**NOTA**: Here, there is something fishy. I suspect a mistake in my routine. Go and see *this link* for a correct calculation of FSLE.

The backward-in-time finite-size Lyapunov exponents (FSLE), representing the unstable manifolds, have been recomputed using more accurate products for the sea surface heigth (SSH); These product contain, in particular, the *time-mean* SSH which was not the case for the previous product used. In this animation, the upper left panel shows the FSLE averaged from a distance around the float, the lower left panels shows the upper 200-m of nitrate measured by the float and the right panel shows the horizontal distribution of FSLE at the specific time shown by the vertical dashed white bars in the left panel. The SSH product used for this animation is “DT-ref” (delayed in time, referenced) and is supposed to be the most accurate and consistent SSH product.

Fig. 1 shows the time series for the two other SSH products, “DT-upd” and “NRT”, the latter being the “near-real-time” product which covers the entire period of available observations.

Fig. 2 shows the time series of FSLE averaged within 0.2° from the float position for the different SSH products, as well as the time series of nitrate averaged between 150 and 200 m.