Inferring animal population distributions from individual tracking data: theoretical insights and potential pitfalls

Animal population-level phenomena are often inferred from large tracking data sets obtained from only a few individuals. Two key challenges are to understand how these two scales are related, and to identify the factors that influence the extent to which small samples consisting of a few individuals can predict spatial patterns at the population scale. We used a simple spatially explicit theoretical model to explore some of the factors that affect inferences made at the population level from individual tracking data. We adopted a 'mixtures of correlated random walks' approach to simulate two discrete movement modes with different step lengths and turning angles in a hypothetical ungulate population with contrasting population sizes and sampling intensities. Movement state was assumed to be influenced by habitat type (patch or matrix) and social cues. We explored the predictive power of a tracked population subsample by regressing the space-use map generated by a few randomly chosen individuals against the map generated by the entire population (the 'true' map) for different scenarios (e.g. random and clumped habitat distributions) and parameter values. We show that the predictive power of the tracking sample varies nonlinearly and often counter-intuitively with factors such as habitat preference, the spatial context of the landscape and the importance of social interactions. We suggest that movement models coupled with individual tracking data can be used with Monte Carlo simulations to improve tracking studies by better understanding the links between detailed individual movement data and population distributions.