With a need to act fast in priority landscapes, Ducks Unlimited Canada (DUC) is developing artificial intelligence to help map and classify wetlands in the Prairies.
Canada is home to approximately 25 percent of the world’s wetlands, yet we risk losing these habitats more quickly than we can measure. Wetlands are the Earth’s most threatened ecosystems, vanishing globally at a rate three times faster than forests.
To address this urgency, Ducks Unlimited Canada (DUC) is using artificial intelligence (AI) and machine learning to map some of the country’s most at-risk wetlands and quickly pinpoint conservation needs before it’s too late.
In a project supported by RBC Tech for Nature , the non-profit conservation organization is mapping roughly 30,000 square kilometres of priority habitat in Alberta that may be under consideration for development or agriculture.
“Our work with RBC right now is focusing on high-priority waterfowl areas that have never been mapped appropriately in detail,” says Lyle Boychuk, DUC’s manager of geographic information systems (GIS) and inventory programs for the Prairie Region, who oversees the project. These areas were identified as hot spots for breeding waterfowl based on more than 30 years of research and modelling, he says, including areas around Buffalo Lake, Pine Lake, Eastern Plains, Eastern Irrigation District and Calgary.
Traditional, manual mapping methods were deemed either too costly or too slow to effectively update maps at such a large scale.
Instead, DUC remote-sensing specialists have developed new processes to automate the most time-consuming part of the work. They’re applying machine learning to analyze images and classify wetlands, combining multiple current data from satellites with three-dimensional LiDAR observations (which measure the Earth’s surface using light in the form of a pulsed laser).
Summer satellite image captured in late July in pothole landscape in Eastern Plains of Alberta.
Canada’s wetlands are diverse and dynamic. They can be categorized as marshes, bogs, fens, swamps or open water, in bodies large and small.
Not only are wetlands critical habitats for wildlife, they’re also being increasingly valued as a natural buffer against the effects of climate change . They filter water and help mitigate flooding. They also absorb and store large amounts of carbon in their soils.
The target landscapes for DUC’s project in Alberta are full of what geographers call “potholes”: the shallow depressions formed by glaciers 12,000 years ago. Today, many are disappearing as demand for arable land grows.
“The Prairies are characterized by very, very small wetlands—and there are literally millions of them. There is either an abundance or a scarcity of water, depending on the year or season. We go through cycles of dry–wet, wet-dry, and the wetlands change; their vegetation changes. Some appear for weeks or months, or may be completely dry for extended periods in a drought scenario,” says Boychuk.
To assess accurately, the conditions of these Prairie wetlands must be viewed over different time periods and in three dimensions. DUC has proved that a machine learning process could be used to analyze massive amounts of data and imagery from multiple sources and fuse it into a reliable representation of wetlands on the ground.
“What the algorithm is really good at is establishing a baseline, or a ‘normal.’ It can incorporate multi-temporal imagery, from different years and different times of year. And we can do that consistently—much more than you could by interpreting a single aerial photo,” says Boychuk.
The work underway in Alberta will have continental significance, says Boychuk.
These wetlands are part of North America’s “Prairie Pothole Region,” known to be an essential habitat for waterfowl and migrating birds as well as an estimated one-third of all bird species at risk in Canada.
“Obviously, we target waterfowl. That’s our game. But there’s a lot of benefit to our work, of course, other than just growing ducks,” he says.
“We’re building data that will go into the public domain and will have value for others, as well as inform our own programs. It’s ultimately contributing to continental waterfowl goals, and biodiversity benefits,” says Boychuk.
No one really knows the current condition of Canada’s wetlands—there isn’t a comprehensive national inventory or monitoring system that keeps track (although DUC is actively advocating for this ).
DUC started using aerial photography and satellite imagery in the 1980s to interpret and classify millions of acres of wetland to support conservation programs.
In 2002, DUC—in collaboration with Environment Canada, the Canadian Space Agency and the North American Wetlands Conservation Council—built a database, the Canadian Wetland Inventory (CWI), with the support of 150-plus conservation partners.
The CWI database is only partially completed but continues to grow project by project. (You can view progress on the ducks.ca website.)
The information is used to focus conservation efforts where they’re most needed, assess changes in wetlands, and provide scientific data to governments as well as conservation and industry groups to develop their land-use policies and protocols.
Boychuk, who began his career as a computer programmer in the 1990s, says there has never been a more exciting time to work in this field.
DUC and its partners in the Prairies and across Canada are applying leading-edge tools and technology to guide conservation programs, from mobile GIS applications for field work and by citizen scientists, to flying drones to design wetland restoration projects.
“We’re doing things that, four or five years ago, I wouldn’t say were ‘unthinkable,’ but they would’ve been unreachable because of the cost. It was as if habitat wasn’t seen as a big enough social priority to warrant the expenditures,” Boychuk says.
There is a lot of enthusiasm for developing AI and machine learning as tools to solve environmental challenges, but the data itself is just half the solution. People still need to care about wetlands.
“Of course, there’s more to this than just algorithms. You have to know a lot about wetlands to make these models work. So, there’s our technical knowledge, but also, we have our corporate knowledge and access to our science and wetland experts. We have operational people to help on the ground.”
“We’re bringing a lot of experience, over and above the technology,” says Boychuk.
Committed to a better future. Learn how RBC Wealth Management supports the communities where we live and work.