ArticlePDF Available

Tracking Animal Location and Activity with an Automated Radio Telemetry System in a Tropical Rainforest

Article

Tracking Animal Location and Activity with an Automated Radio Telemetry System in a Tropical Rainforest

Abstract and Figures

How do animals use their habitat? Where do they go and what do they do? These basic questions are key not only to understanding a species' ecology and evolution, but also for addressing many of the environmental challenges we currently face, including problems posed by invasive species, the spread of zoonotic diseases and declines in wildlife populations due to anthropogenic climate and land-use changes. Monitoring the movements and activities of wild animals can be difficult, especially when the species in question are small, cryptic or move over large areas. In this paper, we describe an Automated Radio-Telemetry System (ARTS) that we designed and built on Barro Colorado Island (BCI), Panama to overcome these challenges. We describe the hardware and software we used to implement the ARTS, and discuss the scientific successes we have had using the system, as well as the logistical challenges we faced in maintaining the system in real-world, rainforest conditions. The ARTS uses automated radio-telemetry receivers mounted on 40-m towers topped with arrays of directional antennas to track the activity and location of radio-collared study animals, 24 h a day, 7 days a week. These receiving units are connected by a wireless network to a server housed in the laboratory on BCI, making these data available in real time to researchers via a web-accessible database. As long as study animals are within the range of the towers, the ARTS system collects data more frequently than typical animal-borne global positioning system collars (∼12 locations/h) with lower accuracy (approximately 50 m) but at much reduced cost per tag (∼10X less expensive). The geographic range of ARTS, like all VHF telemetry, is affected by the size of the radio-tag as well as its position in the forest (e.g. tags in the canopy transmit farther than those on the forest floor). We present a model of signal propagation based on landscape conditions, which quantifies these effects and identifies sources of interference, including weather events and human activity. ARTS has been used to track 374 individual animals from 38 species, including 17 mammal species, 12 birds, 7 reptiles or amphibians, as well as two species of plant seeds. These data elucidate the spatio-temporal dynamics of animal activity and movement at the site and have produced numerous peer-reviewed publications, student theses, magazine articles, educational programs and film documentaries. These data are also relevant to long-term population monitoring and conservation plans. Both the successes and the failures of the ARTS system are applicable to broader sensor network applications and are valuable for advancing sensor network research.
Content may be subject to copyright.
© The Author 2011. Published by Oxford University Press on behalf of The British Computer Society. All rights reserved.
For Permissions, please email: journals.permissions@oup.com
doi:10.1093/comjnl/bxr072
Tracking Animal Location and Activity
with an Automated Radio Telemetry
System in a Tropical Rainforest
Roland Kays1,2,, Sameer Tilak3, Margaret Crofoot2,4, Tony Fountain3,
Daniel Obando2, Alejandro Ortega2, Franz Kuemmeth5, Jamie Mandel6,
George Swenson7, Thomas Lambert8, Ben Hirsch1,2and Martin Wikelski2,4
1New York State Museum, Albany, NY, USA
2Smithsonian Tropical Research Institute, Balboa, Panama
3California Institute for Telecommunications and Information Technology (Calit2), University of California,
San Diego, CA, USA
4Max Plank Institute for Ornithology, Radolfzell, Germany
5E-Obs, Gruenwald, Germany
6Princeton University, Princeton, NJ, USA
7University of Illinois, Urbana, IL, USA
8Frostburg State University, Frostburg, MD, USA
Corresponding author: rkays@mail.nysed.gov
How do animals use their habitat? Where do they go and what do they do? These basic questions are
key not only to understanding a species’ ecology and evolution, but also for addressing many of the
environmental challenges we currently face, including problems posed by invasive species, the spread
of zoonotic diseases and declines in wildlife populations due to anthropogenic climate and land-use
changes. Monitoring the movements and activities of wild animals can be difficult, especially when
the species in question are small, cryptic or move over large areas. In this paper, we describe an
Automated Radio-Telemetry System (ARTS) that we designed and built on Barro Colorado Island
(BCI), Panama to overcome these challenges. We describe the hardware and software we used to
implement the ARTS, and discuss the scientific successes we have had using the system, as well as
the logistical challenges we faced in maintaining the system in real-world, rainforest conditions. The
ARTS uses automated radio-telemetry receivers mounted on 40-m towers topped with arrays of
directional antennas to track the activity and location of radio-collared study animals, 24 h a day,
7 days a week. These receiving units are connected by a wireless network to a server housed in
the laboratory on BCI, making these data available in real time to researchers via a web-accessible
database. As long as study animals are within the range of the towers, theARTS system collects data
more frequently than typical animal-borne global positioning system collars (12 locations/h) with
lower accuracy (approximately 50 m) but at much reduced cost per tag (10X less expensive). The
geographic range of ARTS, like all VHF telemetry, is affected by the size of the radio-tag as well
as its position in the forest (e.g. tags in the canopy transmit farther than those on the forest floor).
We present a model of signal propagation based on landscape conditions, which quantifies these
effects and identifies sources of interference, including weather events and human activity.ARTS has
been used to track 374 individual animals from 38 species, including 17 mammal species, 12 birds, 7
reptiles or amphibians, as well as two species of plant seeds. These data elucidate the spatio-temporal
dynamics of animal activity and movement at the site and have produced numerous peer-reviewed
publications, student theses, magazine articles, educational programs and film documentaries. These
data are also relevant to long-term population monitoring and conservation plans. Both the successes
and the failures of the ARTS system are applicable to broader sensor network applications and are
valuable for advancing sensor network research.
Keywords: sensor networks; animal tracking; environmental observing systems
Received 7 January 2011; revised 1 July 2011
Handling editor: Damianos Gavalas
The Computer Journal,2011
The Computer Journal Advance Access published August 12, 2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
2R. Kays et al.
1. INTRODUCTION
Sensor networks have the potential to revolutionize our
understanding of the natural and man-made environment
by providing fine grained spatio-temporal data. This paper
describes a sensor network designed to automatically,
continuously and simultaneously track the locations and
activities of radio-tagged wild animals living in a tropical rain
forest. The developed system is not an in-laboratory research
prototype, but a real-world working system that has been
gathering science-quality data for over 6 years. This system
is able to monitor the behavior of these wild animals at a
much higher resolution than would be possible using traditional
observational methods or other tracking technologies, including
global positioning system (GPS) tracking. For this reason, we
believe that the Automated Radio-Telemetry System (ARTS)
system represents a significant advance in the use of sensor
networks for animal monitoring.
2. SCIENCE MOTIVATION
Many important moments in an animal’s life are difficult to
study because they are rare, cryptic or occur over large spatial
or temporal scales. One of the first obstacles any researcher
studying wild animals must overcome is how to monitor and
observe the behavior of mobile organisms.Attaching radio-tags
to animals has been a primary method for studying animals
in their natural environment for 50 years [1] and that vastly
improved the quality and quantity of data that the biologists can
collect [2,3]. For example, our understanding of reproduction
rates in many species is tied to an ability to find and monitor
females during the birthing season, a feat that is often possible
only through the use of radio telemetry [46]. In addition,
much information on juvenile dispersal, a critical but poorly
understood life stage, has come from radio-tracking studies
[79]. Population density is notoriously difficult to quantify,
and although dozens of methods are used to count individual
animals, tracking is critical to any density estimate because
it is the best way to quantify the areas used by the censused
population. Finally, causes of mortality can best be determined
by finding animals soon after their death, which typically
requires animal tracking [1012].
Radio-telemetry has greatly improved our ability to study
rare behaviors and shy species. Traditional methods of radio-
tracking, however, are inherently limited by the manpower
that can be devoted to following study animals, and thus may
not be adequate for addressing certain types of questions. In
addition, many events such as predation are known to occur
outside an animal’s normal activity period and therefore are
likely to be missed in traditional hand tracking studies [13]. The
development of new tracking technologies capable of remotely,
continuously and simultaneously monitoring the movements
of a large number of study animals would provide a solution
to these problems, allowing scientists to address previously
intractable questions. This in turn improves our knowledge of
the dynamics of the natural world.
3. BACKGROUND
There are two basic ways to record animal motion [14]. The
Lagrangian approach monitors a specific organism and records
all the locations it passes over, while the Eulerian approach
monitors a specific location and records the movement of all
organisms across it. Eulerian studies are sometimes preferred
because they do not require the capture of an animal, and so are
less invasive [15]. However, they typically provide much less
detailed data and therefore are more restrictive in the questions
they can address. A Lagrangian approach (hereafter: tracking
data), on the other hand, repeatedly records the locations of
an animal moving through space. Observing the movement of
an animal without tagging, it is rarely practical; so scientists
have relied heavily on sensor technologies, especially radio-
telemetry, and the GPS and Argos satellites (Table 1).
Radio-telemetry was the first technique developed to find
and track free-ranging animals [1], and remains the most
common method because of its low cost (300$) and lightweight
transmitters (>0.2 g [16]). Small transmitter size not only
extends the range of species that can be tracked using radio-
telemetry, but also minimizes the impact that the radio-tag has
on the behavior of the study animal [2]. Because traditional
radio-telemetry is collected manually, it is limited in the
intensity and scale that data can be collected, typically <50
data points per day [2]. Automated tracking from satellites
provides global coverage, but requires larger (usually >10 g)
more expensive (few 1000s $) tags, thus limiting the variety
and number of animals that can be studied [17]. GPS- and
satellite-based tracking technologies overcome many of these
limitations, but Wikelski et al. estimated that 66% of mammal
species and 81% of bird species were too small to be tracked
by the smallest GPS tags available (Fig. 1)[17]. In contrast,
VHF telemetry tags are so much smaller that they can be
used on nearly all mammal species (Fig. 1). In addition, an
obscured view of the sky by trees or mountains may limit the
functionality of satellite-based systems [19]. The automation
of data collection from VHF transmitters offers the potential of
increasing the data resolution and scale of tracking projects [20]
(Table 1). However, the promise of these systems has not
been fully realized, in part, because of the difficulties in data
acquisition and management.
3.1. Objectives
Here we describe an ARTS operating at the Smithsonian
Tropical Research Institute field station on Barro Colorado
Island (BCI), Panama. The ARTS uses standard VHF radio-
tracking technology to monitor the movements of radio-tagged
study animals automatically, continuously and simultaneously.
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
Tracking Animal Location and Activity with an ARTS 3
TABLE 1. Primary automated animal tracking methods.
Smallest tags, Automated data Interference by
Tracking method animal weight collection? vegetation
Traditional radio telemetry 0.2 g, 4g No Low
GPS satellite tracking 10 g, 400g Yes Medium
Satellite tracking (ARGOS) 10 g, 200g Yes High
Automated radio telemetry (ARTS) 0.2 g, 4g Yes Low
1200
1000
800
600
400
200
00.6 7.86.65.44.23.01.8
lo
g
10mass
No of obs
Can be tracked with >0.2g VHF telemetry tags
Can be tracked with >10g GPS tags
FIGURE 1. The body mass of Mammalia showing which species
could be tagged by a 0.2 g VHF telemetry tag (animals >4g) and those
that could be tracked with a 10 g GPS tag (animals >200 g). Please
refer to data from [18].
Because it relies on VHF technology, it can be used to track
animals that are too small to be fitted with GPS transmitters.
Additionally, ARTS is better able to track animals through the
dense vegetation of a tropical rainforest than satellite-based
tracking systems, which rely on UHF signals. In this paper, we
provide a technical description of the ARTS system and report
on its utility for studying the movements and activity of a variety
of rainforest animals. The ARTS system has been used to track
374 individual animals from 38 species, including 17 mammal
species, 12 birds, 7 reptiles or amphibians, as well as two species
of plant seeds. The gathered data elucidates the spatio-temporal
dynamics of animal activity and movement at the site. It is also
relevant to long-term population monitoring and conservation
plans.
