The Argos Data Collection and Location System: Preparing for the Future

Deborah J. Shaw

Service Argos, Inc., 1801 McCormick Drive, Suite 10,
Largo, Maryland 20774, USA


The Argos system has provided satellite-based data collection and geo-location to its user community for more than 20 years. The Argos system has capabilities that make it suitable for use by biologists operating in remote or harsh environments. Beginning with the launch of NOAA-K in May 1998, Argos has enhancements scheduled well into the future. Argos will continue to seek out user requirements for advancing satellite telemetry.


Advances in space technology allow new observational strategies, as satellites provide unique opportunities for observing the Earth. The Argos Satellite Data Collection and Location System first met the challenge of tracking remote mobile platforms and collecting sensor data in 1978. Since then, more than 20,000 transmitters have been deployed, with more than 6,500 operational today. Much of the worldwide environmental data collected through the Argos System would not be possible to obtain any other way.

The Argos system is dedicated to environmental observation and protection. Today, Argos is the only fully integrated global system that can locate a transmitter and relay the sensor data from anywhere in the world to a user’s desktop anywhere else in the world (Figure 1).

Argos instrument packages are currently flown on National Oceanic and Atmospheric Administration (NOAA) polar-orbiting environmental satellites (POES). The orbits are circular and sun-synchronous, and thus each provides daily world-wide coverage. NOAA maintains two satellites in operation, plus one or more in standby, resulting in at least 10 passes per day at mid-latitudes and more than 20 in the polar regions.

Argos-generated positioning information and sensor data are updated after every pass of the satellites for users to access or receive automatically via telecommunication networks. These data are also archived at the Argos processing centers and made available on floppy disk or CD-ROM.

Figure 1. Argos system components.

Two global processing centers, in Toulouse, France, and Largo, Maryland, USA, are fully redundant to maximize reliability. More than 45 countries use Argos for a wide variety of applications (Figure 2).


The Argos system was developed to meet certain global environmental data collection requirements.

On-Board Receiver Sensitivity

The high sensitivity of the Argos on-board receiver is a key capability of the system. The uplink data rate (400 bits/sec), cleanliness of the frequency band (401.650 MHz), and remote area operation (without electromagnetic interference) coupled with the high receiver satellite sensitivity provide the unique capability to operate with small, lightweight, and very low power transmitters. The low power consumption of Argos transmitters contributes to platform longevity – often two years or more without human intervention.

Figure 2. Argos platform types.


Intelligent transmitters can be equipped with microcomputers which have sophisticated duty cycles so that data are collected at pre-selected times. Intelligent also means transmitting only when a satellite is overhead. To achieve this, the Argos transmitter is linked to a receiver tuned to the satellite VHF downlink frequency. The transmitter comes on only when it detects that signal, resulting in a net benefit to the power budget.

Small, low-power Argos transmitters continue to open up a range of new applications, especially in animal tracking. Some transmitters now weigh as little as 15 g, including the electronics, batteries and antenna, and consume as little as 20 to 40 mW. This means even birds can be tracked for months over thousands of kilometers.

Argos Doppler and GPS

Argos location capability (based on Doppler effect) is an important characteristic of the system. Argos has in place a global network of reference platforms to enhance the Doppler location capability. Doppler location also contributes to simple, low-power, platform operation, as the calculation is performed at the global processing centers.

Global Positioning System (GPS) positions are also transmitted through the Argos system. As GPS receivers continuously recalculate position fixes, higher temporal resolution is possible. For example, samples taken every 15 minutes can now be associated with a location. The results are integrated with other Argos data; GPS and Argos locations are presented in the same format (a flag indicates whether the location is Argos or GPS based). Transmitters with built-in GPS receiver electronics are available from a number of manufacturers worldwide.

System Capacity

Continuity is especially critical in long-term programs. Due to the highly redundant Argos ground processing systems, coupled with the dependability and redundancy of the NOAA/Argos space segment:

Two fully operational POES plus three standby satellites are in orbit as of August 1998.

All NOAA POES have exceeded their design lifetimes by at least 50%,

No Argos on-board instrument has ever failed before a satellite was deactivated.

Worldwide Data Distributions

The Argos system was originally used to gather oceanographic and meteorological data, such as atmospheric pressure and sea surface temperature. To make it easy for scientists to transmit their data directly onto the Global Telecommunication System (GTS), Argos established a GTS processing subsystem. The system provides flexibility in processing all types of sensor data, which has increased the quantity and quality of data on the GTS. These data are provided to science and forecast centers around the world.

An Automatic Distribution System (ADS) provides a specially developed data distribution system that supplies results automatically, either at fixed times or whenever new data become available. The user specifies the most appropriate distribution method. For example, in the USA, many users employ the Internet to receive their data. There is no need to interrogate Argos on-line, as data can be delivered directly to the user's desktop.

The Argos Global Processing Centers in Largo, Maryland, USA, and Toulouse, France, are operated around-the-clock, seven days a week. Operators and engineers are continuously available for problem solving or to immediately alert users of transmitter alarms or other anomalous situations. There also are five regional processing offices around the world to help users with routine matters such as data processing modifications, or to assist them with data sharing arrangements with colleagues.

Real-Time and Stored Data

Timeliness is often critical in environmental studies. Today, more than 60% of data processed at the Argos global processing centers is done in near real time. When the satellite is in view of a ground receiving station, it will instantly rebroadcast any message for subsequent transfer for processing at one of the Argos global processing centers. In areas served by near real-time coverage, 30% of results are available in less than 15 minutes, and 90% within an hour. Service Argos continues to extend near real-time coverage by adding receiving stations (Figure 3).

Figure 3. Real-time coverage of the Argos System.

International Cooperation

The Argos system provides an important scientific tool to serve the global environmental observation community. Because of its unique history and position in this community, the Argos system serves as a catalyst for the formation of many international scientific agreements, to date, involving 45 countries. Agreements have extended beyond data telemetry to include other aspects of data collection activity such as platform design and construction, platform deployment, and shared post-processing of data obtained via Argos. Argos' orientation to serve the environmental science community results from the system's governmental origins.


The Argos system has evolved along with the requirements of its user community. Plans include additional satellites, increased data throughput capacity, higher receiver sensitivity, and the addition of a downlink messaging function (two-way capability). Implementation of these enhancements will be oriented to the requirements of the Argos User Community and thereby, further add to the unique capabilities of Argos.

Together with NOAA, Argos has been making plans for future programs, and is taking its system users' comments into account. For example, beginning with the 1998 launch of NOAA - K, the Argos onboard processing units were doubled. This had the effect of nearly quadrupling the number of platforms that could be received at the same time. Other changes to the instrument on that spacecraft included:

Increasing the bandwidth from 24 kHz to 80 kHz, to increase data volume and also to allow the frequency band to be apportioned among platforms having different transmission characteristics, e.g., low power, normal, and short period.

