5 
T~i - f's^tidx 
with R iy the normalized auto-correlation function 
R t - f ( t) Uj ( t+T) dt 
u, v zonal and meridional component of V 
lPu' zonal variance 
u r \P covariance 
v ' v ' meridional variance 
V mean current vector 
Progressive vector diagrams 
The progressive vector diagrams are processed from unfilte 
red data. The start position is marked by a big cross (+). Time 
* 
marks fsmall crosses) are given every 30 days. The numbers near 
the tracks are the sampling depth in metres. 
Time series 
The time series have been processed using partly modified 
functions of the MATLAB Signal Processing and Optimisation Tool 
box. The data have been low-passed with a 10th order IIR-Butter- 
worth filter (infinite-duration impulse response). Filtering in 
both directions led to a zero phase lag.