Automatic detection of oceanographic structures at SHOM: a review

Mesoscale phenomena such as eddies, thermal fronts, or jet currents modify acoustic propagation. Their knowledge is a major issue for (the use of) naval sensors. At SHOM (the French Hydrographic and Oceanographic Office), the Center of Military Oceanography (CMO) aims at providing the naval forces with information about marine environment. In particular, the mesoscale activity is operationally described via two services at CMO, an oceanic prevision model (called SOAP) and a daily reporting of thermal fronts from sea surface temperature (SST) images. Besides during special naval operations over a theater, the oceanographic mesoscale situation is also assessed within an oceanographic bulletin elaborated from model outputs and spatial observations (SST, ocean color and altimetry). In this context, the use of automatic or semi-automatic algorithms for the detection and characterization of mesoscale structures is very helpful for decreasing computation times and obtaining more reliable and objective results (regardless of the operator). Also such techniques could support statistical studies about oceanic phenomena occurring in an area, thus contributing to a better knowledge of oceanic processes. Automatic detection tools can be applied on observation or modelisation data: things are much easier on the latter since numerical models outputs have good properties which observation data have not: continuity and derivability, describing equations, reliable physical, numerical and statistical knowledge of the phenomena (described by the model). This presentation aims at reviewing processing algorithms for the automatic or semi-automatic detection of mesoscale features currently used or simply studied at the CMO over the last decade. Oceanic structures will be divided into 2 categories: the linear structures (such as thermal fronts) or opened contours, and the 2D structures (such as eddies) or closed contours. The algorithms presented were developed either in studies led by SHOM in the late 90's (in the frame of programs for the development of oceanographic systems), or recently within a European project which ended late 2004 (SOFOCCLE). A new European project (SCOOBIDOO) will study further on image processing techniques for automatic detection of mesoscale phenomena on satellite data. The first chapter will present the algorithms implemented current operational applications at SHOM. Then the algorithms studied for detection of linear and 2D structures will be tackled. Finally the results will be discussed and some prospects will be given.