Increasing the signal reception sensitivity by 3 dBm to allow the possibility that weak or environmentally constrained platforms will be received by the satellite, or to permit transmissions at a lower power level (one-half the power in some cases).

Beyond the instrument enhancements scheduled for the NOAA satellites, Argos now has an agreement with Japan to fly an expanded instrument on ADEOS II, a polar orbiting satellite. This instrument will have a downlink messaging capability to allow users to modify the characteristics of their platforms. This new instrument will transform Argos into a two-way communications system. Additionally, an understanding has been negotiated to include Argos on EUMETSAT's METOP 1 and METOP 2 satellites, which are scheduled for launch in the early to mid 2000s (Figure 4).


Figure 4. Argos launch manifest

The downlink messaging capability will be added to Argos-2 with the Japanese ADEOS II satellite in 2000. For the European METOP-1 satellite and NOAA-N, plans are underway for Argos-3, which will incorporate further enhancements, such as another increase of the sensitivity. Plans are also being discussed to allow "channeling" of the bandwidth into sub-bands, which could be allocated to different types of applications. For example, different channels of the bandwidth could be used for low-power applications, high data rate applications, or high sensitivity applications. Channelization will enable Argos to serve diverse user applications in a more specialized way, and thereby improve results for the user community.


In conclusion, Argos has been serving the scientific community for more than 20 years, in a spirit of worldwide cooperation. The best example of this is the way the system continues to evolve as a partnership between its users and operators. As in the past, Argos system users are urged to participate in discussions with Argos management to provide input to program requirements. The enhancements described will allow the user community to satisfy increasingly difficult data relay needs. With a proven, reliable and robust data collection system, Argos continues its observation and monitoring mission well into the next century (Table 1).

Table 1. Argos enhancements.


Operational from

Increased onboard sensitivity Now (with NOAA-K)
Extended bandwidth Now (with NOAA-K)
Bandwidth channeling Early 2000
Internet interface/database management system Early 2000
Three times greater system capacity with two satellites Early 2000 (NOAA-M)
Downlink messaging (two-way capability) End 2000 (ADEOS-2)
4.5 kbps high data rate channel 2003 (METOP1)

Feasibility of Tracking Arctic Foxes in Northern Alaska Using the Argos Satellite System:
Preliminary Results

Erich H. Follmann

Institute of Arctic Biology, University of Alaska Fairbanks,
Fairbanks, Alaska 99775, USA

Philip Martin

United States Fish and Wildlife Service, 101 12th Avenue, Box 19,
Fairbanks, Alaska 99701, USA


Ten foxes were tracked with Platform Transmitter Terminals (PTTs) to evaluate their effectiveness on a small mammal. A total of 4,183 locations were obtained. Mean signal strength of the PTTs never exceeded the Argos threshold of -120 dB. There was no significant relationship between duty cycle and mean number of locations or with mean duration of satellite overpasses. Satellite coverage was excellent due to the high latitude of the study area.


Despite the many benefits that have been achieved following the advent of radio telemetry techniques at mid-century, there remain situations where standard VHF telemetry is still insufficient to answer important ecological questions about wild animal populations. One such problem is tracking in the Arctic where animals can range over vast distances including on the pack ice of the Arctic Ocean. VHF telemetry attempted in past years for polar bear (Ursus maritimus) studies proved inadequate because of the enormous area of ice over which bears can range both in summer and in winter, the problems of short days and storms in winter, and the great expense of chartering twin-engine aircraft to conduct surveys. Experimentation with satellite tracking of bears in the 1970s (Kolz et al. 1978, Larsen et al. 1983) has yielded an impressive array of information regarding the ecology and behavior of polar bears throughout the circumpolar Arctic. Currently the Argos satellite system is being used to track bears, continuing to add significantly to our knowledge about these animals.

The success of satellite tracking of animals such as bears and other vertebrates (Harris et al. 1990) was facilitated by the large size of these animals. Because the PTTs were quite heavy and bulky, only animals of bear or large ungulate size could carry these units without seriously affecting their behavior. Despite improvements in miniaturizing circuitry, there were still problems with the batteries being quite large and/or numerous in order to allow for an output of about 1 W, yet allowing for studies of sufficient length to meet project objectives. Further miniaturization of circuitry, reduction of power output below 1 W, utilization of solar panels, and programming duty cycles which reduced the time the transmitter was operating, all aided in further decreasing PTTs to sizes that would allow use on smaller vertebrates (Taillaide 1992).

Despite a satellite radio collar being available in a size easily carried by an arctic fox (Alopex lagopus), which ranges in weight from 2.5-5 kg, there were still two important questions that needed to be addressed before a large-scale project on arctic foxes could be undertaken. As with any radio transmitter there is an effect of the substrate on signal transmission that must be considered. There was less effect the greater the height of the transmitting antenna was above the ground. On a fox, where that height is on the order of 30 cm, a significant effect could occur thereby potentially reducing signal output.

Another potential problem related to the batteries used to power the PTT. Batteries with lithium as the anode are used extensively in wildlife studies, but not all materials used as the cathode function effectively at very cold temperatures. Lithium batteries with cathode material that functions well at low temperatures were not available in a size suitable for use in the fox PTT. Therefore, it was thought that very cold temperatures, reaching –40oC and lower, might affect the battery to the extent that insufficient energy was available to transmit a signal powerful enough to reach the satellite. The posture of the fox at the time of satellite overpass would influence the severity of both the temperature problem and the antenna position problem noted above. If the fox is standing the PTT case will be exposed to ambient air more than if the animal is curled in a resting position (Follmann 1978). In the latter case the transmitter could be quite warm but the antenna would be near the ground where signal attenuation might result.

In an effort to resolve these two questions, a pilot study was initiated in November 1998 to instrument 10 arctic foxes in northern Alaska. The biological objectives of the study were to determine whether the Prudhoe Bay Oilfield contributed to maintaining an artificially high population of foxes during the winter as a result of anthropogenic foods being available in the form of kitchen wastes and garbage.


The PTT selected for this study was the Model ST-10 PTT manufactured by Telonics (Mesa, Arizona, USA). Because of the need to limit weight of the unit, no VHF transmitter was included that would have allowed ground tracking of foxes. A duty cycle of 8 h on/64 h off was selected to achieve a life expectancy of about 6 months. No special sensors were incorporated into the PTT to avoid adding weight and to extend the functional life of the PTT. The total weight of the unit was about 190 g and the power output about 1/2 W.

