Abstract

Topic: Ungulate species interactions and management

Strategic use of camera traps and AI for improved moose management in Sweden

Lars Hillström¹, Marcus Larsson², Petter Hillborg³, Anna Göransson³

University of Gävle, Gävle, Sweden
Swedish Forest Agency, Gävle, Sweden
Department of Electrical Engineering, Mathematics and Natural Sciences, Academy of Engineering and Sustainable Development, University of Gävle, Gävle, Sweden

Estimating the size and dynamics of moose populations is crucial for sustainable wildlife management. Traditional methods such as moose observations and scat inventories have limitations in objectivity and accuracy. This study presents an innovative method that combines camera traps with artificial intelligence (AI) to automatically record and analyze spatial and temporal variation in local moose populations. The method has been tested in two areas in Gävleborg County, Ockelbo moose management unit (OMMU) and an area that was on fire in 2018. From August 2020 to February 2022, we recorded 2,475 moose observations in the area with forest fire and 3,698 in OMMU, using a camera density of one camera/850 hectares. AI-based analysis enables a detailed classification of individuals based on sex and number of calves born. This has the potential to improve decision-making for hunting and forest management. The study shows that this method offers a more continuous and objective monitoring of the moose population compared to traditional methods. Automated analysis of image data reduces subjectivity and improves the precision of population estimates. Data collected immediately after calving and in the following months provide more accurate estimates in moose management than current methods such as moose observations and droppings inventory. Having current data from the period immediately after calving in the same year as culling is to be determined, offers significantly better and more up-to-date information than current methods. Objective and well-distributed image material makes it possible to better determine local variations in the game population. If the cameras are placed equally between different areas at salt rocks, good comparison opportunities are created, and through several years of experience, the method can ensure variations between areas and changes in the population over time. Further development of the AI tools can improve data quality, even if individual recognition of animals is not possible with current technology. The focus on distribution of males and females and calf survival provides a more reliable picture of moose population dynamics. Through collaboration with hunting teams and forest owners, this method can pave the way for more transparent and scientifically based wildlife management in Sweden.