Sensitivity of numerical tracer trajectories to uncertainties in OGCM velocity fields

Three different data sets of numerical drifters are obtained with degrading the time sampling (1 day, 1 month and 1 year) of the Eulerian velocity field computed from a Mediterranean general circulation model. The Finite-Scale Lyapunov Exponent (FSLE) technique is used to characterize, for each of the three data sets, Lagrangian dispersion properties in relation to the time resolution of the field. In particular, we are interested in measuring th eunpredictability of trajectories due to the uncertainly in the knowledge of the velocity field. Our data analysis indicates that surface relative dispersion of the Mediterranean Sea has two regimes; exponential spreading due to chaotic advection at small scales (~mesoscale) and super-diffusion at larger scales (up to ~sub-basin scales). In this scenario, it is shown that trajectory evolution is most sensitive to the time sampling of the field at small spatial scales, while, at scales larger than (~100 km, it is essentially independent from the details of the models. Also, FSLE is employed to visualize the geographical regions characterized by high Lagrangian unpredictability. The relation of FSLE with common oceangraphic observables (e.g., local shear, velocity variance) is duscussed. © 2002 Published by Elsevier Science Ltd.