Foxes were trapped in the Prudhoe Bay Oilfield using large wire box traps with bait. Following capture, foxes were chemically immobilized with a combination of xylazine hydrochloride/ketamine hydrochloride. Radio collars were attached and, following recovery from anesthesia, foxes were released near their site of capture. The protocol for the study was reviewed and approved by the University of Alaska Fairbanks Institutional Animal Care and Use Committee.

Data were transmitted from Landover, Maryland to Fairbanks, Alaska, via electronic mail. Data were processed through a program developed for SAS that evaluated locations obtained during one duty cycle and selected the best location. Analyses for this paper included determining number of locations/duty cycle, mean signal strength/duty cycle, duration of satellite overpasses/duty cycle, and number of locations in each quality category. Duty cycle was used as a surrogate for time since activation of the PTT in three analyses of transmitter function. Because foxes were instrumented over a 3-day period and released at different times, the duty cycles for the 10 foxes differed according to when the magnet was removed and the transmitter placed on each fox.


Upon release of foxes at their site of capture, all reverted to normal behavior and did not exhibit any ill effects from the collar. In fact, foxes released near other foxes began to interact within a very short time including vigorous chases that were initiated by either the instrumented fox or another. There was no suggestion of ill effects from the radio collar. Due to the short (about 5 cm) transmitting antenna extending vertically at the nape of the neck and the long winter fur of the foxes, only about 3 cm of antenna were visible above the hair.

Foxes were trapped, instrumented and released during November 11-13, 1998. All PTTs but one, from which data were no longer received after December 31, continued to operate through 18 April, the date used as a cut-off for preparation of this paper.

A total of 4,183 locations were recorded for the ten foxes (Table 1). The maximum number of locations obtained for a fox was 727 and the minimum 206, this for Fox 10 whose PTT stopped on December 31. The PTT of Fox 8 stopped sending signals between January 6-21 and between January 21- February 26, 1999, but began to function again and continues to do so. The reason for this hiatus remains unknown. Because these PTTs did not function for the entire 5-month period dealt with in this paper, data from them were not included in the analyses for relationship between duty cycle and mean number of locations and mean duration of satellite overpasses.

Table 1. Radio locations obtained from arctic foxes at Prudhoe Bay, Alaska, including Argos location quality information, 1998-99. Location categories are defined in the text.























































































































3+2+1 %




Argos provides location quality in 7 categories of accuracy as follows: 3 (<150m), 2 (150<350m), 1 (350<1000m), 0 (>1000m) and A, B and Z with no location accuracy (ARGOS 1996). The number of locations in each category varied greatly between foxes (Table 1). Those categories with accuracy of <1000 m constituted 45% of the total locations obtained during the study, whereas those with no accuracy information accounted for 31%.

ARGOS evaluates signal reception based on strength with a threshold of -120 dB (ARGOS 1996), presumably with values below that being of marginal quality. The average signal strength for all PTTs was -129.82+/-0.12 SE (range -126.42 to -131.39 dB), and never exceeded the threshold. Despite this result adequate locations of sufficient quality were obtained to meet study objectives.

The number of locations per duty cycle was consistent during the study for the eight foxes included in this analysis, and began to decline after about 5 months (Figure 1). This reflects the usual depletion curve of a lithium battery (Ko 1980) and indicates that the PTTs are approaching the end of their projected functional life.

Figure 1. Mean number of radio locations received for Foxes 1-7 and 9 during each duty cycle, Prudhoe Bay, Alaska, 1998-99.

The mean duration of time (sec) that a satellite was in communication with the PTTs did not vary significantly with duty cycle (Figure 2). This was consistent with the minimal variation over time of signal strength received by the satellite. Due to the high latitude (about 700 North) of the study area, there was excellent coverage by satellites during the 24-h day.


Despite the premature failure of one PTT and the erratic performance of one other about 2-3 months following deployment, our results indicate that the PTTs used in the study are quite suitable for use on fox-sized animals where a high degree of location accuracy is not required. Over the 5-month period included in this analysis about 1,900 locations were suitable for locating foxes with an accuracy of <1 km. Although this degree of accuracy would be inadequate for specific home range analyses including habitat utilization, it is perfectly acceptable for investigations where significant movements could occur over wide areas. This is particularly the case in arctic regions where sea ice constitutes a significant enlargement of substrate over which foxes can travel during winter months.

Figure 2. Mean duration of satellite overpasses for Foxes 1-7 and 9 during each duty cycle, Prudhoe Bay, Alaska, 1998-99.

It appears that both the number of locations/duty cycle and the duration of satellite overpasses/duty cycle diminished during January and early February. This perhaps resulted from low temperature effects although these data have not as yet been compared. Temperature effects could have been via the battery or perhaps on the oscillator. These factors could explain the erratic performance of the PTT on Fox 8, yet this unit has worked well since late February. As noted above, there are also considerations of posture and antenna distance above the ground that could have additive or canceling effects on the effective transmission of the PTT. It may not be possible to resolve that issue.

The numbers of locations obtained during the study, although quite variable among foxes, were adequate and of sufficient quality to allow determination of home ranges. These data are as yet not analyzed but will be completed when all PTTs have stopped. We think that this pilot study demonstrated that the PTT used provided adequate data for a field investigation of arctic foxes in winter. Despite only 45% of radio locations providing an accuracy of 1 km or less, the total number of locations obtained with a duty cycle of 8 h on/64 h off was sufficient to provide very useful information on fox movements.


We wish to thank Shellie Colgrove and Tom Manson of ARCO Alaska, Inc. for their generous support while trapping foxes in the Prudhoe Bay Oilfield. The assistance provided by Del Sandvik of Alaska Clean Seas was invaluable in trapping and temporarily housing foxes at Prudhoe Bay. David Douglas of the United States Geological Service Biological Research Division is gratefully acknowledged for assistance with data management.


ARGOS. 1996. User’s Manual. Service Argos, Inc., Landover, Maryland, USA.

Follmann, E. H. 1978. Behavioral thermoregulation of arctic foxes in winter. Pages 171-174 in Klewe, H.-J. and H. P. Kimmich (eds.), Biotelemetry IV, Proc. of the 4th International Symposium on Biotelemetry, 28 May-2 June 1978, Garmisch-Partenkirchen, Germany.

Harris, R. B., S. G. Fancy, D. C. Douglas, G. W. Garner, S. C. Amstrup, T. R. McCabe and L. F. Pank. 1990. Tracking wildlife by satellite: current systems and performance. U. S Dept. of the Interior, Fish and Wildlife Service Tech. Rep. 30.

Ko, W. H. 1980. Power sources for implant telemetry and stimulation systems. Pages 225-245 in Amlaner, C. J. and D. W. Macdonald (eds.), A Handbook on Biotelemetry and Radio Tracking. Pergamon Press, Oxford.

