Preventing Dangerous Encounters with the Endangered (or Vulnerable)
- Dec 17, 2025
- 5 min read
Updated: Apr 7
AI Bear Monitoring Before They Come Knocking On the Door
As the Arctic warms faster than any other region on Earth, the rapid loss of sea ice is forcing polar bears to spend more time on land, bringing them into closer proximity to human communities. This shift has led to an increase in potentially dangerous encounters, underscoring the need for systems that detect approaching bears before conflicts occur.
Polar bears are officially listed as "vulnerable" to extinction by the IUCN due to rapid Arctic sea-ice loss from climate change, which threatens their hunting grounds. While not currently classified as "endangered" globally, they are listed as threatened under the US Endangered Species Act. The primary danger is the melting of sea ice, which limits their ability to hunt seals, leading to starvation and reproductive failure.
In recent years, Geoff York, Senior Director of Science and Policy at Polar Bears International (PBI), has led the development of an AI-enabled radar system designed to remotely detect polar bears and alert residents in time to take preventive action. According to Kieran McIver, manager of Churchill Field Operations at PBI, the system must be able to distinguish between bears, humans, vehicles, and other wildlife—an essential requirement to avoid false alarms. The approach combines a radar capable of scanning large areas with AI software for object classification.
By providing early warnings, tools such as Bear-dar allow communities to anticipate polar bear presence and respond before encounters occur. With advance notice, residents can adjust daily activities and deploy non-lethal deterrents, reducing the likelihood of conflict and improving safety for both people and wildlife.
From Concept to Cold Reality: Building Bear-dar
Bear-dar is a medium-range radar system developed to monitor expansive open areas and accurately detect polar bears. Researchers at PBI are training their AI to recognize polar bear movement patterns and body shapes, a complex task given the variability of a bear’s radar signature. This variability poses a central challenge for system training, as the AI must maintain accurate classification across a wide range of movement patterns and conditions. In addition, the radar’s high sensitivity increases the risk of false positives, requiring the system to filter out signals from humans, vehicles, and other wildlife.
Remote radar towers could also be equipped with strobe lights or loudspeakers that activate automatically as a bear approaches, providing an additional deterrent.
Once detected, the system can track a bear’s movements and send alerts via text message or email to people in the area. After fine-tuning, the radar will be deployed in locations with a high risk of bear–human interactions. Bear-dar is part of a broader suite of emerging tools that integrate deep learning, environmental forecasting, and community knowledge. For instance, a deep learning model created by the British Antarctic Survey and WWF partners now predicts sea-ice conditions up to three months in advance—twice the range previously possible. These forecasts help Indigenous hunters, shipping operators, and conservation planners make safer decisions. Similar advances in passive acoustic monitoring and satellite data analysis are improving visibility into Arctic ecosystems, revealing whale movements beneath sea ice and documenting long-term glacier retreat. Together, these innovations help communities and scientists respond more quickly to the region’s accelerating transformations.
A Shield for Communities and Conservation
With advance warning, communities can make immediate safety decisions that reduce the risk of bear–human encounters. When alerts are issued, people can delay routine activities, such as having children walk to school, until it is safe. Raymond Friesen, a research support specialist at PBI, emphasizes that early detection reduces the number of times a bear must be killed for public safety. By warning residents, Bear-dar reduces the likelihood of conflict and ultimately protects both humans and bears.
Communities need additional safety tools to manage this shift, and early-warning systems are among the most practical solutions. With advance notice, individuals can deploy nonlethal deterrents—such as noisemakers or flares—to redirect bears away from populated areas. This proactive approach helps ensure bears are not habituated to human settlements, thereby reducing long-term risks.
To test the system in real-world conditions, PBI deployed Bear-dar near Churchill, Manitoba—a location where large numbers of polar bears gather each fall as they wait for the sea ice to return. Churchill already hosts PBI’s Polar Bear Cams and Tundra Connections webcasts, providing infrastructure for testing. Community support has been strong, with residents recognizing the potential safety benefits. These tests help refine the system and ensure it can respond to the unpredictable behavior of wild bears.
Field Testing and the Challenge of the Unseen
Testing the radar in Churchill revealed some of the challenges of tracking polar bears. In some cases, a bear may lie down for minutes or even days, causing the radar signal to disappear. To address this, the team conducted additional trials at the Assiniboine Park Zoo in Winnipeg, where the radar could monitor bears under more controlled conditions. Under McIver’s supervision, the system collected data that would otherwise be difficult to capture in the wild, helping improve the AI’s ability to recognize variations in bear behavior.
This work fits within PBI’s broader innovation vision: technology must serve a clear conservation purpose and should never be pursued merely for novelty. In practice, this means combining scientific tools with local expertise and aligning technological development with community needs. Technology, in this framework, is meant to amplify human knowledge—not replace it.
While many environmental AI systems are being developed—from deep-learning models that analyze sea ice to acoustic tools that detect whales or algorithms that monitor glacier changes—Bear-dar stands out because it addresses a direct and urgent safety issue. These other projects illustrate the broader growth of AI in environmental research, but Bear-dar’s purpose is more specific: helping communities respond before encounters occur.
By focusing on real-time detection rather than large-scale environmental patterns, the system fills a gap in existing Arctic monitoring technologies. Unlike many research dashboards designed to visualize climate trends, Bear-dar is intended to operate as a practical, on-the-ground tool that can help protect people before an encounter occurs.
The Future of AI Wildlife Monitoring
York emphasizes that conservation organizations have a responsibility to support people who share their environment with large predators. Building tools that enhance coexistence is central to PBI’s mission, and Bear-dar reflects this commitment by addressing real safety concerns with practical technological solutions. By providing communities with early warning of approaching bears, the system fosters safer interactions and reduces the likelihood of conflict.
McIver and Friesen note that the underlying technology is not limited to polar bears. With proper training, similar systems could be adapted to track other species or support conservation research in different regions. They envision a future in which radar units are permanently installed in northern environments to provide year-round monitoring of bear movements.
Ultimately, the Bear-dar project illustrates how AI and radar technology can be integrated into conservation strategies grounded in local knowledge and community partnerships. As Arctic conditions continue to evolve, tools like these will be essential for helping people and wildlife coexist safely across a landscape undergoing rapid transformation.