|dc.description.abstract||Kaikoura, New Zealand is one of a few places worldwide where sperm whales can be found near-shore year-round. However, diving depths and underwater movements of the sperm whales here have remained largely a mystery due to technological limitations, lengthy dive durations, and deep bathymetry typical of dive locations. The overall goal of this research was to quantify sperm whale acoustic and diving behaviour in Kaikoura, in order to better understand the ecology of this globally distributed species. Multihydrophone recordings and emerging signal processing techniques were used to obtain information on whale growth, vocal behaviour, underwater movements, and foraging habits.
Software written to compute acoustic size estimates formed the basis of a new nonlethal method for measuring growth of resident sperm whales. Growth rates were measured for 32 whales that had been recorded repeatedly between 1991 and 2009. All whales with multiple recordings more than six months apart showed an increase in size, and the largest changes were present in whales recorded over longer timespans. To investigate underwater movements and vocal behaviour, a passive sonar array was designed and built for tracking diving sperm whales in three dimensions. The system consisted of four free-floating buoys, each with a hydrophone, depth sensor, built-in recorder, and GPS, and one vertical stereo hydrophone array deployed from a 6 m boat. Recordings were synchronised using a GPS timing pulse modulated onto one track of each recorder.
Passive acoustic detection and localisation algorithms were developed and applied to recordings to obtain diving profiles. Detection algorithms, which included two different energy detectors both with and without a matched filter, all performed well showing few false positives at low signal-to-noise ratios. Closely spaced hydrophones on the vertical array were used to estimate bearings to individual whales, and these bearings formed tracks that could distinguish a single whale when multiple whales were vocalising. Rhythm analysis was used to match vocalisations among recording platforms, and cross-correlation was used to compute time of arrival differences among matched vocalisations. Sensitivity analysis based on hyperbolic localisation was used to obtain probability distributions for the whale’s 3D location. These localisations were compared to those computed by a particle filter algorithm, isodiachronic sequential bound estimation, which had higher accuracy.
Location data consisted of 3D probability clouds, which differ from 2D point-locations typically obtained in animal tracking studies. An analysis method based on a speed-filter and smoothing spline interpolation was used to convert 3D localisation clouds into regularly sampled 3D tracks. While diving, whales in Kaikoura had a mean swimming speed of 1.43 m/s, and, on average, dived to a depth of 432 m. Tracking whales during creak vocalisations revealed that whales foraged throughout the water column from sea surface to sea floor, but 67% of all creaks occurred between depths of 275 - 725 m. Dive characteristics from 78 interpolated 3D tracks from Kaikoura were broadly similar to those computed from one- and two-dimensional tracks of sperm whales studied in other oceans.||