Kolz, A. L., J. W. Lentfer and H. G. Fallek. 1978. Polar bear tracking via satellite. Int. ISA Biomed. Sci. Instrumentation Symp. 15:137-144.

Larsen, T., C. Jonkel and C. Vibe. 1983. Satellite radio-tracking of polar bears between Svalbard and Greenland. Int. Conf. Bear Res. and Manage. 5:230-237.

Taillade, M. 1992. Animal tracking by satellite. Pages 149-160 in Priede, I. G. and S. M. Swift (eds.), Wildlife Telemetry, Ellis Horwood Ltd., West Sussex, England.

GPS Tracking of Wildlife:
A Convergence of Technologies

Stanley M. Tomkiewicz, Jr.

Telonics, Inc., 932 East Impala Avenue, Mesa, Arizona 85204, USA


Radio telemetry began to emerge as a dominant and critically important tool used in the developing sciences of wildlife management and ecology in the early 1950s. Used first to determine positional information, radio telemetry capabilities expanded to include the relaying of important physiological and behavioral data to research biologists from free-ranging animals. For many years, the basic technology, now often referred to as conventional VHF telemetry, was refined and expanded, and now represents a reliable tool used to study many species.

In the late 1970s, the use of the Argos Data Collection and Location System to track wide-ranging migratory species opened a new door to the understanding of long-range movements of animals. In the 1990s, the incorporation of reduced powered Global Positioning System (GPS) units into subsystems capable of being placed on free-ranging animals further opened the door to different lines of research requiring repeatable high-accuracy positioning. This paper discusses the development and coordinated integration of various wildlife telemetry technologies.


The topic I present today deals with the integration of technologies that has been occurring in the telemetry field. Several examples have been presented during this symposium that demonstrate the integration and convergence of numerous biotelemetry technologies as applied to tracking fish. This paper will deal more extensively with technologies that are used to track birds and mammals. It represents an overview of technologies, beginning with the development of conventional VHF systems in the 1950s and progressing to the development of GPS systems of the 1990s.


Beginning in the 1950s, radio telemetry began to emerge as a dominant and critically important tool used in the developing sciences of wildlife and fishery management and ecology (Patric et al. 1982). Used first to determine positional information, radio telemetry capabilities expanded to include the relaying of important physiological and behavior data to research biologists studying free-ranging animals. For many years, the basic technology (now often referred to as conventional VHF telemetry) was refined and expanded. In the early days it was not uncommon to see the "classical telemetry report" wherein a researcher purchased ten transmitters, two did not work upon arrival, one more failed before the units left his laboratory for the field, and two more units failed in the first week while deployed on animals. Another unit failed after a month, one unit’s signal was lost after only three months in operation, and three other units lasted about six months. Often in the conclusion of the paper there was a statement to the effect that "radio telemetry was a viable tool for the study of unrestrained animals." It appears the conclusion was more wishful thinking than scientific assertion.

Somewhere between those early developmental reports in the 1950s and 1960s and the present, telemetry has developed into a sophisticated and reliable scientific tool for research. This technology has been used to study hundreds of species ranging in size from passerine birds and mice to elephants and whales. The technology has allowed researchers to repeatedly find individual animals and monitor the animal’s location, determine if the animal is alive or dead, establish an activity level, monitor body temperatures, and collect numerous behavioral observations. Conventional VHF technologies were refined through the 1960s, 1970s and 1980s; however, the basic systems remained virtually the same throughout this time. The use of discrete and integrated circuit hardware allowed the development of a few new sensors (i.e., delayed time mortality/activity sensors), but options were limited and the development of hardware solutions was time consuming and expensive (Garshelis et al. 1982). The real changes in system functionality arrived with the development of small microprocessor-controlled units that allowed functions like seasonal duty cycling to extend the operational life of systems, the ability of the unit to send ID codes and to process data from an array of sensors. These advances occurred in the 1990s and revolutionized the technology into a much more versatile and powerful tool for research (Tomkiewicz 1990).

Besides advances in transmitter functionality and sophisticated receiver-scanner technology, and the development of data acquisition, technology has progressed to allow data to be acquired and stored in the field at unattended stations. In some cases, data is linked between remote sites and the laboratory through other advancing communications technologies, i.e., cell phone, satellite telemetry, or other alternatives.

For many years, VHF telemetry for terrestrial and avian species and ultrasonic telemetry for fish were the only tools available to the research community. In the early 1970s the use of the Argos system to track wide-ranging and migratory species opened a new door to understanding the long-range movements of animals (Mate 1987, Harris et al. 1990). Migration timing, routes, and stopovers were determined with greater frequency and with resolution not previously possible. The studies had direct impact on basic biological research, wildlife management in general, the protection of endangered species and resource development issues (Harris et al. 1990). The availability of low power microprocessors to control sophisticated satellite transmitters had a secondary outcome in allowing the development of new sensor technologies (including pressure sensors) that could be used to determine such things as surfacing times, depth of dive, and even dive profiles of marine mammals. Since the microprocessor was already on board, specialized sensors were more easily developed and incorporated into transmitting subsystems (National Marine Fisheries Service 1992, Bryzek and NovaSensor 1993).

In addition to this new "smart" control capability, microcontrollers allowed the development of on-board processing and data reduction techniques that reduced the volume of data transferred. In fact, the incorporation of low power, low current microprocessors in this technology is what actually led to the development and crossover of these microprocessor-controlled systems into conventional VHF telemetry a few years later.

In the 1990s the incorporation of low powered GPS units into subsystems capable of being placed on free ranging animals further opened the door to the development of numerous novel and divergent lines of research requiring repeatable high accuracy positioning. The 1990s have seen the proliferation of GPS systems that can store positional data on board the unit for later download after recovery of the unit. Data links were also developed in order to allow GPS and additional supplementary data to be recovered directly across a radio link or indirectly via satellite systems such as Argos.

There now are three basic GPS systems: The first is the Store-on-Board (SOB) system. In this case the data are GPS positions specifically obtained and stored in nonvolatile memory for later download. In this system the unit must be physically recovered to obtain the data. This system does not have a radio frequency (RF) link. These SOB systems are usually physically smaller and weigh less than systems that require a data transmitter to implement an RF link. A second, related system that has been developed is often called the GPS-FM system. In this system, GPS positions are obtained and stored in memory. This system can then transfer information across an FM data link on command. A third system is the GPS/Argos system. In this system, GPS positions are obtained and stored in memory and the data system transfers the data through the Argos-NOAA satellite system. In this case information can be obtained without the researcher being in the field, and be linked back through a satellite system from very remote locations (Tomkiewicz et al. 1996).

