ArticlePDF Available

Using tri-axial accelerometers to identify wild polar bear behaviors

Authors:

Abstract and Figures

Tri-axial accelerometers have been used to remotely identify the behaviors of a wide range of taxa. Assigning behaviors to accelerometer data often involves the use of captive animals or surrogate species, as accelerometer signatures are generally assumed to be similar to those of their wild counterparts. However, this has rarely been tested. Validated accelerometer data are needed for polar bears Ursus maritimus to understand how habitat conditions may influence behavior and energy demands. We used accelerometer and water conductivity data to remotely distinguish 10 polar bear behaviors. We calibrated accelerometer and conductivity data collected from collars with behaviors from video-recorded captive polar bears and brown bears U. arctos, and with video from camera collars deployed on free-ranging polar bears on the sea ice and on land. We used random forest models to predict behaviors and found strong ability to discriminate the most common wild polar bear behaviors using a combination of accelerometer and conductivity sensor data from captive or wild polar bears. In contrast, models using data from captive brown bears failed to reliably distinguish most active behaviors in wild polar bears. Our ability to discriminate behavior was greatest when species- and habitat-specific data from wild individuals were used to train models. Data from captive individuals may be suitable for calibrating accelerometers, but may provide reduced ability to discriminate some behaviors. The accelerometer calibrations developed here provide a method to quantify polar bear behaviors to evaluate the impacts of declines in Arctic sea ice.
Content may be subject to copyright.
ENDANGERED SPECIES RESEARCH
Endang Species Res
Vol. 32: 19– 33, 2017
doi: 10.3354/esr00779 Published January 12
INTRODUCTION
Knowledge of an animal’s behavior can inform
species conservation and management by revealing
how individuals respond to environmental conditions
(Suther land 1998, Caro 1999, Cooke et al. 2014).
Although visual observation is the most direct
method to study animal behavior, it is impractical for
many species. Innovations in electronic logging and
tracking devices have provided new methods to
© The authors 2017. Open Access under Creative Commons by
Attribution Licence. Use, distribution and reproduction are un -
restricted. Authors and original publication must be credited.
Publisher: Inter-Research · www.int-res.com
*Corresponding author: apagano@usgs.gov
Using tri-axial accelerometers to identify wild
polar bear behaviors
A. M. Pagano1,2,*, K. D. Rode1, A. Cutting3, M. A. Owen4, S. Jensen5, J. V. Ware6,
C. T. Robbins7, G. M. Durner1, T. C. Atwood1, M. E. Obbard8, K. R. Middel8,
G. W. Thiemann9, T. M. Williams2
1US Geological Survey, Alaska Science Center, 4210 University Dr., Anchorage, AK 99508, USA
2Department of Ecology & Evolutionary Biology, University of California, Santa Cruz, CA 95060, USA
3Oregon Zoo, Portland, OR 97221, USA
4Institute for Conservation Research, San Diego Zoo Global, San Diego, CA 92027, USA
5Alaska Zoo, Anchorage, AK 99507, USA
6Department of Integrative Physiology and Neuroscience, Washington State University, Pullman, WA 99164, USA
7School of the Environment and School of Biological Sciences, Washington State University, Pullman, WA 99164, USA
8Wildlife Research and Monitoring Section, Ontario Ministry of Natural Resources and Forestry, Trent University,
Peterborough, ON K9L 0G2, Canada
9Faculty of Environmental Studies, York University, Toronto, ON M3J1P3, Canada
ABSTRACT: Tri-axial accelerometers have been used to remotely identify the behaviors of a wide
range of taxa. Assigning behaviors to accelerometer data often involves the use of captive animals
or surrogate species, as their accelerometer signatures are generally assumed to be similar to
those of their wild counterparts. However, this has rarely been tested. Validated accelerometer
data are needed for polar bears Ursus maritimus to understand how habitat conditions may in -
fluence behavior and energy demands. We used accelerometer and water conductivity data to
remotely distinguish 10 polar bear behaviors. We calibrated accelerometer and conductivity data
collected from collars with behaviors observed from video-recorded captive polar bears and
brown bears U. arctos, and with video from camera collars deployed on free-ranging polar bears
on sea ice and on land. We used random forest models to predict behaviors and found strong
ability to discriminate the most common wild polar bear behaviors using a combination of
accelerometer and conductivity sensor data from captive or wild polar bears. In contrast, models
using data from captive brown bears failed to reliably distinguish most active behaviors in wild
polar bears. Our ability to discriminate behavior was greatest when species- and habitat-specific
data from wild individuals were used to train models. Data from captive individuals may be
suitable for calibrating accelerometers, but may provide reduced ability to discriminate some
behaviors. The accelerometer calibrations developed here provide a method to quantify polar
bear behaviors to evaluate the impacts of declines in Arctic sea ice.
KEY WORDS: Activity · Behavior · Polar bear · Ursus maritimus · Acceleration · Accelerometer ·
Brown bear · Ursus arctos
O
PEN
PEN
A
CCESS
CCESS
Endang Species Res 32: 19– 33, 2017
study the behavior, movement, physiology, energetic
rates, and environmental conditions of wildlife that
may otherwise be difficult or impossible to monitor
(Ropert-Coudert & Wilson 2005, Cooke 2008, Wilson
et al. 2008, Bograd et al. 2010, Costa et al. 2010,
Wilmers et al. 2015).
Tri-axial accelerometers, which collect high fre-
quency measures of acceleration in the form of gra -
vitational and inertial velocity (Brown et al. 2013),
have provided a means to remotely identify animal
behaviors (Yoda et al. 1999, Watanabe et al. 2005).
Accelerometers have been particularly useful in
studying widely dispersed animals or those occurring
in remote habitats, such as marine mammals and
birds (Brown et al. 2013). Once calibrated, tri-axial
accelerometer data from wild animals can be used to
remotely identify behaviors such as resting, walking,
running, and even feeding events (Yoda et al. 2001,
Shepard et al. 2008, Wilson et al. 2008, Watanabe &
Takahashi 2013, Williams et al. 2014). Calibration
typically involves time-synchronizing behavioral ob -
servations with their associated accelerometer read-
ings, which often necessitates the use of captive ani-
mals or surrogate species (e.g. Yoda et al. 2001,
Shepard et al. 2008, Nathan et al. 2012, Campbell et
al. 2013). Alternatively, animal-borne video cameras
can be used to directly calibrate accelerometers (e.g.
Watanabe & Takahashi 2013, Nakamura et al. 2015,
Volpov et al. 2015), but cameras can be expensive
and can only collect data over limited durations.
Polar bears Ursus maritimus typically occupy re -
mote environments, and few quantitative data exist
on their behaviors or activity budgets. Much of what
is known about polar bear behavior on the sea ice
comes from coastal indigenous resident knowledge
(e.g. Nelson 1966, Kalxdorff 1997, Kochnev et al.
2003, Voorhees et al. 2014) and direct observational
research limited to 2 locations over limited time pe -
riods (Stirling 1974, Stirling & Latour 1978, Hansson
& Thomassen 1983, Stirling et al. 2016). Satellite
telemetry has been used to track polar bears in some
subpopulations since the late 1970s (Schweinsburg &
Lee 1982, Taylor 1986) and has helped to identify
important habitats (Ferguson et al. 2000, Mauritzen
et al. 2003, Durner et al. 2009, Wilson et al. 2014).
However, detailed behavioral data in association
with habitat conditions are lacking. Recent declines
in Arctic sea ice have already caused declines in
abundance, survival, or body condition of polar bears
in some subpopulations (Stirling et al. 1999, Regehr
et al. 2007, Rode et al. 2010, 2012, Bromaghin et al.
2015, Obbard et al. 2016) and models project increas-
ing negative impacts in the 21st century (Amstrup et
al. 2008, Hunter et al. 2010, Molnár et al. 2010,
Atwood et al. 2016). In order to better predict the
impacts of projected sea ice loss on polar bears, it will
be important to understand the behavioral and
physio logical mechanisms driving current declines
(Vongraven et al. 2012, Atwood et al. 2016). Ac -
celerometers could be used in combination with
satellite telemetry to better understand the behav-
ioral consequences of sea ice loss. This mechanistic
information would allow for improved assessment of
the relationships between habitat loss, individual
health, and vital rates in polar bear populations.
In this study, we developed a method to quantify
wild polar bear behaviors using accelerometers and
conductivity sensor data, validated through animal-
borne video camera data. Additionally, we evaluated
the effectiveness of using accelerometer data from
captive polar and brown bears U. arctos to predict
behaviors of wild polar bears. Though it is generally
assumed that accelerometer signatures of captives or
surrogates are similar to those of their instrumented
wild counterparts (Williams et al. 2014, McClune et
al. 2015, Wang et al. 2015, Hammond et al. 2016), this
has rarely been tested. Captive individuals may
exhibit different behaviors and/or kinematics than
wild counterparts (McPhee & Carlstead 2010), which
could potentially influence accelerometer signatures.
Because polar bears use both sea ice and terrestrial
habitats and because differences in habitat substrate
or gradient could also affect accelerometer signa-
tures (Bidder et al. 2012, Shepard et al. 2013,
McClune et al. 2014), we examined data from wild
polar bears in both of these habitats. Lastly, because
sampling frequency affects the longevity of accelero -
meters during deployment as well as computational
power for analyses, we evaluated the ability of acce -
lero meters to predict wild polar bear behaviors using
3 different sampling frequencies (16, 8, and 4 Hz).
MATERIALS AND METHODS
Accelerometer recordings on captive bears
We deployed collars with archival loggers (TDR10-
X-340D; Wildlife Computers) on 3 adult female polar
bears Ursus maritimus housed at the Alaska Zoo,
Oregon Zoo, and San Diego Zoo, USA, as well as 2
adult female brown bears U. arctos housed at the
Bear Research, Education, and Conservation Center
at Washington State University (WSU; Table 1), USA.
Ar chi val loggers recorded tri-axial acceleration (m
s−2) at 16 Hz (range: ±20 m s−2), time-of-day, and
20
Pagano et al.: Accelerometers identify polar bear behaviors
wet/dry conductions (via an on-board conductivity
sensor; Fig. 1). Conductivity data were sampled at
1 Hz. Bears at the Oregon and San Diego Zoos were
trained to voluntarily place their heads into crates in
which collars could be applied or removed, and wore
collars for 1 to 4 h sessions. Bears at the Alaska Zoo
and WSU were anesthetized for collaring, with a
combination of tiletamine HCl and zolazepam HCl
(Telazol®; Pfizer Animal Health) and dexmedeto -
midine HCl (Dexdomitor®; Pfizer Animal Health)
(Teisberg et al. 2014). Following collar placement, the
effect of the anesthetic were reversed with atipame-
zole HCl (Antisedan®; Pfizer Animal Health). We
used release mechanisms (Lotek Wireless) to remove
collars from bears at the Alaska Zoo and WSU. We
matched accelerometer recordings to
the behaviors of captive bears while
they moved freely around enclosures
based on visual examination of time-
stamped video recordings (Sony cam-
corder model DCR-TRV280 or OpenEye
Digital Video Security Solutions).