4. SYSTEM DESIGN GOALS
Our field work was conducted at the BCI (910N, 7951W)
research station, which is managed by the Smithsonian Tropical
Research Institute. BCI is a completely forested 1567 ha island
that was formed when the Chagres River was dammed to create
Lake Gatun and complete the Panama Canal. Animals continue
to move between the island and the surrounding national park
land, which are separated by a distance of <400 m.
Tropical and sub-tropical environments are difficult for
running a system of automatic electronic sensors because of
the high rainfall and humidity. Although the temperatures on
BCI are relatively stable throughout the year, there is a major
variation in rainfall. The island receives an average of 2632 mm
of rain per year, although roughly 90% of this falls during
the 8 month wet season (May–December). Because of the
dense vegetation, humidity remains high year-round.At ground
level in the forest, where most electronics are kept, relative
humidity averages 80.6% in the dry season and 93.1% in the wet
season [21]. Above the canopy, where some tower-based sensors
are fixed, humidity is slightly lower (69.2% dry season, 80.2%
wet season). In addition to the general risk of rain and humidity
to electronics, rainy season storms bring an increased risk of
lightning strikes to above-canopy towers. Finally, the increased
cloud cover that characterizes the rainy season also decreases the
potential for generating electricity from solar panels mounted on
the towers. Sunny dry-season months average 19.5MJ/m2day
of solar radiation while rainy season months average only
14.0 MJ m2day [21]. While these weather conditions are wetter
than most temperate zone systems, they are typical of many
tropical conditions, and represent a challenge to any tropical
sensor network.
The ARTS system was designed to operate in the challenging
conditions of BCI over multiple years with as little human
intervention as possible. In designing the ARTS, we had seven
primary design goals:
Robustness: The goal was to build a system that would
operate effectively under these natural conditions (high rainfall,
humidity and insect activity) and is robust to noise, signal-loss,
multi-path effects due to forest-mountain environment.
Meet application-specific accuracy in harsh outdoor
environments: In animal tracking studies, high accuracy of
location estimates is always preferred. However, no remote
tracking system is error-free. This system was designed to
track a variety of animals that move over large areas (species
that move multiple kilometers in a day). An accuracy of
less than 50 m was desired to investigate both the large-scale
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
4R. Kays et al.
space-use patterns of study species and their finer-scale patterns
of movement. InARTS, the radio transmission range is typically
few hundreds to a thousand meter.
Small form-factor for transmitters: The system was designed
to track a wide variety of species, from insects to tapirs. We
wanted to make sure that the behavior of a study animal is not
altered by transmitter size or weight. Therefore, one of the goals
was to use transmitters that are as inexpensive and light-weight
as possible.
Scalability: An animal tracking study typically involves
multiple animals being tracked at the same time. The system
was designed to handle from one to at least twenty animals being
tracked simultaneously. Because different users have different
tracking requirements, the system needs to be programmable,
allowing users to select sampling rates appropriate for their
study question and study organism.
Spatial extensibility: Another design goal was the ability to
change and augment the spatial coverage over time through the
temporary deployment of additional receivers.
Remote command and control: Because sensors are spread
throughout hilly terrain, the system was designed to stream live
feed to monitor the performance of equipment. Additionally,
we wanted to have an ability to modify programming from the
central laboratory. In essence, the ability to remotely debug and
control the system is a desirable property.
Multi-user functionality: Because a wide variety of
researchers work on BCI, the system should be designed as
a multi-user infrastructure to simultaneously and continuously
track the movements of many radio-collared animals tagged for
different studies. Each study might have different accuracy and
power requirements.
5. AUTOMATED RADIO TELEMETRY SYSTEM
INFRASTRUCTURE
In this section, we describe hardware and software elements
of ARTS.
5.1. ARTS hardware
The ARTS system uses automated receivers to track the location
and activity of transmitters mounted on animals, and relays these
data to the laboratory in real time (Fig. 2). Biologists tracking
animals with ARTS use standard radio transmitters (Fig. 3)
available from a variety of commercial sources. However, each
study must customize the design of the transmitters to not only
be small enough to be carried by an animal without affecting
its behavior, but also to securely attach to an animal that will
probably try to remove it. Many mammals can wear a collar,
which typically has few negative side effects. However, other
FIGURE 2. Schematic overview of the ARTS system. Radio
transmitters placed on study animals emit 148–168 MHz radio signals,
which are received by automated receiving units (red stars) connected
to arrays of six wide-band Yagis antennas atop 40 m towers. The
automated receiving units record the strength of the signal across the
antennas array. The data are passed from the automated receiving unit
to a filter box at the tower, which then sends it via a 900MHz wireless
network to the ARTS computer bank at the Smithsonian field station on
BCI, Panama. The raw data (i.e. signal strengths), which are displayed
real-time, are passed into SQL database on the ARTS server, which
converts it into bearing estimates. These bearing data are then available
to ARTS researchers anywhere in the world via the the Internet (art by
Patricia Kernan).
FIGURE 3. VHF Transmitters are tracked with theARTS. (a) A small 1gVHF radio collar designed to track small mice. (b) A Central American
Agouti tagged with VHF tracking devices.
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
Tracking Animal Location and Activity with an ARTS 5
solutions must be found for animals that cannot wear collars
(e.g. anteaters, birds), which sometimes include gluing the tags
to the animals, attaching with a harness or even implanting them
in the body cavity. Individual studies must carefully consider the
anatomy and behavior of a particular species when customizing
tracking tags.
The signals from the animal-borne transmitters are detected
by one or more Automated Receiving Units (ARU), Sparrow
Systems (http://www.sparrowsystems.biz/), Table 2, Fig. 4b,
which comprise the core of the ARTS system. As shown
in Fig. 4a, seven ARU’s are deployed on 36m radio towers
constructed in the rainforest from Rohn 25 tower. These guyed
towers manufactured by ROHN Products, LLC. are constructed
with high strength steel tubing, and hot-dip galvanized after
fabrication. The ARUs scan a user-selected list of radio-
frequencies that correspond to the radio tags being worn by
study animals, and record the signal strength of each frequency
from each of six directional antennas in an array. The receivers
are capable of searching over 200 channels in the frequency
range, but time typically constrains the list to many fewer.
The scan rate depends on the pulse rate of the transmitter.
We typically set it to 1.5 s per antenna, or 9 s per frequency.
TABLE 2. Technical specifications for the ARU used in the
ARTS system.
ARU specification Values
Dimensions 15 ×15 ×15 cm
Weight 1000 g
Current 6V 33 mA (1/5W)
Current 12V 35 mA (2/5W)
Frequency range 148–170 MHz
Dynamic range Variable standard calibration
Antenna input BNC 50 Ohm, Unbalanced
An internal clock provides time stamps accurate to the nearest
second over periods of many weeks. The receivers can record
signal amplitude for continuous transmitters, in addition to
pulse interval and pulse width from pulse transmitters. These
receiving units are housed in water-proof containers to minimize
the damage from the humid environment. TheARUs are located
at the bottom of each of the 40-m radio towers that bear the
antenna arrays. ARUs are connected to the antenna array on top
of the tower by coaxial cables. A single ARU is sufficient to
obtain temporal activity patterns, but at least three are needed
to triangulate a location. In addition to streaming live data to
the laboratory, data are recorded in exchangeable flash-memory
modules, which we exchange every 1 or 2 weeks.
Each tower supports an above-canopy network of two sets
of six log-periodic antennas vertically stacked 1.2 m apart,
with six azimuth directions separated by 60. These antennae
allow the scanning of the frequency range of 148–170 MHz.
This is the typical range for VHF animal tracking because
the signals are able to penetrate the dense vegetation better
than higher frequencies. Directional VHF antennas provide the
strongest signal when pointed directly at the source, dropping
off steeply when angled away. The specific directionality pattern
of our antennas are shown in (Fig. 5). When manually tracking
animals, these antenna are physically rotated while listening
to the signal to determine the compass bearing from the
receiver to the transmitter. To automate this process without
requiring the mechanical movement of antennas, we use six
stationary antenna, but space these out evenly across 360
so that their directionality patterns overlap. Comparisons of
the two strongest signals allow us to estimate the compass
bearing to the transmitter. For example, in Fig. 5a, an animal
sitting at a bearing of 100from the receiver would have its
signal registered most strongly by antenna 6 and its second
strongest signal from antenna 5. The ratio of these two signal
strengths can thus be used to estimate compass bearings to radio
transmitters.
FIGURE 4. BCI hardware infrastructure-based around ARU. (a)An aerial view of anARTS tower extending above the rainforest canopy on BCI,
near the shore of the Panama Canal. (b) An automated receiving unit with connections for up to eight separate antenna (we use six for the ARTS).
The unit scans each antenna to record the strength of a radio-tagged animal’s signal from each one.
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
6R. Kays et al.
15
25
35
45
55
65
(a) (b)
0 5 0 100 150 200 250 300 350
signal strength (dB)
radial angle (degrees)
FIGURE 5. Directionality patterns for VHF antennas used by the ARTS towers showing the strength of signal received relative to the orientation
of the transmitter to the compass bearing of the antenna. (a) The overlap of directionality patterns for the six antennas as they are mounted on the
ARTS towers, pointing in 6 different directions. The bearing to a transmitter can be estimated by comparing the strength of the signal received from
the two strongest antennas. (b) The overlap of directionality pattern of six antennas showing their consistency and symmetrical decline to either
side of 0.
Data from the receivers are streamed live to the laboratory
via a 900 MHz network using waterproof FreeWave radios
(www.freewave.com). We use a filter box between the receiver
and the FreeWave radio to manage communications by buffer-
ing the output of the ARUs and tagging them with a unique
tower identifier. The FreeWave radios are mounted on our
above-canopy towers and set up as a multi-point network feed-
ing into a master radio at the laboratory. Upon reaching the
laboratory, data are received by a Linux machine that automat-
ically loads them into a Web-accessible PostgreSQL database
and it is also sent to a Windows machine, where it is plotted in
real time (Fig. 6a). The researchers can monitor these real-time
plots of radio signal strengths to determine approximately
where an animal is, and recognize when a malfunction that
stops data collection has occurred.
The electronics used at each ARTS tower (receiver, filter box,
FreeWave) are powered by standard 12V car batteries located
at the bottom of each tower. Without a charger, these 12V car
batteries last about a month with just the ARU recording data to
exchangeable flash memory, or a week with the added draw of
the FreeWave network and filter box. However, solar panels
mounted on the towers above the forest canopy are able to
recharge the batteries and allow uninterrupted operation of the
equipment. We also supplement this tower network with two
movable understory receiving units consisting of one antenna
array and ARU components mounted 3–6 m above the ground
on one or two tower sections. Understory units do not have
the same range as tower-mounted antenna, but are useful to
obtain more detailed coverage of small areas. For example, we
set up two understory units to supplement one nearby tower
in tracking frogs visiting a localized breeding pond in the
wet season [22]. Solar panels are not functional in the shady
understory; thus it is necessary to manually recharge the power
supply for these portable sub-units. The 900 MHz radio link can
typically connect to a nearby tower, thus relaying live data from
these understory units.
5.2. Software infrastructure
Here we outline the software components used to program
the receivers and do the initial data processing to convert
radio signals into usable information about animal activity
and location. The ARUs are programmable through a
manufacturer-provided software suite (Sparrow Systems,
http://www.sparrowsystems.biz/), which allows users to set the
sequence of frequencies to be searched by the ARUs. Once
designed using their software, these programs can be loaded into
the receivers via the flash memory, or live over the FreeWave
Network (Over The Air Programming). The data recorded by
ARUs can also be either extracted from the flash memory units
or retrieved live through the FreeWave network. In both cases,
the real-time data are then loaded into a PostgreSQL database.
