No satellite-based SSS observations are used in the assimilation, see
which is for GOFS 3.0 but I think still relevant to the observation sources in GOFS 3.1.
Which compares the top 500 m salinity against unassimilated profiles.
GOFS 3.1 does relax SSS towards monthly climatology providing the model SSS is within 0.5 psu of climatology (otherwise it does no relaxation, e.g. near river mouths).