In the vast landscapes of the western United States, where over 1 million kilometres of fencing crisscrosses the terrain, barriers originally erected to contain livestock now pose a significant threat to wildlife. Species such as pronghorn, deer, and elk often find their migratory routes blocked by these fences, leading to starvation and death, particularly in harsh winter conditions. Recognising this issue, scientists are turning to artificial intelligence (AI) to help wildlife managers locate and mitigate the impacts of these barriers.
A pioneering project led by Wenjing Xu, a postdoctoral researcher at the Senckenberg Biodiversity and Climate Research Centre, and Zhongqi Miao, an applied research scientist with Microsoft AI for Good Lab, aims to use AI to identify fences from aerial images. Their focus is on southwestern Wyoming, a critical area for migratory pronghorn and mule deer. By training AI to recognise fences in aerial photographs, they have achieved a 70% success rate in identifying these barriers, a promising start that could revolutionise conservation efforts.
Fences have long been a silent yet deadly obstacle for wildlife across the West. For example, a recent study by biologist Hall Sawyer tracked the movements of pronghorn in Wyoming’s Red Desert. During the winter of 2023, severe weather conditions led to the deaths of half of the collared pronghorn, many of which were trapped by fences while trying to escape the harsh environment. The study highlighted the urgent need for more effective management of fencing to protect vulnerable wildlife.
The AI-driven approach by Xu and Miao represents a significant advancement in this area. By automating the process of fence identification, conservationists can more effectively target areas for fence removal or replacement with wildlife-friendly alternatives. These efforts are crucial, as research has shown that removing fences can increase access to high-quality forage for pronghorn by up to 38% across both public and private lands in regions such as Alberta, Saskatchewan, and Montana.
However, the challenge of systematically removing or replacing fences is compounded by the lack of accurate data on their locations. Public records are incomplete, and on-the-ground surveys are time-consuming and limited in scope. Xu and Miao’s work to map over 7,000 kilometres of fence in Wyoming using AI is a critical step towards creating a comprehensive inventory of these barriers.
As AI technology continues to evolve, the potential to refine these models and expand their application globally is immense. Xu and Miao plan to extend their methods to include satellite images, which could help identify fences in other regions of the world, such as Tibet, Australia, and Kenya. This global perspective aligns with the broader goals of the Global Society, which seeks to address environmental challenges through innovative solutions and international collaboration.
The use of AI in conservation is not without its challenges. Early-stage projects, such as this one, typically achieve a 70% success rate, which is promising but still leaves room for improvement. For example, the AI model occasionally mistakes roads for fences, highlighting the need for higher-resolution images and more sophisticated training data.
Despite these limitations, the potential benefits of AI-driven conservation are clear. By providing wildlife managers and conservation groups with better tools to locate and manage fences, AI can play a crucial role in reducing wildlife deaths and promoting sustainability. As Xu and Miao continue to refine their model and expand its application, their work represents a significant step forward in the global effort to protect our planet’s biodiversity.
The connection between AI, conservation, and sustainability is increasingly recognised as vital to addressing complex environmental challenges. As this project demonstrates, the integration of advanced technology with ecological research can lead to innovative solutions that support the goals of sustainable development and global cooperation.
Youtube credits: https://www.youtube.com/@drapernaturalhistorymuseum3506