Software seeks to save the whale

Edited by the Engineeringtalk editorial team Nov 5, 2002

A research team working for the International Fund for Animal Welfare is using The MathWorks' software as part of a project to determine the whereabouts of endangered North Atlantic right whales.

A research team working for the International Fund for Animal Welfare (IFAW) is using The MathWorks' software to process real whale and synthetic (whale-like) signals as part of a project to show if the whereabouts of endangered North Atlantic right whales can be determined acoustically.

The ultimate objective of the project is to monitor the whales' seasonal habitat continuously and use information on their presence and movements to reduce collisions with vessels and entanglements in fishing gear.

IFAW's research team plans to have prototype detection software ready within 18 months.

The current North Atlantic right whale population is about 300, and is predicted to become extinct within 200 years at today's mortality rates.

At the moment, the main focus of the project is to examine and compare different ways of detecting right whales, and to distinguish their calls from other ocean sounds, including vocalisations from other species of whale and man-made sounds such as those from seismic exploration guns.

To do this, IFAW is examining different ways of estimating the parameters of a flexible model for whale signals.

It is working closely with the Bioacoustics Laboratory at Cornell University, which has developed archival ocean-bottom recording devices that are being used in the right whale's spring habitat off the east coast of the USA.

Right whales have been recorded in different places and at different times of the year to provide sample signals for analysis.

Matlab is being used to develop both triggers (to detect sound) and classifiers (to identify the sound) and to generate artificial 'training' sound data sets for testing the triggers and classifiers.

The model that is used for the signals has phases that are polynomials or splines.

These are used to generate variable signals that are 'whale-like'.

IFAW is examining several methods of estimating the parameters (ie the polynomial coefficients) of these signals, including maximum-likelihood, polynomial (or smoothing spline) regression, and the 'discrete polynomial transform'.

Once established, the best method will be used to estimate the sounds of real whales.

If the project is successful, there are two ways in which IFAW hopes it will help right whales.

The first is that acoustic detections will provide reliable data about the distribution and movements of whales over different times of the year (at present the only information comes from incidental sightings from boats and from aerial surveys, which rely on daylight and calm weather).

The second, and more ambitious idea, is that a network of buoys could be placed at sea to collect and analyse acoustic data in something like real-time and transmit this information to shore.

In both cases the information would be used to try and reduce the risk to whales from collisions with vessels or entanglement in fishing gear by redirecting these activities away from the whales.

From 1970-1999, 19 (42%) of reliably documented right whale deaths were confirmed as caused by ship collisions or entanglement in fishing gear.

This figure is probably an underestimate of the actual number, as some whales which die offshore, or whose bodies sink, are never recovered.

In addition to these fatalities, there is evidence of many other harmful or potentially fatal incidents.