Tourist Photos From Antarctica May Help Map Penguin Colonies

RACHEL FELTMAN: We’ve got a few minutes left in the hour, so how about we take a quick little trip to Antarctica? If you’re lucky enough to make it down there, you’ll probably aim to snag a classic photo op, a colony of hundreds of penguins set against the barren landscape. Countless versions of that picture taken over the years could turn out to be a valuable source of ecology data.

That’s thanks to a technique described this week in the journal PLOS One. It uses a computer model to pinpoint those images to a specific location in the Antarctic habitat. Joining me now to talk about it is Dr. Heather Lynch. She’s the Institute for Advanced Computational Science Endowed Professor of Ecology and Evolution at Stony Brook University, and she studies penguins in Antarctica. Welcome to Science Friday.

HEATHER LYNCH: Oh, thank you for having me.

RACHEL FELTMAN: So first of all, would you describe your standard penguin colony for us?

HEATHER LYNCH: Oh, sure. I’m not sure there is a standard one, because actually, they range quite a bit in size. So some might only be a couple dozen penguins, and some of them will have hundreds or thousands of penguins nesting there, so it can be quite variable. And you could have just one species there or multiple species in the same location.

RACHEL FELTMAN: This new approach lets you map out the boundaries of a penguin colony. How does it actually work?

HEATHER LYNCH: Sure. So one of the big things that my lab has been focused on is trying to track how populations of penguins are changing over time in Antarctica, but scientists can only be in so many places at once. And yet we actually have a very large Antarctic tour industry, and every passenger to the Antarctic has a camera in their pocket. So we wanted to find a way to harness all of this data that are being collected that otherwise we wouldn’t have access to.

So the idea is that we would want to take a picture of a penguin colony that’s taken from the ground and given no other information about the camera or where the picture was taken to extract the boundary of the penguin colony and to actually place that in three dimensional space to figure out where on the world were the penguins, so that we could compare that with a photo that was maybe taken by someone else three or five or 10 years later, and we could track how the population has changed between those two time periods.

RACHEL FELTMAN: Very cool. So what is this approach give you that, say, a satellite image or pictures from a drone might not provide?

HEATHER LYNCH: Yeah. Well, we are very invested in using both drones and satellites. Both are tremendous sources of information and have really revolutionized the way we study penguin colonies. But the drone imagery that we have in particular is only available from the last few years, and satellite imagery only really goes back maybe 10 or 15 years if we’re talking about the highest resolution imagery that we’re now using.

So what we wanted to do was develop a method that would allow us to go back to historic photos of the Antarctic and locate those– we call it georeferencing. We want to georeference those images so that we could compare the data that we’re collecting now from drones and satellites to those photographs that may have been taken before those other sources of data were available to scientists.

RACHEL FELTMAN: We’re talking about the changes in boundaries of penguin colonies. How much are they really changing in position over time?

HEATHER LYNCH: So one of the things that’s remarkable about penguins is that they are very site faithful, and we know that from comparing some historic air photo data from the earliest days of mapping Antarctica to satellite imagery now, that we know that the population, if there’s the same number of penguins, they really are nesting in roughly the same locations as their parents and grandparents and great grandparents.

But the populations, unfortunately, in Antarctica are changing quite a bit due to climate change and other threats. And so when the population is shrinking or growing, those are the changes that we can track now using photographs in addition to the other sources that we’ve talked about, such as drone imagery or satellites.

RACHEL FELTMAN: And I’m assuming that not every tourist penguin pic will work for this, so what will folks need to make sure is in the image to make sure that it’s going to be useful?

HEATHER LYNCH: So your classic close up photo of an individual penguin is not going to do us a lot of good, because we won’t have enough context to know where that penguin was. And at the same time, if it’s too much of a landscape photo and the penguins are in the deep, deep distance, it’s going to be too far away.

So ideally, what we have is a photograph that shows enough of the colony that we can actually extract its boundary, but also enough of the context, like ridges and coastline– it helps the computer models locate that image in three dimensional space relative to our models of these islands.

RACHEL FELTMAN: And are there any other limits on how well this can work? Does snow impact the data at all?

HEATHER LYNCH: Well, snow is a really interesting problem and one of the reasons why this has been such a challenge in the Antarctic. This idea of creating three dimensional models of locations based on photographs is actually not particularly new, and there’s a word for that. It’s called photo tourism, where people take these ad hoc tourist photos and could reconstruct, say, the city of New York City. But in the Antarctic, we don’t have buildings or other sharp edges that would make it easier for the computers to reconstruct the scene in three dimensions.

And as you mentioned, the snow is constantly changing in the Antarctic, so every photograph taken of the Antarctic, even the same location two days in a row, will look quite different. And so that’s one of the reasons why this was a challenging computer vision problem. How do we accommodate those changes in what the scene looks like? And so we had to rely very much on the fact that the terrain is quite static. The island surface is not changing between days, even if a lot of the things on the surface, like the snow, is changing.