This ability to remotely program and control the ARUs as well
as the data acquisition is crucial since working in the field is
not always easy due to challenges posed by the harsh physical
environment.
To estimate animal locations, we must first convert signal
strength into bearings, and then use the bearings to triangulate
animal locations. We calculate bearings from signal strength
data with a database trigger programmed using a formula
derived from the directionality pattern of our antenna array
(Fig. 5). Antennas receive their strongest signals when pointed
directly at the transmitter, and this decreases at a predictable
rate as the source moves away from the central direction
of the antenna. Therefore, by comparing the relative signal
strength of the two strongest antenna, we can estimate the true
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
Tracking Animal Location and Activity with an ARTS 7
FIGURE 6. ARTS live data stream at BCI (a) allows us to quickly detect the death of a tagged animal and go investigate the cause. In this case,
we placed a motion-sensitive camera (b) at the carcass of a recently killed agouti to capture the return of the predator, an ocelot. (a) Real-time data
stream received at the laboratory depicting the change of radio-signal (y-axis) received by one ARU over time (x-axis showing time of day). This
example is from a transmitter attached to an agouti that died during the night. The signal is relatively dynamic during the day (<20:00) compared
with the resting animal (20:00–3:45). At approximately 3:45, the animal was killed by a predator, and the radio-signal changes very little because
the collar is laying on the forest floor. (b) Motion-triggered cameras set at the remains carcasses show ocelots to be the primary of agoutis.
bearing of the source of the signal as an offset to the central
direction of these two antennas. The following equation shows
our bearing calculation formula.
Bearing =ψ±(2.9015 s)+30.0.(1)
Where ψis the angle of strongest antenna, sis the difference
in signal strength between the two strongest antenna, and ±is
determined by which side (right or left) of the strongest antenna
lies the second strongest signal. Numbers 2.9015 and 30 are
based on directionality patterns of antennas [23] (Fig. 5).
We estimate animal locations by triangulating with bearings
from at least three different towers using standard wildlife
software (LOAS Ecological Software Solutions). Bearing data
from towers are often noisy, with erroneous bearings caused
by interference and multipath propagation. These problems
are not unique to automated systems, but plague all radio
telemetry projects [2]. However, because the ARTS collects data
constantly over long periods of time, we have a greater ability
to extract the real signal out of a messy, scattered data set.
Our signal processing approach makes use of a hand-
smoothing technique commonly used in physiological stud-
ies [24], and involves visualizing the bearing data from each
animal, from each tower, and interpolating a line through the
weighted center of the points (Fig. 7). This manual process is
analogous to calculating a running median, which is a more
appropriate measure of the true signal than the mean bearing
because it is relatively uninfluenced by outliers caused by sig-
nal bounce and interference. Although this approach to signal
processing is labor-intensive (It takes about 30min to smooth 1
week of data for one animal), the circular nature of bearing data
make automating a running median function difficult, and we
have found human pattern recognition to perform better than
any of the automated smoothing techniques we have tried. This
not only removes outliers, but also fills small gaps in the data
and improves triangulation [25].
ARTS data are also useful for estimating the activity of an
animal. Activity can be monitored using data from a single
receiver and antenna, making it useful even for animals living
where ARTS coverage is not sufficient to obtain data from
the three towers needed to estimate locations. The strength of
signals received by the automated receiving units from active
animals are dynamic, while those from resting animals are
relatively constant. This is obvious to the human eye when
inspecting plots of live data (Fig. 6a), and also simple to
calculate to quantify animal activity levels [26,27]. The strength
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
8R. Kays et al.
FIGURE 7. The smothing of bearings (black dots) from one tower to
one radio-collared animal over one day. A red line is drawn through the
central data cloud to smooth the bearings and eliminate the erroneous
bearings caused by interference and multipath propagation.
of sequential signals scan be used to determine whether an
animal is active or not. We use a query to extract the strongest
signal for a given animal from a given tower over a period of
time, use standard statistical techniques to produce the absolute
value of the difference between two signals and then code each
time point as active (1) if |s|> T hreshold or inactive(0)
if |s|<Threshold. These values can then be averaged
over different time periods to produce an Activity Index that
can be compared between species or between experimental
treatments.
5.3. System design space and ARTS design choices
In this section, we describe related research on localization and
tracking and explain the motivation behindARTS system design
choices. Great Duck Island [28] project was a pioneering effort
for the habitat and environmental monitoring using a large-
scale network. However, their deployment consisted of static
sensors. The ZebraNet system includes custom tracking collars
that include GPS, Flash memory, wireless transceivers and a
small CPU; essentially each node is a small, wireless computing
device [29] . These collars form a peer-to-peer network to
deliver logged data back to researchers. Since it uses a GPS,
it is mainly limited to tracking large mammals. Recently, Dyo
et al. [30] proposed a RFID–WSN hybrid system to monitor
European badgers (Meles meles) in a forest. This is a RFID-
based monitoring solution and has a very short range such
that it is only useful for knowing, for example, if an animal is
within 1 m of a sensor, such as on a nest or not. The ARTS is a
unique solution for collecting a dense set of activity and location
data in real time for small animals. The physical environment
(heavy rains, dense tree canopy etc.) and science requirements
(ability to track small-size animals) had a heavy influence on
our infrastructure design. For example, we used the following
asymmetric approach for hardware design. Because animals
are very hard on the tags attached to their body, and because
of weight limitations, we use very light-weight, low-cost and
dumb devices (tags) that fit on animal bodies. Since towers
have relatively lower constraints on power (because of solar
panels) and form factor, we adopted a centralized architecture
in which the receivers transmit data to a base station. This
approach allows us to estimate animal locations in a place where
computation power is easily available.
In this project, we focus on tracking a mobile sensor
(animal), rather than locating a stationary sensor. We reduce
the problem of tracking mobile sensor to a sequence of
location estimation problem for a nearly-stationary sensor.
Localization has received a lot of attention in the context of
static sensor networks. We now mention some of the state-of-
the-art techniques that can be used for localization for static
networks. He et al. [31] have classified existing localization
techniques into two categories: range-based and range-free.
In range-based techniques, information such as distances (or
angles) of a receiver are computed for a number of references
points using one of the following signal-strength- or timing-
based techniques and then position of the receiver is computed
using some multilateration technique [32]. However, range-
free techniques do not depend upon the presence of any such
information.
Localization techniques typically require some form of
communication between reference points (nodes with known
coordinates) and the receiver (node that needs to localize). Some
examples of communication technologies are RF-based and
acoustic-based communication. In RADAR system [33], RF-
based localization is suggested, where distance is estimated
based on the received signal strength. Cricket [34] uses
concurrent radio and ultrasonic sounds to estimate distance.
Some researchers have used time-based techniques such
as Time-of-Flight (TOA) [32], Time-Difference-of-Arrival
(TDOA) [34,35] between reference point and the receiver
node as a way to estimate distance. Niculescu and Nath [36]
proposed using angle-of-arrival to estimate position. Recently
He et al. [31] proposed range-free techniques for localization.
In our work, we use the Received Signal Strength Indicator
(RSSI) to localize. The RSSI-based method has advantage
since it is readily available in practically all the receivers
in market. However, its accuracy is severely hampered by
nonlinearities, noise, interference and absorption due to walls
in indoor environments. Since our target deployment is an
outdoor environment, we believe that the RSSI-based approach
is a reasonable choice for localization. However, forests are
notoriously difficult environments for radio-based signaling
networks, due to issues of signal attenuation and multipath.
A key question we seek to answer is: how accurate a signal-
strength-based localization/tracking system will perform in this
environment? Our results show that within the range of ARTS
towers the signal-strength-based approach meets the accuracy
requirement of our application.
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
Tracking Animal Location and Activity with an ARTS 9
Although acoustic ranging systems [37] provide precise
range estimates, we believe that they are not a good fit for
our application due to the nature of the physical environment
(e.g., animal sounds, environmental noise due to rain
etc.). Bulusu et al. [38] studied signal-strength-based and
connectivity-based techniques for localization in outdoor
environments. Nodes localize themselves to the centroid of their
proximate reference points. Their outdoor testbed consisted of
five Radiometrix RPC 418 (radio packet controller) modules
connected to a Toshiba Libretto running RedHat Linux 6.0.
Although the paper presented several insights, the work of
adapting their localization method to noisy environments and
large-scale deployments was left for future work. In addition,
the paper did not focus on mobile sensor networks.
Stoleru et al. [39] proposed Spotlight—a novel localization
system for high-accuracy (sub-meter localization error) and
low-cost localization in WSNs. The Spotlight system also
employs an asymmetric architecture in which field-deployed
sensors do not include any additional hardware for localization,
and all sophisticated hardware and computation resides on
a single Spotlight device. The authors assume that sensor
nodes are deployed from an unmanned aerial vehicle. Upon
deployment, nodes run a time-synchronization protocol and
self-organize. An aerial vehicle such as a helicopter equipped
with Spotlight device then flies over the network and generates
light events. Sensors report these events along with timestamps,
which are then used for calculating their locations. Once the
sensors have determined their own location, they could be used
analogous to the ARTS towers, e.g. the animals would still be
tagged by radio transmitters, but instead of sending their signals
to the towers, the tags would send to the neighboring sensors.
This requires that the sensors are equipped with additional
hardware, which in turn raises the cost and increases the form
factor of sensor nodes. Also, if sensors on ground are used, the
advantages offered by the tower height are lost. In addition,
sensor nodes are likely not designed to withstand the high
humidity in a rainforest environment such as BCI. Long-term
monitoring is a key goal of ARTS. We believe that the Spotlight
system is suitable for a campaign-style deployment rather than a
long-term deployment for the following two reasons: (1) sensors
can be easily moved compared with guyed towers. Given the
nature of the physical environment, we do not believe that once
deployed the sensors will remain unperturbed for extended time
periods. (2) the Spotlight system uses battery-operated sensors,
whereas the ARTS system uses electricity generated from solar
panels mounted on the towers.
Solutions that require RSSI and do not need beacon nodes
essentially use a mobile beacon node [4042]. Sensors that hear
the beacon node use various localization algorithms. Beacon
nodes can range from human operators to unmanned vehicles.
Given the nature of the physical environment (heavy rains, dense
tree canopy etc.) and the frequent mobility of sensors, these
protocols are not suitable for our environment. The design space
of a sensor-network-based tracking application is quite rich.
It has many dimensions such as modes of tracking (active vs.
passive), placement of computation (centralized vs. distributed),
placement of functionality (smart collars and dumb receivers vs.
dumb collars and smart receivers) etc. The science requirements
and real-world challenges led to the aforementioned design
goals, which in turn provided the basis for our architectural
design decisions, technology selections and system deployment.
6. EXPERIMENTAL RESULTS
In this section, we describe the results of 6 years system
deployment of the ARTS designed to track animal movements
and activity patterns. We evaluated the ARTS system across
various dimensions including accuracy, data quality, scalability
etc. We now describe some of the key questions that we
used during our system evaluation. How does ARTS compare
to the traditional localization/tracking technologies such as
GPS? What are the tradeoffs between tracking accuracy,
cost and energy efficiency? What are the true ecological
and infrastructure limitations for various animal tracking
technologies? What is the impact of the physical environment
(rain, tree canopy), animal size (large tags vs. small tags) and
position (tags in the canopy, on the forest floor, in underground
burrows etc.) on system performance? What are the major
bottlenecks in operating the system in real-world over an
extended time period? Finally, does the system generate science
equality data and enable new research that was not possible
before? We now describe our experiments that seek to address
the above questions.