Much as been learned about collecting and processing GPS data from remote sites and from animal- borne packages over the past five years. As with most engineering efforts, actual field trials and deployments provide more information than any other source. For example, our techniques for dealing with power management have been substantially refined. The new second-generation GPS units consume less power and have a longer operational life as compared with early units. Further, new generation units also incorporate smaller, lower power GPS receiver technologies capable of more rapid time to first fix. This advancement further extends the operational field life of contemporary units.

Today’s GPS tracking units incorporate either a user interface serial port or RF link as a means to provide the researcher the ability to reprogram user changeable parameters. This capability allows the alteration of duty cycles and GPS fixed time schedules as well as other parameters.

In addition, the memory available on board the newer GPS units has been greatly expanded to allow this storage of additional position "fixes." In many units there is sufficient memory to handle up to 5,500 differentially correctable positions in memory for later download and post-processing. New GPS receiving antennas are smaller, consume less power and can be packaged in smaller housings than earlier GPS antennas.

In conclusion, the most significant trend in the biotelemetry field is the increasing reliance upon microprocessor technology in all of our contemporary telemetry subsystems. The reduced power consumption and low voltage operation of these new microprocessors has lead to the development of "smart" and versatile VHF transmitting subsystems. In addition, it has led to even "smarter" receiver technology. Animal-borne technologies are being used in conjunction with one another to create highly customized equipment capable of aiding the research community in answering new and important research questions, such as the combination of ultrasonic and radio transmitter technologies used in an integrated configuration in tracking fish. In addition, there are systems that contain Argos satellite telemetry and VHF systems, GPS systems with VHF backup beacons, and GPS systems with Argos satellite telemetry capability, GPS systems with Argos satellite telemetry capability with VHF being used as a backup beacon function.

These technologies are being integrated in novel ways that allows numerous combinations of modules to assure a system that can meet the requirements of research. It is the convergence of all these technologies that continues to make the biotelemetry field an exciting area for research and development. As new technologies feed into this field, equipment gets smaller, more intelligent, and more versatile. This trend should extend well into the next millennium.


Bryzek, J. and L NovaSensor. 1993. The Evolution Of Smart Sensor And Transducer Design: Examination of signal conditioning requirements for piezoresistive pressure sensors; interfacing between sensor and digital sensor signal processing electronics; and evolving design configurations of smart transducers. Sensor: October volume.

Garshelis, D. L., Quigley, H. B., C.R. Villarrubia, and Michael R. Pelton. 1982. Assessment of telemetric motion sensors for studies of activity. Canadian Journal of Zoology 60(8):1800-1805.

Harris, R. B., S.G. Fancy, D.C. Douglas, G.W. Garner, S.C. Amstrup, T.R. McCabe, and L.F. Pank. 1990. Tracking Wildlife by Satellite: Current Systems and Performance. U. S. Department of Interior, Fish and Wildlife Service Technical Report 30. National Information and Technical Service. Washington D.C. 52 pp.

Mate, B.R. 1987. Development of Satellite-Linked Methods of Large Cetacean Tagging and Tracking In OCS Lease Areas – Final Report. Oregon State University, Hatfield Marine Science Center, Newport, OR.

National Marine Fisheries Service. 1992. Workshop on Tagging and Tracking Technology at the Airlie House, Warrenton, Virginia: The difficulties associated with collecting reliable information on marine mammals – their movements, habitat use, activity patterns, ecology, and behavioral and physiological responses to environmental and anthropogenic activities. Northeast Fisheries Science Center Reference Document 93-08. Woods Hole, Massachusetts, USA. 79 pp.

Patric, E.F., G.A. Shaughnessy, and G.B. Will. 1982. A Bibliography of Wildlife Telemetry and Radio Tracking: Department of Forest and Wildlife Management, College of Resource Development, University of Rhode Island.

Tomkiewicz, Jr., S. 1990. Technical Note: Advances In Satellite And Radio Location Technology. Pages 229-230 in Proceeding of the Eleventh Annual Information Transfer Meeting of Mineral Management Service. Gulf of Mexico OCS Region, New Orleans, Louisiana, USA.

Tomkiewicz, S., R. Degler, T. Rios, G. Jones, G. Swinford, P. Beaumont, M. Beaumont, T. Hansen, and R. Wagner. 1996. GPS Applications For Wildlife–A Review. Telonics Quarterly V. 9(1):1-12.

Evaluation of a GPS Used in Conjunction with Aerial Telemetry

Edward M. Olexa and Peter J. P. Gogan

United States Geological Survey - Biological Resources Division, Greater Yellowstone Research Group, Montana State University,
Bozeman, Montana, USA

Kevin M. Podruzny

Department of Biology, Montana State University, Bozeman, Montana, USA


We investigated the use of a non-correctable Global Positioning System (NGPS) in association with aerial telemetry to determine animal locations. Average error was determined for 3 components of the location process: use of a NGPS receiver on the ground, use of a NGPS receiver in a aircraft while flying over a visual marker, and use of the same receiver while flying over a location determined by standard aerial telemetry. Average errors were 45.3, 88.1 and 137.4 m, respectively. A directional bias of <35 m was present for the telemetry component only. Tests indicated that use of NGPS to determine aircraft, and thereby animal, location is an efficient alternative to interpolation from topographic maps. This method was more accurate than previously reported Long-Range Navigation system, version C (LORAN-C) and Argos satellite telemetry. It has utility in areas where animal-borne GPS receivers are not practical due to a combination of topography, canopy coverage, weight or cost of animal-borne GPS units. Use of NGPS technology in conjunction with aerial telemetry will provide the location accuracy required for identification of gross movement patterns and coarse-grained habitat use.


Aerial radio tracking is typically used in areas where direct observation or triangulation is not possible due to rugged terrain, large study area size, or extensive movement of study animals. Location estimates from aerial radio tracking are usually determined by plotting positions on maps and interpolating coordinates subsequent to the flight. Maximum accuracy averages £ 100 m (Krausman et al. 1984, Miller 1986), but can be ³ 1,400 m (Garrott et al. 1987, Walsh et al. 1992). This level of accuracy may meet study requirements but direct plotting can be very labor intensive, particularly when large samples of instrumented animals are involved.

LORAN-C, a navigation system that utilizes land-based radio transmitter stations, has been evaluated as an alternative means of acquiring location data for aerial survey and radio telemetry work (Boer et al. 1989, Leptich et al. 1994, Viljoen and Retief 1994). Average accuracy of £ 385 m has been reported (Patric et al. 1988, Boer et al. 1989, Rhoades et al. 1990, Leptich et al. 1994, Carrel et al. 1997) with linear errors 5.1-6.9 times greater than reported for direct plotting (Jaeger et al. 1993). Accuracy is dependent on the position of the receiver relative to the transmitters, the position solution, and user experience (Jaeger et al. 1993). Accuracy limitations minimize its utility in association with radio telemetry and precluded further consideration.

