Echo level segmentation on echo-integration of fisheries acoustics data
Résumé
In fisheries acoustics the analysis of data usually often concern biomass assessment mainly for small pelagic fish stocks using the well-known echointergration approach. Other can concern the analysis of single fish using their target strength (TS in dB) and more seldom analysis can also be done with the fish school descriptors using e.g. shoal extraction method (Movies+, Ifremer Software). In the framework of the Preface project we have focused on the micronektonic layers observed by scientific echosounder. Matecho, a friendly automatized processing method to extract information and perform echo-integration, fish shoal extraction and also performs a segmentation, on each zone of a cruise with a constant twilight, of the echointegrated echogram from an echo level threshold fixed by user to extract micronektonic layers in the water column. Here we describe this methodology which allows an accurate description of the spatial organisation and structuration of the marine ecosystem. The process is based on three main steps which consist in : (i) adjust the echo level threshold in dB, (ii) the extraction of the echoes inside each contours and the calculation of the layer descriptors, (iii) and then the correction of the extraction. Finally the echo segmentation, setup to extract micronektonic sound scattered layer, allows to get 34 layers descriptors, e.g., minimum/maximum depth (m), geographical position in 3D, maximum depth width (m), duration of the layer, surface covered by the layer, mean volume backscattering strength 'Sv' (dB re 1 m-1)': mean nautical area scattered coefficient 'Sa' (or NASC m2 nmi-2), to characterise their spatial position in the water column and acoustics properties. Moreover, a second class of descriptors, classified by elementary sampling unit (ESU), are estimated e.g. number of layer per ESU, layer depth per ESU. An innovative descriptor is also computed using this methodological approach: the water column fulling rate per layer and per ESU. Both classes of descriptors are then available for ecological studies.