A fine-grained behavior-based approach to estimating the probability of collision between moving vehicles and birds

Ryan B Lunn , Bradley Blackwell, Esteban Fernández-Juricic

1. Collisions between animals and vehicles contribute to biodiversity loss, threaten human safety, and have economic consequences. Escape responses of wildlife to vehicles are a critical factor in determining whether a collision occurs. However, presently species-specific vulnerability estimates do not consider the species escape behavior, potentially resulting in inaccurate mortality estimates.

2. Recently, a mathematical model was proposed to estimate whether a collision occurs considering fine-grained properties of both an approaching vehicle and the species escape response. Herein, we expanded upon an existing model and applied it for the first time to estimate the probability of collision in a behavioral experiment where a UAS directly approached Canada geese with different light (light-off, light-on steady, light-on pulsing) and approach type (level, descending) treatments.

3. The probability of collision for the light-on steady and light-on pulsing treatment increased by 9.92% and 25.47%, respectively, during level approaches but decreased during descending approach treatments by 34.38% and 37.24%, respectively, relative to the light-off treatment. We attribute this interaction effect to differences in the ratio of UAS to LED light surface area.

4. We examined the role of the different parameters in the model and found that flight-initiation distance had the largest effect size, followed by escape trajectory, and UAS altitude. Specifically, longer flight-initiation distances, an increase in an away escape trajectory, and higher UAS altitude all contributed to reducing the probability of collision.

5. Synthesis and applications. Using a fine-grained behavioral approach, our model can provide estimates of the probability of an animal-vehicle collision when observing a collision is either logistically challenging or unethical. We also demonstrate that our model can provide quantitative guidance on UAS based hazing strategies to improve animal welfare. Lastly, our model can be used to quantitatively estimate how many animal-vehicle interactions were avoided due to the escape response of the animal, thus enabling conservation strategies to not just focus on reducing collisions but also promoting successful avoidance.