CTD data analysis
All time series were processed applying a post-processing methodology, based on a routine package created with MATLAB, to obtain cleaned and despiked data. The method of despiking consists of a first visual check, and a later check of the consistency of the data. In particular, we verified that the data belong to given instrumental and physical ranges and checked stationarity. Assuming for the time-series a normal distribution, we accepted all values within the interval µ ± 3σ, where µ is the mean value and σ is the standard deviation . Furthermore, we applied a low-pass filter with a cut-off period at 33 hours to the hourly data in order to obtain sub-inertial non-tidal flow. The overall accuracies were within ±0.002°C for temperature and ±0.005 for salinity.