Satellite based systems, such as Argos, broadcast a UHF radio signal to satellites (Harris et al. 1990) and use Doppler shift of a transmitter’s carrier to determine location (Keating 1995). Ninety-fifth percentile accuracies of 1-3 km are reported (Harris et al. 1990, Keating et al. 1991, Keating 1995). This accuracy limitation renders the Argos system an unacceptable alternative to aerial telemetry.

Global Positioning Systems (GPS) have made coordinate determination via satellites orbiting the earth common. GPS technology has been miniaturized to the point at which receivers are now used either alone or in conjunction with VHF telemetry to determine animal locations (Rodgers and Anson 1994, Rempel et al. 1995, Moen et al. 1996). Optimal non-differentially corrected accuracy in civilian GPS receivers is routinely estimated at 40 m 50% of the time, and 100 m at 95% circular error probable, but greater accuracy has been reported (Wells et al. 1986, Jones 1989, Gibbons 1992). Mean errors of £ 26 m were reported during tests of a hand-held GPS receiver (Ardö and Pilesjö 1992). Leptich et al. (1994) recorded mean position errors of 50 m while testing the accuracy of a non-differentially corrected GPS receiver used on a small fixed-wing aircraft. Signal degradation by the U.S. Department of Defense causes most of the introduced error. Differential correction can remove this error and increase accuracy to within 1-10 m, but this typically requires access to a base station operating at a previously surveyed location £ 300 km from the mobile unit (Rempel et al. 1995). Many basic GPS receivers lack this ability. Despite recent advances, GPS receivers still can be carried on only the largest animals, are more expensive than VHF transmitters, and require line-of-sight contact between satellites and the receiver, which may reduce the frequency and/or accuracy of locations in areas with well-developed forest canopies and/or steep topography.

We evaluated the accuracy of an airborne NGPS used in conjunction with standard aerial VHF radio telemetry under a simulated scenario but real field conditions. We simulated situations encountered when locating radio-collared adult mule deer (Odocoileus hemionus) does on northern Yellowstone herd summer ranges. We simulated the two most likely outcomes of attempting to locate radio-collared deer, determining a location upon sighting a deer once signal strength indicated the animal's location or determining a location based solely on signal strength. Our objectives were to quantify the accuracy of an aircraft mounted GPS unit in acquiring position coordinates for instrumented animals and to identify the source and magnitude of introduced error from each component of the location process.


Our study area was a portion of a summer range utilized by radio-collared mule deer and was approximately 97 km wide along its northern boundary, 36 km wide along the southern boundary, and 47 and 57 km wide along the western and eastern edges respectively. The area included portions of Yellowstone National Park (YNP) and the adjacent Gallatin National Forest (GNF).

Elevations in the study area ranged from 1,575 m along the Yellowstone River to 3,360 m at Electric Peak, Montana. Test site elevations ranged from 1,577 to 2,556 m. In the GNF portion of the study area, broad river corridors separated abrupt mountain ranges dissected by steep drainages with narrow riparian zones. High rhylotic plateaus with broad river valleys at lower elevations characterize the YNP portion of the study area. Abrupt mountain ranges were present along the northern and western boundaries where YNP borders GNF.

Coniferous forests covered most of the study area with lodgepole pine (Pinus contorta) being the predominant tree species. Douglas fir (Pseudotsuga menziesii), juniper (Juniperus spp.), and aspen (Populus tremuloides) occurred at lower elevations. Whitebark pine (Pinus albicaulis), Engelmann spruce (Picea engelmannii), and subalpine fir (Abies lasiocarpa) commonly occurred above 2,700 m. Houston (1982) and Despain (1990) described the topography and vegetation present in YNP, while Davis and Shovic (1996) provided similar information for the GNF.

Site Selection

Twenty test sites representative of the range of topography and over- and understory vegetation utilized by instrumented mule deer does were selected in areas accessible on foot. Sites were evenly distributed between forested and open areas. These areas were marked. However, neither the site markers nor the flagging used to aid in locating sites during ground visits was visible from the air.

Data Collection

Precise coordinates were obtained by positioning the external antenna of a differentially correctable Trimble Pathfinder Professional gps unit at each site and recording 250 coordinate pairs. These data were differentially corrected using the program Pfinder and data collected at a base station 130-184 km away at Jackson, Wyoming. The differentially corrected position estimates had an estimated accuracy of 2-5 m circular error probable, but were treated as the true location of each site.

Three types of locations were collected to identify the error associated with each component of the location process. Non-differentially corrected ground locations were collected to identify the location error associated with our specific GPS receiver. Non-differentially corrected locations were collected from the air while flying over visual makers placed on the ground to quantify the location error associated with using our GPS receiver in an airplane. The GPS receiver was also used in the airplane to record coordinate data for locations of non-visible targets to quantify the location error associated with using our GPS in conjunction with standard aerial radio telemetry.

Using a TransPak II receiver, we recorded the date, time, and Universal Transverse Mercator (UTM) coordinates of each location estimate. Satellite and pseudo-range information were not collected. Both the ground and the airborne locations were estimated using identical settings on the GPS receiver to ensure consistency. All coordinates were based on the North American 1927 datum. The receiver was set to record 3-dimensional fixes only during all sampling. Thus, it required contact with a minimum of 4 satellites to calculate location estimates.

To determine the error associated with a NGPS, site locations were estimated from the ground by positioning the external antenna of the Trimble TransPak II receiver at each site. Five locations were recorded at ³ 60-sec intervals. Five location estimates also were recorded at ³ 10-min intervals to evaluate the affect of varying satellite configurations over time. This was repeated at each site on each of three days.

Radio transmitters were placed at both open and forested sites, and open sites were marked to be visibly detectable from the air. Markers consisted of two pieces of laminated white paper each measuring 0.28 by 1.52 m (11" by 5') arranged to form an X and were secured in place by either weighing down with small rocks or nailing to the ground. The sites with visual reference points simulated situations where animals were detectable from the air. Sites with radio collars simulated situations where animals were not visible from the air. At sites with both radio collars and visual markers, the radio-tracking tests were completed prior to placing the ground markers.

Location flights were conducted in a PA-18 Super Cub, typically flying £ 200 m above ground level (AGL) at an air speed of ³ 65 km/hr. Angle of approach and altitude varied as terrain required. Test locations were unknown to the pilot, but were known to the observer. The observer directed the pilot to the general area of a radio collar or visual marker. The pilot then determined test locations via standard aerial telemetry techniques (Mech 1983). The observer obtained the position of the aircraft when the pilot verbally indicated it was directly over the radio collar as determined by receiver signal overload or center of the marker as determined visually. The aircraft's GPS location was obtained from a Trimble TransPak II receiver that was connected to an antenna mounted to the top of the windscreen between the pilot and the observer. Five aerial fixes were obtained sequentially as rapidly as possible to minimize environmental and satellite variability at each of the radio-collared or visually marked sites. The procedure was repeated on each of three separate days.

