Geotags:

CanadaBritainNunatsiavutNunavut

A novel predictive AI to save Canadian Arctic caribou

A study combines satellite data, indigenous knowledge, and algorithms to anticipate migrations and protect dolphins in Nunavut.

Caribou: Northern populations that cross the sea ice during their annual migrations, a symbol of Arctic biodiversity, put at risk by melting ice, unstable routes and increasing human activity.
A herd of barren-ground caribou crosses the snowy landscape of Wapusk National Park, Manitoba: their long migrations mark the natural rhythms of the Great North, but the future of these herds appears increasingly uncertain due to habitat loss and extreme climate changes.
(Photo: Peter Ewins/WWF Canada)

Every autumn, along the southern coast of Victoria Island in the Canadian Arctic, thousands of Dolphin-Union caribou gather, waiting for the sea to freeze again. Only then, having reached a solid, firm surface, do they begin their journey to the mainland of Nunavut, where they spend the winter.

For centuries, this natural cycle has ensured the survival of a species that is fundamental not only to the ecological balance, but also to the native Inuit and Inuvialuit communities, who have always relied on caribou for food, materials, and cultural continuity.

Today, however, this delicate ritual is under siege. Global warming, occurring four times faster in the Arctic than elsewhere, has made seasonal ice more unstable and delayed. Caribou, forced to wait longer on the coasts, lose precious energy, and when they attempt the crossing, they often encounter fragile ice sheets that give way under their weight. No less dangerous is the threat posed by icebreakers, which breach channels and can interrupt migratory routes at crucial moments.

The consequences are serious. According to the Committee on the Status of Endangered Wildlife in Canada (COSEWIC), the Union dolphin population has plummeted from an estimated 34.000 individuals in 1997 to just under 4.000 in 2020, a decline of approximately 89 percent in just over twenty years. In 2017, the species was formally classified as "endangered." This dramatic decline has prompted scientists and local organizations to seek new solutions that combine scientific rigor and traditional knowledge.

Caribou: A migratory Arctic species that depends on seasonal sea ice to move between islands and the continent, now threatened by climate change, delayed freezing, and new environmental pressures.
The autumn migration of Dolphin and Union caribou between Victoria Island and the mainland represents one of the most delicate passages for the survival of the species: the delay in the formation of sea ice makes this journey increasingly uncertain, increasing the risk of mortality during the seasonal crossing.
(Graph: modified from Poole et al. (2010), edited by B. Fournier, GNWT-ENR 2016)

IceNet, a predictive model for sea ice

The answer came from an international team led by the British Antarctic Survey, in collaboration with the Alan Turing Institute, WWF, and the Government of Nunavut. The result is IceNet, an artificial intelligence model capable of predicting sea ice formation and concentration up to three months in advance. The study, published in May 2025 in the journal Ecological Solutions and Evidence by Ellen Bowler and colleagues, demonstrated that the algorithm can provide an early warning system regarding the most likely times for caribou crossings.

The first author stressed the urgency of the research, explaining that

Many Arctic animals have life cycles deeply intertwined with the seasonal changes in the ice, which is increasingly unstable today. The combination of fragile pack ice and increasing naval activity risks having fatal consequences for caribou. By combining AI forecasts and historical data, we can anticipate the timing of migration, giving local communities the ability to intervene at critical moments.

The principle is simple yet revolutionary: linking historical caribou movement data, collected via satellite collars from 1996 to 2019, with ice state observations provided by the OSI-SAF and AMSR2 satellites. From this correlation, the researchers identified a key threshold: caribou begin to move when the ice concentration (SIC) averages 98,8 percent. Based on this, IceNet was trained to generate probabilistic maps capable of indicating when and where the ice will become suitable for migration.

Caribou: resilient animals of the Arctic ecosystem that follow ancient migratory patterns, whose survival depends on accurate climate forecasts, habitat protection, and reduced human impact on the ice.
A male barren-ground caribou from the Qamanirjuaq herd emerges powerfully from the ice of Nunavut: this subspecies, fundamental to Inuit culture, today faces the challenges posed by climate change, which alters ice cycles and puts the delicate balance of Arctic ecosystems at risk.
(Photo: John E. Marriott/WWF Canada)

Science, local communities and shared management

A key aspect of this project is its multi-stakeholder nature. It's not just an academic exercise: researchers involved regional biologists, territorial governments, hunting and fishing organizations, software engineers, and ice experts. Local Inuit and Inuvialuit populations were actively involved, providing field observations, suggestions, and ongoing feedback.

As early as 2019, workshops organized by the Victoria Island Waterways Safety Committee highlighted the need to manage navigation-related risks. This led to the creation of the Notice to Mariners (NOTMAR) system, which requires vessels transiting the channels to report their passage and comply with recommendations such as reducing speed or avoiding the creation of multiple parallel ice cracks. Integrating IceNet into this communication system would make these measures even more effective, transforming simple warnings into a truly predictive tool.

