Counting Whales

As it turns out, migration patterns of fin and blue whales throughout the world’s oceans are still not fully understood.  Which makes our seismic dataset really interesting – we have continuous data collected at various locations over several years.  Here’s an image of the locations where we currently have data:

Ocean Bottom Seismometer locations. Blue triangles are KECK instruments, and red circles are NEPTUNE Canada instruments.

The site that has been studied most extensively (by our lab group, at least) has been the Endeavour site, which is at the Endeavour segment of the Juan de Fuca Ridge. At this location, there are 8 seismometers. Dax has worked on this data for the last couple of years, and has resolved over a hundred fin whale track! Because of course, if you measure the same call on at least 3 sensors, you can find a location (if you fix the depth, which is what has been done so far).

One of Dax's fin whale tracks, showing the whale travelling through the network over several hours.

One of the things I’ve been working on is developing an automatic detection algorithm for fin whale calls in our seismic data. Since we have such a large amount of data, it’s not very efficient to count whales manually – there are hundreds of thousands of calls! Dax developed a supervised detection algorithm, so I started with that as a baseline, and eventually implemented and compared a couple of different automated methods: matched filtering and spectrogram correlation.