Evaluation of some processing techniques applied to acoustic recordings from two krill fishing vessels

Acoustic data from krill fishing vessels have the potential to provide useful contributions to assess local krill abundance and distribution, and even as input to the CCAMLR feedback management system under development. The possibility to collect such data from the fleet has been demonstrated through a CCAMLR proof of concept, but processing is in most cases essential before any use of the data can be made. Procedures for data processing have not yet been addressed in detail within CCAMLR. Here we show examples of processing with a prerequisite that abundance and distribution of krill are our main interest, and that data of interest include both those collected during fishing operations and during surveying. We have worked with datasets covering most of the fishing season 2015/2016, one from the Chinese vessel Fu Rong Hai and one from the Norwegian vessel Saga Sea, both transect data and data collected during fishing operation. Spike noise and background noise were the most severe source of unwanted backscatter, but could be dealt with adequately using different noise removal techniques. Other sources of noise like integration of bottom echo or false bottom noise could not be removed using the automatic noise removal techniques we had at hand. We also evaluated the use of delineation and integration of only detected swarms for assessment of abundance, and the results are promising both for swarm integration as a robust noise excluding technique, and as a krill identification technique in case data have been collected without one or more of the CCAMLR prescribed three frequencies for target identification.