Testing a key assumption of using drones as frightening devices: do birds perceive drones as risky?

Wildlife managers have recently suggested the use of unmanned aircraft systems or drones as nonlethal hazing tools to deter birds from areas of human-wildlife conflict. However, it remains unclear if birds perceive common drone platforms as threatening. Based on field studies assessing behavioral and physiological responses, it is generally assumed that birds perceive less risk from drones than from predators. However, studies controlling for multiple confounding effects have not been conducted. Our goal was to establish the degree to which the perception of risk by birds would vary between common drone platforms relative to a predator model when flown at different approach types. We evaluated the behavioral responses of individual Red-winged Blackbirds (Agelaius phoeniceus) to 3 drone platforms: a predator model, a fixed-wing resembling an airplane, and a multirotor, approaching either head-on or overhead. Blackbirds became alert earlier (by 13.7 s), alarm-called more frequently (by a factor of 12), returned to forage later (by a factor of 4.7), and increased vigilance (by a factor of 1.3) in response to the predator model compared to the multirotor. Blackbirds also perceived the fixed-wing as riskier than the multirotor, but less risky than the predator model. Overhead approaches mostly failed to elicit flight in blackbirds across all platform types, and no blackbirds took flight in response to the multirotor at either overhead or head-on approaches. Our findings demonstrate that birds perceived drones with predatory characteristics as riskier than common drone models (i.e. fixed-wing and multirotor platforms). We recommend that drones be modified with additional stimuli to increase perceived risk when used as frightening devices, but avoided if used for wildlife monitoring.We did not include trials where the behavioral response occurred during or after a piloting error (e.g., drone crashed or significantly deviated from the flight path; n = 7). For vigilance, we only considered birds that foraged from the perch a full 30 s. Thus, we omitted birds from analysis that did not forage for a full 30 s or perched in a manner that was not comparable to other birds (i.e. directly on the food tray as opposed to the perch). These values are "NA" in the CSV files. We are also missing a single value for ambient sound intensity, because the sound recorder was turned off during a single trial. Overview description of variables in each data set: 1.) Eganetal_ArchivedDataset.csv contains the following variables: Treatment = Combination of platform type and approach type that each bird was exposed to during trials Platform = Platform type that each bird was exposed to during trials (i.e, predator model, fixed-wing, multirotor) Trajectory = Approach type that each bird was exposed to during trials (i.e., head-on or overhead) Type = Fixed-wing and Predator model coded as "1", Multirotor coded as "0" Latency = Latency to resume foraging (seconds) LogLatency = Log transformed latency to resume foraging (seconds) Speed = Drone speed per trial (meters per second) Deprived = Length of time birds were food deprived prior to trials (hours) Wind Origin = Direction wind originated from while being measured Wind = Wind origin combined to reflect headwind, downwind, or crosswind OriginCat = Direction wind orginated from, convered to #s for coding Windspeed = Windspeed (kilometers per hour) Light = Ambient light intensity (µmol m-2 s-1) Temperature = Temperature (Celsius) PUP1 = Proportion of time (30 s) birds spent head-up versus head-down prior to drone approaches PUP2 = Proportion of time (30 s) birds spent head-up versus head-down following drone approaches UPdiff = The difference between PUP1 and PUP2 Approach = Approach time (seconds) of the drone between launch and the time it passed above the enclosure Sound = Ambient sound intensity (decibels) during trials Alerttime = Individual alert time (seconds) per trial LogAlert = Log transformed invidual alert time (seconds) per trial Flush = Did the bird flush the perch in response to drone flights? 1 = Yes, 0 = No Call = Did the bird alarm call in response to drone flights? 1 = Yes, 0 = No Flight = Individual flight-initiation time (seconds) per trial 2.) Eganetal_MixedModelData.csv contains the following variables: ID = Identifiation number assigned to each individua bird Treatment = Combination of platform type and approach type that each bird was exposed to during trials Platform = Platform type that each bird was exposed to during trials (i.e, predator model, fixed-wing, multirotor) Trajectory = Approach type that each bird was exposed to during trials (i.e., head-on or overhead) Speed = Drone speed per trial (kilometers per hour) Light = Ambient light intensity (µmol m-2 s-1) Time = Time of measurement relative to drone flight. Before = before drone flight, After = after drone flight Vigilance = Proportion of time (30 s) birds spent head-up versus head-down Funding provided by: Animal and Plant Health Inspection ServiceCrossref Funder Registry ID: http://dx.doi.org/10.13039/100009168Award Number: USDA-APHIS-WS NWRC; #7438-0020-CA; QA-2731Funding provided by: Animal and Plant Health Inspection ServiceCrossref Funder Registry ID: http://dx.doi.org/10.13039/100009168Award Number: Interagency Agreement DTFACT-14-X-400007See publication for full methods: Conor C. Egan, Bradley F. Blackwell, Esteban Fernández-Juricic, Page E. Klug. 2020. Testing a key assumption of using drones as frightening devices: do birds perceive drones as risky? The Condor: Ornithological Application