Epsilon Archive for Student Projects

Abstract

In wildlife management, accurate monitoring of animal populations and estimation of population densities are essential for informed decision-making. In Sweden, traditional moose monitoring uses hunter observations during the hunting season, a system called Älgobs. While cost-effective, this method is constrained by the time lag between counting in autumn and implementing the management plans the next year. This study investigates the effectiveness of camera trap deployments, both systematic and targeted, in estimating moose reproductive outcome estimates, specifically the proportions of females without calves, females with one calf, and females with two calves in comparison to the proportions found in the counts of the Älgobs method. Using static multi-state occupancy models, I analysed data from systematically deployed cameras, alongside targeted cameras.
My results indicate that systematically deployed cameras provide reproductive outcome ratio estimates comparable to those of the proportions calculated with the Älgobs counts. The naïve occupancy estimates, despite not accounting for detection probabilities, yielded ratios similar to those from the multi-state occupancy models. This suggests that naïve occupancy can still be useful for detecting population differences. However, the heterogeneous data sources and deployment methods presented challenges, with two datasets showing large standard errors and wide confidence intervals, highlighting the need for enough cameras for sufficient data and consistent methodologies.
This study underscores the potential of camera traps in providing reliable data on moose population ratios, particularly when combined with occupancy modelling. It shows the possible solution for the limitation of traditional method Älgobs, the time lag between moose counts and the production of management plans. Future research should focus on optimizing camera placement strategies and providing clear guidelines for volunteers to enhance data quality and model robustness. The inclusion of AI tools like MegaDetector can further streamline the classification process and could improve the efficiency and accuracy of wildlife monitoring.