Comparison and evaluation of methods for distinguishing among several functional classes of zooplankton and fish using acoustic backscatter data at four frequencies

Zooplankton net samples data were collected at 30 locations near the Pribilof Islands, Alaska in September 1999 using a MOCNESS. Simultaneous volume backscatter data were collected with an HTI echosounder system at: 43, 120, 200, and 420 kHz. Three algorithms were compared for their ability to classify acoustic data into the dominant zooplankton functional/size groups (euphausiids, copepods, and ostra cods) and separating them from fish. The algorithms tested include: (1) combining morphological image processing and image differences to identify patches in different size ranges, and using the forward problem calibrated to plankton found in MOCNESS hauls to ascribe size/functional groups to the different patches; (2) application of canonical correlation between acoustics (backscatter volume and target strengths) and plankton biomass captured by the MOCNESS; and (3) application of inverse techniques to estimate the number of individuals in set size classes using multiple theoretical models of backscatter volume compared to plankton biomass captured by a net. Knowledge of the strengths and weaknesses of each technique allows us to better interpret broad-scale acoustic survey results from the same cruise, thus providing a synoptic view of the fish and their plankton prey which are in the size and density range to be acoustically detected.