Quality Flags

1
Quality Flags

• Primary Standards for Consideration
• ARGO
• CSIRO XBT
• IGOSS
• IRD -Noumea Sea Surface Salinity codes (included
  just for discussion of the "incremental flags"
  approach)
• ODV
• SeaDataNet
• WOCE
• WOD

2
Quality Flags
3
Quality Flags
4
Quality Flags

• IRD -Noumea Sea Surface Salinity codes
• Code 1   visual detection/correction of errors
• Code 2  Code 1 comparison with climatological
  mean and standard deviation.
• Code 3  Code 2 utilisation of simultaneous
  water samples and/or CTD measurements
• Code 4  Code 1 utilisation of the pre/post
  calibration coefficients of the sensors
• Code 5  Code 2 utilisation of the pre/post
  calibration coefficients of the sensors
• Code 6  Code 4 utilisation of simultaneous
  water samples and/or CTD measurements
• Code 7  Code 5 utilisation of simultaneous
  water samples and/or CTD measurements

5
Quality Flags
Ocean Data View Quality Codes 0 Good 1 Unknown 4
Questionable 8 Bad
6
Quality FlagsSeaDataNet
7
Quality Flags
8
Quality Flags
9
QUALITY CONTROL FLAGS The obvious advantage of
flagging data is that users can choose to accept
or ignore all or part of the flags assigned to
data values. The most important flags those that
are set based on unusual features produced during
objective analyses of the data at standard
levels. Data from small-scale ocean features such
as eddies and/or lenses may not be representative
of the large-scale permanent or semi-permanent
features and may cause unrealistic features such
as bulls-eyes to appear. As noted by Levitus
(1982), it is not possible to produce one set of
data analyses to serve the requirements of all
possible users. A corollary is that it is not
possible to produce one set of quality control
flags for a database that serve the exact
requirements of all investigators. As data are
added to a database, investigators must realize
that flags set for having violated certain
criteria in an earlier version of the database
may be reset solely due to the addition of new
data which may change the statistics of the
region being considered. Even data that have
produced unrealistic features may turn out to be
realistic when additional data are added to a
region of sparse data. Conkright et al.