Edge of Innovation

Citizen Scientists and AI Help Save the Whales

Conservationists use data collecting technologies and artificial intelligence to turn whale watching into scientific research, gathering photos of whale encounters from citizen scientists to help protect the mammals.

After centuries of whaling and habitat destruction, saving whales is one of modern science’s great ambitions. Researchers are turning to technology like artificial intelligence (AI), cloud computing and crowd sourced data to better understand and protect these magnificent, mysterious creatures.

Increasingly, research biologists use digital technologies to help track, document and analyze animal populations and their migrations. And they’re not alone in watching over vulnerable whale pods. The technology even enables weekend whale watchers to become citizen scientists in whale tracking.

“Many whale species are poorly understood,” said John Calambokidis, a senior research biologist at Cascadia Research, which studies marine mammal and bird biology. “There are whales swimming around in the ocean that do not yet have a scientific name that are being discovered.”

Calambokidis first started studying cetaceans in the 1980s, long before digital technologies were introduced. Today, he’s working with software engineers at Wildbook who are using AI to study whales and document their struggle for survival.

He said killer whales around the San Juan Islands are probably the best studied whale population in the world, but they’re vulnerable. The black and white orcas swimming in the Puget Sound near Washington state have been listed as endangered since 2005. Named the Southern Residents, the whale pod population dropped in recent years to only 78 whales, according to Calambokidis. One reason is a diminishing food supply of salmon.

“We’re at a critical moment, with the health of the oceans and the survival of so many species at stake,” Calambokidis said. “Technology is opening new worlds of understanding about whales, and we can put these tools to work for the benefit of the whales.”

Using AI to ID Whales

Advances in AI are making computer vision and photographic identification of whales easier. Now crowd sourcing websites like Happywhale are helping identify and track whales from photos submitted by citizen scientists.

Algorithms can analyze photos and identify individual animals based on images of fins or parts of the whale visible above water. This allows researchers to identify individual whales as well as their movement, population and trends.

“These algorithms use pattern recognition of the markings on the underside of flukes as well as contours riding on the trailing edge,” said Calambokidis, whose organization sponsors Happywhale. “The combination is very effective.”

whale tales
Using pattern recognition, an individual whale can be identified by its tail fluke from three photos taken years apart. Photo courtesy of Cascadia Research.

Happywhale allows people to make meaningful contributions by sharing their whale photos, but Calambokidis warns against the rise in whale paparazzi.

“While a boat getting close may not sound serious, we’ve seen 50 to 100 boats following the resident orcas near the San Juan Islands,” he said.

“It’s not a good situation. People used to have SLRs with telephoto lenses when they went whale watching. Now, boats get closer to the whales because smartphones aren’t great at magnification.”

The latest algorithms from the Wildbook project make lower quality photographs more useful, which encourages people to take pictures from farther way.

Changing Role of Citizen Scientists

The idea of being noninvasive with marine animals was the inspiration for Wildbook, a platform using an image analysis server based on deep convolutional neural networks. Like a social media profile, it manages the identities of individual whales by collecting photos, information about where and when the whale was seen, and its behavior at the time.

“The system is trained to detect animals in pictures and then identify individual animals,” said Jason Holmberg, the project’s information architect.

But Wildbook takes this task to a mind-bending level. It not only relies on the photos of contributing citizen scientists, it also acts as a bot, searching the internet for user-generated content on whales to add to the database.

Holmberg’s first subject when developing the technology was whale sharks. The Wildbook AI agent downloads every YouTube video titled or tagged whale shark. And through natural language processing, determines information such as when and where the animal was sighted.

“The agent will extract the high quality key frames of a whale shark and send them to the image analysis server. It’s working beautifully,” he said.

Southern Resident Killer Whale
An endangered Southern Resident whale breaches near Henry Island in the Puget Sound.

The AI collects far more images and data than contributors alone can submit. Wildbook said that’s freeing researchers from tasks like curating images, estimating population sizes and ingesting more data. Instead, biologists can engage more people through whale education.

“If the AI grabs a post on YouTube from someone swimming and diving with whale sharks in the Maldives, we’ve even created the post back automation so that we can circle back to them: ‘Hey, we saw whale shark XM-1195 in your video.’”

Where natural language processing fails, Wildbook can ask people for more data.

“It’s changing the concept of a citizen scientist,” he added.

Whale Trail for Coastline Watchers

While Wildbook serves as a sort of Facebook for animals, The Whale Trail is being developed as a digital guided tour of whales from the beach. The organization has erected a series of whale watching signs along the west coast of North America and is developing an app to create an educational experience that works in concert with each sign.

Currently under development, the app will share information (as does Wildbook) with Conserve.iO, an organization that collects wildlife research data.

“The signs are just the starting point,” said Donna Sandstrom, founder of The Whale Trail. “With a phone or tablet you could hear stories about the waters you’re looking at, learn what whales were recently sighted there, or contribute your own pictures and stories.”

“For the average person there’s a lot you can see from shore,” she added. “I’ve been standing here at Alki Beach (located in southwest Seattle) and suddenly a pod of orcas just swam by,” said Sandstrom.

Resident orcas are fish eaters that live in a tightly bonded family group organized around the mother. The range of Southern Resident orcas stretches from the southern end of Vancouver Island to Northern California. Between October and February, they return to Puget Sound near Seattle, following salmon runs.

Girl reading whale trail sign.
The Whale Trail features educational signs along the Pacific Coast. Photo credit: Jason Lopez.

“Lime Kiln State Park on San Juan Island is one of the best places in the world to watch whales,” Sandstrom said. It was part of what inspired the former Adobe software project manager to start The Whale Trail.

Partnering with scientific organizations such as NOAA Fisheries and the Seattle Aquarium, the organization has installed more than 90 interpretive whale watching signs on coastal locations in Washington, Oregon, California and British Columbia.

She said the signs are informative and send an important message about the future of orcas living in Puget Sound.

“Maybe people living in west Seattle will see the signs and then pull weeds in their garden rather than use herbicides, because those toxins end up in runoff and harm the whales and other creatures,” she said.

Saving wildlife requires a public effort to help supply marine researchers with more data on whales. The contributions of citizen scientists  —  those who knowingly submit their data and others who unknowingly share their vacation photos and videos on social media for smart bots to gather — are saving whales .

“Big data suddenly scales for wildlife research,” Holmberg said.

 

iQ newsletter signup

Share This Article

Read Full Story