Can general circulation models be assessed and their output enhanced with drifter data?

Drifter data from the Gulf of Mexico are used to assess and enhance the out put of a primitive equation general circulation model. The analysis is made in a 450 km x 450 km open subdomain encompassing a Loop Current ring. The model velocity field is compared with position data from four drifters at t he drogue depth of 50 in using geometrical orthogonal functions (GOF). An E ulerian velocity field is reconstructed from the model velocity field and d rifter velocities. This reconstructed velocity improves 8-day numerical tra jectories relative to the model field by at least an order of magnitude, as quantified by two Lagrangian error metrics referenced to the real drifter paths. An Eulerian metric that compares the two fields, however, does not e xceed 7% for the 20-day assessment period. Thus the drifter data may be rep roduced with modest impact on the model velocity. Enhancement of the model velocity field is determined by two tests: the ability of the GOF velocity field to (1) improve the forecast of drifter positions using only a posteri ori data and (2) improve the forecast of withheld drifter data. Using a pos teriori data, the 20-day temporal mean of the position error is improved fo r all drifters by 87-89% for 6-hour and 26-38% for 30-hour forecasts. For 6 days, a withheld drifter is 35-40 km from a drifter whose velocity is used in the reconstructed velocity field. The temporal mean of the position err or during this period is improved by 20% for 6-hour and 26% for 30-hour for ecasts.