Google’s New Bio-Acoustics Machine Learning Project could Save Salish Sea Killer Whales
Fisheries and Oceans Canada (DFO) Partners with Google AI to Save Marine Mammals in the Salish Sea Killer Whales
AI and Machine learning can be used to do many things — but letting them track the whales– Well, that’s a great initiative. Let’s find out what’s happening with AI ML engineering to track orcas.
Salish Sea orcas are among the most endangered species in the Canadian marine ecosystem. If you were asked to find a solution to save the 73 orcas that remain in their natural habitat, you may give up even before trying. But, not Google AI team! Google AI, in sync with Rainforest Connection (RFCx), is partnering with the Fisheries and Oceans Canada (DFO) to not only protect the killer whales in the Salish Sea, but also study their various other life-threatening conditions using bioacoustics monitoring tools.
Rainforest Connection, which is a grantee of the ‘Google AI Impact Challenge’, has built a mobile app and a web interface to generate live alerts on the whale movement.
Why the Salish Orcas Face Extinction?
The Salish Sea orcas are nature’s delightful creatures found only in the oceans near Vancouver and Washington State’s San Juan Islands and British Columbia’s Gulf Islands. These whales are genetically separated from their sub-group cousins from the Canadian Pacific.
Despite the passing of Regulations Amending the Marine Mammal Regulations: SOR/2018-126, harassment of marine mammals continues.
While poaching and whaling continue in the deep seas where these orcas live, another major factor leading to their possible extinction is its social and genetic isolation from other species in the group.
Read Also: Using AI to Map Marine Environments
Man-made threats are also hampering orcas’ natural habitat and their free movement in the seas has been hugely disturbing.
For example, Resident Killer Whales are living in 1500x times more polluted waters today than in the 1980s. These are polluted with plastics, contaminants, and other chemicals.
In addition, sonar disturbances also lead to acoustic and physical disturbance. Ship strikes have been reported to kill or maim these beautiful orcas. In short, the Southern Resident Killer Whale population is much likely to be extinct by 2022, knowing climate changes are getting more real and catastrophic than in the last 5 years.
Adding to the already tough climatic patterns, these orcas are known to be part of a complex social structure which results in late sexual maturity and low birth rate.
How AI and Machine learning-based Bioacoustics Applications Come to Rescue
As part of Google’s intimate AI for Social Good campaign, the Google AI team decided to apply machine learning applications to protect these Salmon Whales or Chinook Whales, as they are locally known.
AI and Machine Learning applications are used to clearly identify the exact population density of the orcas in the Canadian Seas. This information would be matched to ascertain the impact of natural and man-made factors on their population. For example, geospatial data and water analysis data can be used to find the degree of contaminants in the water, or find the level of scarcity of food, particularly Chinook salmons, and to find out the disturbing activities like motor boating, and off-shore drilling.
In partnership with Google AI, DFO and Rainforest Connection use deep neural networks to track, monitor and observe the orcas’ behavior in the Salish Sea. This highly-classified information is then sent to the Canadian authorities.
Mammal Managers, as they are called within the community, study the information to monitor, track and treat whales that are injured, sick or distressed.
For example, the mammal managers use an AI-based oil spill detection system to accurately locate the orcas and their prey. They use a piece of “specialized equipment to alter the direction of travel of the orcas to prevent exposure.”
Google AI team closely worked with the DFO to build a machine learning that recognizes orca sounds and other acoustic models to track this species. DFO provided close to 2000 hours of sub-sea audio with 68,000 labels to identify the exact origin of the sound. The machine learning system now analyzes live sounds based on the DFO data across 12 locations known to be core SR killer Whales’ habitat.
The AI-based mobile app provides live movement alerts on the whales’ location. It is still working on a way to distinguish the three sub-groups in the Canadian orcas- Southern Resident Killer Whales, Northern Resident Killer Whales and Biggs Killer Whales. The exact distinction would enable ‘mammal managers’ to better monitor SR Killer Whales and study their health and breeding patterns.
Organizations working to save the Killer Whales: wildwhales.org, DFP, the Government of Canada, Google, and RFCx.
(To share your insights, please write to us at firstname.lastname@example.org)