search

IQuOD / Formats

Data format development
1 stars 1 forks source link

QC flag scheme #6

Open mhidas opened 6 years ago

mhidas commented 6 years ago

It's been suggested that we could have a two-tiered flagging scheme:

  1. A simple good/bad(/maybe) flag that most users would use to select good data for processing;
  2. A "reason" flag to record which QC checks (auto or expert) failed.

Both of these could be applied on a per-profile and a per-level basis.

tcarval commented 6 years ago

Hello Marty,

Argo is using such a scheme, although it does not appear as a two-tiered flagging.

Explanations in Argo user's manual, version 3.2, September 11th 2017 http://dx.doi.org/10.13155/29825

  1. A QC flags scale : see page 78 §3.2 “Reference table 2: Argo quality control flag scale”

  2. A history section that records the list of tests performed and the list of passed/failed tests: see page 122 §5 “Using the History section of the Argo netCDF Structure”

Best regards,

Thierry

De : Marty Hidas notifications@github.com Envoyé : mercredi 18 avril 2018 12:36 À : IQuOD/Formats Formats@noreply.github.com Cc : Subscribed subscribed@noreply.github.com Objet : [IQuOD/Formats] QC flag scheme (#6)

It's been suggested that we could have a two-tiered flagging scheme:

  1. A simple good/bad(/maybe) flag that most users would use to select good data for processing;
  2. A "reason" flag to record which QC checks (auto or expert) failed.

Both of these could be applied on a per-profile and a per-level basis.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/IQuOD/Formats/issues/6 , or mute the thread https://github.com/notifications/unsubscribe-auth/AS5znADuKdyoLgkShKHn0gT8SOCYznrwks5tpxcsgaJpZM4TZ045 . https://github.com/notifications/beacon/AS5znDn60rQoA0En_oERuf3KlitKiqsRks5tpxcsgaJpZM4TZ045.gif

BecCowley commented 6 years ago

The scheme recommended by Tim is the IODE flagging scheme. Here is the link: https://www.iode.org/index.php?option=com_oe&task=viewDocumentRecord&docID=10762 I suggest we need to continue with the simple scheme followed in Argo and GTSPP. Can we fit it with the IODE scheme? I think we should avoid using history records if possible. The reason for choice of flag should be clear in the flags that we select.