Abstract
Species identification is “the grand challenge” (MacLennan and Holliday, 1996) for acoustic methods used to estimate fish abundance (Simmonds and MacLennan, 2005). Acoustic survey methods are continuously improved to increase the accuracy of acoustic classification and thereby reduce the uncertainty of abundance estimates.
Most commonly, single-frequency acoustic data are classified using echogram features and biological samples. First, the echogram data are scrutinized (analysed, corrected, and classified) e.g. by checking for errors, removing noise, thresholding, and setting analysis depth layers. Then, target features are delineated by lines, rectangles, or polygons and ascribed to species using expert knowledge resulting from relevant biological and oceanographic samples. In most surveys, the aim is to identify echoes from one or two species, with other echoes considered less important. Acoustic target classification can be improved by using multifrequency data and exploiting its inherent information.
The target audiences for this report are:
- users who conduct surveys and derive abundance estimates;
- those who understand what can and cannot be done using existing and modified processing tools, but may not be familiar with the theory underlying acoustic target classification methods;
- developers who use and modify existing tools and develop advanced tools;
- those with advanced theoretical knowledge.