Balancing detection probability and survey effort in multistate occupancy models: A camera trap simulation analysis

Abstract

Camera trapping has become crucial in wildlife research, enabling detailed observations of elusive and nocturnal species with limited human interference. The use of occupancy modeling to analyze camera trap data is rapidly increasing, aiding in the assessment of species distribution, multispecies dynamics, and the presence of different states of a species (e.g., reproducing or non-reproducing), while considering imperfect detection. Multistate occupancy models, which capture these different states, are particularly effective tools. However, the design of camera trap studies—typically involving large grids with a limited number of cameras and animal observations—often results in sparse data and low detection probabilities, impacting model performance (e.g., convergence) and inference reliability (e.g., accuracy and precision) in basic occupancy models. The effect of these factors on more complex models (e.g., multistate occupancy models) remains largely unexplored. Here, we conducted a series of simulations with varying detection probabilities, numbers of sites, and survey periods for both single- and multistate occupancy models, to evaluate the impact of these factors on model performance and reliability. Our results revealed that multistate models require higher detection probabilities compared to the single-state models. Additionally, minimum needed detection probabilities decreased as the number of surveys increased for all models. Furthermore, the number of sites required was substantially higher for multistate models compared to single-state models. We conclude that when detection probabilities are low, occupancy models encounter difficulties in fitting and produce unreliable results. Strategies such as deploying clustered cameras, targeted camera placement (e.g., at frequent wildlife paths) or using bait to increase detection rates could be used to address these issues but may introduce other biases. The gained model performance from higher detection probabilities might outweigh these biases. Moreover, different data aggregation strategies in combination with increasing the length of the study can increase detection probabilities, addressing reliability issues; however, this is not always feasible due to time constraints (e.g., season-based research questions). This study highlights key thresholds and considerations for improving the use of multistate occupancy models using camera trap data, aiding in the design of more effective wildlife research studies.