Quality control of ocean temperature and salinity profiles — Historical and real-time data

A new automated quality control system for oceanic temperature and temperature–salinity profiles is presented. Substantial development was needed for some of the quality control algorithms although the checks were based on documented procedures used elsewhere where appropriate. A new automated ship track check was developed: the results of an undetected position error can be very damaging to ocean analyses. Also important is a check against a gridded background, this can be a climatology but near the surface it is advantageous to use an estimate that is evolving over time. Bayesian probability theory is used in the background check and the associated check against nearby observations (buddy check). The system was used to process archive data for 1956–2004. As a by-product monthly model-free objective analyses for this period were produced. Versions of the system are used for near-real time ocean analysis and for initialising both short-range ocean forecasts and seasonal atmosphere–ocean forecasts. The main features of the oceanic observing systems are presented along with quality control statistics, examples of errors that can occur and some additional problematic cases.