Accelerometer recordings on
free-ranging polar bears
We deployed GPS-equipped video
camera collars (Exeye) and archival log-
gers (TDR10-X-340D; Wild life Comput-
ers) on 4 adult female polar bears and 1
subadult female polar bear captured on
the sea ice of the southern Beaufort Sea
in April 2014 and 2015 (hereafter ‘ice bears’) and 2
subadult polar bears (1 male and 1 female) captured
on land on Akimiski Island, Nunavut, Canada, in
September 2015 (hereafter ‘land bears’; Table 1).
Video collars, including archival loggers and release
mechanisms, weighed 1.6 to 2.1 kg (0.8 to 1.5% of
body mass of bears in this study). We captured polar
bears by injecting them with immobilizing drugs
through projectile syringes fired from a helicopter.
On the sea ice, we anesthe tized bears using a combi-
nation of tiletamine HCl and zolazepam HCl (Tela-
zol®) with no reversal (Stirling et al. 1989). On land,
we anesthetized bears with a combination of medeto-
midine (Domitor®; Pfizer Animal Health) and tileta-
mine HCl and zolazepam HCl (Telazol®) and
reversed with atipamezole HCl (Antisedan®) (Cattet
et al. 1997). Archival loggers were attached to collars
in the same location and orientation as captive
deployments (Fig. 1) and similarly recorded tri-axial
acceleration at 16 Hz (range: ±20 m s−2), time-of-day,
and wet/dry conductions (via an on-board conductiv-
ity sensor). Conductivity data were sampled at 1 Hz.
Video cameras were programmed to record at vary-
ing frequencies during daylight periods (see
Table S1 in the Supplement at www.int-res.com/
articles/ suppl/ n032 p019 _ supp. pdf) and programmed
to turn off if the temperature of the collar fell below
−17°C to protect video equipment. Collars deployed
on ice and land bears were recovered 4 to 23 d fol-
lowing deployment, either by recapture of the indi-
vidual or by remote activation of the collar release
and retrieval of the dropped collar by the field crew.
We matched accelerometer data to behavior of ice
and land bears based on visual examination of the
time-stamped video recordings from the collar.
21
Species Sex Age class Body Location
mass (kg)
Polar bear Female Adult 288 Alaska Zoo
Polar bear Female Adult 212 Oregon Zoo
Polar bear Female Adult 237 San Diego Zoo
Brown bear Female Adult 126 Washington State University
Brown bear Female Adult 126 Washington State University
Polar bear Female Adult 173 Southern Beaufort Sea
Polar bear Female Adult 176 Southern Beaufort Sea
Polar bear Female Adult 199 Southern Beaufort Sea
Polar bear Female Adult 172 Southern Beaufort Sea
Polar bear Female Subadult 141 Southern Beaufort Sea
Polar bear Male Subadult 186 Akimiski Island
Polar bear Female Subadult 140 Akimiski Island
Table 1. Polar bears Ursus maritimus and brown bears U. arctos wearing col-
lars with tri-axial accelerometers that were video recorded (captive bears) or
that wore video-equipped collars (wild bears)
Fig. 1. Orientation of an archival logger containing a tri-
axial accelerometer attached to a collar for use on polar
Ursus maritimus and brown bears U. arctos
Endang Species Res 32: 19– 33, 2017
Behaviors
Behaviors were annotated based on the video data
on a per second basis. For bears that were anesthe -
tized, we excluded behaviors on the day of capture
and during retrieval of the collar. Resting behaviors
included standing, sitting, and lying down. Head
movements while standing, sitting, or lying down
were included as resting behaviors, but limb move-
ments were treated as transitionary behaviors (Knud-
sen 1978, Williams 1983). Swimming included
surface swimming and diving. We excluded from
ana lyses any behaviors that were not indicative of
natural movements in captive bears (e.g. stereotypic
behaviors), were rare (e.g. fighting, breeding, drink-
ing), were transitionary, or were non-descript.
Modeling
We derived summary statistics from the accelerom-
eter data and linked the accelerometer data with cor-
responding behaviors of interest (SAS version 9.3;
SAS Institute). We converted accelerometer meas-
ures from m s−2 to g(1 g= 9.81 m s−2). We calculated
magnitude (Q) as a fourth dimension, where
(Nathan et al. 2012). We used a 2 s running mean of
the raw acceleration data to calculate static accelera-
tion (gravitational acceleration) and subtracted the
static acceleration from the raw acceleration data to
calculate dynamic acceleration (Wilson et al. 2006,
Shepard et al. 2008). We calculated overall dynamic
body acceleration (ODBA) as the absolute sum of
dynamic acceleration across the 3 axes (Wilson et al.
2006). We used a Fast Fourier Transform to calculate
the dominant power spectrum (dps) and frequency
(fdps) for each axis (Watanabe et al. 2005, Shamoun-
Baranes et al. 2012). In total, we derived 25 predictor
variables based on previous accelerometer studies
(e.g. Watanabe et al. 2005, Nathan et al. 2012,
Shamoun-Baranes et al. 2012, Wang et al. 2015). Pre-
dictor variables were extracted from the accelerome-
ter data over 2 s intervals; mean conductivity data
(wet/dry) was also extracted over 2 s intervals using
program R (R Core Team 2014) (Table 2). Video-
linked behaviors that lasted less than 2 s were
excluded from analyses. We used a random forest
supervised machine learning algorithm (Breiman
2001) in R (‘RandomForest’ package) to predict polar
bear behaviors. Random forest models use multiple
classification trees from a random subset of predictor
variables and then replicate this process over multi-
ple iterations using a subset of the data for each iter-
ation to determine the best variables for making pre-
dictions (Breiman 2001). An estimate of error is
derived by using the remaining data not used in each
iteration to test the predictive ability of the model,
which is termed the ‘out-of-bag’ (OOB) error rate
Q heave +surge +sway
222
=
22
Parameter Label Definition
Static acceleration (g) staticX, staticY, staticZ, staticQ Mean static acceleration along the surge,
heave, sway, and magnitude axes
Maximum dynamic body mdbaX, mdbaY, mdbaZ, mdbaQ Maximum dynamic body acceleration along the
acceleration (g) surge, heave, sway, and magnitude axes
Standard deviation dynamic stdbaX, stdbaY, stdbaZ, stdbaQ Standard deviation dynamic body acceleration
body acceleration (g) along the surge, heave, sway, and magnitude axes
Overall dynamic body odbaX, odbaY, odbaZ, ODBA Mean dynamic acceleration body acceleration along
acceleration (g) the surge, heave, and sway axes. ODBA = odbaX +
odbaY + odbaZ
Dominant power spectrum dpsX, dpsY, dpsZ, dpsQ Maximum power spectral density of dynamic accelera-
(g2Hz−1) tion along the surge, heave, sway, and magnitude axes
Frequency at the dominant fdpsX, fdpsY, fdpsZ, fdpsQ Frequency at the maximum power spectral density of
power spectrum (Hz) dynamic acceleration along the surge, heave, sway,
and magnitude axes
Mean wet/dry wetdry Mean conductivity determination of whether the tag
is wet or dry (range: 0−255)
Table 2. Parameters extracted from tri-axial accelerometer and conductivity data and used in random forest models to predict
wild polar bear Ursus maritimus behaviors. Respective acceleration measures from the surge (X), heave (Y), sway (Z), and
magnitude (Q) axes
Pagano et al.: Accelerometers identify polar bear behaviors
(Breiman 2001, Liaw & Wiener 2002). The random
forest algorithm has previously shown high accuracy
(>80%) for predicting animal behaviors from
accelerometer data (Nathan et al. 2012, Resheff et al.
2014, Graf et al. 2015, Lush et al. 2015, Rekvik 2015,
Wang et al. 2015, Alvarenga et al. 2016). We fit 500
classification trees to each training dataset and used
a random subset of 5 predictor variables for each split
in the tree.
Analyses
Unbalanced datasets can bias the predictive ability
of classification algorithms toward the most dominant
classes (Chen et al. 2004). Therefore, we performed 3
initial analyses to test the effect of uneven distribu-
tions on predictive ability. The first analysis used an
uneven distribution in which for ice and land bears,
we randomly selected 70% of each behavior for the
training dataset and used the remaining 30% to test
the predictive ability of the random forest algorithm
(e.g. Nathan et al. 2012, Alvarenga et al. 2016). For
captive polar and brown bears we used the entire
datasets to train the random forest algorithm. The
second analysis used a subsampling approach in
which we attempted to reduce the uneven distribu-
tion of more frequent behaviors (e.g. resting) in our
training dataset. To reduce the uneven distribution of
behaviors in the dataset from ice bears, we randomly
selected 5% of the resting behaviors, 30% of the
walking behaviors, and 70% of each of the remain-
ing behaviors for training the random forest algo-
rithm. We used the remaining data from ice bears for
testing predictions. To reduce the uneven distribu-
tion of the dataset from land bears, we randomly
selected 5% of the resting behaviors and 70% of
each of the remaining behaviors for training and
used the remaining data to test predictions. To re -
duce the uneven distribution of the datasets from
captive polar bears and brown bears, we randomly
selected 10% of the resting behaviors, 30% of the
walking behaviors, and 100% of each of the remain-
ing behaviors for training the random forest algo-
rithm. The third analysis used a completely balanced
distribution in which we used identical sample sizes
of 500 observations for each behavior in the training
dataset and the remaining observations to test and
excluded behaviors with less than 500 observations.
Based on these 3 analyses, we used the sampling
distribution (i.e. uneven, subsampled, or balanced
distribution) with the greatest predictive ability for
further analyses.
To evaluate our ability to predict behaviors of ice
bears, we used 3 different datasets to train the ran-
dom forest models and evaluated the ability of each
of these models. First, we used a random subset of
the data from ice bears as the training dataset and
the remaining data from ice bears to test predictions
(testing dataset). Second, we used the data from cap-
tive polar bears as the training dataset. Third, we
used the data from captive brown bears as the train-
ing dataset.
To evaluate our ability to predict behaviors of land
bears, we conducted 4 additional analyses. First, we
used a random subset of the data from land bears as
the training dataset and the remaining data from
land bears to test predictions (testing dataset). Sec-
ond, we used the training data from ice bears as the
training dataset. Third, we used the training data
from captive polar bears as the training dataset.
Fourth, we used the training data from captive brown
bears as the training dataset.
To examine the effect of sampling frequency on
our ability to discriminate behaviors, we subsampled
our 16 Hz accelerometer data to lower data acquisi-
tion rates of 8 and 4 Hz using SAS, and repeated the
predictive analyses above for both ice and land
bears.
Predicted behaviors were categorized as true posi-
tive (TP) if they correctly matched the actual behav-
ior, true negative (TN) if they correctly identified as a
different behavior, false positive (FP) if they incor-
rectly identified the behavior, and false negative
(FN) if they incorrectly identified as a different be -
havior. We evaluated the predictive abilities of these
models based on Matthews’ correlation coefficient
(MCC; e.g. Basu et al. 2013, Martins et al. 2016), the
percent precision, recall, and F-measure. We used
MCC in place of accuracy due to the unbalanced
nature of our dataset.
