LB 5-38 - Using direct observations to validate movement-based behavioural inferences: a case study using GPS-video collars on black bears (Ursus americanus)

Background/Question/Methods:

Increasingly fine-resolution wildlife GPS-tracking technology has led to an influx of movement models that infer animal behavioural states from patterns in the step lengths and turning angles of consecutive GPS locations. For example, longer step-lengths and smaller turning angles may be inferred as interpatch movements, while shorter step-lengths and larger turning angles may be inferred as foraging behaviour. However, movement-based inferences are rarely validated with direct observation. Novel GPS-video wildlife collars present a unique opportunity to validate the accuracy of movement-based behavioural inference by recording direct video observations of animal behaviour paired with fine-scale GPS data. Here, we use black bears (Ursus americanus) fitted with GPS-video collars as a case study to verify whether different observed behaviours result in characteristic movement patterns distinguished by step lengths and turning angles. In an industrial landscape in northeastern Alberta, Canada, the behaviours of four adult, male black bears were classified for 10-second video clips, recorded every 10-minutes from dusk until dawn. We assessed whether the mean step-lengths and turning angles varied significantly among behaviours commonly inferred from movement-based analyses: resting, foraging, travelling, and foraging while travelling.

Results/Conclusions:

Our preliminary results demonstrate bear step-lengths are smallest while resting, but are comparable between foraging, travelling, and foraging while travelling behaviours. Turning angles were significantly smaller in travelling than non-travelling behaviours, consistent with the high tortuosity predicted for foraging behaviour and the GPS-error of resting locations. In popular movement-based analyses, e.g., integrated step-selection analysis, the interest is typically on the movement of an animal through the environment, and the exclusion of non-moving behaviours is often a critical first step. Our results suggest that movement-based models can accurately infer resting, foraging, and travelling behaviours of black bears using step-lengths and turning angles. However, our inability to discern between travelling and travelling while foraging behaviours suggests that movement-based inference using only step-lengths and turning angles may not appropriately define behavioural states depending on the research question. For example, longer step lengths of black bears while on industrial linear features has been inferred as exclusively travelling behaviour, with implications for encounter rates with threatened boreal caribou (Rangifer tarandus caribou). However, if black bears are using linear features for both foraging and movement efficiency, the focus of linear feature habitat restoration efforts may need adjustment.