Categorising salmon migration behaviour using characteristics of split-beam acoustic data

Abstract

Milling behaviour is problematic when using hydroacoustics to estimate the number of migrating fish in rivers. Milling behaviour was observed for adult sockeye salmon migrating upstream in the Wannock River, to their spawning grounds in the tributaries of Owikeno Lake in the central coast area of British Columbia, Canada. We classified the acoustic salmon tracks to separate the milling fish from the actively migrating fish in an attempt to obtain an estimate of sockeye salmon flux as they migrate to their spawning grounds. We used discriminant function analysis and found that three variables measured on each track were sufficient for the classification of the acoustic tracks into milling and non-milling categories with an approximate classification accuracy of 98%. The method we present can also be used to separate tracks of targets of interest from noise or debris tracks that occur in the acoustic data, if discrete track characteristics are ascertained.