GODAE Quality Control Pilot Project

1
GODAE Quality Control Pilot Project

• An initiative aimed at improving the
  effectiveness of automated quality control
  procedures for GODAE prototype systems
• History
• Apr 2001 - need for systematic approach to ocean
  data quality control in support of GODAE
  recognized at US GODAE Workshop
• Jul 2001 - need reiterated at GODAE IPRC workshop
  at U. Hawaii
• Dec 2001 - model and data assimilation quality
  control monitoring and evaluation project
  formulated at the IGST VI meeting
• May 2002 - became integral part of the GODAE
  implementation plan
• Jun 2002 - QC workshop held in conjunction with
  GODAE symposium in Biarritz
• Oct 2003 - initiate QC methods and outcomes
  intercomparison project
• Biarritz Workshop Report www.bom.gov.au/bmrc/oce
  an/GODAE/frames.html

2
Ocean Data Quality Control

• Data quality control is a fundamental component
  of any ocean data assimilation system
• accepting erroneous data can cause incorrect
  analysis
• rejecting extreme data can miss important events
  and anomalous features
• Quality control involves checking data against a
  pre-established set of standards
• requires clearly documented procedures
• depends on the reliability of the standards and
  the choice(s) made for measuring goodness of fit
• assumptions concerning the distribution and
  coherences of observation errors relative to the
  standards need to be routinely examined

3
Goals GODAE Quality Control Pilot Project

• Document ocean data quality control procedures in
  use today
• real-time or delayed-mode
• fully automated or man-machine mix of operations
• any data type that directly impacts ocean state
  estimation process (in situ and
  remotely sensed)
• Initiate intercomparison projects among groups
  performing QC
• routine sharing of QC results (real-time and
  delayed mode)
• define QC outcomes (metadata) that can be shared
  among data users
• describe QC test procedures needed in the future
  (e.g. specific
  instrumentation error checks)
• Develop partnerships with data providers
• feedback on model-data differences
• data acceptance, data rejection rates
• identification of systematic problems in
  observing systems

4
Participants GODAE Quality Control Pilot Project

• US Navy (NRL - FNMOC, NAVOCEANO) - J. Cummings
• real-time, fully automated
• all operational data sources (profile and in situ
  off GTS, satellite data from NAVOCEANO)
• Met Office (FOAM, ENACT) - M. Huddleston, B.
  Ingleby
• real-time, delayed mode, automated and manual
  (visual inspection)
• profile, in situ data (GTS)
• MEDS (GTSPP) - B. Keeley
• real-time, delayed mode, automated and manual
  (visual inspection)
• profile, in situ data off GTS
• Coriolis Center (SOAP/MERCATOR) - T. Carval
• real-time, delayed mode, automated and manual
  (visual inspection)
• profile, in situ data (GTS)

5
Status GODAE Quality Control Pilot Project

• QC Pilot Project web page created on GODAE server
• QC system documentation available - MEDS,
  FOAM/ENACT, Coriolis, NRL
• QC data on GODAE server
• NRL QC data files updated daily (all sources, in
  situ and satellite)
• MEDS GTSPP files updated every 3 days
• links to Coriolis Center Mercator/SOAP data
• code to read different formats and produce common
  ASCII text format for intercomparison of QC
  outcomes under development
• formed liaison with CLIVAR GSOP/DPM (SIO, Feb
  2004)
• access to additional scientific QC data for
  intercomparison
• access to delayed mode/scientific QC procedures
  can delayed mode QC procedures be automated and
  run in real-time?

6
Plans GODAE Quality Control Pilot Project

• Focus on Argo data for initial intercomparison
  study
• existing, well defined data management strategy
• real-time and delayed mode QC in place with
  feedback mechanisms
• observations are readily available and widely
  used
• Reanalysis effort using expertly QCd Indian
  Ocean data from CSIRO
• compare with Enact dataset, NRL R/T system, World
  Ocean Database, others?
• examine regional and/or temporal variability in
  patterns of rejection
• detailed cross-evaluation of profiles rejected by
  one system and accepted by another (knowing
  truth from expert)
• test down-stream effects of QC decisions
• what is added value of a model to the process
• what is effect of differential profile
  acceptance/rejection on derived ocean products
• define factors that influence both R/T and
  scientific QC acceptance and rejection rates