David K. Mellinger

Listening for Large Whales in the Offshore Waters of Alaska

Abstract In 1999, the first phase of a multiyear program was initiated at the National Oceanic and Atmospheric Administrations National Marine Mammal Laboratory and Pacific Marine Environmental Laboratory to advance the use of passive acoustics for the detection and assessment of large whales in offshore Alaskan waters. To date, autonomous recorders have been successfully deployed in the Gulf of Alaska (1999–2001), the southeastern Bering Sea (2000–present), and the western Beaufort Sea (2003–2004). Seasonal occurrences of six endangered species (blue, fin, humpback, North Pacific right, bowhead, and sperm whales) have been documented on the basis of call receptions in these remote ocean regions. In addition, eastern North Pacific gray whale calls were detected in the western Beaufort Sea from October 2003 through May 2004. Here we provide an overview of this suite of research projects and suggest the next steps for applying acoustic data from long-term recorders to the assessment of large whale populations.

Seasonal variability and detection range modeling of baleen whale calls in the Gulf of Alaska, 1999–2002

Five species of large whales, including the blue (Balaenoptera musculus), fin (B. physalus), sei (B. borealis), humpback (Megaptera novaeangliae), and North Pacific right (Eubalaena japonica), were the target of commercial harvests in the Gulf of Alaska (GoA) during the 19th through mid-20th Centuries. Since this time, there have been a few summer time visual surveys for these species, but no overview of year-round use of these waters by endangered whales primarily because standard visual survey data are difficult and costly. From October 1999-May 2002, moored hydrophones were deployed in six locations in the GoA to record whale calls. Reception of calls from fin, humpback, and blue whales and an unknown source, called Watkins whale, showed seasonal and geographic variation. Calls were detected more often during the winter than during the summer, suggesting that animals inhabit the GoA year-round. To estimate the distance at which species-diagnostic calls could be heard, parabolic equation propagation loss models for frequencies characteristic of each of each call type were run. Maximum detection ranges in the subarctic North Pacific ranged from 45 to 250 km among three species (fin, humpback, blue), although modeled detection ranges varied greatly with input parameters and choice of ambient noise level.

PAMGUARD: SEMIAUTOMATED, OPEN SOURCE SOFTWARE FOR REAL-TIME ACOUSTIC DETECTION AND LOCALISATION OF CETACEANS

PAMGUARD is open‐source, platform‐independent software to address the needs of developers and users of Passive Acoustic Monitoring (PAM) systems. For the PAM operator—marine mammal biologist, manager, or mitigator—PAMGUARD provides a flexible and easy‐to‐use suite of detection, localization, data management, and display modules. These provide a standard interface across different platforms with the flexibility to allow multiple detectors to be added, removed, and configured according to the species of interest and the hardware configuration on a particular project. For developers of PAM systems, an Application Programming Interface (API) has been developed which contains standard classes for the efficient handling of many types of data, interfaces to acquisition hardware and to databases, and a GUI framework for data display. PAMGUARD replicates and exceeds the capabilities of earlier real time monitoring programs such as the IFAW Logger Suite and Ishmael. Ongoing developments include improved real‐time l...

Cetacean population density estimation from single fixed sensors using passive acoustics

Passive acoustic methods are increasingly being used to estimate animal population density. Most density estimation methods are based on estimates of the probability of detecting calls as functions of distance. Typically these are obtained using receivers capable of localizing calls or from studies of tagged animals. However, both approaches are expensive to implement. The approach described here uses a MonteCarlo model to estimate the probability of detecting calls from single sensors. The passive sonar equation is used to predict signal-to-noise ratios (SNRs) of received clicks, which are then combined with a detector characterization that predicts probability of detection as a function of SNR. Input distributions for source level, beam pattern, and whale depth are obtained from the literature. Acoustic propagation modeling is used to estimate transmission loss. Other inputs for density estimation are call rate, obtained from the literature, and false positive rate, obtained from manual analysis of a data sample. The method is applied to estimate density of Blainvilles beaked whales over a 6-day period around a single hydrophone located in the Tongue of the Ocean, Bahamas. Results are consistent with those from previous analyses, which use additional tag data.

Classification of echolocation clicks from odontocetes in the Southern California Bight

This study presents a system for classifying echolocation clicks of six species of odontocetes in the Southern California Bight: Visually confirmed bottlenose dolphins, short- and long-beaked common dolphins, Pacific white-sided dolphins, Rissos dolphins, and presumed Cuviers beaked whales. Echolocation clicks are represented by cepstral feature vectors that are classified by Gaussian mixture models. A randomized cross-validation experiment is designed to provide conditions similar to those found in a field-deployed system. To prevent matched conditions from inappropriately lowering the error rate, echolocation clicks associated with a single sighting are never split across the training and test data. Sightings are randomly permuted before assignment to folds in the experiment. This allows different combinations of the training and test data to be used while keeping data from each sighting entirely in the training or test set. The system achieves a mean error rate of 22% across 100 randomized three-fold cross-validation experiments. Four of the six species had mean error rates lower than the overall mean, with the presumed Cuviers beaked whale clicks showing the best performance (<2% error rate). Long-beaked common and bottlenose dolphins proved the most difficult to classify, with mean error rates of 53% and 68%, respectively.

A method for detecting whistles, moans, and other frequency contour sounds

An algorithm is presented for the detection of frequency contour sounds-whistles of dolphins and many other odontocetes, moans of baleen whales, chirps of birds, and numerous other animal and non-animal sounds. The algorithm works by tracking spectral peaks over time, grouping together peaks in successive time slices in a spectrogram if the peaks are sufficiently near in frequency and form a smooth contour over time. The algorithm has nine parameters, including the ones needed for spectrogram calculation and normalization. Finding optimal values for all of these parameters simultaneously requires a search of parameter space, and a grid search technique is described. The frequency contour detection method and parameter optimization technique are applied to the problem of detecting boing sounds of minke whales from near Hawaii. The test data set contained many humpback whale sounds in the frequency range of interest. Detection performance is quantified, and the method is found to work well at detecting boings, with a false-detection rate of 3% for the target missed-call rate of 25%. It has also worked well anecdotally for other marine and some terrestrial species, and could be applied to any species that produces a frequency contour, or to non-animal sounds as well.

Patterns in the vocalizations of male harbor seals

Comparative analyses of the roar vocalization of male harbor seals from ten sites throughout their distribution showed that vocal variation occurs at the oceanic, regional, population, and subpopulation level. Genetic barriers based on the physical distance between harbor seal populations present a likely explanation for some of the observed vocal variation. However, site-specific vocal variations were present between genetically mixed subpopulations in California. A tree-based classification analysis grouped Scottish populations together with eastern Pacific sites, rather than amongst Atlantic sites as would be expected if variation was based purely on genetics. Lastly, within the classification tree no individual vocal parameter was consistently responsible for consecutive splits between geographic sites. Combined, these factors suggest that site-specific variation influences the development of vocal structure in harbor seals and these factors may provide evidence for the occurrence of vocal dialects.