Uncovering behaviors of cryptic species: Latent behavioral states, activity budgets, and habitat associations of giant armadillos (Priodontes maximus) in the Brazilian Pantanal

Joshua Cullen and Denis Valle, School of Forest Resources and Conservation, University of Florida, Gainesville, FL, Nina Attias and Arnaud L.J. Desbiez, Instituto de Conservação de Animais Silvestres (ICAS), Campo Grande, Brazil, Arnaud L.J. Desbiez, Instituto de Pesquisas Ecológicas (IPÊ), Nazaré Paulista, Brazil, Arnaud L.J. Desbiez, Royal Zoological Society of Scotland (RZSS), Edinburgh, United Kingdom

Background/Question/Methods
Recent advances in biologging have facilitated a greater understanding of how animals interact with their environment, especially for cryptic species that are infrequently sighted during surveys. To fully understand animal movement, it is necessary to account for behavior since space and resource use are directly linked to an animal’s internal state. Giant armadillos (Priodontes maximus) are the largest extant species of armadillo, but are rarely encountered due to their fossorial and nocturnal behavior. Recent work on giant armadillos has elucidated habitat selection patterns and space-use across the Brazilian Pantanal, but there is a paucity of information on the behavior during its brief nightly activities. Therefore, we investigated the behavioral states exhibited by individuals tagged with GPS transmitters and accelerometers, examining how environmental variables impacted the likelihood of exhibiting one state over another. Activity budgets were also evaluated on an hourly basis to discern diel patterns. Behavioral states were estimated on measures of speed, turning angle, and changes in acceleration using a non-parametric Bayesian mixture model that provides flexible inference on latent state estimation. Generalized additive mixed models (GAMMs) were used to investigate relationships between pairs of latent states and land use/land cover (LU/LC) with random effects for individuals.

Results/Conclusions
The mixture model estimated four latent behavioral states: ‘Burrow Vigilance and Reconstruction’ (BVR), ‘Foraging’, ‘Exploratory’, and ‘Transit’. The first two states shared similar state-dependent distributions apart from turning angles, whereas the latter two exhibited greater speed, increasingly straight turning angles, and increasing levels of accelerometry activity. The activity budget varied over the night, where the first and last hours of activity presented relatively high proportions of the BVR state while the majority of the activity period was dominated by ‘Foraging’ and ‘Exploratory’ states. The post-hoc pairwise comparisons found that certain land classes influenced the probability of exhibiting one state over another. The BVR state occurred more frequently when in greater proportions of forest cover compared to all other states. The ‘Foraging’ state was more likely to occur within greater proportions of forest cover compared to the ‘Exploratory’ state and in greater proportions of closed savanna compared to the BVR state. Additionally, the ‘Transit’ state was more likely to occur when in greater proportions of floodable habitat compared to ‘Foraging’ and ‘Exploratory’. These findings support previous observations of giant armadillo behavior and support the critical role of biologging for the monitoring and conservation of cryptic species.