Suggestions of Acoustic Monitoring Equipment for a Whale Sanctuary

Question

There are a few sanctuaries being setup to care for cetaceans that were in harsh captivity settings. One is in Iceland and another to be in Nova Scotia. I am not currently related to any sanctuary project. I am aware that some locations are not ethical sanctuaries and my questions pertains to the ethical and well managed ones. The one in Iceland that I know of, as well as the coming one in Nova Scotia, appear to be well managed. And there is a great need for many more well managed sanctuaries as the multiple cetaceans in captivity need to be protected and given better living conditions.

Acoustic monitoring equipment are important for ensuring the health and safety of sanctuary residents.

In order to support excellent new cetacean sanctuaries, I am taking the initiative to ask this community to suggest acoustic monitoring equipment. I am hopeful that these suggestions will eventually reach the right people. The suggestions could included generic equipment descriptions and could be more detailed, with specific models. Costs should be minimized. Off-the-shelf equipment may be preferable, as opposed to customized. Ease of use is another factor to consider. The typical people using the suggestion may have an academic background in biology, or no academic background, and would most probably not be bioacousticians.

The main purpose of the monitoring equipment is to acquire vocalization patterns that can be used eventually for automatically detecting health issues in the residents, such as unusual stress situations, infections, other disorders. Upon anomalous health situation detection, the apparatus can notify caretakers (e.g., via email or text) and present some relevant data. The caretakers may be remote or sleeping. Once alerted, the caretakers are expected to do the rest of the investigation with other tools. The apparatus could also emit a short acoustic pattern to notify the cetacean residents that caretakers have been notified. Eventually such notifications to cetaceans can easily be learned without much efforts by anyone.

There may be two somewhat different processes and therefore sets of equipment involved in this kind of project.

1. The identification of the acoustic signals associated with a potential health or security issue.
2. The automated recognition of the signals identified in item 1. This process includes that automated notification to caretakers.

These two processes may take place continuously over some time and in parallel with each other.

I invite participants in this discussion to propose improvements to the overall components, as well as to describe actual implementations that can be existing or that may require to be developed.

Given that the goal is using acoustic monitoring equipment for ensuring the health and safety of sanctuary residents, consider the question to be:

"How would you do it?"

These are the sanctuaries that are probably ethical based on my judgement (certainty varies from 75% to 90%):

1. Umah Lumba Center, in Bali, Indonesia; in operation; maybe 3 Tursiops: https://www.dolphinproject.com/campaigns/indonesia-campaign/bali-sanctuary/
2. Beluga Whale Sanctuary, Iceland; in operation; 2 belugas; room for many more: https://belugasanctuary.sealifetrust.org/en/
3. Whale Sanctuary, Nova Scotia, Canada; construction to start in 2022; 2 belugas identified; possibly a few orcas to be added, separated: https://whalesanctuaryproject.org/

Related publication: https://www.researchgate.net/publication/343779074_Introducing_NMMF_WAMS_an_open-source_PAMGuard_plug-in_and_some_pilot_data_for_its_use_as_a_welfare_acoustic_monitoring_system/fulltext/5f3f20b8299bf13404d8ba43/Introducing-NMMF-WAMS-an-open-source-PAMGuard-plug-in-and-some-pilot-data-for-its-use-as-a-welfare-acoustic-monitoring-system.pdf

Answer 1

Here's is how I would do it, as a first draft. Many members of this exchange can do a better job than I, no doubt.

1. For the identification of health and security issue signals:

• 1.1. review literature, ask knowledgeable community members and outsiders;
• 1.2. collect signals from current residents and select candidate signals; collect additional occurrence data on these signals;
• 1.3. convert candidate signals in a form usable by the recognition system;
• 1.4. software to be evaluated: PAMGuard, and/or DC Dolphin Communicator paired with recording software, others TBD
• 1.5. hardware to be evaluated: low cost laptop or tablet, low-cost hydrophone (high quality recordings are not required)

2. For the real-time signal recognition and notification system:

• 2.1. evaluate DC Dolphin Communicator on tablet with low-cost hydrophone (as used in the identification system in item 1), DC should be able almost as-is to recognize issue signals and notify by email; DC is at https://play.google.com/store/apps/details?id=sm.app.dc (disclosure: I wrote it)
• 2.2. if the option to notify the residents is enabled then we'll need an inexpensive projecting transducer (underwater speaker), and maybe it could be in air, or both in water and air;
• 2.3. evaluate PAMGuard (on laptop): How hard is it to have it recognize signals? Can it send notification email when an issue signal is recognized? Probably can be modified to do that;
• 2.4. are there others components to be evaluated?

The equipment part is relatively easy to define.

The hard parts are:

1. The identification of the issue signals.
2. The encoding of the selected issues signals into a form used by the recognition system, for example, if an artificial neural net (aka. machine learning) is attempted, then a large amount of examples of the issue signals would be required to train the ML and obtain a valid model. A sufficiently large amount of such issue signals may not be available or may be only obtainable over a long period of time which would hinder the benefits of the system. For this reason, a machine learning approach may not be applicable, but this remains to be confirmed.