2.3 Animal Movement
To best quantify error rates, only one animal was simulated within the
landscape so that counts >1 indicated multiple counting,
whereas 0 or average counts <1 were associated with animal
omission. While the simulation of multiple animals on the landscape at a
time would add realism to the natural world, this would have added
complexity to our model and negated our study objective; instead, we
aimed to create a simulation with just enough complexity to isolate our
variables of interest and address our study objectives (Grimm and
Railsback, 2005). For each survey type, the animal was first positioned
randomly on the landscape. Initial validation simulations had no
movement, mimicking a stationary animal for the entire survey duration
as a control to compare to other simulation scenarios that subsampled
the landscape. A moving animal was then simulated with one of three
different movement patterns: (1) random walk, (2) correlated random
walk, and (3) biased random walk. Walks were created by sampling an
exponential step length distribution and varying turning angle
distributions (see Appendix; Duchesne, Fortin and Rivest, 2015). For
each walk type, simulations were run with average animal velocities
representing a spectrum of natural terrestrial animal speeds (2, 4, 6,
8,10 m/s), as animal taxa differ substantially in various locomotion
behaviors that affect speed (walking, running, etc.). To maintain
standardized comparisons within the study purpose for drone surveys, the
simulated animal was designed to only move within the closed landscape
(i.e., no immigration or emigration) and was always available for
detection within the viewing window of the drone (i.e., no occlusion).
Count outputs also assumed that perception and detection probability
during image review was perfect. If an animal reached the border of the
landscape, depending on its programmed movement type, it was randomly
reflected in a new direction and continued its programmed movements
within the simulated landscape area until the drone survey was complete.