MCC, ,
provides a measure of the agreement between the
predicted and actual classifications, where +1 repre-
sents a perfect prediction and −1 represents total dis-
agreement (Matthews 1975). Precision is the propor-
tion of positive classifications that were correctly
classified (TP/TP + FP), recall is the probability that a
behavior will be correctly classified (TP/TP + FN),
and F-measure is the harmonic mean of precision
and recall (2 × precision × recall/precision + recall).
We used 2 sample t-tests to evaluate whether MCC,
precision, and recall differed significantly using a
16 Hz sampling frequency compared to either an 8 or
TP TNFP FN
TP+FP TP+FN TN+FP TN+FN
××
()
×
()
×
()
×
()
23
Endang Species Res 32: 19– 33, 2017
4 Hz sampling frequency based on the ice and land
datasets.
RESULTS
Behavior on the sea ice
Video collars on ice bears Ursus maritimus re -
corded 14 to 55 h of video (x
= 38 h, SD = 17 h, n = 5).
For predicting the behavior of ice bears, we collected
a total of 140 h of video-linked accelerometer data
from ice bears, 37 h from captive polar bears, and
72 h from captive brown bears U. arctos. We identi-
fied 10 different behaviors from ice bears, with rest-
ing, walking, and eating being the most prevalent
(Table 3). Ice bears ate recently killed adult, sub -
adult, or pup ringed seals Pusa hispida, seal carcas -
ses, bowhead whale Balaena mysti cetus carcasses, or
unidentifiable carcasses. Captive polar bears con-
sumed fish, and captive brown bears ate dry omni-
vore chow. Captive brown bears also grazed on
grass, which was excluded from analyses predicting
behaviors of ice bears, but was included as eating for
predicting behaviors of land bears.
Our models using an uneven distribution of behav-
iors in which we used 70% of each behavior from ice
bears and all of the available data from captive polar
or brown bears exhibited 5% greater predictive abil-
ity overall compared to the subsampled distribution,
and 7% greater predictive ability overall compared
to the balanced distribution based on
F-measure (see Table S2 in the Sup-
plement at www. int-res. com/ articles/
suppl/ n032 p019 _ supp. pdf). In particu-
lar, the data sets with an uneven distri-
bution ex hibited greater ability to dis-
criminate less frequent behaviors such
as swimming, eating, and running
(Table S2). Therefore, we used the
datasets with uneven distributions for
subsequent analyses (Table 3).
Our model with training data from
ice bears had an OOB error rate of
2.0% and exhibited the greatest pre-
dictive abilities for all 10 behaviors
(Fig. 2) compared to all other models
tested. Our models with training data
from captive polar bears and brown
bears had OOB error rates of 3.7 and
0.5% respectively, indicating that both
models performed well in discriminat-
ing captive behaviors. Both the ice
bear and captive polar bear models exhibited strong
predictive ability for identifying resting and walking
behaviors in wild bears (>90% MCC, precision, re -
call, and F-measure; Table 4 & Table S3 in the Sup-
plement). Predictive abilities for other behaviors var-
ied, with swimming and head shaking exhibiting
strong predictive ability using the ice bear model
(>75% MCC, precision, recall, and F-measure), but
lower predictive ability for eating, running, pounc-
ing, grooming, digging, and rolling (Fig. 2, Tables 4
& 5). The model from ice bears had particularly
greater ability than the captive polar bear model for
24
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Head
shake
Roll
F-measure
Rest Walk Swim Eat Run GroomPounce Dig
Behavior
Ice bears
Captive polar bears
Captive brown bears
Fig. 2. Ability (F-measure) of the random forest model to predict 10 behaviors
of polar bears Ursus maritimus on the sea ice from 3 different training
datasets of accelerometer data. Ice bears: polar bears on the sea ice of the
southern Beaufort Sea
Behavior Captive Captive Wild Wild
polar brown ice land
bears bears bears bears
Rest 53656 104838 163301 43 132
Walk 8962 33059 36 083 958
Swim 423 0 703 0
Eat 2108 973 2100 1529
Run 0 0 943 0
Pounce 458 0 49 0
Groom 3729 432 1138 289
Dig 405 0 1194 0
Head shake 86 14 435 19
Roll 87 0 1473 0
Table 3. Number of 2 s long behaviors used in random forest
training datasets for predicting behaviors of wild polar bears
Ursus maritimus. Ice bears: polar bears on the sea ice of the
southern Beaufort Sea. Land bears: polar bears on Akimiski
Island, Nunavut
Pagano et al.: Accelerometers identify polar bear behaviors
swimming, pouncing, and digging (Fig. 2, Table S3).
The captive brown bear model provided weaker abil-
ity to distinguish behaviors of ice bears for walking,
eating, and grooming (<65% MCC and F-measure),
but reliably distinguished resting (Fig. 2, Table S4).
Using the model from ice bears, eating had a high
rate of false positive classifications resulting from
digging behavior being incorrectly classified as eat-
ing (Table 5) as well as a high rate of false negative
classifications with eating behavior incorrectly classi-
fied as either resting or walking (Table 5). A post hoc
test using only feeding behavior while eating a
recently killed ringed seal within the training and
testing datasets failed to improve our ability to dis-
criminate eating (MCC = 0.61, precision = 0.67, recall
= 0.56, F-measure = 0.61). Additionally, running was
often misclassified as walking, whereas rolling was
often misclassified as resting (Table 5).
The most important predictors using the model from
ice bears were static acceleration in the heave
(staticY) and surge directions (staticX), wet/dry con-
ductivity (wetdry), and frequency at the dominant
power spectrum in the surge direction (fdpsX; Fig. 3).
Differences in the intensity of behaviors were dis-
cernible in the ODBA measures, with head shaking
having the greatest ODBA and resting having the
lowest (Table S5). Eating and swimming showed simi-
25
Behavior MCC Precision Recall F-measure
Rest 0.973 0.992 0.997 0.994
Walk 0.971 0.964 0.989 0.976
Swim 0.887 0.957 0.823 0.885
Eat 0.674 0.677 0.677 0.677
Run 0.709 0.835 0.604 0.701
Pounce 0.700 0.833 0.588 0.690
Groom 0.417 0.658 0.266 0.379
Dig 0.532 0.712 0.400 0.513
Head shake 0.818 0.839 0.798 0.818
Roll 0.754 0.821 0.696 0.753
Table 4. Performance of a random forest model using ac-
celerometer data from polar bears Ursus maritimus on the
sea ice to predict behaviors of bears on the sea ice as veri-
fied by video data. MCC: Matthews’ correlation coefficient
Rest Walk Swim Eat Run Pounce Groom Dig Head shake Roll
Rest 69760 31 33 99 1 0 281 28 0 111
Walk 115 15 295 10 111 153 4 15 102 26 41
Swim 6 2 246 1001001
Eat 35 51 2 608 1 0 45 145 0 11
Run 04401243 00111
Pounce 0000010 0200
Groom 17 2 0 33 0 0 129 1302
Dig 216044029203 09
Head shake 26003002146 15
Roll 451381115210437
Table 5. Cross-validation comparing predicted behaviors (rows) from accelerometer analyses of polar bears Ursus maritimus
on the sea ice to actual behaviors (columns) confirmed by video recordings. Correct classifications are denoted in bold. See
Table 4 for performance statistics in predicting behaviors
Fig. 3. Variable importance plot from the random forest
model of accelerometer data from polar bears on the sea ice.
The importance plot provides a relative ranking of parame-
ters in which higher values indicate parameters that con-
tributed more toward classification accuracy. Mean de-
crease in accuracy is normalized by dividing by the standard
errors of the parameters (i.e. z-score). See Table 2 for de-
scription of parameters
Endang Species Res 32: 19– 33, 2017
lar mean ODBA values, but had differing mean static
acceleration values (Table S5). Eating and grooming
had low values of static acceleration in the heave di-
rection (staticY), which was indicative of a head-down
posture. Walking and running exhibited periodic un-
dulating patterns in static acceleration in the heave
direction (staticY; Fig. 4 & Fig. S1 in the Supplement),
which was indicative of the bear’s head moving up
and down as it stepped. Wet/dry conductivity while
swimming was lower for wild polar bears (x
= 81.9, SD
= 81.5) than captive polar bears (x
= 205.3, SD = 57.8)
and lower than all other behaviors (all x
> 234). A post
hoc test excluding the conductivity variable reduced
the ability of the algorithm to correctly identify swim-
ming be haviors using the training data set for ice
bears (MCC = 0.47, pre cision = 0.77, recall = 0.29, F-
measure = 0.42) with a high rate of swimming behav-
iors misclassified as resting.
Behaviors on land
Video collars on land bears recorded 19 to 36 h of
video (x
= 27 h, SD = 12 h, n = 2) and in total we col-
lected 36 h of video-linked accelero meter data for the
behaviors of interest. We identified 5 different be ha -
viors from land bears, with resting being the most
prevalent, followed by eating (Table 3). Eating on
land consisted of berries, primarily crowberries Em -
petrum nigrum.
Our model with training data from land bears had
an OOB error rate of 0.5% and had the greatest
success in discriminating on-land behaviors (Fig. 5,
Table S6). All behaviors, except for grooming and
head shaking, had MCC, precision, recall, and F-
measure values >90% using the model from land
bears (Fig. 5, Table S6). In particular, the model from
land bears was able to distinguish eating (MCC =
0.95, precision = 0.95, recall = 0.96, F-measure =
0.95), which was not possible with the other datasets.
Our model with training data from ice bears had suc-
cess in discriminating resting behaviors on land
(MCC = 0.60, precision = 0.96, recall = 1.0, F-mea-
sure = 0.98) and walking on land (MCC = 0.82, preci-
sion = 0.89, recall = 0.76, F-measure = 0.82), but eat-
ing was often misclassified as resting or walking (FP).
The captive polar bear model performed similarly to
the model from ice bears for discriminating behaviors
on land (Fig. 5). The captive brown bear model per-
formed less well than the other models for discrimi-
nating walking on land, but otherwise performed
similarly to the models from ice bears and captive
polar bears (Fig. 5).
Sampling frequency
The OOB error rate using the data from ice bears in-
creased from 2.0 to 2.2% using an 8 Hz sampling fre-
quency and to 2.6% using a 4 Hz sampling frequency.
OOB error rate using data from land bears increased
from 0.5 to 0.6% at 8 Hz and to 0.8% at 4 Hz. Predic-
tive ability using an 8 Hz sampling frequency was
nearly identical to 16 Hz among all behaviors using
the dataset from ice bears (t58 = 0.70, p =0.24) and
land bears (t28 = 0.61, p =0.27) based on MCC, preci-
sion, and recall. Predictive ability using a 4 Hz sam-
pling frequency was lower than predictive ability us-
ing 16 Hz for ice bears (t55 = 1.8, p =0.04), but not for
land bears (t27 = 0.59, p =0.28). In particular, the
ability to discriminate the high intensity behaviors of
pouncing and head shaking declined using a 4 Hz
sampling rate (Fig. 6).
