autonomous flight. We used the APM Planner 2.0 software (http://planner.
ardupilot.com/) to program and fly each flight. In brief, each mission was flown
according to the protocol below, but the re was some variation among flight
plans due to weather conditions, distance to the animal, and the ability to
pinpoint the bear’s location (the only means to track the denning bear was
with VHF telemetry).
For each 5-min flight, the UAV was programmed with a GPS fix based on
the last known location of the focal bear obtained from the GPS collars or, in
the case of the VHF-collared bear, the triangulation of the bear’s location.
The UAV was launched and climbed to an altitude of 20 m, and then flew
straight to the programmed GPS fix. Upon reaching this point, the UAV loitered
in place for 1 min before initiating two consecutive large turns, each with a
radius of 20 m (1 min for each turn) around the GPS point. After completing
the turns, the UAV returned to the programmed fix to loiter in place for 1 min.
After completing its mission over the bear, the UAV flew back to the launch
point and automatically landed. Each mission was initiated by an FAA-certified
pilot who armed the quadcopter and increased the throttle to 50%. The pro-
gramed mission commenced automatically at this point, and each flight was
flown and landed fully autonomously with no further user input.
Following each flight we downloaded the data logged by the UAV flight com-
puter (Pixhawk) using APM Planner 2.0. We used PyMAVLink Tools (https://
pixhawk.org/dev/pymavlink) to extract the time stamps, GPS locations,
speed, and absolute altitude of the UAV (height of UAV above the ground)
throughout each flight. These data are logged at 3–5 times per second by
the Pixhawk flight computer. Following their extraction, these data were pro-
cessed so they could be linked with the HR and movement data from each
bear (see Statistical Methods).
Statistical Methods
All statistical analyses were carried out in R [29], an open source statistical pro-
gramming language. We fit linear regression models to the HR data collected
1 hr prior to each UAV flight, using natural cubic regression splines (ns function
in package: splines [29]) with 2 degrees of freedom to account for temporal
trends in the HR values. We used this model to predict the HR values occurring
during an 8-min window covering the time period of the UAV flight and a few
minutes post-flight (see Figure 2A). If two UAV flights occurred over the
same individual, with less than 20 min between each flight, we used the HR
values for the hour prior to the first flight to estimate the predicted values for
the second flight. We formed HR anomalies, representing the increase in HR
beyond what might be expected given the trend in HR for the hour prior to
the flight, as the difference between the observed and predicted HR values
during the 8-min window, divided by the SD of HR values from the hour prior
to the UAV flight.
We generated control observations by repeating this process using HR data
from all dates without a UAV flight (female with cubs of the year: 175 days;
young male: 181 days; hibernating adult female: 79 [winter hibernation days
only]) but collected during the same time of day as the UAV flights. We formed
a null distribution for the empirical distribution function (ECDF), assuming no
effect of the UAV, by repeatedly subsampling these ‘‘control’’ data, keeping
the same number of observations per bear as in the original UAV-flight dataset.
We calculated the ECDF for each of 10,000 subsampled control datasets and
created a 95% simulation envelope to compare to the ECDF of the HR anom-
alies associated with the UAV flights (Figure 2B). An ECDF of the UAV HR flight
data that did not fall within the 95% simulation envelope suggested that the
maximum HR anomaly values from control and experimental conditions
were drawn from two different distributions.
We calculated the recovery time of bear HRs post-flight for each flight and
reported the median and range for each individual. We defined recovery
time as the number of minutes until HR returned to values below the upper
99% confidence interval based on values from 30 min prior to each flight. If
a set of flights occurred such that the second flight began prior to recovery af-
ter the first flight, we considered only recovery after the second flight.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Supplemental Experimental Procedures,
three figures, and one movie and can be found with this article online at
http://dx.doi.org/10.1016/j.cub.2015.07.024.
AUTHOR CONTRIBUTIONS
M.A.D. analyzed the data and wrote the paper. M.A.D., J.B.V., L.K.W., and
J.C.T. designed the study and performed the UAV fieldwork. J.R.F. suggested
and helped develop the statistical approach. P.A.I. and T.G.L. performed the
biologger surgery and consulted on the physiological aspects of the study.
