Assessing sampling error associated with animal movements and distributions across drone monitoring strategies

Long-term animal population monitoring is essential for informing wildlife management practices. Erroneous estimates skew researchers’ understanding of population parameters. Novel survey methods resulting from technological advancements should be assessed to identify potential sampling error before adoption and execution. Drones (i.e., Unoccupied Aircraft Systems) allow for remote monitoring of animal populations. Typical drone flight planning involves large (40–80 %) image overlap to create orthomosaic images. However, this approach may introduce counting errors when monitoring non-randomly dispersed or mobile animals. We evaluated errors associated with counting multiple, moving animals among various drone flight patterns in an agent-based model framework. Scenarios investigated included nine animals in random and clumped spatial distributions exhibiting multiple movement patterns (random walk, correlated random walk, and stationary) to mimic specific, literature-informed animal density, distribution, and movement patterns. Drone flight patterns included lawnmower patterns with and without image overlap, belt transects, and systematic point counts. Drone flight pattern was the most important variable influencing the accuracy of counts of moving animals, although animal distributions had the greatest effect on precision. We recommend the use of spaced transects distributed across and covering at least 50 % of the entire survey area to minimize sampling error among various animal distribution patterns and/or when animals move during the survey period. Understanding the influence of animal movements and distributions on population estimates from drone surveys will aid researchers in deciding whether drones are appropriate for specific aerial wildlife monitoring scenarios and if so, in designing robust surveys.