Localizing North Atlantic Right Whales using a deformable sonobuoy grid

In 2018, a large-scale data collection effort was conducted in the Gulf of St. Lawrence, a known feeding ground for North Atlantic Right Whales (NARW), over two days. On each day, visual surveys were conducted, 32 sonobuoys were deployed to gather directional acoustic time series, and a Slocum glider operated in the area to collect oceanographic data. Following the collection phase, the acoustic data were manually annotated with a focus on NARW vocalizations. This project uses the multi-modal dataset to test and compare the results of three localization algorithms for NARW calls. The first method of localization used the directionality of the calls, where probability density maps were created by overlapping the bearing statistics across each of the relevant sonobuoys, commonly referred to as cross-fixing. To improve the bearing distributions, the signals were isolated with thresholding after using a conditional whitener and power-law statistic on its short-time Fourier transform. In the next approach, localization was performed using a spherical interpolation method to initialize a maximum likelihood time-difference-of-arrival algorithm. Finally, matched-field processing was used to model replica fields at potential source locations and correlate the replicas with the received pressure on the hydrophones.