6.1. Radio propagation in forests
Forests are notoriously difficult environments for the successful
employment of radio-based signaling networks, due to the
issues of signal attenuation and multipath. The first question
we seek to answer is: How much impact does this environment
have on the propagation of radio signal?
A radio wave radiated by a point source, propagating in
free space (an empty space free of reflecting or absorbing
boundaries), loses intensity in proportion to the inverse square
of the distance traveled. This is simply a result of spherical
geometrical dilution. For example, if the distance from the
source to a receiver is doubled, the intensity in the wave (the
power in a square meter of the spherical wave front) is reduced
by a factor of four, or about 6 decibels (dB) (dB). This rule
would apply, for example, for signals propagating above the
forest canopy, as from a flying bird to a tower, or for the free-
space component of a more complex path, which also involves
vegetation or other obstacles.
Within the forested environment, the situation is considerably
more complicated. Not only are radio waves scattered and
absorbed by foliage and tree trunks and branches, but also by
the ground. These effects control the signal strength in a forest
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
10 R. Kays et al.
with a well-established canopy and considerable undergrowth,
such as a lowland tropical rainforest or a temperate deciduous
forest, up to distances from source to receiver of 500 m.
At longer distances the signal strength between two points on
or near the ground beneath the canopy is typically governed by
losses along a path from the transmitter up through the canopy,
then to a more or less horizontal path above the canopy toward
the receiver, then down through the canopy to the receiver. As
may be imagined, the situation is much more complicated than
this simplified scenario, but experiments have shown that at
short ranges the strength of a received signal is dominated by
attenuation caused by the vegetation and the ground, while at
long ranges it is dominated by the inverse square law attenuation
of the path above the canopy [43]. In fact, at distances of more
than a kilometer, say, one might say that the forest has negligible
effect. These arguments apply broadly to frequencies over the
range from 100 to 1000 MHz.
We conducted experiments in two forest types: lowland
tropical rainforest on BCI in Panama and a second growth
cottonwood plantation in Illinois (ref. Table 3). Both had
fairly dense undergrowth. In each case, a transmitter with
antenna at one to two meters height was moved through
the forest while its signal strength was recorded at a fixed
location with an antenna positioned and polarized similarly to
that of the transmitter. The transmitting antenna was moved
to several positions one or two meters apart while multiple
measurements were made and averaged at each geographical
station. Measurements were generally repeatable over intervals
of a few days; there were no significant differences between dry
and wet foliage and no systematic differences between vertical
and horizontal antenna polarizations (though in all cases,
receiving and transmitting antennas were similarly polarized).
In the Illinois tests, there was only minimal difference between
summer and winter.
To permit comparisons with free-space spherical spreading,
these measurements were converted to the losses encountered
when the distance from transmitter to receiver is increased from
100 to 200 m. In free space, this would result in a loss of 6 dB;
the differences (ref. Table 3) are due to the vegetation and
the ground. Given that these results clearly show that radio
signal propagation is significantly impacted by the physical
environment (Table 3), we now investigate how accurate a
signal-strength-based localization/tracking system will perform
in that environment.
6.2. Localization error comparison
When an animal is within the range of three or more towers, its
location can be estimated through triangulation. We evaluated
the accuracy of these location estimates by comparing them
with known locations of test transmitters moved along marked
paths on BCI (Fig. 8a). More specifically, we conducted test
walks to assess the accuracy of the ARTS. A researcher carried
a radio transmitter along trails that had already been geo-
referenced with a handheld GPS unit (Garmin 60CSX) at 100 m
intervals. These bearings were processed as described already,
with bearing smoothed using PV-Wave software, and locations
triangulated from three or more bearings using an Andrew’s M-
estimator [44,45] implemented in LOAS. Since the standard
least-squares method is unstable if there are outliers present
in the data, we used Andrew’s M-estimator. We calculated the
distance between the ARTS estimate and the trail to produce a
measure of system accuracy (Fig. 8a). As shown in Fig. 8a,
ARTS accuracy varied by location on the island: error was
smallest in central portions of the island and largest towards
the periphery. Average error for the entire island was 73 m
(s.d.=71 m, range =0–590 m), but average error within
the center of the island (box drawn around towers was 42m
(s.d.=34 m, range 0 to 200 m. Ref. Fig. 8a). Our tests show
that within the range of ARTS towers, we collect data more
frequently (5 m) than typical animal-borne GPS collars with
lower accuracy (50 m) but at much reduced cost (10×less
expensive animal tags) and power consumption.
6.3. Accuracy of activity estimates
Patterns of activity and inactivity are one of the most basic
measures of animal behavior. To determine an appropriate
threshold for distinguishing between inactivity and activity, we
tested the correlation between a standard definition of activity
and what is detected by the ARTS system. To do this, we
calibrated ARTS estimates of activity levels by having human
TABLE 3. Experimental signal strength loss in forest as distance increases from 100 to 200 m.
Frequency, MHz Environment Season Distance doubling loss, dB
462 Zetek trail, BCI Dry 16
154.6 Zetek trail, BCI Dry 8.2
462 Armour trail, BCI Dry 18
154.6 Armour trail, BCI Dry 10
150 Illinois cottonwoods Spring and Winter 16
302 Illinois cottonwoods Spring and Winter 13
Free space loss would be 6 dB, local vegetation and ground caused the amount to be greater in these tests.
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
Tracking Animal Location and Activity with an ARTS 11
FIGURE 8. Study of Accuracy of Localization and activity measurement of ARTS. (a) Estimated and true locations of a radio transmitter in a
test-walk on trails of BCI. (b) Calibration of activity levels estimated by the ARTS as tested on human subjects following different presubscribed
activity exercises.
subjects (20 total) carry a radio transmitter around their waist
while walking back and forth on a short forest path. Subjects
were active for different amounts (10, 25, 50, 75 or 90%) of
time during subsequent 10-min periods. As shown in Fig. 8b,
the activity index estimated by the ARTS was tightly correlated
with the true activity protocol being executed by the subjects.
Calibration between ARTS activity measures and observations
of animal behavior was likewise highly correlated [27].
6.4. Localization error vs. transmitter and receiver
distance
The primary ARTS infrastructure consists of seven 40-m
towers with receiving units, but can be expanded through
the use of mobile, understory units. We wanted to study the
impact of mobile understory units for improving coverage and
accuracy of the system. We also wanted to investigate the
overall infrastructure needed for higher-accuracy tracking. The
objective of this study was to understand the tradeoff between
accuracy and cost (including labor and maintenance) of the
ARTS tracking system.
The permanent towers of the ARTS, spaced approximately
800 m apart, are well suited to track the location of larger
terrestrial animals (weight greater than 10 kg) or smaller
arboreal animals (weight between 0.5 and 10 kg) but are too
far apart to triangulate the locations of smaller animals. The
size of the animal that can be tracked and the accuracy to which
they can be located is limited by the distance between receiver
stations. We decrease our inter-receiver distances by temporarily
setting up two understory units to fill gaps in our coverage.
This denser network of receivers also inherently has better
accuracy in locating an animal because the effects of angular
error increase with linear distance between the transmitter and
the receiver.
FIGURE 9. Modeled location estimates as a function of transmitter-
receiver distance.
To evaluate the added location accuracy provided by an
even denser networks of receivers, we used the location
error simulation model in the triangulation software LOAS
(www.ecostats.com). Setting an estimated bearing error
of 4(our estimated error after bearing smoothing), we
simulated system configurations with different between-
receiver distances. The results show the extent to which location
accuracy would improve by having more receivers over a
smaller area (Fig. 9). Equally important is the decrease (by
a factor of 4) in the variance of the location error. Location
accuracy improves to as much as <5 m when the receivers are
spaced 50 m apart. Such a resolution would be sufficient to
record the movement of small understory rodents or large insects
at sufficient resolution to map them onto individual tree crowns.
However, achieving such a dense receiver network would be
a major challenge. For example, we estimated that it would take
another 30 understory receiver units, in addition to the existing
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
12 R. Kays et al.
above-canopy towers, to give us this accuracy over about 75ha.
Achieving this density of receivers would be very expensive and
incredibly complex.
6.5. Geographic coverage model
For ARTS to function, transmitters worn by study animals must
be within the range of the ARU and their associated antenna
arrays. This presents a challenge as the transmitters used in
radio telemetry are low-power, and animals can move over large
areas. To maximize coverage, we mounted antennas atop above-
canopy radio towers set on hilltops. Preliminary tests suggested
that placing antennas on towers 10 m above the forest canopy
roughly doubled the range at which a ground level transmitter
could be detected in comparison with antennas in or below the
tree canopy. Thus, tree-mounted antennas would not be nearly
as efficient as tower-mounted ones.
We tested the actual range of our system by walking all the
trails on BCI while holding radio transmitters at 1 m height. We
recorded the strength of signal received from each of the seven
tower-mounted receivers and then related this to the specific
location of the transmitters to test for relationships of elevation,
angle of the slope relative to the receiver, distance to the tower
and distance to the nearest area with line-of-site to the receiver.
Using these data, we made a model to predict the strength
of signal received by an ARTS tower based on the strength of
the transmitter and the surrounding landscape (r20.39, df: 5,
P<0.0001). All variables were included in the final model,
which we used to predict island-wide coverage for different
types of transmitter (Table 4).
Our model predicts the range of animals walking on the forest
floor (Fig. 10a), but does not take into account the effect of their
movement up into trees, or down into underground holes. Our
experience shows that the effect of going underground depends
on exactly how far down an animal goes, but radio-tagged
animals sleeping in very deep holes can sometimes only be
heard within a few dozen meters of the hole. Transmitters in the
forest canopy can be detected from a greater distance than those
at ground level. We quantified this by raising test transmitters
into trees and found an increase in signal strength of 25–35 dB
between the forest floor and 40 m high in a canopy tree (Fig. 11).
This difference can be incorporated into our model to predicting
the coverage for arboreal animals (Fig. 10b).
6.6. Radio interference
Detecting the faint signal of a radio-tagged animal can be
impossible if there is strong interference from other radio
transmitters on or near the same frequency. We discovered that
the two-way communication networks of local taxis, television
stations and other unknown sources of radio-traffic caused
TABLE 4. Variables used in function to predict the strength of signal
received by a tower-mounted receiver from an animal-mounted radio
transmitter, based on local landscape characteristics.
Variable Coefficient
Transmission power of radio-tag (dBm) 1.18055670
Distance to tower-mounted receiver (m) 0.01218720
Distance to nearest area with line-of-site to tower
(meters, 0 if in line of site)
0.00244816
Elevation (meters) 0.03867165
Angle of the slope of the hill related to the receiving
antenna (0 =facing towards tower, 180 =facing
away)
0.02403199
Intercept 98.9454675
FIGURE 10. The predicted range of ARTS towers to detect different types of radio-collars. Triangles represent the location of the towers and the
colors show the number of towers expected to receive the radio signal from an animal at a given location for (a) a monkey-sized collar on the forest
floor or (b) in the trees. (a) Tower overlap when monkey is on the ground. (b) Tower overlap when monkey is in trees.
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
Tracking Animal Location and Activity with an ARTS 13
significant radio-interference. This is a common problem for all
animal tracking studies near human settlements, but in our case
we could use the automated receivers to map this interference
-140
-130
-120
-110
-100
-90
-80
0 5 10 15 20 25
received signal strength (dBm)
height above the ground (m)
FIGURE 11. Change in signal strength recorded by a telemetry
receiver from a radio-transmitter at different heights above the ground.
Each line represents a different location where a radio-collar was
hauled into the rainforest canopy. The biggest increase in signal
strength comes in moving the collar off the ground and up above of
first few meters of understory vegetation.
on BCI. Normal background signal for our system was
130 dB, and anything above this was considered man-made
interference.