DISCUSSION
Our results show that tri-axial accelerometers in
combination with measures of conductivity can reli-
ably distinguish the 3 most common behaviors of
wild polar bears Ursus maritimus (resting, walking,
and swimming; Stirling 1974, Latour 1981, Hansson
& Thomassen 1983, Lunn & Stirling 1985). This will
provide a method to remotely document the activity
budgets of these far-ranging animals, which can be
further linked with location data from satellite collars
to examine the effects of habitat on behavior and
energy expenditure. Our results indicate that differ-
ences among habitats and species can impact the
ability to discriminate behaviors in wild individuals
using accelerometers. We found no loss in predictive
ability using an 8 Hz sampling frequency, which
would allow for twice the battery longevity of a 16 Hz
rate and reduce the computational power needed for
analyses. Although accelerometer studies on smaller
species appear to require greater sampling frequen-
cies (e.g. >30 Hz; Broell et al. 2013, Brown et al.
2013), our results are similar to data obtained by
Rekvik (2015) from captive brown bears U. arctos,
and by Wang et al. (2015) from captive mountain
lions Puma concolor, which both found little loss in
predictive ability at sampling frequencies 8Hz.
Habitat effects
Our results indicate that accelerometer signatures
on sea ice are similar to signatures on land for most
26
Pagano et al.: Accelerometers identify polar bear behaviors 27
Fig. 4. Accelerometer signatures of static acceleration in the surge (X), heave (Y), and sway (Z) directions and overall dynamic acceleration (ODBA) while walking,
swimming, standing, and eating a seal from an adult female polar bear Ursus maritimus on the sea ice of the southern Beaufort Sea
Endang Species Res 32: 19– 33, 2017
behaviors, but eating berries by land bears had a dis-
tinct signature that our ice bear model and captive
bear models misclassified as grooming, resting, or
walking. This highlights the value in linking obser-
vational and acce lero meter data from wild subjects
over multiple time periods and habitats, and the
importance of accounting for as many behaviors as
possible in training datasets. Knowledge of eating
frequency and du ration would provide insight in de -
termining foraging success, an im portant determi-
nant of individual re productive success
and survival (Stirling et al. 1999, Regehr et
al. 2007, 2010). Al though we had success
discriminating eating events by land bears,
we had lower precision and recall in dis-
criminating eating events by ice bears. This
was likely related in part to the movement
pattern of bears eating berries, in which
they typically stood with their head down
and grazed. Conversely, bears eating on
the sea ice exhibited a variety of positions
including standing, sitting, and lying
down, and both tore pieces of food from
seals or gnawed on carcasses. Since most
kill events involve bears pouncing on their
seal prey (Stirling 1988, Derocher 2012),
we may be able to identify successful kills
based on the combination of a pouncing
signature followed by eating signatures
(e.g. Williams et al. 2014), but this requires
further evaluation. Additionally, feeding on
a seal would typically last for a prolonged
period; hence, if the model primarily predicted eat-
ing over a prolonged period this could be used as an
indication of a feeding event, but this also requires
further evaluation.
Use of captive animals and surrogate species
Our ability to discriminate behaviors was greatly
improved by including data from free-ranging polar
28
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Roll
F-measure
Rest Walk Swim Eat Run Pounce Groom Dig Head
shake
Behavior
16 Hz
8 Hz
4 Hz
Fig. 6. Ability (F-measure) of a random forest model to predict behaviors of polar bears Ursus maritimus on the sea ice using
3 different accelerometer sampling frequencies
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Rest Walk Eat Groom Head shake
F-measure
Behavior
Land bears
Ice bears
Captive polar bears
Captive brown bears
Fig. 5. Ability (F-measure) of a random forest model to predict behaviors
of polar bears Ursus maritimus on land from 4 different training datasets.
Ice bears: polar bears on the the sea ice of the southern Beaufort Sea.
Land bears: polar bears on Akimiski Island, Nunavut
Pagano et al.: Accelerometers identify polar bear behaviors
bears rather than using data from captive bears
alone. However, resting and walking could be reli-
ably discriminated using data from either captive or
wild polar bears. This illustrates the value of collect-
ing data from captive individuals when data collec-
tion is difficult or impossible from wild counterparts.
However, data from captive brown bears exhibited
poorer performance for predicting active behaviors
in wild polar bears. This may be related to differ-
ences in walking kinematics between polar and
brown bears as well as potential differences in limb
lengths between the species (Renous et al. 1998).
Additionally, polar bears have longer necks relative
to their body size than other ursid species (DeMaster
& Stirling 1981), which could also affect accelero -
meter signatures from a neck-worn collar. Although
Campbell et al. (2013) proposed the use of surrogate
species to predict the behaviors of other species, our
findings suggest that polar bear accelerometer signa-
tures are likely species- and habitat-specific, at least
for distinguishing specific behaviors. The brown bear
model did reliably distinguish resting behavior in
wild polar bears, which suggests that surrogate spe-
cies could be used to distinguish coarse activity pat-
terns such as active versus inactive (e.g. Gervasi et
al. 2006, Ware et al. 2015).
Our analyses indicate that conductivity measures
are needed to reliably discriminate swimming.
Greater conductivity measures in captive polar bears
that were swimming in fresh water likely caused the
poorer performance for discriminating swimming in
wild polar bears that were swimming in salt water.
For pouncing, captive polar bears pounced on large
plastic barrels, which resulted in similar measures of
ODBA as wild counterparts, but had different signa-
tures of static acceleration (i.e. body posture). Dig-
ging by wild bears, which was often through snow
and ice into subnivean lairs to locate seals, exhibited
greater ODBA measures and slightly different static
acceleration than captive bears digging in snow and
ice. These results suggest that some behaviors of
captive bears may not fully reflect behaviors of their
wild counterparts, which further illustrates the value
of collecting simultaneous observational data (e.g. vi -
deo) from free-ranging individuals to calibrate ac -
celerometer-based behavioral data.
Accelerometer attachment
Regardless of which training dataset was used,
we found lower precision and recall for predicting
5 of the behaviors tested for bears on the sea ice.
Eating, grooming, and rolling had high rates of
misclassifications as resting, whereas running and
digging had high rates of misclassifications as
walking. These re sults suggest the random forest
algorithm could be prone to slightly overestimate
the amount of true resting and walking behaviors
in quantifying activity budgets. Our lower precision
and recall for discriminating some behaviors was
likely due in part to the attachment of the accel -
erometer on a collar. Al though a number of studies
have successfully discriminated behaviors using
accelerometers on collars (Watanabe et al. 2005,
Martiskainen et al. 2009, Soltis et al. 2012, McClune
et al. 2014, Lush et al. 2015, Rekvik 2015, Wang et
al. 2015), many of these studies limited their analy-
ses to 4 or 5 behaviors or documented high misclas-
sification rates for distinguishing some behaviors.
Wang et al. (2015) similarly reported low accuracy
of accelerometers on collars for predicting eating
and grooming by captive mountain lions, and Lush
et al. (2015) reported low ac curacy for predicting
some behaviors, including grooming, in wild brown
hares Lepus europaeus. Attach ment of the accel -
erometer to a collar, as opposed to attachment
directly on the animal, likely introduces noise in
the data due to independent collar motion (i.e. the
collar must be fitted to ensure animals do not
remove it, but loose enough to accommodate
potential changes in body mass) and may reduce
the ability of the accelerometer to detect some low
intensity movements (Shepard et al. 2008). The
effect of independent collar motion is evident in
our large values of ODBA when bears shook their
heads. This behavior may be useful for identifying
the end of a swim, as bears are known to shake
and roll in the snow following a swim. Additionally,
our ability to discriminate head shaking allows for
excluding it from potential energetic analyses using
accelerometers. Use of a higher sampling frequency
than was used in this study (i.e. >16 Hz) could
potentially improve the ability to discriminate some
fine-scale body movements (Nathan et al. 2012)
such as eating, though Wang et al. (2015) sampled
at 64 Hz and had low accuracy in discriminating
eating behaviors of captive mountain lions.
Video calibration
Having video-linked observational data from cam-
era-mounted collars on wild polar bears was the most
practical method to calibrate accelerometers on free-
ranging individuals. However, because the animal’s
29
Endang Species Res 32: 19– 33, 2017
body was not visible in the video, some behaviors
may have been incorrectly classified. For example,
distinguishing walking versus running was often
challenging, as was determining when bears were
actively swimming versus resting in the water. Both
of these could have contributed to the misclassifica-
tions between running and walking and swimming
and resting. Additionally, the models had greater
success discriminating behaviors as sample sizes
increased. Although unbalanced datasets are known
to affect the predictive ability of random forest algo-
rithms (Chen et al. 2004), we found that the inclusion
of larger sample sizes in the training dataset was
more important than imbalance. This highlights the
value of calibrating accelerometers from multiple
individuals over prolonged periods.
CONCLUSIONS
Our results underscore the importance of thor-
oughly validating accelerometers for use in remote
detection of behavior, ideally on a species- and habi-
tat-specific level. The use of tri-axial accelerometers,
as shown here, will enable detailed assessments of
polar bear behaviors to better understand polar bear
habitat use and the implications for energy demands.
For example, measures of acceleration could be com-
bined with measures of oxygen consumption from
captive bears while resting, walking, and swimming
to both quantify activity budgets and estimate the
energetic costs of these behaviors (e.g. Wilson et al.
2006, 2012, Halsey et al. 2009, 2011, Gómez Laich et
al. 2011, Williams et al. 2014). Future advances are
needed that would enable remote transmission of
raw accelerometer data to further enhance the appli-
cability of these devices to animals occurring in
remote environments and obviate the need for sensor
recovery. As declines in sea ice are expected to in -
crease the activity rates of polar bears across much of
their range (Derocher et al. 2004, Molnár et al. 2010,
Sahanatien & Derocher 2012), the use of accelero -
meters provides a method to monitor the impacts of
habitat change on activity and energy budgets to
better understand the implications for body condi-
tion, reproductive success, and survival of this Arctic
apex predator.
Acknowledgements. Procedures were approved by the Ani-
mal Care and Use Committees of the University of Califor-
nia, Santa Cruz; Washington State University; York Univer-
sity; Oregon Zoo; Alaska Zoo; San Diego Zoo; Ontario
Ministry of Natural Resources and Forestry; and the US
Geological Survey, Alaska Science Center. Research in the
USA was permitted under US Fish and Wildlife Service per-
mits MA690038 and MA95406A. Research on Akimiski
Island, Nunavut was approved under Nunavut Wildlife
Research Permit WL 2015-073. This work was supported by
US Geological Survey’s Changing Arctic Ecosystems Initia-
tive. Additional support was provided by Polar Bears Inter-
national; World Wildlife Fund (Canada); Ontario Ministry of
Natural Resources and Forestry; Natural Sciences and Engi-
neering Research Council of Canada; Born Free Foundation;
Helen McCrea Peacock Foundation; Institute for Conserva-
tion Research, San Diego Zoo Global; and the International
Association for Bear Research and Management. Support for
T.M.W. was provided by the National Science Foundation’s
Instrument Development for Biological Research program.
