Skip to content

Better QC of Argo profiles #2052

@guillaumevernieres

Description

@guillaumevernieres

Description

We need to update the current QC for gfsv17: profiles that "look OK" but have probably been wrongly geo-located in the original file or have a constant bias throughout the entire profile are not being QC'ed out. This results in very large, unphysical, localized increments, see example below.

Image

QC update

Addition of a profile backround check:

Remove the entire profile if the RMSE of O-B for the profile is greater than a specific threashold.

      # Reject profiles with large differences from background
      - filter: Profile Background Check
        filter variables:
        - name: salinity
        absolute threshold: 0.5

Reject obs that are below the model's sea floor

      # Reject obs where MetaData/depth > GeoVaLs/sea_floor_depth_below_sea_surface + maxvalue
      - filter: Bounds Check
        filter variables:
        - name: salinity
        test variables:
        - name: ObsFunction/Arithmetic
          options:
            variables: [MetaData/depth, GeoVaLs/sea_floor_depth_below_sea_surface]
            coefs: [1.0, -1.0]   # depth - sea_floor_depth_below_sea_surface
        maxvalue: -100.0         # i.e., depth - sea_floor_depth_below_sea_surface <= 0

Metadata

Metadata

Type

No type

Projects

Status

Todo

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions