Variational analysis of drifter positions and model outputs for the reconstruction of surface currents in the central Adriatic during fall 2002

Abstract

[1] In this paper we present an application of a variational method for the reconstruction of the velocity field in a coastal flow in the central Adriatic Sea, using in situ data from surface drifters and outputs from the ROMS circulation model. The variational approach, previously developed and tested for mesoscale open ocean flows, has been improved and adapted to account for inhomogeneities on boundary current dynamics over complex bathymetry and coastline and for weak Lagrangian persistence in coastal flows. The velocity reconstruction is performed using nine drifter trajectories over 45 d, and a hierarchy of indirect tests is introduced to evaluate the results as the real ocean state is not known. For internal consistency and impact of the analysis, three diagnostics characterizing the particle prediction and transport, in terms of residence times in various zones and export rates from the boundary current toward the interior, show that the reconstruction is quite effective. A qualitative comparison with sea color data from the MODIS satellite images shows that the reconstruction significantly improves the description of the boundary current with respect to the ROMS model first guess, capturing its main features and its exchanges with the interior when sampled by the drifters.

Estimating Lagrangian transport blending drifters with HF radar data and models: Results from the TOSCA experiment in the Ligurian Current (North Western Mediterranean Sea)

“…LAVA is a variational method developed by Taillandier et al (2006a) and applied in a number of cases for both full assimilation of subsurface float data and blending of surface drifters (Chang et al, 2011;Taillandier et al, 2006b;Taillandier et al, 2008). The basic concept behind LAVA is that Lagrangian data are used to correct an available Eulerian velocity field, used as first guess, by minimizing the distance between the observed positions and the positions of numerical trajectories advected in the Eulerian field.…”

“…1). While blending with model results has been tested in a number of previous applications (Chang et al, 2011;Taillandier et al, 2008), blending with HF radar is performed here for the first time, and it deserves a brief discussion to clarify its interpretation. The process of blending two data sets, each one with its peculiarity and uncertainties, is intended to provide an optimized field that takes into account the best aspects of each platform.…”

“…In particular, the LAgrangian Variational Analysis (LAVA) is a method that has been used for both, full assimilation of subsurface floats at 350 m (Taillandier et al, 2006b(Taillandier et al, , 2010 and blending of surface drifters in various and different regions such as the Adriatic Sea (Mediterranean Sea) and the Kuroshio extension (Chang et al, 2011;Taillandier et al, 2008). LAVA combines information from Lagrangian instruments with model velocity fields requiring that the distance between observed trajectories and synthetic trajectories computed from the velocity field is minimized.…”

On the use of a simple primary productivity model to assess the skill of a physical ocean model

“…However, it is rather difficult to assess the skill of a numerical model in producing accurate currents and circulation features, mostly because the verification data available are severely limited (Rixen et al 2009, Kantha 2005, Taillandier et al 2008. The dearth of in situ data forces modelers to rely instead on more readily available remotely sensed data for model skill assessment.…”

Assimilating en-route Lagrangian observations

“…The resulting observation operator in this methodology is then the identity in the instrument coordinate at observation instances, for example, times at which the instrument has surfaced and where it can report its location. Several theoretical and applied studies have successfully assimilated Lagrangian data in the context of ocean modelling (Salman et al, 2006;Taillandier et al, 2006Taillandier et al, , 2008Restrepo, 2008;Salman et al, 2008;Mariano et al, 2011).…”