Improvement of bias detection in Argo float conductivity sensors and its application in the North Atlantic

We propose modifications of the OW method (Owens and Wong, 2009) used to estimate the time varying correction of conductivity measurements from Argo floats. These modifications are necessary to account for large interannual to decadal variability of the large-scale salinity field observed, for instance, in the North-Atlantic Ocean and to provide corrections with realistic error bars. The covariance function used to map reference salinity data at the float profile position was modified in order to minimize the contribution of the oldest reference data to the large-scale salinity field estimate. Mapping error now includes errors in the large-scale field estimates and fit error now takes into account the lateral dependence between climatological profiles. Finally, we used the modified OW method to check the consistency of the Argo salinity dataset available in delayed mode in the North-Atlantic Ocean. Overall, salinity corrections need to be reconsidered for 4.5% of the floats. (C) 2016 Elsevier Ltd. All rights reserved.