CCAMLR

Understanding the equipment used and the data handling capability of the krill fishing vessels is important background information in considering the implementation  of potential data collection protocols for use by the fishery. 

The Scientific Committee has requested SG-ASAM to provide advice on the collection and processing of acoustic data from fishing vessels.  Some krill fishing vessels are equipped with scanning or multi-beam sonars.  Here I highlight a recently published analysis technique - an adaptation of distance sampling theory - that provides a general means for estimating krill density and biomass from vertically oriented multi-beam echo-sounder data.  I describe an extension to this approach that may, subject to the constraints of krill target strength models, acoustic instrument stability, and survey design allow the use of data observed with horizontally oriented multi-beam and scanning sonars to contribute to krill abundance estimation.

Single frequency (38 and 120 kHz) acoustic and associated net data were extracted from US AMLR datasets and used to mimic data that might be collected by the commercial fishery during the fishing seasons. From these data we develop semi-empirical indices of acoustic biomass in a Generalized Linear Modeling framework. We correlate these estimates of biomass with acoustic estimates derived from the 3-frequency method using the CCAMLR protocol for the period between 1996 and 2011 in both the Elephant Island and West Shelf Areas of the US AMLR survey grid. Cross-validation of acoustic estimates between these two areas and the 3-frequency biomass showed that models developed using a wide length frequency distribution (Elephant Island) could be used to estimate biomass from other areas where the length frequency of animals is skewed towards larger animals. We show that it is possible to develop semi-empirical models of krill biomass, at120 kHz frequency that can be used to augment research acoustic surveys if proper survey design and calibration of transducers is maintained and if time series are sufficiently long to average out differences among years.