We thank Mehdi Bakhtiari (Exeye) for developing the video
collars used in this study. We thank the bear keeper and
trainer teams at the Oregon Zoo, San Diego Zoo, and Alaska
Zoo for enabling data collection at their facilities. We thank
Stephen Atkinson, Tyrone Donnelly, Katie Florko, Sarah
Hagey, Tim Moody, Kristin Simac, and Maria Spriggs for
assistance in the field. We thank helicopter pilots Frank Ross
(Soloy Helicopters) and Doug Holtby (OMNRF) for field sup-
port. B. Battaile and 3 anonymous reviewers provided valu-
able input on earlier versions of the manuscript. This re -
search used resources of the Core Science Analytics and
Synthesis Applied Research Computing program at the US
Geological Survey. Use of trade names is for descriptive pur-
poses only and does not imply endorsement by the US Gov-
ernment. This paper was reviewed and approved by the
USGS under its Fundamental Science Practices policy
(www.usgs.gov/fsp).
LITERATURE CITED
Alvarenga FAP, Borges I, Palkovi L, Rodina J, Oddy VH,
Dobos RC (2016) Using a three-axis accelerometer to
identify and classify sheep behaviour at pasture. Appl
Anim Behav Sci 181: 91−99
Amstrup SC, Marcot BG, Douglas DC (2008) A Bayesian
network modeling approach to forecasting the 21st cen-
tury worldwide status of polar bears. In: DeWeaver ET,
Blitz CM, Tremblay LB (eds) Arctic sea ice decline:
observations, projections, mechanisms, and implications.
Geophysics Monograph Series 180. American Geophysi-
cal Union, Washington, DC, p 213−268
Atwood TC, Marcot BG, Douglas DC, Amstrup SC, Rode
KD, Durner GM, Bromaghin JF (2016) Forecasting the
relative influence of environmental and anthropogenic
stressors on polar bears. Ecosphere 7: e01370
Basu I, Graupe D, Tuninetti D, Shukla P, Slavin KV, Metman
LV, Corcos DM (2013) Pathological tremor prediction
using surface EMG and acceleration: potential use in
‘ON-OFF’ demand driven deep brain stimulator design.
J Neural Eng 10: 036019
Bidder OR, Qasem LA, Wilson RP (2012) On higher ground:
How well can dynamic body acceleration determine
speed in variable terrain? PLOS ONE 7: e50556
Bograd SJ, Block BA, Costa DP, Godley BJ (2010) Biolog-
ging technologies: new tools for conservation. Introduc-
tion. Endang Species Res 10: 1−7
Breiman L (2001) Random forests. Mach Learn 45: 5−32
Broell F, Noda T, Wright S, Domenici P, Steffensen JF,
Auclair JP, Taggart CT (2013) Accelerometer tags: de -
30
Pagano et al.: Accelerometers identify polar bear behaviors
tecting and identifying activities in fish and the effect of
sampling frequency. J Exp Biol 216: 1255−1264
Bromaghin JF, McDonald TL, Stirling I, Derocher AE and
others (2015) Polar bear population dynamics in the
southern Beaufort Sea during a period of sea ice decline.
Ecol Appl 25: 634−651
Brown DD, Kays R, Wikelski M, Wilson R, Klimley AP (2013)
Observing the unwatchable through acceleration log-
ging of animal behavior. Anim Biotelem 1: 20
Campbell HA, Gao L, Bidder OR, Hunter J, Franklin CE
(2013) Creating a behavioural classification module for
acceleration data: using captive surrogate for difficult to
observe species. J Exp Biol 216: 4501−4506
Caro T (1999) The behavior-conservation interface. Trends
Ecol Evol 14: 366−369
Cattet MRL, Caulkett NA, Polischuk SC, Ramsay MA (1997)
Reversible immobilization of free-ranging polar bears
with medetomidine-zolazepam-tiletamine and atipame-
zole. J Wildl Dis 33: 611−617
Chen C, Liaw A, Breiman L (2004) Using random forest to
learn imbalanced data. University of California, Berke-
ley, CA. http: //statistics.berkeley. edu/ sites/ default/ files/
tech- reports/666.pdf (accessed Oct 2015)
Cooke SJ (2008) Biotelemetry and biologging in endan-
gered species research and animal conservation: rele-
vance to regional, national, and IUCN Red List threat
assessments. Endang Species Res 4: 165−185
Cooke SJ, Blumstein DT, Buchholz R, Caro T and others
(2014) Physiology, behavior, and conservation. Physiol
Biochem Zool 87: 1−14
Costa DP, Huckstadt LA, Crocker DE, McDonald BI, Goebel
ME, Fedak MA (2010) Approaches to studying climate
change and its role on the habitat selection of Antarctic
pinnipeds. Integr Comp Biol 50: 1018−1030
DeMaster DP, Stirling I (1981) Ursus maritimus. Mamm Spe-
cies 145: 1−7
Derocher AE (2012) Polar bears: a complete guide to their
biology and behavior. Johns Hopkins University Press,
Baltimore, MD
Derocher AE, Lunn NJ, Stirling I (2004) Polar bears in a
warming climate. Integr Comp Biol 44: 163−176
Durner GM, Douglas DC, Nielson RM, Amstrup SC and oth-
ers (2009) Predicting 21st-century polar bear habitat dis-
tribution from global climate models. Ecol Monogr 79:
25−58
Ferguson SH, Taylor MK, Messier F (2000) Influence of sea
ice dynamics on habitat selection by polar bears. Ecology
81: 761−772
Gervasi V, Brunberg S, Swenson JE (2006) An individual-
based method to measure animal activity levels: a test on
brown bears. Wildl Soc Bull 34: 1314−1319
Gómez Laich A, Wilson RP, Gleiss AC, Shepard ELC, Quin-
tana F (2011) Use of overall dynamic body acceleration
for estimating energy expenditure in cormorants: Does
locomotion in different media affect relationships? J Exp
Mar Biol Ecol 399: 151−155
Graf PM, Wilson RP, Qasem L, Hackländer K, Rosell F (2015)
The use of acceleration to code for animal behaviours; a
case study in free-ranging Eurasian beavers Castor fiber.
PLOS ONE 10: e0136751
Halsey LG, Green JA, Wilson RP, Frappell PB (2009)
Accelerometry to estimate energy expenditure during
activity: best practice with data loggers. Physiol Biochem
Zool 82: 396−404
Halsey LG, Shepard ELC, Wilson RP (2011) Assessing the
development and application of the accelerometry tech-
nique for estimating energy expenditure. Comp Biochem
Physiol A Mol Integr Physiol 158: 305−314
Hammond TT, Springthorpe D, Walsh RE, Berg-Kirkpatrick
T (2016) Using accelerometers to remotely and automat-
ically characterize behavior in small animals. J Exp Biol
219: 1618−1624
Hansson R, Thomassen J (1983) Behavior of polar bears with
cubs in the denning area. Int Conf Bear Res Manage 5:
246−254
Hunter CM, Caswell H, Runge MC, Regehr EV, Amstrup
SC, Stirling I (2010) Climate change threatens polar bear
populations: a stochastic demographic analysis. Ecology
91: 2883−2897
Kalxdorff SB (1997) Collection of local knowledge regard-
ing polar bear habitat use in Alaska. US Fish and
Wildlife Service Tech Rep MMM 87-2. Marine Mam-
mals Management, US Fish and Wildlife Service,
Anchorage, AK
Knudsen B (1978) Time budgets of polar bears (Ursus mar-
itimus) on North Twin Island, James Bay, during sum-
mer. Can J Zool 56: 1627−1628
Kochnev AA, Etylin VM, Kavry VI, Siv-Siv EB, Tanko IV
(2003) Traditional knowledge of Chukotka Native peo-
ples regarding polar bear habitat use. Final Report. Pre-
pared for the US National Park Service, Chukotka Asso-
ciation of Traditional Marine Mammal Hunters, Alaska
Nanuuq Commission, and Pacific Fisheries Research
Center (Chukotka Branch), Anchorage, AK
Latour PB (1981) Spatial relationships and behavior of polar
bears (Ursus maritimus Phipps) concentrated on land
during the ice-free season of Hudson Bay. Can J Zool 59:
1763−1774
Liaw A, Wiener M (2002) Classification and regression by
randomForest. R News 2/3: 18−22
Lunn NJ, Stirling I (1985) The significance of supplemental
food to polar bears during the ice-free period of Hudson
Bay. Can J Zool 63: 2291−2297
Lush L, Ellwood S, Markham A, Ward AI, Wheeler P (2015)
Use of tri-axial accelerometers to assess terrestrial mam-
mal behavior in the wild. J Zool (Lond) 298: 257−265
Martins M, Santos C, Costa L, Frizera A (2016) Feature
reduction and multi-classification of different assistive
devices according to the gait pattern. Disabil Rehabil
Assist Technol 11: 202−218
Martiskainen P, Järvinen M, Skön JP, Tiirikainen J, Koleh -
mainen M, Mononen J (2009) Cow behaviour pattern
recognition using three-dimensional accelerometer and
support vector machines. Appl Anim Behav Sci 119:
32−38
Matthews BW (1975) Comparison of the predicted and
observed secondary structure of T4 phage lysozyme.
Biochim Biophys Acta 405: 442−451
Mauritzen M, Belikov SE, Boltunov AN, Derocher AE and
others (2003) Functional response in polar bear habitat
selection. Oikos 100: 112−124
McClune DW, Marks NJ, Wilson RP, Houghton JDR and
others (2014) Tri-axial accelerometers quantify be -
haviour in the Eurasian badger (Meles meles): towards
an automated interpretation of field data. Anim Biotelem
2: 5
McClune DW, Marks NJ, Delahay RJ, Montgomery WI,
Scantlebury DM (2015) Behavior-time budget and func-
tional habitat use of free-ranging European badger
(Meles meles). Anim Biotelem 3: 7
31
Endang Species Res 32: 19– 33, 2017
McPhee ME, Carlstead K (2010) The importance of main-
taining natural behaviors in captive mammals. In:
Kleiman DG, Thompson KV, Baer CK (eds) Wild mam-
mals in captivity: principles and techniques for zoo man-
agement, 2nd edn. University of Chicago Press, Chicago,
IL, p 303−313
Molnár PK, Derocher AE, Thiemann GW, Lewis MA (2010)
Predicting survival, reproduction and abundance of
polar bears under climate change. Biol Conserv 143:
1612−1622
Nakamura I, Goto Y, Sato K (2015) Ocean sunfish rewarm at
the surface after deep excursions to forage for siphono -
phores. J Anim Ecol 84: 590−603
Nathan R, Spiegel O, Fortmann-Roe S, Harel R, Wikelski M,
Getz WM (2012) Using tri-axial acceleration data to iden-
tify behavioral modes of free-ranging animals: general
concepts and tools illustrated for griffon vultures. J Exp
Biol 215: 986−996
Nelson RK (1966) Alaska Eskimo exploitation of the sea ice
environment. Arctic Aeromedical Laboratory, Fort Wain-
wright, AK
Obbard ME, Cattet MRL, Howe EJ, Middel KR and others
(2016) Trends in body condition in polar bears (Ursus
maritimus) from the Southern Hudson Bay subpopula-
tion in relation to changes in sea ice. Arctic Sci 2: 15−32
R Core Team (2014) R: a language and environment for sta-
tistical computing. R Foundation for Statistical Comput-
ing, Vienna
Regehr EV, Lunn NJ, Amstrup SC, Stirling I (2007) Effects of
earlier sea ice breakup on survival and population size of
polar bears in western Hudson Bay. J Wildl Manag 71:
2673−2683
Regehr EV, Hunter CM, Caswell H, Amstrup SC, Stirling I
(2010) Survival and breeding of polar bears in the south-
ern Beaufort Sea in relation to sea ice. J Anim Ecol 79:
117−127
Rekvik GES (2015) Identification of five brown bear (Ursus
arctos) behaviors using tri-axial accelerometers. MSc
thesis, Telemark University College
Renous S, Gasc JP, Abourachid A (1998) Kinematic analysis
of the locomotion of the polar bear (Ursus maritimus,
Phipps, 1774) in natural and experimental conditions.
