Autonmous Real-time Marine Mammal Detections

Woods Hole Oceanographic Institution

Chukchi Sea, Arctic, Late Summer 2013


Slocum glider we167 (deployed Sep 10, 2013, recovered Sep 19, 2013)

Study objectives

A single glider was deployed in the Chukchi Sea off the northwest coast of Alaska to study the occurrence of several species of marine mammals, including fin, humpback, bowhead, right, beluga, and killer whales, as well as bearded seals and walrus. This project is primarily designed to test the capabilities of the autonomous marine mammal detection and classification system for sub-Arctic and Arctic species, but we are also interested in examining relationships between marine mammal distribution and oceanographic conditions monitored by the glider. This is our first deployment in the Arctic with the DMON/LFDCS, and while the technical details of collecting and delivering the DMON/LFDCS data are worked out, interpretation of the results is not. Specifically, the call types for each species should be considered preliminary, as we have not yet been able to assess false detection rates (one of the primary goals of this study). Therefore, the summary classification data shown on this website should be viewed skeptically, and interpretation of the results should be done with caution. A discussion of the real-time results by Mark Baumgartner and Kate Stafford is included in the study chronology below to provide some context and interpretation based on our knowledge of the detection system and of Arctic acoustic ecology.

Principal Investigators: Peter Winsor (University of Alaska Fairbanks), Kate Stafford (University of Washington), Mark Baumgartner (WHOI) and Dave Fratantoni (WHOI)

Collaborator: Janet Clarke (Aerial Surveys of Arctic Marine Mammals, BOEM/NMFS)


Please email Mark Baumgartner at For a general desciption of the detection system and the autonomous platforms, visit

Study chronology
Sep 10
A single glider (we167) was deployed today by Dr. Peter Winsor (UAF) in the Chukchi Sea west of Wainwright, Alaska. Within an hour of the deployment, the glider began to report a unique pattern of detections: pitch tracks that were stacked on top of one another in frequency (see example). This pattern of pitch tracks can be produced by broadband sounds, such as noises made by the glider. However, the DMON/LFDCS has an algorithm to remove broadband sounds prior to pitch tracking. We think it is more likely that these are biologically produced sounds; in particular, these may be "bell" sounds produced by walrus. The sounds are being classified as walrus sounds by the DMON/LFDCS, and having access to the pitch tracks provides us with additional information to evaluate the sounds. In addition to detecting sounds in real time, the DMON/LFDCS is also continuously recording audio so that we will be able to review these sounds after the glider is recovered to determine with much greater certainty whether these sounds are indeed walrus "bell" calls.

Sep 11
We observed the same stacked pattern of pitch tracks that my be indicative of walrus "bell" sounds today (see example), although very late in the evening today the total number of these detections decreased significantly. We also observed a very interesting 200-100 Hz downsweep that we do not have in our current call library (see examples between 00:10:05 and 00:10:40); we unfortunately do not know which species produces this sound. Interestingly, upsweep calls very similar to right whale upcalls were observed by the glider during the afternoon today (see examples at 16:15:28 and 16:15:42). The glider also observed some low-frequency moans that are being classified as humpback whale calls (call type 9; see example).

Sep 12
After some analysis today, we have determined that the putative walrus "bell" calls are triggering numerous false detections in two humpback whale call types (call types 7 and 11) and a low-frequency killer whale call type (call type 34). We have removed these call types from the plots and summary tallies reported on this website. Prior to deployment of the glider, we knew that some call types were constructed very generally, and as such, were susceptible to high false detection rates. The website was also upgraded today to include diagnostic information, as well as a plot of background noise reported every hour by the glider as a 1-minute averaged spectrum. There has been high background noise at very low frequencies since the deployment, while a 100-1500 Hz band of noise is likely caused by the calls we are interpreting as the walrus "bell" calls. Our colloborator, Peter Winsor, was at sea in the vicinity of the glider yesterday (Sep 11), and he reported clear skies and calm seas, so the low-frequency noise is unlikely to be caused by precipitation, wind or waves.

The putative walrus "bell" calls returned around midnight, but then subided shortly thereafter (this can be seen in both the summary histogram of walrus calls as well as the background noise plot). The day has been characterized by low calling rates, with only a few narrowband calls showing up in the pitch tracks. There are a few interesting upsweeps reminiscent of right whale upcalls (see example 1 at 06:10:54 and 06:11:30 and example 2 at 08:19:28 and 08:24:12). While it certainly would be exciting if these were North Pacific right whale upcalls, it is more likely that these are humpback or bowhead calls. In fact, there were a few bouts of sporadic narrowband calling that are quite indicative of humpback whale presence (see example 1 and example 2). Also of note, we have finally observed some higher frequency calling behavior likely produced by killer whales around mid-day (see example at 11:15:03 and 11:15:58).

