A virtual ecologist approach for improving methods in movement ecology

As human encroachment into natural systems continues, we must develop ways to reconcile human and animal needs while protecting biodiversity and human health. Studying animal movement provides incredible insights into their requirements [1, 2] as well as highlighting anthropogenic impacts [3, 4] and potential conflict areas [5,

6]. Insights from movement ecology are frequently incorporated into conservation plans [7]; therefore, it is imperative we fully grasp the generalisability and robustness of these findings. Low sample sizes, localised studies, and a growing diversity of analytical approaches could all be limiting generalisability and replicability in movement

ecology, but assessing the replicability of ecological studies is costly and disincentivised by the current publication system [8], despite a clear agreement on replication's value. Movement ecology in particular, with its rapid growth and reliance on expensive telemetry devices, has yet to be subjected to rigorous

replicability efforts. To avoid further costs to both animals and researchers associated with repeat studies, we can leverage a virtual ecologist approach to explore a key component implicated in undermining replicability of results - researcher degrees of freedom [9]. Researcher degrees of freedom, or analytical flexibility, can be used to generate

findings that better fit within the publishing incentive system, namely narrativelycoherent significant results. However, these incentives can reward underpowered extreme results over robust, more reliable but less glamorous studies [10, 11]. A virtual ecology approach will reveal what options are available to researchers when

assessing movement ecology data, and via the implementation of a multiverse analysis [12], identify what choices can markedly impact results from a known simulated truth. Research Objectives The foundation of this project is the development of an agent-based model to simulate animal movement, integrating several key components: spatial covariates

(e.g., habitat and connectivity), central tendency (i.e., home range), behavioural states (e.g., resting, foraging, and dispersal), and interspecific & intraspecific relationships (e.g., attraction & avoidance). The model will then be used to explore how researcher choice impacts findings, and how these findings deviate from

parameters used to characterise the simulated data. Once the impacts of analysis choice are known, the student will apply them to a real world scenario and illustrate the utility of multiverse analysis in displaying the robustness of results.