Data Analysis

Error was calculated as the Cartesian distance (m) between the estimated and true site coordinates. Average error was calculated for each set of 5 location estimates taken at a site on a given day. A 5-location series average was treated as a single error estimate associated with a trial. Thus, for all statistical tests, the 5-location series average was the sampling unit. For each set of measurements, the skew and kurtosis of the distributions were evaluated to determine the presence or absence of normality (Wilkinson et al. 1996). Error estimates were log transformed to approximate a normal distribution. Mean error and 95% confidence interval (CI) values were calculated for the log-transformed data. These were then back-transformed to produce the reported mean error and 95% CI. Single factor ANOVA with Bonferroni comparisons (Zar 1984) was used to test for differences among the mean of the 3 trials.

Errors in non-differentially corrected ground locations collected at 1- and 10-min intervals were compared after log transformation using a paired t-test (Zar 1984). Reported mean errors and 95% CI were calculated using back transformation.

We tested for differences in mean error between log transformed data sets created from the first location in a 5-location series and the series average using Wilcoxon’s signed rank test (Zar 1984, Leptich et al. 1994). Reported mean errors and 95% CI were calculated using back transformation. Rayleigh’s z test with Dunn-Sidak corrections for multiple comparisons was used to determine if locations were uniformly distributed around the true location for each trial (Zar 1984). SYSTAT (Wilkinson et al. 1996) and Lotus 1-2-3 were used for all statistical analyses.


Visual inspection of all estimated locations relative to true locations indicated an increase in error from GPS ground trials to aerial trials on visible targets to aerial telemetry trials on non-visible targets (Figure 1). Errors in locations for all 20 sites collected with GPS on the ground at 1-min intervals (n = 60) ranged from 15.0 m to 181.2 m and averaged 45.3 m (95% CI: 39.4 - 52.1 m). Errors in locations determined for 10 visual reference points from the air (n = 30) ranged from 54.4 m to 203.3 m and averaged 88.1 m (95% CI: 80.2 - 97.9 m) while locations at 20 sites with radio collars determined by aerial telemetry (n = 60) ranged from 68.7 m to 280.8 m and had an average error of 137.4 m (95% CI: 127.0 - 148.8 m). The latter 3 means were all significantly different (F2, 147 = 110.7, P < 0.0001).

Figure 1. Distributions of all GPS position estimates relative to true locations for each of three accuracy tests conducted in the Yellowstone northern range. Trials included GPS receiver on the ground, use of a GPS receiver in a aircraft while flying over a visual marker, and use of the same receiver while flying over a location determined by standard aerial telemetry (n = 300, 150, and 300 respectively). Gaussian bivariate 95% ellipses for the samples are shown in each trial plot.

Although three of 20 test sites examined during the ground trial had biased position estimates (P < 0.02), bias was not evident when all sites were pooled (Rayleigh’s z = 0.92, P > 0.20). Thus, pooled samples were used for each trial to test for bias. Estimates obtained during the visual trial showed no bias (Rayleigh’s z = 1.45, P > 0.05). However, bias was detected during the telemetry trial (Rayleigh’s z = 3.99, P < 0.02). Based on the pooled sample, position estimates were biased by 34.6 m in a direction of 325.6o.

Errors in locations for all 20 sites collected with GPS on the ground at 10-min intervals (n = 60) ranged from 14.9 m to 103.7 m and averaged 44.6 m (95% CI: 40.4 - 49.1 m). These were not significantly different (t = 0.235, P = 0.815) from the above ground fixes collected at 1-min intervals.

Error estimates based on the first telemetry location of each 5-location series (n = 60) ranged from 14.4 m to 493.5 m and averaged 121.7 m (95% CI: 103.8 – 142.3 m). The average of the 5-location series (n = 60) ranged from 68.7 m to 280.8 m and had a mean of 137.4 m (95% CI: 127.0 – 148.8 m). There was no significant difference (Z = 1.259, P = 0.208) between the 2 groups.


Our research supports the utility of GPS used in conjunction with aerial telemetry. Mean location error of 137.4 m for fixes without visual confirmation compares favorably with the mean of 122 m reported by Carrel et al. (1977) and the 100 m maximum accuracy reported with direct mapping. Our mean error was half that calculated by Garrott (1987) while using direct mapping and less than the 200 m calculated by Leptich et al. (1994) using LORAN-C.

Error estimates generated by the telemetry trial may differ from errors encountered during routine aerial telemetry. Normally, aerial telemetry will result in one estimate of location while our trial used the average of 5 fixes. Failure of accuracy to improve with additional flight passes mirrors that of Leptich et al. (1994) and is likely due to strong correlation of position error over brief periods of time and not due to a systematic bias. This temporal correlation is also a likely explanation for the lack of any difference when average error was compared for data sets collected at 1- and 10-min intervals.

An explanation for the presence of a directional bias in the telemetry trial eludes us given the lack of bias for both the ground and visual trials. Small sample size and/or lack of independence among fixes collected in a short time may have influenced bias testing. Given that weather and topography dictated flight speed, altitude, and orientation, systematic bias appears unlikely.

The importance of testing GPS receivers under the conditions in which they are utilized cannot be overemphasized. Our increase in error from the ground trial to the aerial telemetry trial provides important insight into the accuracy that one can reasonably expect given varying operating constraints.

The use of GPS in conjunction with aerial telemetry provides a level of accuracy similar to that produced by direct plotting, but does not require time-consuming interpolation from maps and is not affected by the amount of topographic relief in an area or an observer’s map reading prowess. The combination of GPS and aerial telemetry provides accurate location data in areas where ground deployed GPS collars may be ineffective due to dense canopy cover and/or steep topography, or where GPS collars are impractical due to study animal size. Use of airborne GPS in conjunction with standard VHF transmitters provides the accuracy required to identify gross movement patterns and coarse-grain habitat use without utilizing GPS collars.


The skills and interest of pilot W. S. Chapman made this study possible. Early versions of this manuscript benefited from review by K. A. Keating and D. Ouren.


Ardö J. and P. Pilesjö. 1992. On the accuracy of the global positioning system – a test using a hand-held receiver. Int. J. Remote Sensing 13:3229-3233.

Boer, A., H. G. Redmond, and T. J. Pettigrew. 1989. Loran-C: A navigation aid for aerial surveys. J. Wildl. Manage. 53:228-230.

