Exploitation of large calibrated single beam echosounder datasets using a Bayesian geoacoustic inversion

Many governments use single beam echosounders (SBES) to assess fish populations on their continental shelves. While these data typically catalogue the seafloor, it is often overlooked for seafloor characterization. A calibrated EK60 scientific SBES dataset from the National Center for Environmental Information (NCEI) is one such example. The geographic extent of this dataset is immense, and therefore the potential for exploitation is profound. To this end, a geoacoustic inversion is performed on a subset of these data which produces posterior probability distributions of seafloor sediment characteristics including density, porosity, and sound speed ratio, mean grain size, and descriptions of seafloor roughness and volume scattering. The parallel tempered, Metropolis-Hastings, Markov-Chain Monte Carlo Bayesian inversion algorithm is coupled with an established seafloor backscatter model and augmented with a depth–density gradient reflection coefficient model and improved empirical sediment relationships. Using co-located marine sediment samples from the US Geological Survey and University of New Hampshire, the inverted mean grain size results are shown to agree with ground truth samples for sandy sediments, while predicting a finer mean grain size than ground truth samples for muddy sediments. Ongoing efforts for continental-shelf-scale sediment characterization are discussed. [Work sponsored by ONR N000142312819]