Neth J Zool 48: 145−167
Resheff YS, Rotics S, Harel R, Spiegel O, Nathan R (2014)
AcceleRater: a web application for supervised learning
of behavioral modes from acceleration measurements.
Mov Ecol 2: 27
Rode KD, Amstrup SC, Regehr EV (2010) Reduced body size
and cub recruitment in polar bears associated with sea
ice decline. Ecol Appl 20: 768−782
Rode KD, Peacock E, Taylor M, Stirling I, Born EW, Laidre
KL, Wiig Ø (2012) A tale of two polar bear populations:
ice habitat, harvest, and body condition. Popul Ecol 54:
3−18
Ropert-Coudert Y, Wilson RP (2005) Trends and perspec-
tives in animal-attached remote sensing. Front Ecol Env-
iron 3: 437−444
Sahanatien V, Derocher AE (2012) Monitoring sea ice habi-
tat fragmentation for polar bear conservation. Anim Con-
serv 15: 397−406
Schweinsburg RE, Lee LJ (1982) Movement of four satellite-
monitored polar bears in Lancaster Sound, Northwest
Territories. Arctic 35: 504−511
Shamoun-Baranes J, Bom R, van Loon EE, Ens BJ, Ooster-
beek K, Bouten W (2012) From sensor data to animal
behaviour: an oystercatcher example. PLOS ONE 7:
e37997
Shepard ELC, Wilson RP, Quintana F, Gómez Laich A and
others (2008) Identification of animal movement patterns
using tri-axial accelerometry. Endang Species Res 10:
47−60
Shepard ELC, Wilson RP, Rees WG, Grundy E, Lambertucci
SA, Vosper SB (2013) Energy landscapes shape animal
movement ecology. Am Nat 182: 298−312
Soltis J, Wilson RP, Douglas-Hamilton I, Vollrath F, King LE,
Savage A (2012) Accelerometers in collars identify
behavioral states in captive African elephants Loxodonta
africana. Endang Species Res 18: 255−263
Stirling I (1974) Midsummer observations on the behavior of
wild polar bears (Ursus maritimus). Can J Zool 52:
1191−1198
Stirling I (1988) Polar bears. University of Michigan Press,
Ann Arbor, MI
Stirling I, Latour PB (1978) Comparative hunting abilities of
polar bear cubs of different ages. Can J Zool 56:
1768−1772
Stirling I, Spencer C, Andriashek D (1989) Immobilization of
polar bears (Ursus maritimus) with Telazol® in the Cana-
dian Arctic. J Wildl Dis 25: 159−168
Stirling I, Lunn NJ, Iacozza J (1999) Long-term trends in the
population ecology of polar bears in western Hudson Bay
in relation to climatic change. Arctic 52: 294−306
Stirling I, Spencer C, Andriashek D (2016) Behavior and
activity budgets of wild breeding polar bears (Ursus
mari timus). Mar Mamm Sci 32: 13−37
Sutherland WJ (1998) The importance of behavioral studies
in conservation biology. Anim Behav 56: 801−809
Taylor MK (1986) Movements of Alaskan polar bears instru-
mented with satellite transmitters. Int Conf Bear Res
Manage 6: 103−104
Teisberg JE, Farley SD, Lynne Nelson O, Hilderbrand GV
and others (2014) Immobilization of grizzly bears (Ursus
arctos) with dexmedetomidine, tiletamine, and zola -
zepam. J Wildl Dis 50: 74−83
Volpov BL, Hoskins AJ, Battaile BC, Viviant M and others
(2015) Identification of prey captures in Australian fur
seals (Arctocephalus pusillus doriferus) using head-
mounted accelerometers: field validation with animal-
borne video cameras. PLOS ONE 10: e0128789
Vongraven D, Aars J, Amstrup S, Atkinson SN and others
(2012) A circumpolar monitoring framework for polar
bears. Ursus 23: 1−66
Voorhees H, Sparks R, Huntington HP, Rode KD (2014) Tra-
ditional knowledge about polar bears (Ursus maritimus)
in northwestern Alaska. Arctic 67: 523−536
Wang Y, Nickel B, Rutishauser M, Bryce C, Williams TM,
Elkaim G, Wilmers CC (2015) Movement, resting, and
attack behaviors of wild pumas are revealed by tri-axial
accelerometer measurements. Mov Ecol 3: 2
Ware JV, Rode KD, Pagano AM, Bromaghin J and others
(2015) Validation of mercury tip-switch and accelerome-
ter activity sensors for identifying resting and active
behavior in bears. Ursus 26: 86−96
Watanabe YY, Takahashi A (2013) Linking animal-borne
video to accelerometers reveals prey capture variability.
Proc Natl Acad Sci USA 110: 2199−2204
Watanabe S, Izawa M, Kato A, Ropert-Coudert Y, Naito Y
(2005) A new technique for monitoring the detailed
behaviour of terrestrial animals: a case study with the
domestic cat. Appl Anim Behav Sci 94: 117−131
32
Pagano et al.: Accelerometers identify polar bear behaviors
Williams TM (1983) Locomotion in the North American
mink, a semi-aquatic mammal. I. Swimming energetics
and body drag. J Exp Biol 103: 155−168
Williams TM, Wolfe L, Davis T, Kendall T and others (2014)
Instantaneous energetics of puma kills reveal advantage
of felid sneak attacks. Science 346: 81−85
Wilmers CC, Nickel B, Bryce CM, Smith JA, Wheat RE,
Yovovich V (2015) The golden age of bio-logging: how
animal-borne sensors are advancing the frontiers of eco -
logy. Ecology 96: 1741−1753
Wilson RP, White CR, Quintana F, Halsey LG, Liebsch N,
Martin GR, Butler PJ (2006) Moving towards acceleration
for estimates of activity-specific metabolic rate in free-
living animals: the case of the cormorant. J Anim Ecol 75:
1081−1090
Wilson RP, Shepard ELC, Liebsch N (2008) Prying into the
intimate details of animal lives: use of a daily diary on
animals. Endang Species Res 4: 123−137
Wilson RP, Quintana F, Hobson VJ (2012) Construction of
energy landscapes can clarify the movement and distri-
bution of foraging animals. Proc R Soc B 279: 975−980
Wilson RR, Horne JS, Rode KD, Regehr EV, Durner GM
(2014) Identifying polar bear resource selection patterns
to inform offshore development in a dynamic and chang-
ing Arctic. Ecosphere 5: 136
Yoda K, Sato K, Niizuma Y, Kurita M, Bost CA, Le Maho Y,
Naito Y (1999) Precise monitoring of porpoising behav-
iour of Adélie penguins determined using acceleration
data loggers. J Exp Biol 202: 3121−3126
Yoda K, Naito Y, Sato K, Takahashi A and others (2001) A
new technique for monitoring the behaviour of free-
ranging Adélie penguins. J Exp Biol 204: 685−690
33
Editorial responsibility: Rory Wilson,
Swansea, UK
Submitted: April 25, 2016; Accepted: October 13, 2016
Proofs received from author(s): December 3, 2016
... Polar bears exhibit a diversity of behaviors to successfully exploit the spatiotemporally dynamic sea-ice habitat ( Fig. 1; [104,138,161]). In areas of seasonal sea ice, polar bears migrate between the terrestrial refugia and on-ice foraging grounds [11,22]. They exhibit philopatry to their summering grounds and compensate for sea ice motion in their navigation and for station keeping [5,34,74,90]. ...
... Recent advances in data acquisition [99] and analytical methods (e.g., [94,155]) have enabled the identification of more intricate behaviors and research at an increasingly large scale and resolution. Pagano et al. [104] described the use of accelerometers to identify up to ten fine-scale behaviors, and Pagano et al. [107] used a combination of accelerometer data and conductivity sensors to identify resting, walking, and swimming. Unfortunately, most existing telemetry datasets do not lend themselves to many of the newer analytical methods because they lack necessary auxiliary data (i.e., they only estimate tag location). ...
... Polar bears exhibit a high degree of behavioral plasticity and diversity [104,138,151,161], however the remote and dynamic nature of their habitat has made it difficult to study their behavior, particularly during the critical winter foraging period. We used remote tracking data to investigate the spatiotemporal distribution of three movement states representative of three important behaviors (stationary/drifting, area-restricted search, and olfactory search) and to examine what factors may promote them. ...
Article
Full-text available
Background Change in behavior is one of the earliest responses to variation in habitat suitability. It is therefore important to understand the conditions that promote different behaviors, particularly in areas undergoing environmental change. Animal movement is tightly linked to behavior and remote tracking can be used to study ethology when direct observation is not possible. Methods We used movement data from 14 polar bears ( Ursus maritimus ) in Hudson Bay, Canada, during the foraging season (January–June), when bears inhabit the sea ice. We developed an error-tolerant method to correct for sea ice drift in tracking data. Next, we used hidden Markov models with movement and orientation relative to wind to study three behaviors (stationary, area-restricted search, and olfactory search) and examine effects of 11 covariates on behavior. Results Polar bears spent approximately 47% of their time in the stationary drift state, 29% in olfactory search, and 24% in area-restricted search. High energy behaviors occurred later in the day (around 20:00) compared to other populations. Second, olfactory search increased as the season progressed, which may reflect a shift in foraging strategy from still-hunting to active search linked to a shift in seal availability (i.e., increase in haul-outs from winter to the spring pupping and molting seasons). Last, we found spatial patterns of distribution linked to season, ice concentration, and bear age that may be tied to habitat quality and competitive exclusion. Conclusions Our observations were generally consistent with predictions of the marginal value theorem, and differences between our findings and other populations could be explained by regional or temporal variation in resource availability. Our novel movement analyses and finding can help identify periods, regions, and conditions of critical habitat.
... Cattle [18,19], other livestock such as sheep [10,20] and chickens [21], and pets such as dogs [22,13] are popular in behavior recognition research because of the ease of data collection and the size of the market, where a variety of behaviors are targets for recognition: foraging, ruminating, resting, traveling, and other behaviors for cattle [19]; barking, chewing, digging, drinking, and other 12 behaviors for dogs [22]; and pecking, preening, and dust bathing for chickens [11]. However, to the best of our knowledge regarding bears, there are no other studies that deal with more complex behavior recognition than discriminating between active and inactive states, except for the work by Pagano et al. [23]. In their work, 10 behaviors of wild polar bears, i.e., resting, walking, swimming, eating, running, pouncing, grooming, digging, head shaking, and rolling, were recognized by a random forest classifier with data obtained from a three-axis accelerometer and conductivity sensor. ...
