Aller au contenu principal

    NET-BASED VERIFICATION OF ACOUSTIC TECHNIQUES USED TO IDENTIFY ANTARCTIC KRILL

    Demander un document de réunion
    Numéro du document:
    WG-EMM-08/54
    Auteur(s):
    J. Watkins and S. Fielding (United Kingdom)
    Résumé

    We validate the acoustic target classification protocols developed for the Stochastic Distorted-Wave Born Approximation (SDWBA) model using three frequency acoustic data and concurrent net hauls that were collected during two cruises to the South Georgia region in 1996. For each krill aggregation sampled by net we calculated the difference between acoustic backscatter at 120 and 38 kHz (Sv120-38) and at 200 and 120 kHz (Sv200-120). We considered the performance of 4 different acoustic target identification algorithms for krill: (i) ‘3 freq model’ - using the SDWBA to set the acceptance windows for both Sv120-38 and Sv200-120, (ii) ‘2 freq model’ - using the SDWBA to set the acceptance window for just Sv120-38, (iii) ‘2 freq 2-16’ - where the Sv120-38 window was fixed at 2-16 dB and (iv) ‘2 freq 2-12’ - where the Sv120-38 window was fixed at 2-12 dB. The overall aggregation dB difference for 120 – 38 kHz for every net fell within the SDWBA model derived target id window, however, for 200 - 120 kHz the SDWBA model derived target id window only identified krill in 6 of the 16 nets correctly. The ‘2 freq 2-16’ algorithm attributed more than 90 % of the total backscatter to krill in all but 1 aggregation with the ‘2 freq model’ using a smaller window but still attributing more than 90 % of the total backscatter to krill in 12 out of the 16 nets. The ‘2 freq 2 – 12’ window only attributed more than 90 % of the total backscatter to krill in 6 nets while the ‘3 freq model’ attributed only just greater than 50 % of the backscatter to krill in only 2 aggregations, and in 6 aggregations attributed less than 10 % of the total backscatter to krill. Therefore the SDWBA(11,4) using 38, 120 and 200 kHz in the present configuration to set variably sized windows is likely to substantially underestimate krill. In contrast the SDWBA(11,4) used at 38 and 120 kHz identifies very well the krill detected during these net hauls and because it uses a window substantially smaller than the fixed 2-16 window, will at the same time reduce the amount of bycatch that may occur when targets other than Antarctic krill are present in the water column.