Strategies for assimilation of drifter trajectory data into
a numerical ocean model (Eulerian) are considered.  Converting
the drifter trajectories into surface-layer velocities seems
to be a simple, practical method, which is examined here.
Twin experiments are conducted with a 20-km grid, double-gyre
circulation model (a 4-layer configuration of Miami Isopycanl
Coordinate Ocean Model).  The ROIF (reduced-order information
filter) implementation of Kalman filter is used for data
assimilation.  A small number of clusters of tracers are
released to simulate drifter trajectories, from which
surface velocities are derived for the assimilation runs.
Preliminary results indicate that the spatial coverage of 
the derived velocity measurements is an important limiting
factor in reproducing meso-scale features through assimilation.