## 1. Introduction

[2] Horizontal transport and mixing processes are central to the study of the physical, chemical, and biological dynamics of the ocean. Correct understanding and precise modelling of them are relevant from a theoretical viewpoint and crucial for a range of practical issues, such as plankton dynamics or the fate of pollutant spills. In this regard, the last few years have seen the appearance of interesting new developments [*Mariano et al.*, 2002] on the Lagrangian description of transport and mixing phenomena, many of them coming from the area of nonlinear dynamics. Such approaches do not aim at predicting individual tracer trajectories, but at locating spatial structures that are known from dynamical systems theory to act as templates for the whole flow [*Ottino*, 1989; *Wiggins*, 1992]. Such structures (attractors, saddles, manifolds,…) have been used since many years ago for classifying the evolution of trajectories in abstract dynamical systems. However, when put in a fluid dynamics context, they gain a new and direct physical meaning, corresponding for instance to avenues and barriers to transport, vortex boundaries, or lines of strong stretching. In particular, chaotic motions such as the ones occurring in the turbulent ocean are characterized by complex intersection of stretching and contracting manifolds at the so called *hyperbolic points*: regions of fluid initially compact in the proximity of these points become elongated along the stretching directions and then folded, leading to the typical filamental and convoluted structures that are common in satellite pictures of water temperature or chlorophyll, as well as in laboratory experiments. From the point of view of transport, such regions are characterized by a strong mixing: trajectories of initially close particles are quickly separated along the stretching directions, and fluid of different origins is inserted in between.

[3] Until recently, the power of these novel Lagrangian approaches has been mainly relegated to mathematical systems or simplified workbench models, since the required detailed knowledge of the velocity field was not readily available in real geophysical situations. However, in the last decades the situation has dramatically changed, with a rapidly increasing amount of data available from Lagrangian drifters [*Mariano et al.*, 2002], satellite measurements [*Halpern*, 2000], and especially from detailed computer models [*Haidvogel and Beckmann*, 1999; *Dietrich*, 1997]. This has paved the way to a growing number of geophysical applications, among which we mention studies of the Loop Current in the gulf of Mexico [*Kuznetsov et al.*, 2002], or the characterization of dispersion properties in the Adriatic Sea from drifter experiments [*Lacorata et al.*, 2001]. The methodology is not restricted to ocean circulation, but it is also being developed in atmospheric dynamics [*Joseph and Legras*, 2002], and mantle convection [*Farnetani and Samuel*, 2003]. Indeed, new Lagrangian approaches are very appealing for geophysical applications, since they can be used as a tool for automatically extracting transport structures underlying raw Eulerian velocity data.

[4] In this Letter we characterize mixing strength at the mesoscale in different areas of the Mediterranean Sea by means of a Lagrangian technique, the Finite Size Lyapunov Exponents method. The technique also identifies dynamical objects that organize the transport, and relevant coherent structures. To our knowledge, this paper shows for the first time the optimality of the method to identify mesoscale structures in an oceanic context.