Lagrangian data are an important component of WOCE, providing extensive 
sampling in both the horizontal and the vertical. They are often used to 
estimate ocean mean flows, using the "binning" technique where averaging is 
performed over certain spatial bins and certain time periods. These 
"pseudo-Eulerian" estimates are known to be possibly contaminated by biases due 
to correlations between drifter concentration and velocity (e.g. Davis, 1991).

Here we study statistical errors of pseudo-Eulerian mean flow estimates using 
synthetic drifter trajectories, computed in a numerical flow from a high 
resolution MICOM run in the North Atlantic (Garraffo et al., 2000). The full 
data set is substantially larger than the WOCE requirement, and allows to study 
the effects of subsampling in space and time. The estimates from Lagrangian data 
are compared to "true" Eulerian averages over 1 degree square bins. Sampling 
errors, related to subgrid scale variability and finite sampling, are studied 
first. Biases effects are then considered, indicating that pseudo-Eulerian 
estimates tend to underestimate (overestimate) the velocity in the eastern 
equatorial regime (western boundary currents). An analysis of the results 
suggest that these biases are primarily due to mesoscale divergence processes 
resulting in nonzero correlation between instaneous drifter concentration and 
velocity.