Artificial intelligence, or AI has quickly become part of everyday life between virtual assistants like Apple’s Siri and generative AI like ChatGPT. It’s even used in a variety of research settings.
Sea otters are a big part of the Alaskan marine ecosystem. The population was decimated from the fur trade in the 20th century but scientists estimate that number is going back up. There’s a lot of unknowns about how those numbers are changing though.
One of the main ways to count sea otters is through aerial surveys, where trained biologists look for and count sea otters up in a low flying plane. Christina Bonsell is a marine ecologist with the Bureau of Ocean Energy Management, or BOEM, a federal agency. She said the current way of counting sea otters has its challenges.
“This method, if you have really trained staff, is pretty effective, but you know, you have to get people into planes, which is expensive, flying at low altitudes has safety concerns,” she said. “so we're really looking for ways to improve on that method.”
But there’s a new method to count sea otters that makes studying them cheaper and safer for scientists. Since 2020 Bonsell has coordinated a study with the National Parks Service, U.S. Fish and Wildlife Service and U.S. Geological Survey that is using AI to help count sea otters in lower Cook Inlet.
This new method flies a plane on set paths called transects while a camera continuously takes photos during the flight. Those photos are later analyzed by AI software called See Otter that’s trained to identify otters. Humans then verify those counts and help scientists get a better idea of how many otters are in an area, as well as where they can be found.
Bonsell said the research allows them to better understand how and why sea otter populations are changing in the region.
“Sea otter populations in Cook Inlet are changing. We're seeing increases in the number of sea otters in Cook Inlet. And it's sort of still unknown whether that's an increase in the population that's there or whether that's migration and movement of otters,” Bonsell said.
Nicole LaRoche is a biologist with the U.S. Geological Survey who worked on the study. She said using the new system has many benefits.
“It's exciting to to get a little bit more safety when we're doing these overwater flights, because, you know, Cook Inlet is about 60 miles wide at the widest part,” she said, “and that's a lot of open water to fly in a tiny, single engine plane at 300 [feet].”
This software’s been used to do whale surveys in the past, and is redesigned to look for sea otters. LaRoche said this new method can streamline the process to identify sea otters. Scientists usually need a lot of training to find otters during surveys, which can be really difficult to spot.
“If they're wrapped up in kelp, a lot of times, it's hard to see them and or if they're near rocks or other things that could look like sea otters from shore, like logs and birds and several other things,” LaRoche said, “it's really great to have the computer machine learning side of it help us out with saving us a lot of time looking at all of those images.”
Benjamin Weitzman is a wildlife biologist with the U.S. Fish and Wildlife Service. He’s done similar research in Southeast Alaska and helped design the paths planes would take in this study. He said that while the software has many benefits and is constantly improving, it still has its limitations.
“While we've confirmed that photo surveys work, and that AI systems are capable, we've also confirmed that they cannot be trusted on their own,” Weitzman said.
He said that developing the software also requires skills that many biologists don’t have, and that the pool of people to work with is quite small.
“At a federal level, I see the agencies starting to realize that this is a whole new expertise and skillset that we don't yet have the capacity for,” he said, “so how do we build that capacity? And how do we find partners that have that capacity? Because I think these technologies are extremely promising, but we need the right minds working with them.”
As the study wraps up, the team will get a report to BOEM by the end of the year before presenting their findings to the science community.