There have been numerous fin whale detections since the glider was deployed. At this point, we do not know if these detections are real or false. Fin whales often call in distinct repetitive patterns, making a 20-Hz pulse every 10-15 seconds. We have not observed those patterns in the pitch tracks reported by the glider. However, there is a substantial amount of background noise present at very low frequencies which could be triggering false detections if the noise varies over short time scales. It is also possible that the background noise at low frequencies is caused by incessant fin whale calling, and the system cannot pick out individual pulses regularly enough to show us a repetitive calling pattern. If this were the case, we will expect to observe some temporal variability in the background noise, as the fin whales change their calling behavior or the glider moves away from the animals - we have not observed this kind of temporal variability. Another possibility is that the low-frequency noise is produced by flow noise over the DMON hydrophone; however, we have detected fin whale calls from gliders before (see our recent Outer Fall results), so we think this is unlikely.

Sep 13
The putative walrus "bell" calls were numerous around midnight and the early morning hours, but were reduced significantly by noon. They appeared again in the afternoon and remained constant for the remainder of the day. These calls continue to be detected as stacked pitch tracks (see example) - we are extremely eager to get the audio back when the glider is recovered to investigate these sounds! Fin whale 20-Hz call detections continued at the same modest rate observed since deployment. The fact that there is very little temporal variability in the detection rates suggests that these may be false detections. There were fewer humpback detections today than in previous days with a few tonal calls detected (see example). We also may have heard a beluga whistle or two (see example at 04:13:10). The higher frequency whistles of odontocetes like killer whales or belugas do not travel as far in water as the low frequency calls of fins, humpbacks, and bowheads, so the glider needs to be relative close (likely within 1-2 km) to detect these vocalizations.

Sep 14
Except for a short puase just before noon, the putative walrus "bell" calls were numerous all day today. There were also higher frequency calls detected sporadically that were suggestive of beluga whistles (see example from 20:14:15 to 20:14:26). We learned today that the residents of Barrow will begin fall whaling very shortly. Since the glider is near Barrow, we will need to retrieve the glider so that our science operations do not affect the hunt. While we anticipate that the glider would have no impact on bowhead whales, the presence of the R/V Norseman II could. We will recover the glider in the next day or two.

Sep 15
The rate of putative walrus "bell" calls decreased overnight and were comparatively quite low at sunrise. We observed some higher frequency calls suggestive of beluga whistles in the early morning hours (see example at 00:17:16 and 00:18:04). Some very interesting, patterned low-frequency downsweeps appeared in the night (see example), and again just after sunrise (see examples 1, 2, and 3). While its tempting to identify these as whale calls, its probably more likely that these are being produced by air guns (i.e., they are loud man-made sounds used for oil and gas exploration). The residents of Barrow began fall whaling today and they landed two bowheads east of Barrow. Because the hunt has begun, Peter Winsor and his team on the R/V Norseman II recovered the glider just after noon today. Peter will reposition the ship to an area well to the southwest of Barrow closer to Wainwright and redeploy the glider. Final recovery is planned for Thursday, 3 days from now.

Sep 16
Sea conditions were too rough for a safe redeployment of the glider today, so Peter will try again tomorrow morning.

Sep 17
Conditions improved today, and the glider was re-deployed at 15:20 local time. The stacked pitch tracks returned shortly after the glider was deployed and they were received at a fairly constant rate throughout the rest of the day. Aside from a few higher frequency squeaks (see example), no other sounds were detected in great abundance.

Sep 18
More of the putative walrus "bell" calls (stacked pitch tracks) were detected throughout the day today. Some interesting higher frequency sounds were also detected (see example 1 at 11:01:23 and 11:01:53 and example 2 at 20:04:10).

Sep 19
Despite less than ideal sea conditions, the glider was safely recovered today at 9:50 am local time. We are glad to have it back in good shape, and are eager to examine all of the recorded audio.


At sea assistance was provided by the captain and crew of the R/V Norseman II. Support for the development of the Arctic marine mammal call library and preparation of the DMON/LFDCS for this study was provided by the Alaska Ocean Observing System. Glider deployment was supported by the Bureau of Ocean Energy Management. The DMON instrument was developed by Mark Johnson and Tom Hurst at WHOI. Mark Johnson was responsible for developing the application programming interface (API) for the DMON, and coded the initial DMON implementation of the pitch tracking algorithm described in Baumgartner and Mussoline (2011). Support for the development, integration, and testing of the glider DMON/LFDCS was provided by the Office of Naval Research and the NOAA National Marine Fisheries Service Advanced Sampling Technologies Working Group in collaboration with the Northeast Fisheries Science Center's Passive Acoustics Research Group (leader: Sofie Van Parijs). NOAA funding was provided through the Cooperative Institute for the North Atlantic Region.