We first made a broad scale survey scanning 33 channels from
148 to 164 MHz across one entire day (Fig. 12a). This showed
an extensive interference at higher frequencies which we later
tracked to a television repeater, which does not transmit early
in the morning. We also conducted a more fine-grained scan
for interference at BCI across the 4-MHz that we primarily use
for research (Fig. 12b). Most channels are clear, although there
are scattered frequencies that we avoid when purchasing radio-
transmitters to attach to animals.
7. REAL-WORLD DEPLOYMENT CHALLENGES
7.1. Tower repair and management
Table 5enumerates the technological and maintenance
challenges that we have faced with the ARTS system, as well
as the solutions we have employed. The environment has a
significant impact on the system ranging from lightening strikes,
to tree falls and corrosion (Fig. 13a). However, so far, we have
-150
-140
-130
-120
-110
-100
-90
-80
149.0
149.2
149.4
149.6
149.8
150.0
150.2
150.4
150.6
150.8
151.0
151.2
151.4
151.6
151.8
152.0
152.2
152.4
152.6
152.8
dBm
MHz
Tower 1
Tower 2
Tower 3
Tower 4
0:00
3:00
6:00
9:00
12:00
15:00
18:00
21:00
148.0
149.5
151.0
152.5
154.0
155.5
157.5
159.0
160.5
162.0
163.5
Time of Day
frequency MHz
--120---110
--130---120
--140---130
FIGURE 12. The background radio interference at BCI, levels above 130 dBm interfere with our ability to track radio-tagged animals. (a) shows
how interference varies over a broad-band (16MHz) across the course of one day. (b) shows a more fine-scale view of interference over the 4 MHz
used for most animal tracking, and how this varies across four different tracking towers. (a) Broad scale radio interference over the course of one
day from one tower. (b) Fine-Grained-Interference at BCI from four different ARTS towers.
TABLE 5. ARTS: repairs and maintenance details.
Equipment Problem type Frequency Solution
Towers (7) Large tree fall on guy wires Two times total Replacement of bent tower sections ($500)
Towers (7) Small branches growing on guy wires Once per year per tower Trim branches once per year
Towers (7) Corrosion at guy wires, bolts and nuts performed 6 years General tower maitenance ($6500)
ARU (7) Lightning strike Three times total Units repaired by vendor ($250)
FreeWave 900 MHz radios Lightning strike Four times total Units must be replaced ($1000)
RG-8 Cable Broken in middle One time total Cable was replaced ($500)
RG-59 Cable (42) Connectors go bad Replaced after 4 years Cable replaced ($4 each)
PL-259 connectors (64) Corrosion and bad connectivity Rreplaced after 4 years Replaced with custom built aluminum mounts
Solar panel mounts (7) Corrosion Replaced after 4 years Connectors replaced ($5)
Antennae (42) Corrosion Replaced after 4 years all antenna replaced
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
14 R. Kays et al.
FIGURE 13. Real world challenges of tracking animals in the rainforest include towers damaged by falling trees (a) and the limited detection
range for small animals on the forest floor (b). (a) Tower at BCI broken by tree falling on guy wires and (b) ARTS tracking a rat species (animal
with small body size).
been quite satisfied with ARTS in terms of its maintenance effort
and cost.
7.2. System power consumption
Power management in a real-world deployment is always a
challenging issue. Each ARTS tower is equipped with one solar
panel as a renewable source of energy. We now describe the
system power consumption statistics. The ARU Drain Current
is 37 mA. The Filter Box power consumption is 20 mA and
the Freeway radio power consumption is 64mA. We typically
use 90Ah rated batteries. At 75% of discharge (90 0.75 =
67.5 Ah), the battery can last in the field for 76 days
(67.5Ah/37 mA 24 h). As batteries get older or if they get
fully discharged and charged again, they tend to accumulate
less charge.
If the subunit also has a radio (in addition to anARU discussed
already), then the total power consumption is as follows:
EARU +EFilter Box +EFW.(2)
This translates to 23 days of battery life
(67.5Ah/121 mA 24 h). If we constantly monitor the
batteries and avoid them to get fully discharged (aka its voltage
does not go below 10V), they could last up to 2 years. How-
ever, in practice this level of monitoring is unrealistic, and the
batteries last less than this.
As mentioned before, in addition to the above-canopy tower,
we also have understory units. These units do not have solar
panels, as a renewable source of energy since sunlight can
hardly penetrate through dense tree canopy, and use batteries
as their sole source of energy. We typically change the subunit
batteries every 3 weeks. The system had been engineered to
ensure a positive energy balance, and never experienced a case
of insufficient power in 6 years of operation.
7.3. Deployment barrier
One drawback of theARTS system is that it is large and complex.
The complete ARTS system is best suited for use at well-
funded, easily accessible research sites, although components
of the system may also be useful in smaller temporary studies.
However, once operational, the system is able to quickly collect
huge quantities of high quality data, and thus promises to
provide an enormous leap forward in our ability to describe
and understand the ecology of animal movement.
7.4. Tracking error for animals with small body size
or large home range
Using ARTS for tracking small animals is challenging. For
example, Fig. 13b shows shows the predicted signal reception
by the ARTS system for a rat-sized transmitter. This is due to
the fact that since small animals cannot carry large batteries,
transmission power has to be kept low, which in turn generates
a weak signal. This weak signal, coupled with environmental
factors such as dense tree canopy, makes it hard to detect. For
the cases where localization and tracking from at least 3 towers
is not possible, determining whether an animal is active or not
is still very feasible using signal strength changes observed at
one tower. In addition, tracking animals with very large home
ranges is challenging since the animals can come and can go in
and out of the coverage area. Nonetheless,ARTS is still a viable
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
Tracking Animal Location and Activity with an ARTS 15
and attractive option for tracking small animals compared with a
GPS-based system, which are too large to attach to most species
(Fig. 1).
8. SUMMARY OF SCIENTIFIC IMPACTS
Many behaviors that have a major impact on the survival
and reproduction of wild animals are difficult to study
because they are infrequent, cryptic or occur over large
spatial or temporal scales. Our lack of data on important
biological processes, including predation, dispersal, migration
and intergroup competition, is slowing the development and
testing. Compared to traditional methods, including direct
observation and manual telemetry, ARTS allows scientists to
collect more data on the activity and locations for a greater
number of study animals for longer periods of time, and over
larger areas. In addition, the availability of live data allows
scientists to make more efficient use of their field time, and
recognize and act on rare but important events. These functions
have facilitated new scientific discoveries on topics that have
proved difficult to study, using traditional observational or
telemetry methods.
Traditional methods of observation are limited in studying
many important biological processes, including predation,
dipsersal, migration and intergroup competition. Our lack of
data on these is slowing the development and testing of
ecological theory, and interferes with our ability to develop
effective conservation and management strategies in the rapidly
changing world.
8.1. Activity and mortality of animals
The activity of an animal is one of the most basic descriptors
of animal behavior, revealing daily rhythms that are one of the
three basic dimensions of an ecological niche (along with diet
and locomotory mode) [46]. In addition, variations in activity
patterns are related to individual physical and social condition.
Because the ARTS records data constantly and continuously,
these data can provide details on both daily activity (Fig. 6a)
as well as long-term patterns (Fig. 14). These document one
key dimension of an animal’s ecological niche, and have also
tiger
clock time
0 12 24 36 48 0 12 24 36 48
Aug 1, 2006
(a) (b)
Sep 1
Oct 1
Nov 1
Dec 1
Jan 1, 2007
Feb 1
jaws
clock time
Aug 1, 2006
Sep 1
Oct 1
Nov 1
Dec 1
Jan 1, 2007
Feb 1
FIGURE 14. Actograms showing the patterns of daily activity over 7 months for a three-toed sloth with almost no daily rhythm (a) and a two-toed
sloth with strong nocturnal activity (b). Each row represents two continuous days of data, with dark color indicating activity and white indicating
inactivity at 4-min intervals. The long term data also allows the determination of finer-scale patterns such as the early AM rest period of the
three-toed sloth, and a regular effect of the lunar cycle on the nocturnal activity of two-toed sloths. (a) Actogram of a three-toed sloth. (b)Actogram
of a two-toed sloth.
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
16 R. Kays et al.
been used to model how environmental variables, such as
weather and food availability, affect animal activity [27]. As
with the mortality data, results are often surprising. For example,
Holland et al. [47] used ARTS to find that Pallas’ Mastiff
Bats were active flying and foraging for only 82 min per night,
spending the rest of their time hiding in a roost. This suggests
that this species is a very efficient forager, and is probably not
limited by food abundance.
8.2. Movement ecology
Being able to determine how animals use their habitat is critical
for understanding their ecology, behavior and evolution, as
well as for planning effective conservation strategies [25].
Animal movements are influenced by a wide range of factors,
including the distribution of important resources such as food
and water, and interactions with members of their own and
other species. Investigating how these factors interact to shape
patterns of space-use and resource access has been problematic
because of the logistical challenges involved in simultaneously
tracking the movements of many animals. ARTS provides a
means of overcoming this obstacle. For example, simultaneous
and continuous tracking of six white-faced capuchin (Cebus
capucinus) social groups over a six month period demonstrated
that areas that were shared by multiple groups (i.e. areas
of home-range overlap) were underused compared with non-
shared areas [48]. This heterogeneous pattern of space use
was not necessarily the result of intense intergroup aggression,
but could also be explained by the economics of memory-
based foraging [49]. Simultaneous tracking of multiple social
groups also elucidated the factors that determined the balance
of power between neighboring social groups. As expected,
large group size conferred a competitive advantage in this
population, but surprisingly, this effect was less important in
determining the outcome of aggressive inter-group encounters
than the location of the fight [25]. This strong home-field
advantage may be the key to understanding how small social
groups are able to persist in the face of intense, between-group
competition.
In another study, Crofoot et al. [49] tested a critical
assumption of animal behavior studies: that habituated groups
do not move more when followed by a human than they
do when left alone. This has been assumed by thousands of
studies, but never tested empirically. ARTS movement data
allowed them to compare the distance and speed moved by
habituated animals accompanied by a ground-based observer
with other days when they were not being followed. No effect
was found, offering the first real field confirmation of an age-old
assumption.
8.3. Species interaction studies
The seeds are most commonly removed by a 2–4 kg caviomorph
rodent, the Central American agouti (Dasyprocta punctata).
Agoutis typically scatterhoard seeds under a few centimeter of
dirt which protects the seeds from parasitic insects and other
seed predators. During periods of food scarcity, agoutis return to
their cached seeds to consume the seed. We developed a tagging
method, which allowed researchers to turn the transmitters off
of buried seeds without digging up the seed [50]. This has
allowed us to quantify secondary dispersal of seeds, and plot
where individual seeds have moved over time. This method
also ensured that the transmitters turned on when the seed was
moved or eaten by an animal. The ARTS allowed researchers
to know if seed transmitters were active in real time, which led
to a major reduction in field effort and increased the life of the
transmitters.
9. CONCLUSIONS
Networks of sensors deployed in natural areas are increasingly
being used to collect data at the scales and rates needed to
address modern environmental challenges [5156]. However,
these Lare typically focussed on measuring abiotic conditions,
or attributes of sessile plants. Here we describe a unique
ARTS designed to track the movements and activity patterns
of animals. The results of this 6 years system deployment
are relevant not only because of the biological results, but
also because of the more general challenges of maintaining
electronic sensors in real-world rainforest conditions. The
strength of the ARTS for science application lies in its ability
to simultaneously stream real-time data on animal location
and activity from dozens of animals tagged with inexpensive
radios as small as 0.2 g. GPS offers an alternative tracking
technology to ARTS style systems that is not limited by range or
accuracy, but suffers from other shortcomings including large,
expensive tracking tags and complicated data retrieval options.
