Ed Miller, Mary Nguyen, Jorden Benner

Confronted with the growing complexity of conserving wildlife in a rapidly changing environment, researchers and managers are adopting innovative technologies to improve understanding of species dynamics and long-term trends. The application of computer-based technologies, including AI models, my provide both opportunities and challenges, shaping both the knowledge produced and its application to bear conservation. We are bridging advances in conservation technology, specifically camera traps and machine learning, with Indigenous knowledge and land stewardship to expand the suite of information available to decision makers. Our program works with Nanwakolas Council member First Nations to assess the spatiotemporal distribution and movements of individual grizzly bears (Ursus arctos) among First Nation territories, in the southern Great Bear Rainforest (BC, Canada).
Our project to develop a robust, automated method for individual re-identification of grizzly bears from images has highlighted key trade-offs and lessons that can guide both the practical use of this technology and its adaptation to other bear species. Using camera trap videos collected from 2018-2024 to train and test re-ID models, we identified how results can vary based on sample size, data partitioning, individual bears, season and image characteristics. These insights are important for evaluating the reliability of models used in conservation decision making and for extending their application to new populations and environments. Similarly, providing tangible ways to access models for practical use is essential to move these technologies beyond academic development. Despite these challenges, re-ID and other machine learning models offer significant opportunities to speed-up data processing, expand ecological insights, and support management decisions, such as in human-bear conflicts.