It turns out it's possible to count polar bears accurately by identifying them in satellite images, according to U.S. and Canadian researchers, led by Seth Stapleton, a postdoctoral researcher at the University of Minnesota.
Stapleton and his collaborators tested the technique on Rowley Island in Nunavut, then compared their satellite counts to counts made from a helicopter over the same period. The two counts were "remarkably similar," they reported in the journal PLOS ONE earlier this month.
"I'm very happy with the results," said Nicolas Lecomte, who worked on the study when he was a Nunavut government biologist. "Seth and myself, we were really surprised it worked so well."
"It's just a taste, but it looks very promising," said Dirkus Gissing, director for wildlife management for the government of Nunavut's Department of Environment, which was one of groups collaborating on the study.
Gissing said accurate estimates of the polar bear population are very important, since the species is hunted by humans.
"Knowing exactly how many there are gives us information about how many we can harvest on an annual basis at a sustainable level."
Concerns over existing methods
But monitoring polar bears isn't easy – many live in remote areas that are logistically difficult and expensive to access by air. Noisy, low-flying aircraft used to count the bears may also disturb them and other wildlife.
Gissing said Inuit communities have also raised concerns about methods used to keep track of polar bears (such as tagging) that involve close encounters between bears and humans, which aren't ideal for either species.
In many places, these challenges mean that counting is done very infrequently.
The new method has a lot of advantages, Gissing said: "It's cheaper, you can do it more frequently… You don't have to handle the animals."
Satellite imagery has previously been used to count seals, penguins and whales. Stapleton thought it might be worthwhile to give it a try for polar bears. He chose Rowley Island as a test site because it is flat, uninhabited by humans, and has a high density of polar bears.
The researchers obtained satellite imagery from the company Digital Globe, and compared images from two different days in August, when the snow had all melted.
"Really, what we were looking for are white spots on the [brown] landscape that changed between the two sets of images," Stapleton said in a Skype interview.
The researchers reasoned that bears would move, while white rocks and gravel patches that might otherwise be mistaken for polar bears would stay in the same place. The resolution of the images was about 0.5 metres, and each full-grown polar bear took up roughly 6 to 10 pixels.
The counting part wasn't as easy as it sounds, as Rowley Island is 1,100 square kilometres, and Stapleton and his University of Minnesota colleague Michelle LaRue did it manually.
"I think between Michelle and myself we spent combined about 100 hours staring at a computer screen looking for those little while dots that changed between images," he recalled.
Automatic counting needed
Because Rowley Island is small compared to polar bears' entire range, expanding the use of the technique would require computer automated counting. Stapleton said he's exploring that possibility, which involves subtracting the two images from each other to bring out changes.
The satellite count was compared to those of two separate observers counting independently in a helicopter as it flew at a constant speed and height along parallel lines several kilometres apart. Those counts were adjusted to account for the fact that bears are more likely to be visible closer to the helicopter than farther away.
The preparations for the study illustrates just how challenging such aerial surveys can be. The researchers had to cache enough fuel on the island to finish the survey, with the help of a helicopter and Twin Otter plane, recalled Lecomte, who is now a professor at the University of Moncton in New Brunswick.
At one point, Lecomte recalled, "the pilot was stuck deep in mud with the twin otter and there were roughly 100 bears around them."
It took a week to dig the plane out, but no one was hurt.
Aerial surveys cost $1,500 per hour
Besides the logistical issues, aerial surveys are very expensive. Lecomte estimated they cost about $1,500 per hour. Meanwhile, he satellite images run closer to $5 per square kilometre, meaning that the entire island could be surveyed by satellite for less than the cost of a one-hour aerial survey, if counting were automated and no one had to be paid to do the counting.
However, there are disadvantages to satellite counting too, Lecomte said. At the moment, aerial surveys can provide additional information such as the sex and age of the bears, while satellite counting can't. That may soon change – U.S. regulations recently changed to allow higher resolution satellite imaging, and DigitalGlobe will soon launch a satellite capable of resolutions of 0.25 metres.
The satellite counting technique also hasn't yet been shown to work on ice or steeper terrain, although the researchers are gearing up to test it in those conditions.
Ultimately, Lecomte envisions the two methods complementing each other, with satellite counts being done frequently and aerial surveys augmenting those counts every few years. And he thinks the technique can be applied to more than just polar bears.
"I'm very hopeful we can expand that to other species in the north — seals, caribou, muskox — there's a lot of hope for that."