CORA-03

It is updated on a yearly basis vith several processing steps :

• Addition of new profiles- duplicate check
• Update of existing profiles (raw replaced by adjusted parameters)
• Additional Quality control by CLS and Coriolis Scientists
• Specific processing: XBT fall rate correction

The Quality control is done in 2 steps:

1. Automatic tests (CLS) :
   • values outside regional limits, values=0, constant along a profile
   • comparison to climatology ( > 10 σ) (WOA09) + systematic bias compared to climatology
   • Large salinity gradient at the surface
   • All suspicious profiles are visually checked and QC flags are modified if necessary
2. Checks of the whole database

• Anomaly Method test : Bias > 7 sigma over at least half of the profile values
• Objective Analysis test : statistical test involving an objective analysis method and based on residual analysis (Gaillard et al., 2009) §
• Argo platform control : Argo floats with recurrent alerts as well as Argo floats pointed out by the altimetry test (Guinehut et al., 2009) are visually checked over all their life period

Specific processing : correction of XBT fall rate: within the corrections are available , the choice was made on the home made one: empirical correction developed by M. Hamon and submitted to JGR June 2011

Empirical correction of XBT data ,M. Hamon, G. Reverdin, and P.-Y. Le Traon 1.92 MB

This method consists of: :

• Collocation method between XBT and ARGO or CTD profiles from CORA
• Statistical corrections computed for each year and different classes of probes:
  1. Correction of a thermal bias : Txbt = Txbt-Toff (due to possible errors on the calibration of the probe and obtained with selected profiles with a weak T gradient in the upper layer)
  2. Correction of the deph bias : Ztrue= Zobs( 1-A-B*Zobs)-Zoff (due to too simple approximation of the XBT fall rate)

Compared to 2008 realease the coverage has been extended to 1990-2010

• CORA3.2 = more than 5 million profiles
• First challenge : to gathered all the individual profiles !
• Second challenge : to qualify them !
• Third challenge : to be up-to-date with common corrections/adjustments
• In-situ data sets: alive datasets : new validation tools, detection of a failure on a sensor, on a datasets … delayed time processing needed to correct drift, offset …

CORA database not only contains the raw parameters but also the adjusted parameters if it exists. Especially for Argo floats, the adjusted parameter is mainly the salinity corrected in delayed mode by the PI of the float (Owens and Wong method) – but also the pressure. About 75% of Argo Data is processed in delayed mode in CORA3.2