... In this study, signals from the three-axis accelerometer sampled at 8 Hz were used to model bears behavior based on the force and gravity given to the sensor. Pagano et al. [23] showed that there is no difference between the predictability at 8 and 16 Hz even though the predictability at 4 Hz was lower than that at 8 Hz. Thus, we believe that 8 Hz is reasonable for our case. ...
... By referring to the work on behavior classification of polar bears [23] and taking into account the ecology of bears, we classified their behaviors into seven classes (Table 1), where each class contained concrete behaviors. For example, "resting" may indicate that a bear is sleeping, grooming, scratching, or having a break. ...
Chapter
The miniaturizations of sensing units, the increase in storage capacity, and the longevity of batteries, as well as the advancement of big-data processing technologies, are making it possible to recognize animal behaviors. This allows researchers to understand animal space use patterns, social interactions, habitats, etc. In this study, we focused on the behavior recognition of Asian black bears (Ursus thibetanus) using a three-axis accelerometer embedded in collars attached to their necks, where approximately 1% of data obtained from four bears over an average of 42 d were used. A machine learning was used to recognize seven bear behaviors, where oversampling and extension of labels to the period adjacent to the labeled period were applied to overcome data imbalance across classes and insufficient data in some classes. Experimental results showed the effectiveness of oversampling and a large difference in individual bears. Effective feature sets vary by experimental conditions. However, a tendency of features calculated from the magnitude of the three axes contributing to classification performance was confirmed.
... While direct comparisons of accelerometer-derived behavioural signatures between species are rare in the literature, our findings are consistent with other studies showing minor and largely negligible influence of body size and device attachment on accelerometer readings [31,56]. For example, the observation that terrestrial activity in Blanding's turtles was separated by greater thresholds than in Painted turtles is likely due to the larger carapace of the former species. ...
... High classification performance at low frequencies could allow longer battery life, increased memory capacity, and, thus, longer field deployment duration [33,62]. In addition, low-frequency accelerometer data require lower computational power for processing and analysis [56]. It is important to note, however, that species exhibiting behaviours with complex and fast kinematics, may require high-frequency accelerometry for reliable inference and representation [22,54]. ...
Article
Full-text available
Research in ecology often requires robust assessment of animal behaviour, but classifying behavioural patterns in free-ranging animals and in natural environments can be especially challenging. New miniaturised bio-logging devices such as accelerometers are increasingly available to record animal behaviour remotely, and thereby address the gap in knowledge related to behaviour of free-ranging animals. However, validation of these data is rarely conducted and classification model transferability across closely-related species is often not tested. Here, we validated accelerometer and water sensor data to classify activity states in two free-ranging freshwater turtle species (Blanding’s turtle, Emydoidea blandingii , and Painted turtle, Chrysemys picta ). First, using only accelerometer data, we developed a decision tree to separate motion from motionless states, and second, we included water sensor data to classify the animal as being motionless or in-motion on land or in water. We found that accelerometers separated in-motion from motionless behaviour with > 83% accuracy, whereas models also including water sensor data predicted states in terrestrial and aquatic locations with > 77% accuracy. Despite differences in values separating activity states between the two species, we found high model transferability allowing cross-species application of classification models. Note that reducing sampling frequency did not affect predictive accuracy of our models up to a sampling frequency of 0.0625 Hz. We conclude that the use of accelerometers in animal research is promising, but requires prior data validation and development of robust classification models, and whenever possible cross-species assessment should be conducted to establish model generalisability.
... Most applications of animal-borne accelerometers have been to examine behaviours (e.g. [31,[37][38][39][40]), with several resulting in the categorisation of coarse-scale descriptions (i.e. three or four different behaviours) [28,33,40,41]. While several studies have categorised behaviours manually by coarsely examining the acceleration traces generated (e.g. ...
... Whilst the use of such devices has been gaining momentum for decades, interpretation of their outputs for behavioural categorisation is relatively recent, especially when high resolution and precision are desired (e.g. [31,[37][38][39][40]). The cheetah (Acinonyx jubatus) is listed as 'vulnerable' [46] with wild populations purportedly decreasing [47]. ...
Article
Full-text available
Background Extinction is one of the greatest threats to the living world, endangering organisms globally, advancing conservation to the forefront of species research. To maximise the efficacy of conservation efforts, understanding the ecological, physiological, and behavioural requirements of vulnerable species is vital. Technological advances, particularly in remote sensing, enable researchers to continuously monitor movement and behaviours of multiple individuals simultaneously with minimal human intervention. Cheetahs, Acinonyx jubatus, constitute a “vulnerable” species for which only coarse behaviours have been elucidated. The aims of this study were to use animal-attached accelerometers to (1) determine fine-scale behaviours in cheetahs, (2) compare the performances of different devices in behaviour categorisation, and (3) provide a behavioural categorisation framework. Methods Two different accelerometer devices (CEFAS, frequency: 30 Hz, maximum capacity: ~ 2 g; GCDC, frequency: 50 Hz, maximum capacity: ~ 8 g) were mounted onto collars, fitted to five individual captive cheetahs. The cheetahs chased a lure around a track, during which time their behaviours were videoed. Accelerometer data were temporally aligned with corresponding video footage and labelled with one of 17 behaviours. Six separate random forest models were run (three per device type) to determine the categorisation accuracy for behaviours at a fine, medium, and coarse resolution. Results Fine- and medium-scale models had an overall categorisation accuracy of 83–86% and 84–88% respectively. Non-locomotory behaviours were best categorised on both loggers with GCDC outperforming CEFAS devices overall. On a coarse scale, both devices performed well when categorising activity (86.9% (CEFAS) vs. 89.3% (GCDC) accuracy) and inactivity (95.5% (CEFAS) vs. 95.0% (GCDC) accuracy). This study defined cheetah behaviour beyond three categories and accurately determined stalking behaviours by remote sensing. We also show that device specification and configuration may affect categorisation accuracy, so we recommend deploying several different loggers simultaneously on the same individual. Conclusion The results of this study will be useful in determining wild cheetah behaviour. The methods used here allowed broad-scale (active/inactive) as well as fine-scale (e.g. stalking) behaviours to be categorised remotely. These findings and methodological approaches will be useful in monitoring the behaviour of wild cheetahs and other species of conservation interest.
... Shuert et al., 2018), which for many species, such as marine mammals and cryptic, wide-ranging carnivores, presents obvious logistical challenges. However, in recent years, other forms of groundtruthing using animal-borne devices have become more popular (see Sections 6.2.2.2 and 6.2.2.3) and have helped alleviate these issues (Pagano et al., 2017). By simultaneously recording accelerometery data alongside time-stamped behavioural data, acceleration profiles may be identified and associated with different behaviours, such as resting, Raw accelerometery data can be used to differentiate between behaviours that have distinct acceleration profiles (adapted from Fehlmann et al., 2017). ...
... This has yet again been driven by the increasing accessibility of devices to researchers (both in terms of reductions in cost and complexity) and the increasing miniaturization of these technologies. Today, animal-borne cameras are being used to study a wide range of topics on both marine and terrestrial species (Pagano et al., 2017;Watanabe et al., 2019). Considered alongside tracking data, they can help assign behaviours to specific locations (Bruce et al., 2019), capture data on the surrounding environment (e.g. ...
Chapter
The use of animal-borne devices in wildlife ecology and conservation has expanded in recent decades. Animal-borne devices allow a suite of data to be collected, including locational, acoustic, and video. They have revolutionized our ability to collect measurements from animals and the environments that they inhabit, as well as promote the conservation of many species in their natural habitats. However, the use of these devices can also carry animal welfare, ethical, and privacy implications, device costs can limit the deployment of sufficient units, and the complexity of suitable analytical methods can limit inferences made. This chapter covers these topics by discussing how animal-borne devices, and particularly tracking technologies, are currently used in wildlife conservation and ecology. In particular, we discuss how and why animal-borne devices have advanced the study of wildlife, review key animal-borne devices in use today, and discuss the challenges and limitations of their use, as well as future opportunities, by drawing upon real-world examples from research and conservation.
... In recent years, the application of technologies to remotely monitor animal activity (e.g., accelerometers) have made it possible to increase sampling capacity, with minor sampling effort and observer bias [1][2][3][4][5]. Tri-axial accelerometers are sensors that simultaneously measure the acceleration of an object/organism in space along three dimensions, X, Y and Z [6]. ...
Article
Full-text available
Accelerometers are a technology that is increasingly used in the evaluation of animal behaviour. A tri-axial accelerometer attached to a vest was used on Tamandua tetradactyla individuals (n = 10) at Biodiversity Park. First, the influence of using a vest on the animals’ behaviour was evaluated (ABA-type: A1 and A2, without a vest; B, with a vest; each stage lasted 24 h), and no changes were detected. Second, their behaviour was monitored using videos and the accelerometer simultaneously (experimental room, 20 min). The observed behaviours were correlated with the accelerometer data, and summary measures (X, Y and Z axes) were obtained. Additionally, the overall dynamic body acceleration was calculated, determining a threshold to discriminate activity/inactivity events (variance = 0.0055). Then, based on a 24 h complementary test (video sampling every 5 min), the sensitivity (85.91%) and precision (100%) of the accelerometer were assessed. Animals were exposed to an ABA-type experimental design: A1 and A2: complex enclosure; B: decreased complexity (each stage lasted 24 h). An increase in total activity (%) was revealed using the accelerometer (26.15 ± 1.50, 29.29 ± 2.25, and 35.36 ± 3.15, respectively). Similar activity levels were detected using video analysis. The results demonstrate that the use of the accelerometer is reliable to determine the activity. Considering that the zoo-housed lesser anteaters exhibit a cathemeral activity pattern, this study contributes to easily monitoring their activities and responses to different management procedures supporting welfare programs, as well as ex situ conservation.
... Acceleration was frequently used among non-human mammals, mainly in domesticated and captive animals with a bias towards cattle [1]. However, there is also great interest in classifying behaviour of wild, possibly free-living species [e.g., 7,8,9]. Because the waveform of acceleration data is often characteristic for a behavioural pattern, acceleration should be collected with a sampling rate high enough to allow the resolution of even high frequencies. ...
Preprint
Full-text available
Background: We classified the behaviour of wild boar, kept under semi natural conditions in a large outdoor enclosure, using acceleration data. The prediction was made using a machine learning algorithm, specifically a random forest model in the open software h2o. Remarkably, highly accurate prediction was possible using ear-tag acceleration sensors that sampled data only at a frequency of 1 Hz. This measurement device was used to minimise the potentially harmful effects caused by the repeated capture of wild animals to exchange batteries. Results: Overall accuracy was as high as 94.7 %, but sensitivity ranged from 1.0 for common behaviours like foraging to 0.0 % in the case of rare events (e.g., scrubbing). Results show that static features of unfiltered acceleration data, as well as of gravitation- and orientation filtered data, were used in the prediction of behaviour. The waveform of certain behaviours in the sampled frequency range played no important role. Conclusions: Low-cost, slow accelerometers can be used to correctly classify behaviour in the wild boar. The positively identified behaviours, such as food intake and lactation, could be of interest for wildlife managers attempting to control population growth in this pest species. In addition to the model predicting behaviour, we provide several universal R-scripts that ease the generation of input files for machine learning.