D.L.G. was the lead researcher for winter fieldwork and consulted on interpre-
tation of bear behavior. All authors reviewed the final version of the manuscript.
ACKNOWLEDGMENTS
The Institute on the Envi ronment (University of Minnesota) and the International
Association for Bear Research and Management provided financial support.
We thank T. Baker, L. Dillard, and B. Taylor of the University of Minnesota
for advice and technical help. T. Iles, H. Martin, H. Severs-Wilkerson, and M.
McMahon assisted with fieldwork. J. Huener and K. Arola of the Minnesota
Department of Natural Resources and G. Knutsen of the USFWS allowed us
to use their facilities for fieldwork.
Received: June 1, 2015
Revised: July 7, 2015
Accepted: July 9, 2015
Published: August 13, 2015
REFERENCES
1. Anderson, K., and Gaston, K.J. (2013). Lightweight unmanned aerial vehi-
cles will revolutionize spatial ecology. Front. Ecol. Environ 11, 138–146.
2. Marris, E. (2013). Drones in science: fly, and bring me data. Nature 498,
156–158.
3. Lucieer, A., Turner, D., King, D.H., and Robinson, S.A. (2014). Using an un-
manned aerial vehicle (UAV) to capture micro-topography of Antarctic
moss beds. Int. J. Appl. Earth Obs. Geoinf. 27, Part A, 53–62.
4. Vermeulen, C., Lejeune, P., Lisein, J., Sawadogo, P., and Bouche
´
,P.
(2013). Unmanned aerial survey of elephants. PLoS ONE 8, e54700.
5. Vincent, J.B., Werden, L.K., and Ditmer, M.A. (2015). Barriers to adding
UAVs to the ecologist’s toolbox. Front. Ecol. Environ 13, 74–75.
6. Koh, L.P., and Wich, S.A. (2012). Dawn of drone ecology: low-cost auton-
omous aerial vehicles for conservation. Trop. Conserv. Sci. 5, 121–132.
7. Barasona, J.A., Mulero-Pa
´
zma
´
ny, M., Acevedo, P., Negro , J.J., Torres,
M.J., Gorta
´
zar, C., and Vicente, J. (2014). Unmanned aircraft systems
for studying spatial abundance of ungulates: relevance to spatial epidemi-
ology. PLoS ONE 9, e115608.
8. Shahbazi, M., The
´
au, J., and Me
´
nard, P. (2014). Recent applications of un-
manned aerial imagery in natural resource management. GIsci. Remote
Sens. 51, 339–365.
9. Whitehead, K., and Hugenholtz, C.H. (2014). Remote sensing of the envi-
ronment with small unmanned aircraft systems (UASs), part 1: a review of
progress and challenges. J. Unmanned Veh. Syst. 2, 69–85.
10. Ratcliffe, N., Guihen, D., Robst, J., Crofts, S., Stanworth, A., and Enderlein,
P. (2015). A protocol for the aerial survey of penguin colonies using UAVs.
J. Unmanned Veh. Syst. Published online March 31, 2015. http://dx.doi.
org/10.1139/juvs-2015-0006.
11. Mulero-Pa
´
zma
´
ny, M., Stolper, R., van Essen, L.D., Negro, J.J., and
Sassen, T. (2014). Remotely piloted aircraft systems as a rhinoceros
anti-poaching tool in Africa. PLoS ONE 9, e83873.
12. Weisenberger, M.E., Krausma n, P.R., Wallace, M.C., Young, D.W.D., and
Maughan, O.E. (1996). Effects of simulated jet aircraft noise on heart rate
and behavior of desert ungulates. J. Wildl. Manage. 60, 52–61.
13. Ward, D.H., Stehn, R.A., Erickson, W.P., and Derksen, D.V. (1999).
Response of fall-staging brant and Canada geese to aircraft overflights
in southwestern Alaska. J. Wildl. Manage. 63, 373–381.
14. Tracey, J.P., and Fleming, P.J.S. (2007). Behavioural responses of feral
goats (Capra hircus) to helicopters. Appl. Anim. Behav. Sci. 108, 114–128.
2282 Current Biology 25, 2278–2283, August 31, 2015 ª2015 Elsevier Ltd All rights reserved