Coastal Current Intrusions from Satellite Altimetry

Abstract

The use of satellite-based data in coastal regions for the monitoring of fine-scale ocean dynamics, impacting marine ecosystems, is a difficult challenge. A random forest algorithm to detect slope current intrusions into the Gulf of Lion, Mediterranean Sea, has been developed using both improved coastal altimetry data and 10 year-long numerical simulations. The results have been compared to an independent dataset of in situ measurements from a bottom-moored Acoustic Doppler Current Profiler. The algorithm results are very promising: 93% of slope current intrusions have been correctly identified, and the number of false alarms is moderate. The dependence of the algorithm on several environmental factors is discussed in the paper. From the oceanographic point of view, our results confirm the strong impacts of horizontal winds in the dynamic of the intrusion events in the study area. Our methodology combining numerical modeling, in situ data and new machine-learning tools proves effective in improving the capabilities of ocean remote sensing in coastal areas.

Keywords

• altimetry
• coastal circulation
• numerical modeling
• in situ measurements
• machine learning algorithm