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

• Published in: Progress in oceanography 2014-11, Vol.128, p.15-29
• Main Authors: Berta, Maristella, Bellomo, Lucio, Magaldi, Marcello G., Griffa, Annalisa, Molcard, Anne, Marmain, Julien, Borghini, Mireno, Taillandier, Vincent
• Format: Article
• Language: English
• Subjects: Drift; Estimates; Indication; Marine; Mathematical models; Radar; Sciences of the Universe; Sea currents; Transport; Uncertainty
• ISSN: 0079-6611; 1873-4472
• DOI: 10.1016/j.pocean.2014.08.004

•Lagrangian transport estimates in Northwestern Mediterranean Sea.•Application of the LAVA algorithm to blend drifters with HF radar or model.•Application of LAVA to reconstruct velocity fields from drifters only.•Remarkable improvement of transport estimates correcting velocity fields with LAVA.•LAVA capability to reduce search range in operational and forecasting applications. Lagrangian transport estimates are investigated using results from HF radar, model and drifter data during a dedicated experiment in the Ligurian Current in the Toulon area (North Western Mediterranean Sea). Uncertainty estimates on particle position, D(t), are computed and compared to absolute dispersion, D0(t), that provides an indication of the uncertainty in case of zero prior knowledge. In agreement with previous studies, radar results show that D(t)≈1/2D0(t) (i.e. ≈6km after 24h). Model results are less reliable, as it can be expected in highly nonlinear coastal flows without local data assimilation. The central result of this paper is that when drifters are promptly deployed in an area of interest, their data can be used to significantly improve transport estimates using the Lagrangian blending algorithm LAVA with velocity fields from models or radar. Uncertainty can be reduced to ≈1/6D0(t), (i.e. ≈2km after 24h) for both radar and model, implying a much reduced search range in case of operational applications. The method is also found to have some forecasting skills with uncertainty ≈1/2D0(t) during the first ≈6h. Sensitivity tests provide indications on relevant time and space scales of predictability and provide suggestions for appropriate drifter sampling strategies.