * add test of profiles obs to write behaviour
|
Thanks for making the changes. I believe we do need DEPTH_QC to be filled out. This is used by the feedback file reader in NEMO. At the moment we are getting away with it not filled out because it has a negative fill value and the code checks for flag <= 2, but probably we shouldn't rely on this as other code might not work this way. Having said that, in the OPS feedback writer it looks like it is filled with -99999 where there is no valid depth or 1 if there is a valid depth, so it is actually quite similar to what we already have. |
|
Great, thanks @s-good. Would it be better to add on a DEPTH_QC change here or merge this promptly and make the change in a subsequent PR? How about JULD_QC and POSITION_QC? Are we best off just setting all of these to a default of 0? |
Just a note to say that we discussed this earlier and decided that these variables should be set according to the contents of new diagnostic flags (e.g. a PositionReject diagnostic flag). Even if these are not being set in the current processing yaml configuration, it gives us the option of using them in the future if we want to. |
this is in the latest commit now. |
Description
Quality control information was missing or wrong for whole-profile variables such as
OBSERVATION_QC,POTM_QC, andPSAL_QC. I have also added information forDEPTH_QC,JULD_QCThis change fixes these variables as well as adding test output for writing a sub-set of profile observations.
There still remains missing QC information for the QC_FLAGS:
As far as I know, these fields were not used by the old OPS or by NEMO/NEMOVAR. At least some of these fields have no meaning for JOPA, however it is possible we might want to fill out the depth quality control flags, as depth is considered an observed variable at the moment and the QCFlagsData does exist.
Tests