As Scott Hosking, mission director for Environmental Forecasting at the Alan Turing Institute, recalled,

“The Arctic is warming at a staggering rate, and only by combining new technologies with local knowledge can we reduce the risks to animals and communities.”

This approach, which combines predictive models and daily observations of hunters, represents a concrete example of participatory science.

Caribou: resilient animals of the Arctic ecosystem that follow ancient migratory patterns, whose survival depends on accurate climate forecasts, habitat protection, and reduced human impact on the ice.
A small, solitary cabin stands on a frozen lake on the southern coast of Victoria Island, surrounded by tundra and ice ridges: structures of this kind bear witness to human presence in an extreme territory, where survival depends on millennia-old knowledge of the Arctic environment.
(Photo: Terry Milton/Government of Nunavut)

Data, results and protection prospects

The study of Ellen Bowler provided key numbers that help understand the model's effectiveness. Analyses show that IceNet's forecasts have an average error of about 5,5 percent in ice concentration, compared to 12 percent for the physical model SEAS5, which until recently was considered one of the most reliable. This means that, at least in some areas and periods, artificial intelligence has already outperformed traditional climate models.

The most interesting aspect is the system's ability to produce probabilistic maps, with which experts can choose different confidence intervals (for example, 10%-90% of the herd moving). This approach does not replace human judgment, but rather supports it with dynamic tools. For this very reason, the authors emphasize that the system must remain "human-in-the-loop": a technology available to communities, not an automatic system imposed from above.

There are numerous development opportunities. Researchers are already working to increase the model's spatial resolution by integrating variables such as ice thickness, weather conditions, wind, and snow. Another area of ​​focus concerns spring migration: so far, IceNet has focused on autumn, but the risks associated with early ice melt could also compromise the return of caribou to Victoria Island for the calving season.

Caribou: Northern populations that cross the sea ice during their annual migrations, a symbol of Arctic biodiversity, put at risk by melting ice, unstable routes and increasing human activity.
A pair of caribou advance across the snow cover, following ancestral paths that connect the Arctic grazing areas: these seasonal movements, essential for the survival of the species, depend on the formation of sea ice and the availability of lichens, their main food source in winter.

From the North Pole a signal for the entire world

The value of this research extends beyond Nunavut's borders. The Arctic is a global laboratory for climate change: what happens to the Dolphin-Union caribou is a preview of processes that will affect other species and ecosystems. IceNet, if further developed, could be used to predict the arrival of polar bears near coastal communities, manage Arctic whale routes, or prevent incidents in large walrus gatherings.

Rod Downie, Chief Polar Advisor at WWF UK, made clear what is at stake:

Arctic wildlife, including caribou, is feeling the full force of the climate crisis. The changes we're seeing today will create a very different Arctic than we're used to. But every fraction of a degree matters: we must urgently address the climate crisis and plan for change so that people and nature can continue to thrive. AI-based tools will play an important role in achieving this goal.

The Canadian experience demonstrates that synergy between Indigenous communities, international science, and digital tools can produce concrete solutions to seemingly insurmountable problems. This message, in an era of manifest global warming, holds value far beyond the Arctic context.

Caribou: A migratory Arctic species that depends on seasonal sea ice to move between islands and the continent, now threatened by climate change, delayed freezing, and new environmental pressures.
Winter distribution maps of Dolphin and Union caribou show changes between the late 80s and early 2000s: data collected through radio collars highlight how the evolution of sea ice has forced the species to modify its wintering areas in the Canadian Arctic.
(Graph: data from Poole et al. (2010); figure reproduced from SARC (2013), edited by B. Fournier, GNWT-ENR 2016)

A bridge of algorithms over the frozen sea

Ultimately, the image that best describes the project is that of an invisible bridge, built over the icy waters of the Dolphin-Union Strait in CanadaIt is not made of wood or steel, but of satellite data, machine learning algorithms, and knowledge passed down through generations of native Inuit. It is a fragile bridge, because it depends on an ecological balance that climate change is challenging daily, but it is also a solid bridge, because it unites different worlds in a common goal: ensuring a future for the caribou and those who depend on them.

If Arctic ice continues to disappear, at least science and communities are demonstrating that it is possible to react, adapt, and innovate. Artificial intelligence, in this context, appears not as an abstract or distant technology, but as a tool that can make a difference in the survival of a species, and perhaps, indirectly, in the memory of our planet.

Caribou migration in Coronation Bay during sea ice formation in 2021

Here are three insights that might interest you:

Hudson Bay Polar Bears Don't Have Enough Food
The Secrets of the Greenland Shark's Longevity
Alaska: eDNA and Sailing Drones to Study Marine Ecosystem

Caribou: A migratory Arctic species that depends on seasonal sea ice to move between islands and the continent, now threatened by climate change, delayed freezing, and new environmental pressures.
During the autumn, Dolphin and Union caribou migrate between Victoria Island and the mainland, crossing the sea ice of Coronation Gulf: in 2021, this fragile passage was documented at a critical time, with late pack ice formation putting the entire moving herd at risk.
(Photo: Mathieu Dumond)

Location

COMMENTS

Leave a comment