Sammendrag
Acoustic methods may oversample mesopelagic fishes due to uncertainties in resonance, the proportion of backscatter due to siphonophores and population characteristics. On the other hand, nets may under sample due to trawl avoidance and poor retention of small individuals. In the work presented, a stereo camera system (Deep Vision) mounted in the extension of a pelagic trawl was used in combination with a self-contained wide-band echosounder (SIMRAD WBAT) in the trawl’s opening as a tool to compare catch and acoustic estimates and improve understanding of the mesopelagic ecosystem. Estimates from the WBAT (counting acoustic targets) and the Deep Vision (counting the organisms in the pictures) show similar patterns in vertical distribution and densities over time for the mesopelagic organisms encountered, providing an enhanced sampling method for mesopelagic species by verifying acoustic targets in situ and establish their vertical distribution. Images and the trawl’s path through the water column from the Deep Vision system can also be imported directly into the LSSS (Large Scale Survey System) software to verify acoustic scatterers in echograms. Finally, machine learning methods are being developed and tested to automate species identification and length measurement from the Deep Vision stereo images.
Vis fullstendig beskrivelse