In fact, our 6-year long experience shows that ARTS system
collects data more frequently than typical animal-borne GPS
collars (12 locations/h) with slightly lower accuracy (50 m)
but at much reduced cost per tag (10X less expensive). The
success of ARTS across the relatively small scales of Barro
Colorado Island shows the scientific importance of real-time,
high-resolution tracking data and why it is important to continue
to refine tracking technology to make these studies possible on
the smallest animals, at the largest scales [17]. We hope that the
experience gained and lessons learned during our deployment of
the ARTS system are applicable to the broader sensor network
applications and help push for the advancement of the animal
tracking technology.
FUNDING
This work was partially supported by the US National Science
Foundation (Award Number: 0756920, 0201307), Smithsonian
Tropical Research Institute, Frank Levinson Family Foundation
and National Geographic Society.
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
Tracking Animal Location and Activity with an ARTS 17
REFERENCES
[1] Lord, R., Bellrose, F. and Cochran, W. (1962) Radio telemetry of
the respiration of a flying duck. Science,137, 39–40.
[2] Kenward, R.E. (2001) A Manual for Wildlife Radio Tagging.
Academic Press.
[3] Millspaugh, J. and Marzluff, J. (2001) Tracking and Animal
Populations. Academic Press.
[4] Garshelis, D., Gibeau, M. and Herrero, S. (2005) Grizzly bar
demographics in and around banff national park and kananaskis
country, Alberta. J. Wildlife Manag.,69, 277–297.
[5] Lockwood, M.A., Griffin, C.P., Morrow, M.E., Brandel, C.J.
and Silvy, N.J. (2005) Survival, movements, and reproduction
of released captive-reared attwater’s prairie-chicken. J. Wildlife
Manag.,69, 1251–1258.
[6] White, P.A. (2005) Maternal rank is not correlated with cub
survival in the spotted hyena, crocuta crocuta, Behav. Ecol.,16,
606–613.
[7] Mech, L.D., Adams, L.G., Meier, T.J., Burch, J.W. and Dale,
B.W. (1998) The Wolves of Denali. University of Minnesota Press,
Minneapolis, MN.
[8] Nathan, R., Perry, G., Cronin, J.T., Strand, A.E. and Cain, M.L.
(2003) Methods for estimating long-distance dispersal. OIKOS,
103, 261–273.
[9] Winkler, D.W., Wrege, P.H., Allen, P.E., Kast, T.L., Senesac, P.,
Wasson,M.F., Llambias, P.E., Ferretti, V. and Sullivan, P.J. (2004)
Breeding dispersal and philopatry in the tree swallow. Condor,
106, 768–776.
[10] Kraus, C. (2004) Where have all the cavies gone? Causes and
consequences of predation by the minor grison on a wild cavy
population. Oikos,105, 489–500.
[11] Mech, L.D. (1967) Telemetry as a technique in the study of
predation. J. Wildlife Manag.,31, 492–496.
[12] Sinclair, A.R., Mduma, S. and Brashares, J.S. (2003) Patterns of
predation in a diverse predator-prey system. Nature,425, 288–
290.
[13] Aliaga-Rossel, E., Moreno, R., Kays, R. and Giacalone, J. (2006)
Ocelot (leopardus pardalis) predation on agouti (dasyprocta
punctata). Biotropica,38, 691–694.
[14] Turchin, P. (1998). Quantitative Analysis of Movement:
Measuring and Modeling Population Redistribution in Animals
and Plants. Sinauer Associates, Inc.
[15] Long, R., MacKay, P., Zielinski, W. and Ray, J. (2008)
Noninvasive Survey Methods for Carnivores. Island Press.
[16] Naef-Daenzer, B., Früh, D., Stalder, M., Wetli, P. and
Weise, E. (2005) Miniaturization (0.2g) and evaluation of
attachment techniques of telemetry transmitters. J. Exp Biol.,208,
4063–4068.
[17] Wikelski, M., Kays, R., Kasdin, J., Thorup, K., Smith, J., Cochran,
W. and Swenson, G. (2007) Going wild: what a global small-
animal tracking system could do for experimental biologists. Exp.
Biol.,210, 181–186.
[18] Smith, F., Lyons, S., Ernest, S., Jones, K., Kaufman, D., Dayan,
T., Marquet, P., Brown, J. and Haskell, J. (2003) Body mass of
late quaternary mammals. Ecology,84, 3403.
[19] Lewis, J., Rachlow, J., Garton, E. and Vierling, L. (2007) Effects
of habitat on gps collar performance: using data screening to
reduce location error. J. Appl. Ecol.,44, 663–671.
[20] Cochran, W., Warner, D., Tester, J. and Kuechle, V. (1965) Auto-
matic radio-tracking system for monitoring animal movements.
BioScience,15, 98–100.
[21] Paton, S. (2005). Meteorological and hydrological summary for
barro colorado island.
[22] Horan, R.V., Kays, R., Wikelski, M. and Carrol, J. (2010)
Automated Radio Telemetry for the Study of Anuran Dispersal,
Movement and Activity Patterns: Perspectives from Barro
Colorado Island, Panama. Proc. 17th Annual Wildlife Society
Annual Conf., Snowbird, Utah, USA, October.
[23] Cochran, W., Swenson, G. and Pater, L. (2001) Radio direction-
finding for wildlife research. Antenna Applications Symposium
University of Illinois at Urbana-Champaign, IL, USA.
[24] Hutchinson, J. and Alba, J. (1997) Heuristics and biases in the
eyeballing of data: the effects of context on intuitive correla-
tion assessment. J. Exp. Psychol.-Learn. Memory Cogn.,23,
591–621.
[25] Crofoot, M.C., Gilby, I.C., Wikelski, M.C. and Kays, R.W.
(2008) The home field advantage: location affects the outcome of
asymmetric intergroup contests in Cebus capucinus. Proc. Natl.
Acade. Sci.,105, 577–581.
[26] Kjos, C. and Cochran, W. (1970) Activity of migrant thrushes
determined by radio-telemetry. Wilson Bull.,82, 225.
[27] Lambert, T.D., Kays, R.W., Jansen, P.A., Aliaga-Rossel, E. and
Wikelski, M. (2009) Nocturnal activity by the primarily diurnal
central american agouti (Dasyprocta punctata) in relation to
environmental conditions, resource abundanceand predation risk.
J. Trop. Ecol.,25, 211–215.
[28] Szewczyk, R., Mainwaring, A., Polastre, J., Anderson, J. and
Culler, D. (2004) An Analysis of a Large Scale Habitat
Monitoring Application. Proc. 2nd Int. Conf. Embedded
Networked Sensor Systems, New York, NY, USA, SenSys ’04,
pp. 214–226. ACM.
[29] Juang, P., Oki, H., Wang, Y., Martonosi, M., Peh, L.S. and
Rubenstein, D. (2002). Proc. 10th Int. Conf. Architectural
Support for Programming Languages and Operating Systems,
New York, NY, USA, ASPLOS-X, pp. 96–107. ACM.
[30] Dyo, V., Ellwood, S.A., Macdonald, D.W., Markham, A.,
Mascolo, C., Pásztor, B., Trigoni, N. and Wohlers, R. (2009)
Wildlife and Environmental Monitoring usingrfid and wsn
Technology. Poster Abstract: In Proc. Int. Conf. Embedded
Networked Sensor Systems (SenSys), New York, NY, USA.
[31] He, T., Huang, C., Blum, B.M., Stankovic, J.A. andAbdelzaher, T.
(2003) Range-Free Localization Schemes for Large Scale Sensor
Networks. Proc. 9th Annual Int. Conf. Mobile Computing and
Networking, New York, NY, USA, MobiCom ’03, pp. 81–95.
ACM.
[32] Ward, A., Jones, A. and Hopper, A. (1997) A new location
technique for the active office. IEEE Pers. Commun.,4,
42–47.
[33] Bahl, P. and Padmanabhan,V.N. (2000) RADAR: An in-building
RF-based user location and tracking system. INFOCOM,2,
775–784.
[34] Priyantha, N.B., Chakraborty, A. and Balakrishnan, H. (2000)
The Cricket Location-Support System. Proc. 6th Annual Int.
Conf. Mobile Computing and Networking, NewYork, NY, USA,
MobiCom ’00, pp. 32–43. ACM.
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
18 R. Kays et al.
[35] Savvides, A., Han, C.-C. and Strivastava, M.B. (2001)
Dynamic Fine-Grained Localization in Ad-hoc Networks of
Sensors. Proc. 7th Annual Int. Conf. Mobile Computing and
Networking, New York, NY, USA, MobiCom ’01, pp. 166–179.
ACM.
[36] Niculescu, D. and Nath, B. (2003) Ad Hoc Positioning System
(APS) using AoA. Proc. INFOCOM, San Franciso, CA, USA,
March 30–April 3, 2003, pp. 1734–1743.
[37] Girod, L. and Estrin, D. (2001) Robust range estimation using
acoustic and multimodal sensing. IEEE/RSJ Int. Conf. Intelligent
Robots and Systems (IROS 2001), Maui, HI, USA.
[38] Bulusu, N., Heidemann, J. and Estrin, D. (2000) Gps-less low cost
outdoor localization for very small devices. IEEE Pers. Commun.
Mag.,7, 28–34.
[39] Stoleru, R., He, T., Stankovic, J.A. and Luebke, D. (2005)
A High-Accuracy, Low-Cost Localization System for Wireless
Sensor Networks. Proc. 3rd Int. Conf. Embedded Networked
Sensor Systems, New York, NY, USA, SenSys ’05, pp. 13–26.
ACM.
[40] Corke, P., Peterson, R. and Rus, D. (2005) Networked Robots:
Flying Robot Navigation using a Sensor Net. In Chatila, R.
(ed.) Robotics Research, Vol. 15. Springer Tracts in Advanced
Robotics, pp. 234–243. Springer, Berlin/Heidelberg. http://
dx.doi.org/10.1007/11008941_25. Doi: 10.1007/11008941_25.
[41] Pathirana, P.N., Bulusu, N., Savkin, A.V. and Jha, S. (2005)
Node localization using mobile robots in delay-tolerant sensor
networks. IEEE Trans. Mob. Comput.,4, 285–296.
[42] Sichitiu, M.L. and Ramadurai, V. (2004) Localization of wireless
sensor networks with a mobile beacon. IEEE Int. Conf. Mobile
Ad-hoc and Sensor Systems, Fort Lauderdale, Florida, USA,
October, pp. 174–183.
[43] Tamir, T. (1977) Radio wave propagation along mixed paths
in forest environments. IEEE Trans. Antennas Propag.,25,
471–477.
[44] Lenth, R. (1981) Robust measures of location for directional data.
Technometrics,23, 77–81.
[45] Lenth, R. (1981) On finding the source of a signal. Technometrics,
23, 149–154.
[46] Halle, S. (2000) Ecological Relevance of DailyActivity Patterns.
Activity Patterns in Small Mammals: An Ecological Approach,
Springer, Berlin, pp. 68–90.
[47] Holland, R., Meyer, C., Kalko, E., Kays, R. and Wikelski, M.
(2011) Emergence time and foraging activity in pallas? mastiff
bat, molossus molossus (chiroptera: molossidae) in relation to
sunset/sunrise and phase of the moon. acta chiroptera (in press).
[48] Wrangham, R., Crofoot, M., Lundy, R. and Gilby, I. (2007) Use
of overlap zones among group-living primates: a test of the risk
hypothesis. Behaviour,144, 1599–1619.
[49] Crofoot, M.C., Lambert, T.D., Kays, R. and Wikelski M.C. (2010)
Does watching a monkey change its behaviour? Quantifying
observer effects in habituated wild primates using automated
radiotelemetry. Animal Behav.,80, 475–480.
