Acoustic classification of coexisting taxa in a coastal ecosystem

Classifying coexisting taxa in a coastal ecosystem remains an analytic challenge due to the difficulty in verifying species compositions within backscatter data. Multifrequency measurements (38, 70, 120, 200 kHz) were combined with midwater trawls and zooplankton MultiNet tows in Hood Canal, WA, to classify backscatter dominated by single fish species (Pacific Herring, Pacific Hake) or major zooplankton taxa (euphausiids, copepods). Backscatter was categorized into aggregations, single targets, and layers based on morphology. Aggregations and single targets were identified in raw volume backscattering strength (Sv), while layers were classified using differences in mean volume backscattering strength (ΔMVBSi-j = MVBSi - MVBSj, where i and j denote frequency in kHz). Based on a subset of trawl-validated in situ acoustic measurements, backscatter with −16 dB < ΔMVBS200-38 ≤ 2 dB were classified as fish, and 2 dB < ΔMVBS200-38 < 30 dB as zooplankton. Backscatter identified as fish were further classified to hake when ΔMVBS120-38 < −4.8 dB, and herring when ΔMVBS120-38 ≥ −4.8 dB. The classification method was evaluated using a second set of trawl-validated acoustic data, resulting in classification accuracy of fish or zooplankton ranging from 95% to 100%. At the species level, misclassifications of herring and hake were both ∼13%. Removal of aggregations and single targets before calculating ΔMVBS values minimized the possibility of mixed species backscatter within layers. This classification technique provides an approach to separate coexisting aggregations of dominant taxa which are common in mid- and low-latitude coastal ecosystems.