Work in progress

Statistic Method

Before any calculation, salinity data must be corrected, when the value is available, with the tank offset. This value corresponds to an offset measured before float launching. It can reach values around 0.01 psu.

The delayed mode method, used at the Coriolis Data Center in an operational mode, has been developed by Wong et al. (2003) for the Pacific Ocean and adapted to the North Atlantic environment by Böhme and Send (2005). It is based on the comparison of the data with ship-board high-resolution CTDs. Those comparisons are made on deep isotherms and assume that the temperature sensor of the float is stable and that salinity on deep isotherms is steady and uniform. The calibrated historical hydrographic data are interpolated at the float profile position by an objective analysis method. The analysis takes into account the high spatial and temporal variability of the North Atlantic due to ocean processes and bathymetry. A first fit is applied to estimate an individual correction for each profile. Then, assuming that the conductivity sensor drifts slowly over time, a second fit (piece-wise linear) is applied to estimate a time-varying correction over the length of the time series. Calculations are made in potential conductivity space. The result is a set of corrected salinity data with corresponding uncertainties.

However with this method, the smooth correction of a cycle is influenced by the value of 6 month-earlier and 6 month-later data. Therefore, significant corrections can be proposed for cycles that do not need any! To solve this problem, we decided to split the float series into several segments and to calculate the smooth correction of a cycle just considering data that belong to the appropriate segment.

Complementary Tools

In parallel to the statistic method, complementary tools have been developed. These different methods allow us to detect the presence (or not) of an offset or a drift in the float lifetime.

Differences method : This method is inspired by the residuals study from objective analysis. It consists in comparing in situ data with salinity (or temperature) fields values from weekly analysis (at the profile location). We take into account data with flags equal to 1,2 or 3.

Mean fields : We compute the mean of salinity (or temperature) fields from weekly analysis along float lifetime and superpose in situ data.

Other tools: The PI can make some extra plots : technical parameters (battery voltage, surface pressure offset...), evolution of consecutive cycles...

Final correction

Thanks to these different plots, the PI is now able to take a decision concerning the float's behavior and to propose a correction value for each cycle.