[50] Hirsch, B., Kays, R. and Jansen, P. A telemetric thread-tag for
tracking seed dispersal by scatter-hoarding animals. Methods
Ecol. Evol. (in review).
[51] Creon: The coral reef environmental observatory network.
http://www.coralreefeon.org/.
[52] Gleon: The global lake ecological observatory network.
http://www.gleon.org/ (accessed July 2011).
[53] Ocean observatories initiative. http://oceanobservatories.org/
(accessed July 2011).
[54] Estrin, D., Mitchner,W., Bonito, G. and the workshop participants
(2003). Environmental Cyberinfrastructure Needs for Distributed
Sensor Networks: A Report from a National Science Foundation
Sponsored Workshop.
[55] National science foundation, 2008. Cyberinfrastructure for
Environmental Observation Networks (CEON) Workshop
Report. http://proxy.cedarweb.hao.ucar.edu/ceoaf/images/9/9f/
FebCEONWorkshop.pdf.
[56] Porter, J.H., Nagy, E., Kratz, T. K., Hanson, P., Collins, S.L. and
Arzberger, P. (2009). New eyes on the world: advanced sensors
for ecology. BioScience,59, 385–397.
The Computer Journal,2011
by guest on August 15, 2011comjnl.oxfordjournals.orgDownloaded from
... It is therefore not feasible to follow more than a few individuals for long periods each day, so locations are often estimated sparsely and irregularly. Attempts to automate wildlife tracking based on standard radio tags were made as early as the 1960s (Cochran et al., 1965) and more recently using a bounding array of receivers distributed within a specific region (Kays et al., 2011). Locations can be estimated from the data collected by these receivers and has the potential to be accessed almost instantaneously. ...
... Specifically, for the sections of the study that occurred in the most northern part of the mudflat, all detecting receiver stations were located to the south of the tag, and therefore the tag was outside the array of receiving stations. While one receiver station was situated to the north of Griend, on Terschelling, 10 km away, this seems to be outside the range of reception for tags at 1 m above ground, despite another receiver to the West, on Vlieland, (Beardsworth, 2020;Kays et al., 2011). However, the WATLAS system has very few topographical obstacles that limit reception of transmissions by receivers, making it an ideal 'baseline' ...
Article
Full-text available
Fine‐scale tracking of animal movement is important to understand the proximate mechanisms of animal behaviour. The reverse‐GPS system – ATLAS – uses inexpensive (~€25), lightweight (<1g) and low‐power (~0.4mJ/transmission) tags. Six systems are now operational worldwide and have successfully tracked over 50 species in various landscape types. The growing use of ATLAS to track animal movement motivates further refinement of best‐practice application and an assessment of its accuracy. Here, we test the accuracy and precision of the largest ATLAS system, located in the Dutch Wadden Sea, using concurrent GPS measurements as a reference. This large‐scale ATLAS system consists of 26 receivers and covers 1326km2 of intertidal region, with almost no physical obstacles for radio signals, providing a useful baseline for other systems. We compared ATLAS and GPS location estimates for a route (mobile test) and 16 fixed locations (stationary test) on the Griend mudflat. Precision was estimated using standard deviation during the stationary tests. We also give examples of tracked red knots (Calidris canutus islandica) to illustrate the use of the system in tracking small shorebirds (~120g). ATLAS‐derived location estimates differed from GPS by a median of 4.2m (stationary test) and 5.7m (moving test). Signals that were collected by more receiver stations were more accurate, although even 3‐receiver localisations were comparable with GPS localisations (~10m difference). Receivers that detected 90% of the 1Hz transmissions from our test tag were within 5km of their furthest detection but height of both receiver and tag seemed to influence detection distance. The test tag (1Hz) had a fix rate of >90% at 15 of 16 stationary sites. Tags on birds (1/6Hz) on the Griend mudflat had a mean fix rate of 51%, yielding an average sampling rate of 0.085Hz. Fix rates were higher in more central parts of the receiver array. ATLAS provides accurate, regional‐scale tracking with which hundreds of relatively small‐bodied species can be tracked simultaneously for long periods of time. Future ATLAS users should consider the height of receivers, their spatial arrangement, density and the movement modes of their study species (e.g., ground‐dwelling or flying).
... Passive monitoring of amphibians unable to carry heavy GPS tags is also possible. Automated radio telemetry systems (ARTS) use multiple directional antennae attached to various receivers placed in the field to semi-continuously monitor the signal strengths of VHF radio transmitters attached to animals in order to determine their location (Kays et al. 2011). ARTS can monitor multiple individuals and like GPS tags, greatly reduce the disruptions from observers manually triangulating the animals each time position data are required. ...
Article
Full-text available
The use of biotelemetry in amphibian behaviour, migration, dispersal and homing research has increased with the miniaturisation and improved reliability of telemetry equipment. Here we synthesise biotelemetry methods and outcomes from the past two decades of published studies (n = 128; 1 January 2000–31 December 2020). We discuss trends in amphibian study duration, transmitter weight and transmitter attachment methods. The most commonly used harness designs for external transmitter attachment and surgical methods for internal implantation are described, and the benefits and potential problems of these methods are discussed. We recommend improved practices for future studies and address gaps within the current amphibian telemetry literature in regard to accurately reporting study endpoints and study subject survival. Finally, we highlight emerging technologies that could influence the next 20 years of telemetry research on amphibians.
... The detection ranges of stations are usually variable owing to the specification of the transmitters, antennas and receivers, the height of the antennas, the orientation and flight height of the animal relative to the antennas, landscape structure, topography, and weather conditions (Crewe et al., 2019;Kays et al., 2011;Lagerveld et al., 2017). Accordingly, it was not possible to extract information on the flight distance on a small scale, such as between Helgoland and the first receiving station on the mainland accurately. ...
Article
Full-text available
The wind energy-bat conflict is well documented for the onshore sector, with high numbers of casualties, specifically for migratory bat species. Offshore wind turbines might be a threat to bats as well, yet offshore bat migration is poorly documented. Accordingly, potential conflicts between bat conservation and offshore wind energy production are difficult to evaluate. Here, we used automated radio-telemetry to track 50 km continuous offshore movements of two Nathusius' pipistrelles (Pipistrellus nathusii) within the Motus network. After crossing the marine waterbody, tagged bats traveled over several hundred kilometers along the coastline from Germany towards the Netherlands and Belgium. Our study highlights the possibility for migratory bats to collide with offshore and coastal wind turbines. Therefore, we plead for implementing pre- and post-construction surveys and adequate mitigation schemes at offshore wind turbines in sensitive areas of the North and Baltic Sea if not already practised.
... Space use of animals was monitored with automated radio telemetry (ART, e.g., Hoffmann et al., 2018;Kays et al., 2011;Schirmer et al., 2019), and risk taking via radio frequency identification (RFID) systems placed at risky locations. The combination of different methods ...
Article
Full-text available
Individuals of a population may vary along a pace-of-life syndrome from highly fecund, short-lived, bold, dispersive “fast” types at one end of the spectrum to less fecund, long-lived, shy, plastic “slow” types at the other end. Risk-taking behavior might mediate the underlying life history trade-off, but empirical evidence supporting this hypothesis is still ambiguous. Using experimentally created populations of common voles (Microtus arvalis)—a species with distinct seasonal life history trajectories—we aimed to test whether individual differences in boldness behavior covary with risk taking, space use, and fitness. We quantified risk taking, space use (via automated tracking), survival, and reproductive success (via genetic parentage analysis) in 8 to 14 experimental, mixed-sex populations of 113 common voles of known boldness type in large grassland enclosures over a significant part of their adult life span and two reproductive events. Populations were assorted to contain extreme boldness types (bold or shy) of both sexes. Bolder individuals took more risks than shyer ones, which did not affect survival. Bolder males but not females produced more offspring than shy conspecifics. Daily home range and core area sizes, based on 95% and 50% Kernel density estimates (20 ± 10 per individual, n = 54 individuals), were highly repeatable over time. Individual space use unfolded differently for sex-boldness type combinations over the course of the experiment. While day ranges decreased for shy females, they increased for bold females and all males. Space use trajectories may, hence, indicate differences in coping styles when confronted with a novel social and physical environment. Thus, interindividual differences in boldness predict risk taking under near-natural conditions and have consequences for fitness in males, which have a higher reproductive potential than females. Given extreme inter- and intra-annual fluctuations in population density in the study species and its short life span, density-dependent fluctuating selection operating differently on the sexes might maintain (co)variation in boldness, risk taking, and pace-of-life.
... While this does not seem to cause obvious negative effects in animals such as insects or when tags are only fitted for a short amount of time during non-demanding times in an animals' life, relative device mass is one of the most critical parameters in animal tracking [10][11][12][13][14]. Technological advancements providing high-definition tags with small mass may thus allow essential studies of small species with strong conservation concern and reduce the device impact on larger species. Most of the currently available ultra-light tracking devices below one gram do not determine positions based on GPS, but transmit high-frequency radio signals [15][16][17][18] or use light sensors [19]. These technologies require much less energy and can operate on small batteries. ...
Article
Full-text available
GPS-enabled loggers have been proven as valuable tools for monitoring and understanding animal movement, behaviour and ecology. While the importance of recording accurate location estimates is well established, deployment on many, especially small species, has been limited by logger mass and cost. We developed an open-source and low-cost 0.65 g GPS logger with a simple smartphone-compatible user interface, that can record more than 10,000 GPS fixes on a single 30 mAh battery charge (resulting mass including battery: 1.3 g). This low-budget ‘TickTag’ (currently 32 USD) allows scientists to scale-up studies while becoming a ‘wearable’ for larger animals and simultaneously enabling high-definition studies on small animals. Tests on two different species (domestic dog, Canis lupus familiaris and greater mouse-eared bats, Myotis myotis ) showed that our combination of optimised hardware design and software-based recording strategies increases the number of achievable GPS fixes per g device mass compared to existing micro-sized solutions. We propose that due to the open-source access, as well as low cost and mass, the TickTag fills a technological gap in wildlife ecology and will open up new possibilities for wildlife research and conservation.
... The case study conducted in the studied area within July 2020 to September 2020 to identify the recovery system from the negative impact of processed sensor networks [40]. This research was a multi-diversified experiment in connection with sensor technology to augment non-communicable diseases among animals and the human body [41]. The study examined two specimens, one was a dog and another one was a cat among 14 individuals for identification of this misuse application. ...
Article
Full-text available
Background: Coronavirus is the most pressing scientific puzzle in the 21 st century. This is a pandemic spreading globally through exploration of various wireless sensor networks. Yet Medical authorities are facing the obnoxious ever-increasing causes of coronavirus as a very global turning issue. Aims: The study aims to outline the scientific model for recovery of coronavirus disease with comprehensive follow-up and services. Methods: A dynamic scientific model was established in connection with recovery of coronavirus disease. This model identified the COVID-19 patients who need boosted follow up to recover with dynamic community cares. Sensor data were collected from the patient's profile, diagnosis and complication records at light and dark environments. Results: The study demonstrates total 150 patients suffered from coronavirus disease and stayed at home isolation within optical GPS locations. In a light environment, all patients recover from coronavirus disease due to wireless sensor network isolation, changing their GPS locations instantly with tightly closed eyes and wearing anti-radiation sunglasses, and also clothe with black uniforms in the whole body. The obese patients required long time to recover in dark environment in comparison with others. Replication: The findings replicate the coronavirus disease recovery for dynamic health security that the physicians provide on the priority of strategy, mental health service, innovations, potentiality and personalism. Conclusion: Scientific healthcare sensor knowledge is indispensable for recovery of coronavirus disease. The study reveals the implementation of sensor network approach to patients with coronavirus disease recognizing those with augmenting physical, technological and mental healthcare requirements. The study suggests future research trajectories of a new alternative sensor network isolation model to promote global public health security.
