Google is using AI to Save Whales.
The state of Washington and Southern Vancouver is home to a distinct population of killer whales, namely, southern resident Orcas. Discrete black and contrasting white eye patches, each whale stands out from the other. This is usually the sight of serenity for tourists, as whales flock together often mating or hunting for food.
However, off late, these lives have been endangered. This was captured a few summers ago as Tahlequah; a killer whale was found carrying her dead calf, surprisingly so for over a 1000 miles (1,600km).
According to research from the University of Washington, failures to reproduce are linked to nutrition and access to Chinook salmon — which have been in a dramatic decline over recent years.
These animals are often photographed and ID’ed throughout the year, with the hope to better study and understand their behaviour.
Researches have learned that their food supply has been replenishing over the years as more and more salmons disappear due to increased traffic of marine vessels.
Near the coast of Pacific, a team of researchers are studying the coexistence between humans and orcas.
They believe; what happened to Tahlequah wasn’t an anomaly. These whales have been listed as endangered. Back in 2005, they were around 88. Now there are just 72.
From Olympia Washington in the south to the middle of Vancouver Island in the north, much of it is flooded with evergreen wildlife.
Pacific Northwest has been getting crowded lately. As of 2020, Washington’s population was 8 million whereas that of Vancouver topped out at about 2.5 million and growing.
The threat?
Increased ship traffic and recreational boat activity, has proved to be one of the biggest risks that the Orcas face here.
Ships not only pollute the water, but they also create an immensely noisy atmosphere for the thriving marine life.
More the number of boats, more the likelihood of whales getting hit by them. Lesser so for slow-moving ships. If they are hit at around a speed of 10 knots, there are higher chances of survival.
Regulators did initiate slow down directives but what appears to have outweighed such measures are the business conflicts.
As for the ships, a slow or delayed arrival means higher fees and lost revenue.
In theory, it is believed that even slow-moving ships manage to cause acoustic smog that drowns out the Orca’s ability to communicate through echolocation. This is because the sound of the ship is at the same high frequency as that of the Orca’s. The increased sound disturbance for the Orca is equivalent to losing a seat at the dinner table because they are now not able to find food, leading to far-reaching impacts that directly affects their population.
Reduced vocalization hinders communication that in turn affects mating, increasing in stress hormones in whales.
Solution?
- Avoiding routes where these Orcas exist or slowing down near Orca territories. But it isn’t as easy as it sounds. It is often hard for ship Captains to confirm the position of them, especially in severe conditions like Fog and rain.
- Underwater Hydrophones are often used for recording or listening to underwater sound. Most hydrophones are based on a piezoelectric transducer that generates an electric potential when subjected to a pressure change, such as a sound wave.
- Just the way it is hard for the Orca’s to communicate amidst all the chaos, similarly so for someone who is listening for whale sounds, recorded overnight, amongst other marine life.
A smarter solution, maybe then?
The answer lies in the southern Indonesian island of Borneo, Indonesia
The Rainforest Connection, a non-profit organization, is using Big data to save critically endangered animals in the forest.
A weak line of sight, not more than 20 feet in front of you, has forced researchers to rely upon sound even in these ecosystems to understand the behaviour of animals. Walking through the noisy forests bearing the sounds of chainsaws and other human activities has proven to be a significant inefficiency while studying forest wildlife.
Through smart computers and help of an AI model, they managed to track the noise coming from mobile phones perched in trees, picking up sounds coming from chain-saws or any other potential human threat to the ecosystem.
Matt Harvey, based in California, who has been working with his team at Google AI came across this model. He decided to incorporate it to study a data set that included a large set of sound recordings involving different whale species in the Pacific Ocean.
Harvey reached out to Canada’s Department of Fishery and Ocean (DFO) to detect southern resident Orca’s, with the same model used at Rainforest Connection.
The challenge wasn’t to incorporate the model for the alerting system but to collect and run the model on the data. So he designed an AI model that would auto-detect and decipher southern Orca calls from ocean sounds.
To help the model recognize the Orca, the team used audio samples from the Canadian Hydrophones that contained 1800 hours of Orca sounds.
The hydrophones were originally installed to better understand the behaviour of Orca’s in terms of eating, mating and their life cycle. But well before, Harvey, DFO was already listening to different Orca sounds, trying to find a solution, only to realize that it was extremely time consuming.
The system works such that the AI autodetects and programs the lifestream hydrophones and when a mammal is spotted, there is an email alert as well as a dedicated mobile app notification.
Harvey’s model turns audio signals into images of time and frequency, instead of height and width and the brightness of the pixels represent the level of sound.
For instance — Humpbacks have more of a low pitched, upsweeping ‘whoop’ sound and as for Orca’s its more of a squeal.
Following this, captains are alerted on a closed radio network beforehand to change course or slow down in order to avoid the whales and reduce underwater noise.
As of July 28, 2020, Scientists in Washington state made the discovery while recording drone images of Southern Resident killer whales. They said that whales from three pods were pregnant, including Tahlequah.
As advancements in technology gear up, a better-abled system such as Autonomous underwater vehicles or computer vision software is being used to study marine animals and plants and lead to a major increase in data availability.
All this would help speed up study and analysis but not to a level where it stands as a complete replacement for humans.