RACHEL FELTMAN: Yeah. And speaking of the starkness of the landscape, could researchers potentially use this in locations where there’s more going on visually? Trees, vegetation?

HEATHER LYNCH: Absolutely. And in some ways, what we’ve done is solved the hard version of the problem, and I think other locations on the planet will be easier than the Antarctic. So, for example, one of the areas where I’m most interested in applying this, or I think it could make the biggest difference, would be in georeferencing photographs of glaciers. So if we think about, well, how are the glaciers of the Alps changing through time?

This kind of approach would allow us to take photographs from climbers through the decades and to georeference them on the Earth so we could see how the glacier had been changing over decades. Or another application would be looking at the location of flowers in a field. If we wanted to know the timing of flowering or how a particular species of flower had been changing over time, we could use photographs taken from people going to say, National Parks, and we could use this same process there.

So it’s quite a general problem that we’ve solved, but it turns out that in places like the Antarctic, where we don’t have a lot of visual cues that would allow us to figure out where a camera was when it took the photograph, that was the computer science challenge that we were trying to solve. And so this allows us to use it not just in the Antarctic, but pretty much everywhere else where you might have a little bit more information about where you are.

RACHEL FELTMAN: I think a lot of people listening to this will be really interested in how their photos today can be useful for research. But could you theoretically go back to a scan from an old textbook or encyclopedia and use that picture to pull useful information?

HEATHER LYNCH: Absolutely. The estimates are now that there are 5 billion pictures taken every day, but people didn’t just start taking photographs. People have been taking photographs of the planet for 100 years. And so one of the challenges that environmental scientists face is trying to match the current conditions to those conditions that existed before all the modern data collection that we have now.

And so these historic photos that might be in a textbook, or it might be in a museum somewhere, they provide this picture of what the planet looked like before the climate change impacts that we’re seeing now. So it’s really looking back at this archival imagery that I think this approach that we’ve developed here would be most useful.

RACHEL FELTMAN: And what kind of research questions could you try to answer using these older photo sources?

HEATHER LYNCH: In the Antarctic in particular, we’re very interested in how the ice environment has changed, both the depth of the snow, but in particular, the retreat of certain glaciers. We can look at volcanic processes that might have changed the shape of the landscape. We can look at the spread of vegetation.

So there was a really nice paper that was published recently by another group that was talking about the greening of Antarctica. Those are exactly the kinds of questions that we can start to ask now, going back to the historic imagery, because we can say, here’s an image from 1967 or 1985. That plant is not in this location, and we know that because we have a photograph that we can place in real space. And we can see that plant is not there 60 years ago, but it is there now.

RACHEL FELTMAN: And getting back to penguins, I know your lab has another approach for tracking their movements, which is looking at their poop from space. Could you tell us a little bit more about that?

HEATHER LYNCH: Sure. So one of the ways that my lab is studying penguins is actually to use satellite imagery, but we’re not using satellite imagery to see individual penguins. What we’re doing is actually looking at the poop stain that these colonies create on the landscape, and that can be seen from 400 miles away up in space, which is pretty remarkable. So a lot of my research effort is focused on using computer science tools, computer vision, and AI to automate the search for penguins in that satellite image and to use it to actually track their abundance through time.

RACHEL FELTMAN: And what is the big goal here? Why is it important to know where penguin colonies are getting to?

HEATHER LYNCH: Well, these penguin populations are changing rapidly, and we know that is due to a variety of factors that include climate change, but also fishing. We have a large Antarctic tourism industry. These are all of the factors that we’re tracking that might be explaining why we’re seeing declines in some species, we’re seeing growth in other species.

The Antarctic Peninsula is one of the most rapidly warming places on the planet, and we really see that reflected in the Antarctic penguins and their distribution and abundance. So we are using them in many ways as the canary in the coal mine for the Antarctic ecosystem, because we know the ecosystem is changing rapidly, and the penguins are giving us a window into that world.

RACHEL FELTMAN: Very cool. Well, that’s all the time we have for now. Dr. Heather Lynch is the Institute for Advanced Computational Science Endowed Professor of Ecology and Evolution at Stony Brook University in Stony Brook, New York. Thanks so much for joining us.

HEATHER LYNCH: Oh, it was my pleasure.

Copyright © 2024 Science Friday Initiative. All rights reserved. Science Friday transcripts are produced on a tight deadline by 3Play Media. Fidelity to the original aired/published audio or video file might vary, and text might be updated or amended in the future. For the authoritative record of Science Friday’s programming, please visit the original aired/published recording. For terms of use and more information, visit our policies pages at http://www.sciencefriday.com/about/policies/