... Due to the spreading coronavirus, the geopolitical tensions are intensifying, which hampers global public health security. This pandemic is a premeditated shock that is life-threatening to the global order and political leaders [1,17,18,19,20,21,22]. The day of e-vote casting, the democratic leaders suffered from sensor political disease due to tracking with the processed wireless sensor networks at the vote centre or adjacent places with insecure democratic area network control units. ...
Article
Full-text available
Coronavirus is a political agenda through misusing of biosensor weapons worldwide. This pandemic is a tactical shock as the extreme threat to the global orders and politics. Many people, political leaders, animals and the environment have been affected by this disease. These have interrupted the comprehensive economy, political relations, public health security through thought-provoking the strength of humanities for active democracy. Thus, young and old individuals are getting sick and later suffering from depression and sensor diseases, which reflects on national, regional and global democracy. Politicians and others are running towards loss in different ways through misuse of innovative technology, unstable political situation, horrible social relations, violence of human rights, climate crisis and terrible democratic surroundings. Most people use advances in technology, but none can know its impact in political life. The study represents mental health awareness, policy integration and secure network collaboration among all in responsible life. The research focused on the human beings are the root causes of all problems, whereas they are the ways of dynamic solutions due to approaching complications in democracy. Innovative technological communication is indispensable for politicians but such scientific knowledge is below par as democratic users. Lastly, the study suggests future research trajectories of a new alternative secure techno-political approach to protect democracy for a peaceful world with recovery systems that have saved millions of lives.
Chapter
The era of the Information Technology revolution has endeavored mankind with enormous technological comforts. Apart from human health care, cloud computing integrated with other technologies can be exploited for the welfare of endangered wildlife animals that are currently on the verge of facing extinction in the near future. International Union for Conservation of Nature (IUCN) has declared a huge number of wildlife species into red list falling into 'critically endangered', 'endangered' and 'vulnerable' species because of pathogenic diseases, poaching, habitat degradation, and climate change. The aim of the chapter is to propose a cloud computing based wildlife health care model to preserve, sustain and protect the global wildlife heritage and wildlife conservation by identifying and providing treatment to diseased animals, monitoring their health conditions periodically and protecting against poaching. The proposed health care model makes use of various wearable sensors implanted on and within the body of an animal. The sensors perform data gathering about different body-related parameters to be transmitted to a cloud system for analysis and reporting a required timely action. When a cloud system senses emergency with respect to the health-related or threat-related incident, the information is immediately signaled to wildlife physician and wildlife ranger respectively to act accordingly.
Article
Full-text available
Wildlife tracking is used to acquire information on the movement, behaviour and survival of animals in their natural habitat for a wide range of ecological questions. However, tracking and monitoring free-ranging animals in the field is typically labour-intensive and particularly difficult in species that are small, cryptic, or hard to re-capture. In this paper, we describe and evaluate an Internet-of-Things (IoT)-based tracking system which automatically logs detected passive RFID tags and uploads them to the cloud. This system was successfully evaluated with 90 sensor modules deployed in a 30 ha wildlife sanctuary to monitor a small nocturnal mammal of less than 20 g in body size.
Article
Full-text available
The purpose of this data set was to compile body mass information for all mammals on Earth so that we could investigate the patterns of body mass seen across geographic and taxonomic space and evolutionary time. We were interested in the heritability of body size across taxonomic groups (How conserved is body mass within a genus, family, and order?), in the overall pattern of body mass across continents (Do the moments and other descriptive statistics remain the same across geographic space?), and over evolutionary time (How quickly did body mass patterns iterate on the patterns seen today? Were the Pleistocene extinctions size specific on each continent, and did these events coincide with the arrival of man?). These data are also part of a larger project that seeks to integrate body mass patterns across very diverse taxa (NCEAS Working Group on Body Size in Ecology and Paleoecology: linking pattern and process across space, time, and taxonomic scales). We began with the updated version of D. E. Wilson and D. M. Reeder's taxonomic list of all known Recent mammals of the world (N = 4629 species) to which we added status, distribution, and body mass estimates compiled from the primary and secondary literature. Whenever possible, we used an average of male and female body mass, which was in turn averaged over multiple localities to arrive at our species body mass values. The sources are line referenced in the main data set, with the actual references appearing in a table within the metadata. Mammals have individual records for each continent they occur on. Note that our data set is more than an amalgamation of smaller compilations. Although we relied heavily on a data set for Chiroptera by K. E. Jones (N = 905), the CRC handbook of Mammalian Body Mass (N = 688), and a data set compiled for South America by P. Marquet (N = 505), these represent less than half the records in the current database. The remainder are derived from more than 150 other sources. Furthermore, we include a comprehensive late Pleistocene species assemblage for Africa, North and South America, and Australia (an additional 230 species). “Late Pleistocene” is defined as approximately 11 ka for Africa, North and South America, and as 50 ka for Australia, because these times predate anthropogenic impacts on mammalian fauna. Estimates contained within this data set represent a generalized species value, averaged across sexes and geographic space. Consequently, these data are not appropriate for asking population-level questions where the integration of body mass with specific environmental conditions is important. All extant orders of mammals are included, as well as several archaic groups (N = 4859 species). Because some species are found on more than one continent (particularly Chiroptera), there are 5731 entries. We have body masses for the following: Artiodactyla (280 records), Bibymalagasia (2 records), Carnivora (393 records), Cetacea (75 records), Chiroptera (1071 records), Dasyuromorphia (67 records), Dermoptera (3 records), Didelphimorphia (68 records), Diprotodontia (127 records), Hydracoidea (5 records), Insectivora (234 records), Lagomorpha (53 records), Litopterna (2 records), Macroscelidea (14 records), Microbiotheria (1 record), Monotremata (7 records), Notoryctemorphia (1 record), Notoungulata (5 records), Paucituberculata (5 records), Peramelemorphia (24 records), Perissodactyla (47 records), Pholidota (8 records), Primates (276 records), Proboscidea (14 records), Rodentia (1425 records), Scandentia (15 records), Sirenia (6 records), Tubulidentata (1 record), and Xenarthra (75 records).
Article
Full-text available
To study the patterns and determinants of philopatry and breeding dispersal in the Tree Swallow (Tachycineta bicolor) we analyzed the records of 356 males and 1459 females captured in more than one breeding year around Ithaca. New York. Of these captures, only 4% of male and 14% of female breeders dispersed to a new site for breeding. With our combination of intensive study areas in Tompkins County, New York, and the efforts of volunteer banders throughout New York and surrounding states, we could have detected dispersal in excess of 400 km from the initial breeding site. Randomization tests revealed, however, that breeders dispersed much shorter distances than they could have been detected. Detailed analyses of recaptures in Tompkins County showed that over a 22-km range of distances, the chances of dispersal to a new breeding site declined with the distance from the original breeding site. Females that failed to fledge any offspring were much more likely to disperse than females that reproduced successfully, and the probability of dispersal declined gradually with female age. The spatial scale in which swallows gather and process information appears to be much larger than for passerines that defend all-purpose territories.
Book
In the last two decades it has become increasingly clear that the spatial dimension is a critically important aspect of ecological dynamics. Ecologists are currently investing an enormous amount of effort in quantifying movement patterns of organisms. Connecting these data to general issues in metapopulation biology and landscape ecology, as well as to applied questions in conservation and natural resource management, however, has proved to be a non-trivial task. This book presents a systematic exposition of quantitative methods for analyzing and modeling movements of organisms in the field. Quantitative Analysis of Movement is intended for graduate students and researchers interested in spatial ecology, including applications to conservation, pest control, and fisheries. Models are a key ingredient in the analytical approaches developed in the book; however, the primary focus is not on mathematical methods, but on connections between models and data. The methodological approaches discussed in the book will be useful to ecologists working with all taxonomic groups. Case studies have been selected from a wide variety of organisms, including plants (seed dispersal, spatial spread of clonal plants), insects, and vertebrates (primarily, fish, birds, and mammals).
Chapter
Steady technological advancements including subminiaturization of telemetry components, smaller and more accurate global positioning systems (GPSs), and increasingly available geographic information systems (GISs) continue to influence the way researchers collect and analyze wildlife radio-tracking data. These developments have increased the diversity of animals that can be studied and have boosted the number and accuracy of location estimates, within a framework of contemporary conservation issues. This chapter conducts a literature review using the Wildlife Worldwide (WW) database to assess the specific topics and number of studies using radiotelemetry in wildlife research. Databases within the WW search engine include Wildlife Review Abstracts (formerly Wildlife Review), Swiss Wildlife Information Service, Wildlife Database, BIODOC, and the Waterfowl and Wetlands database. The most critical assumption in any radiotelemetry study is that the transmitter has no appreciable effect on the study animal. One can assume that a marked animal behaves, functions, and survives in a manner similar to that of an unmarked animal. Despite this critical assumption, the impact of radio marking of animals has often been subjectively evaluated. Generally, researchers assume that large animals are not affected by radio marking, whereas smaller animals, particularly birds, are impacted more.
Article
Eight snowshoe hares (Lepus americanus) and five cottontail rabbits (Sylvilagus floridanus) were radio-tagged and tracked by an automatic system. Three of the hares, injured upon capture, lived an average of 2.3 days before being killed by predators. The remaining five survived for an average of at least 28.2 days. Of two uninjured rabbits killed, one was preyed upon within an hour after being disturbed; the other evaded predators for 24 days. Red foxes (Vulpes fulva) and owls probably were the main predators. Accounts of each instance of predation are given, including one in which both predator and prey were radio-tagged. The potential of telemetry for predation studies is emphasized, and suggestions are made for setting up such studies.
Conference Paper
Wireless sensor networks have the potential to become the pervasive sensing (and actuating) technology of the future. For many applications, a large number of inexpensive sensors is preferable to a few expensive ones. The large number of sensors in a sensor network and most application scenarios preclude hand placement of the sensors. Determining the physical location of the sensors after they have been deployed is known as the problem of localization. We present a localization technique based on a single mobile beacon aware of its position (e.g. by being equipped with a GPS receiver). Sensor nodes receiving beacon packets infer proximity constraints to the mobile beacon and use them to construct and maintain position estimates. The proposed scheme is radio-frequency based, and thus no extra hardware is necessary. The accuracy (on the order of a few meters in most cases) is sufficient for most applications. An implementation is used to evaluate the performance of the proposed approach.
Article
If an airplane crashes, an emergency locator transmitter (ELT) is activated. The crash site can then be located by taking bearings on the ELT signal. The statistical problem consists of estimating a two-dimensional location parameter, where the data consist of directional bearings observed, with some error, from several known positions. We develop methods for estimating location for several variations of the basic problem: unidirectional and bidirectional observations, biased observations, and finally, techniques that are insensitive to outliers. These methods can easily be implemented on a portable microcomputer, and seem to perform quite well when applied to some actual data.
Article
Established techniques for robust M estimation of a location parameter are adapted for use in directional data. In particular, a periodic version of any of the commonly used ψ functions can be used to define a comparable estimator of angular location. This technique is illustrated with a numerical example and a small simulation study. The proposed estimators appear to perform at efficiency levels similar to those of ordinary M estimators in the linear case.
Article
An automatic radio-tracking system was designed, constructed, and is being used to study the movements of various animals carrying radio transmitters. Two towers spaced 0.5 mile apart support antennas which are continuously rotated at 1⅓ rpm. The signals received from the transmitters are fed, after amplification and conversion, to a 52-channel receiving and recording system located in a building between the towers. Time and bearing data for up to 52 animals can be continuously recorded on 16-mm film. Bearings are taken from the film, as desired, by an operator using a microfilm reader, and locations are determined by triangulation. Ranges of several miles to the radio-tagged animals and accuracies to the nearest degree are obtainable.