... The limited adoption of accelerometers in vocal studies is possibly due to the difficulty of assigning accelerometer data to different behaviors in free-roaming animals (Alonso et al., 2021;Brown et al., 2013;Nathan et al., 2012;Shamoun-Baranes et al., 2012). Observing the behavior of captive conspecifics for the validation of accelerometer data can circumvent the necessity to make field observations of free-living individuals, yet may fail to reliably distinguish between different behaviors in free-roaming populations (Pagano et al., 2017). A particularly challenging species' group for vocal studies consists of highly mobile, medium-and smallsized birds which often have complex vocalizations. ...
Article
Full-text available
Abstract To acquire a fundamental understanding of animal communication, continuous observations in a natural setting and at an individual level are required. Whereas the use of animal‐borne acoustic recorders in vocal studies remains challenging, light‐weight accelerometers can potentially register individuals’ vocal output when this coincides with body vibrations. We collected one‐dimensional accelerometer data using light‐weight tags on a free‐living, crepuscular bird species, the European Nightjar (Caprimulgus europaeus). We developed a classification model to identify four behaviors (rest, sing, fly, and leap) from accelerometer data and, for the purpose of this study, validated the classification of song behavior. Male nightjars produce a distinctive “churring” song while they rest on a stationary song post. We expected churring to be associated with body vibrations (i.e., medium‐amplitude body acceleration), which we assumed would be easy to distinguish from resting (i.e., low‐amplitude body acceleration). We validated the classification of song behavior using simultaneous GPS tracking data (i.e., information on individuals’ movement and proximity to audio recorders) and vocal recordings from stationary audio recorders at known song posts of one tracked individual. Song activity was detected by the classification model with an accuracy of 92%. Beyond a threshold of 20 m from the audio recorders, only 8% of the classified song bouts were recorded. The duration of the detected song activity (i.e., acceleration data) was highly correlated with the duration of the simultaneously recorded song bouts (correlation coefficient = 0.87, N = 10, S = 21.7, p = .001). We show that accelerometer‐based identification of vocalizations could serve as a promising tool to study communication in free‐living, small‐sized birds and demonstrate possible limitations of audio recorders to investigate individual‐based variation in song behavior.
... However, camera traps are limited by their stationary viewpoint, and the large home range of polar bears may produce infrequent observations of behaviours. Other technologies, such as tri-axial accelerometer collars (Pagano et al. 2017) and video camera collars (Pagano et al. 2018), have further advanced our knowledge of polar bear behaviours. For instance, Pagano et al. (2018) used both methods to determine polar bear activity and behaviour to quantify observed variation in field metabolic rates, which lead to a better understanding of polar bear activity/energy budgets. ...
Article
Climate-induced sea-ice loss represents the greatest threat to polar bears (Ursus maritimus), and utilizing drones to characterize behavioural responses to sea-ice loss is valuable to forecasting polar bear persistence. In this manuscript, we review previously published literature and draw on our own experience of using multirotor aerial drones to study polar bear behaviour in the Canadian Arctic. Specifically, we suggest that drones can minimize human-bear conflicts by allowing users to observe bears from a safe vantage point; produce high-quality behavioural data that can be reviewed as many times as needed and shared with multiple stakeholders; and foster knowledge generation through co-production with northern communities. We posit that in some instances drones may be considered as an alternative tool for studying polar bear foraging behaviour, interspecific interactions, human-bear interactions, human safety and conflict mitigation, and den-site location at individual-level, small spatial scales. Finally, we discuss flying techniques to ensure ethical operation around polar bears, regulatory requirements to consider, and recommend that future research focus on understanding polar bears’ behavioural and physiological responses to drones and the efficacy of drones as a deterrent tool for safety purposes.
Article
Predation risk, the probability that a prey animal will be killed by a predator, is fundamental to theoretical and applied ecology. Predation risk varies with animal behavior and environmental conditions, yet attempts to understand predation risk in natural systems often ignore important ecological and environmental complexities, relying instead on proxies for actual risk such as predator–prey spatial overlap. Here we detail the ecological and environmental complexities driving disconnects between three stages of the predation sequence that are often assumed to be tightly linked: spatial overlap, encounters and prey capture. Our review highlights several major sources of variability in natural predator–prey systems that lead to the decoupling of spatial overlap estimates from actual encounter rates (e.g. temporal activity patterns, predator and prey movement capacity, resource limitations) and that affect the probability of prey capture given encounter (e.g. predator hunger levels, temporal, topographic and other environmental influences on capture success). Emerging technologies and statistical methods are facilitating a transition to a more spatiotemporally detailed, mechanistic understanding of predator–prey interactions, allowing for the concurrent examination of multiple stages of the predation sequence in mobile, free‐ranging animals. We describe crucial applications of this new understanding to fundamental and applied ecology, highlighting opportunities to better integrate ecological contingencies into dynamic predator–prey models and to harness a mechanistic understanding of predator–prey interactions to improve targeting and effectiveness of conservation interventions.
Article
Full-text available
Effective conservation planning requires understanding and ranking threats to wildlife populations. We developed a Bayesian network model to evaluate the relative influence of environmental and anthropogenic stressors, and their mitigation, on the persistence of polar bears (Ursus maritimus). Overall sea ice conditions, affected by rising global temperatures, were the most influential determinant of population outcomes. Accordingly, unabated rise in atmospheric greenhouse gas (GHG) concentrations was the dominant influence leading to worsened population outcomes, with polar bears in three of four ecoregions reaching a dominant probability of decreased or greatly decreased by the latter part of this century. Stabilization of atmospheric GHG concentrations by mid-century delayed the greatly reduced state by ≈25 yr in two ecoregions. Prompt and aggressive mitigation of emissions reduced the probability of any regional population becoming greatly reduced by up to 25%. Marine prey availability, linked closely to sea ice trend, had slightly less influence on outcome state than sea ice availability itself. Reduced mortality from hunting and defense of life and property interactions resulted in modest declines in the probability of a decreased or greatly decreased population outcome. Minimizing other stressors such as trans-Arctic shipping, oil and gas exploration, and contaminants had a negligible effect on polar bear outcomes, although the model was not well-informed with respect to the potential influence of these stressors. Adverse consequences of loss of sea ice habitat became more pronounced as the summer ice-free period lengthened beyond four months, which could occur in most of the Arctic basin after mid-century if GHG emissions are not promptly reduced. Long-term conservation of polar bears would be best supported by holding global mean temperature to ≤ 2° C above preindustrial levels. Until further sea ice loss is stopped, management of other stressors may serve to slow the transition of populations to progressively worsened outcomes, and improve the prospects for their long-term persistence.
Article
Full-text available
Sea ice is declining over much of the Arctic. In Hudson Bay the ice melts completely each summer, and advances in break-up have resulted in longer ice-free seasons. Consequently, earlier break-up is implicated in declines in body condition, survival, and abundance of polar bears (Ursus maritimus Phipps, 1774) in the Western Hudson Bay (WH) subpopulation. We hypothesised that similar patterns would be evident in the neighbouring Southern Hudson Bay (SH) subpopulation. We examined trends 1980–2012 in break-up and freeze-up dates within the entire SH management unit and within smaller coastal break-up and freeze-up zones. We examined trends in body condition for 900 bears captured during 1984–1986, 2000–2005, and 2007–2009 and hypothesised that body condition would be correlated with duration of sea ice. The ice-free season in SH increased by about 30 days from 1980 to 2012. Body condition declined in all age and sex classes, but the decline was less for cubs than for other social classes. If trends towards a longer ice-free season continue in the future, further declines in body condition and survival rates are likely, and ultimately declines in abundance will occur in the SH subpopulation.
Article
Full-text available
We quantify the first complete description of breeding behavior and activity budgets of an undisturbed pair of adult polar bears, observed 24 h/d for 13 d from 2 to 15 May 1997, at Radstock Bay, Devon Island, Nunavut, Canada. The male herded the female to an area of 1–2 km 2 , where we observed them throughout the observation period. All behaviors were documented from when the adult female and her 2.5-yr-old cub were first observed being followed by an adult male, through separation of the cub from its mother, a week of intense interactions preceding several days with copulation, after which they parted. They mated for 51, 86, 66, and 150 min on 9–10, 12, 13, and 14 May, respectively, and parted on 14–15 May. The male deterred three challengers. The peak breeding season for polar bears runs from early April through mid-May, although additional mating behavior has been documented in June. Timing of mating and duration of copulations in the wild were similar to reports from zoos. Induced ovulation, male intrasexual competition, female fitness, the mating system, and potential consequences of climate warming are discussed with insights made possible by documentation of the reproductive behavior of wild polar bears.
Article
Activity budgets in wild animals are challenging to measure via direct observation because data collection is time consuming and observer effects are potentially confounding. Although tri-axial accelerometers are increasingly employed for this purpose, their application in small-bodied animals has been limited by weight restrictions. Additionally, accelerometers engender novel complications, as a system is needed to reliably map acceleration to behaviors. In this study we describe newly-developed, tiny acceleration-logging devices (1.5-2.5 grams) and use them to characterize behavior in two chipmunk species. We collected paired accelerometer readings and behavioral observations from captive individuals. We then employed techniques from machine learning to develop an automatic system for coding accelerometer readings into behavioral categories. Finally, we deployed and recovered accelerometers from free-living, wild chipmunks. This is the first time to our knowledge that accelerometers have been used to generate behavioral data for small-bodied (<100 gram), free-living mammals.
Article
Tri-axial accelerometer tags provide quantitative data on body movement that can be used to characterize behaviour and understand species ecology in ways that would otherwise be impossible. Using tags on wild terrestrial mammals, especially smaller species, in natural settings has been limited. Poor battery power also reduced the amount of data collected, which limits what can be derived about animal behaviour. Another challenge using wild animals, is acquiring observations of actual behaviours with which to compare tag data and create an adequate training set to reliably identify behavioural states. Brown hares were fitted with accelerometers for 5 weeks to evaluate their use in collecting detailed behaviour data and activity levels. Collared hares were filmed to associate actual behaviours with tag data. Observed behaviours were classified using Random Forests (ensemble learning method) to create a supervised model and then used to classify hare behaviour from the tags. Increased tag longevity allowed acquisition of large quantities of data from each individual and direct observation of tagged hare's behaviour. Random Forests accurately classified observed behaviours from tag data with an 11% error rate. Individual accuracy of behaviours varied with running (100% accuracy), feeding (94.7%) and vigilance (98.3%) having the highest classification accuracy. Hares spent 46% of their time being vigilant and 25% feeding when active. The combination of our tags and Random Forests facilitated large amounts of behavioural data to be collected on animals where observational studies could be limited, or impossible. The same method could be used on a range of terrestrial mammals to create models to investigate behaviour from tag data, to learn more about their behaviour and be used to answer many ecological questions. However, further development of methods for analysing tag data is needed to make the process quicker, simpler and more accurate.