8/4/2014 7:18:17 PM
Polar Bears from Space
Polar Bears from Space might conjure up images of a bad 1980s sci-fi flick, but our team—which included the U.S. Geological Survey, the University of Minnesota, the Government of Nunavut, and the Polar Geospatial Center—had something slightly different in mind. Could we use high-resolution satellite imagery to monitor polar bears in some of the most remote regions in the world?
Information about how many bears are found in a particular region and where polar bears are located across the landscape are key pieces of data for wildlife managers and decision-makers. Unfortunately, collecting these data can be quite difficult.
Polar bears are located at low densities across the vast and remote Arctic. Working there—generally from a helicopter—is very expensive and can be quite dangerous. And some areas are simply too inaccessible to study. But given the current climate warming in the Arctic, the need to obtain information about polar bears from the far-flung reaches of the region continues to increase.
Over the past several years, wildlife biologists in the Antarctic studying penguins, seals, and whales have successfully tapped space technology—specifically, high-resolution satellite imagery—as a new research tool. My colleagues and I thought that this type of satellite imagery (think Google Earth) might allow us to overcome some of the obstacles biologists face in studying polar bears.
In August and September, 2012, we set out to test this idea. We conducted our work in northern Foxe Basin, Canada, on Rowley Island, a small, flat island west of Baffin Island. It's a place with large numbers of polar bears during the late summer, ice-free season. Based on our previous work there, we knew that polar bears were the only large, white objects on that landscape capable of moving or changing position. Snow and ice are absent at that time, and caribou are significantly smaller, have different coloration, and are largely absent from the island. These factors made the island an ideal setting for examining if satellite imagery could be a useful tool for monitoring bears.
Our concept was quite simple. With resolutions (pixel sizes) of roughly 0.5 meters, an adult polar bear would be roughly six to eight pixels—a small bright speck on the darker, expansive summertime tundra. We compared two satellite images, collected on different days, to count those white spots that were visible on one image but not the other. We also flew a helicopter survey to make sure we were correctly identifying bears; we needed to be certain that what we thought was a large white rock was indeed a large white rock!
The helicopter survey also allowed us to estimate the size of the island's polar bear population using a more-established technique. We could then compare the population estimate from satellite imagery with the results from the aerial survey to get an idea of how well the satellite method was working.
The results have been very encouraging. We are indeed able to distinguish between polar bears and non-target objects like rocks and light-colored patches on the tundra, and the population estimates based on our satellite imagery method and the aerial survey were very similar.
So, what's next? Well, it took us about 100 hours to review the satellite images manually. That's quite a time investment, given that our study island was only about 1,100 km2—a blip in the expansive Arctic.
It's not realistic for people to review satellite images of hundreds of thousands of square kilometers, so we're trying to figure out how to let computers do some of the work for us. This is still a work in progress, but we are optimistic that automating this process with computer-based image differencing (i.e.,identifying what changes between images) and perhaps incorporating information about the spectral signature of bears will speed up the review process. This is a critical step if we are going to attempt to use satellite imagery in those areas where we know very little about polar bears, such as Russia.
High-resolution satellite imagery doesn't provide the same type of detailed information that we can get from physically capturing bears or from flying aerial surveys. In fact, with the current resolutions, we can't consistently identify polar bear cubs; they are just too small. However, satellite imagery does allow us to access the remotest reaches of the Arctic, without associated safety concerns or disturbance to wildlife, and potentially at significantly reduced costs. These factors, combined with improvements in resolution and continued technological advances, suggest that satellite imagery can provide a useful monitoring tool to complement capture and aerial survey programs in the future.