Carrel, W. K., R. A. Ockenfels, J. A. Wennerlund, and J. C. Devos, Jr. 1997. Topographic Mapping, LORAN-C, and GPS Accuracy for Aerial Telemetry Locations. J. Wildl. Manage. 61:1406-1412.

Davis, C. E. and H. F. Shovic. 1996. Soil Survey of Gallatin National Forest, Montana. U.S. For. Serv. And Nat. Resour. Conserv. Serv. U.S. Gov. Printing Off. 209pp.

Despain, D. 1990. Yellowstone vegetation, consequences of environment and history in a natural setting. Roberts Rinehart Publishers, Boulder, CO. 239pp.

Garrott, R. A., G. C. White, R. M. Bartmann, L. H. Carpenter, and A. W. Alldredge. 1987. Movements of female mule deer in northwest Colorado. J. Wildl. Manage. 51:634-643.

Gibbons, G. 1992. The Global Positioning System as a complementary tool for remote sensing and other applications. Photogrammetric Engineering and Remote Sensing 58:1255-1257.

Harris, R. B., S. G. Fancy, D. C. Douglas, G. W. Garner, T. R. McCabe, and L. F. Pank. 1990. Tracking wildlife by satellite: current systems and performance. U.S. Fish and Wildl. Serv. Tech. Rep. 30. 52pp.

Houston, D. B. 1982. The northern Yellowstone elk: Ecology and management. Macmillan Publ. Co., Inc. New York, N.Y. 474pp.

Jaeger, J. R., J. D. Wehausen, V. C. Bleich, and C. L. Douglas. 1993. Limits in the Resolution of Loran- C for Aerial Telemetry Studies. Desert Bighorn Counc. Transactions 37:20-23.

Jones, T. 1989. NAVSTAR global positioning system- status and update. Pages 28-52 in M. Kumar, tech. chairman. Proc. fifth international geodetic symposium on satellite positioning. Las Cruces, N.M.

Keating, K.A. 1995. Mitigating elevation-induced errors in satellite telemetry locations. J. Wildl. Manage. 59:801-808.

Keating, K.A., W. G. Brewster, and C. H. Key. 1991 Satellite telemetry: performance of animal-tracking systems. J. Wildl. Manage. 55:160-171.

Krausman, P. R., J. J. Hervert, and L. L. Ordway. 1984. Radio tracking desert mule deer and bighorn sheep with light aircraft. Pages 115-118 in P. R. Krausman and N. S. Smith, eds. Deer in the southwest: a workshop. Arizona Coop. Wildl. Res. Unit and School of Renewable Nat. Resour.

Leptich, D. J., D. G. Beck, and D. E. Beaver. 1994. Aircraft-based Loran-C and GPS accuracy for wildlife research on inland study sites. Wildl. Soc. Bull.22:561-565.

Mech, L. D. 1983. Handbook of Animal Radio-Tracking. Univ. Minnesota Press, Minneapolis. 107pp.

Miller, G. D. 1986. Feeding ecology of desert bighorn sheep (Ovis canadensis mexicana) in western Arizona. Ph.D. Thesis, Univ. of New Mexico, Albuquerque. 166pp.

Moen, R., J. Pastor, Y. Cohen, and C. C. Schwartz. 1996. Effects of moose movement and habitat use on GPS collar performance. J. Wildl. Manage. 60:659-668.

Patric, E. F., T. P. Husband, C. G. Mckiel, and W. M. Sullivan. 1988. Potential of Loran-C for wildlife research along coastal landscapes. J. Wildl. Manage. 52:162-164.

Rempel, R. S., A. R. Rodgers, and K. F. Abraham. 1995. Performance of a GPS animal location system under boreal forest canopy. J. Wildl. Manage. 59:543-551.

Rhoades, J. D., S. M. Lesch, P. J. Shouse, and W. J. Alves. 1990. Locating sampling sites for salinity mapping. Soil Sci. Soc. Am. J. 54:1799-1803.

Rodgers, A. R. and P. Anson. 1994. Animal-borne GPS: tracking the habitat. GPS World 5:20-32.

U.S. Fish and Wildlife Service. 1994. The Reintroduction of Gray Wolves to Yellowstone National Park: Final Environmental Impact Statement. U.S. Government Printing Office. 601pp.

Viljoen, P. C. and P. F. Retief. 1994. The use of the Global Positioning System for real-time data collecting during ecological aerial surveys in the Kruger National Park, South Africa. Kiedie 37:149-157.

Walsh, N. E., S. G. Fancy, T. R. McCabe, and L. F. Pank. 1992. Habitat use by the Porcupine Caribou Herd during predicted insect harassment. J. Wildl. Manage. 56:465-473.

Wells, D. E., editor. 1986. Guide to GPS positioning. Can. GPS Assoc., Fredericton, N.B. 567pp.

Wilkinson, L., M. Hill, P. Howe, G. Birkenbeuel, J. Beck, and J. Liu. 1996. SYSTAT 6.0 for Windows: Statistics. SPSS, Inc., Chicago, IL. 751pp.

Zar, J. H. 1984. Biostatistical analysis. Prentice-Hall, Inc., Englewood Cliffs, N.J. 718pp.

Telemetry Data Conjoin with GIS
to Manage Big Game

Kreig M. Rasmussen

United States Department of Agriculture, United States Forest Service,
Fishlake National Forest, Utah, USA

Steve Flinders

Utah Division of Wildlife Resources, Southeastern Region


Telemetry data combined with Geographic Information Systems computers help natural resource managers map predator-prey relationships. Harvest objectives and seasonal habitat distribution have been mapped to aid management decisions on the Fishlake National Forest. With telemetry data stored in the computer system, managers are able to access the database when a question arises, enabling better and sounder management decisions to be made to manage wildlife populations and place habitat treatments on the ground.


Conflicts between big game and livestock have precipitated the need to gather more information about big game populations. Arguments about the numbers of elk (Cervus elaphus) arose between livestock permittees and game managers in Central Utah, and, more specifically, on the Fishlake National Forest. These arguments started to damage relationships between individuals, agencies, and working groups.

The ability to accurately model big game populations is an ongoing struggle for natural resource managers. With the help from radio transmitters placed on the species in question, managers can more effectively predict population numbers, migration patterns, seasonal use, and mortality. From 1989-1994, 303 elk, 36 elk calves (1-4 days old), 64 moose (Alces alces), 18 cougars (Felis concolor), and 7 deer (Odocoleius spp.) were outfitted with radiotelemetry transmitter collars on the east half of the Fishlake National Forest. Each species of animal is being studied with different objectives. There is also a desire to understand how the interaction of these species relates to resource management.

Our discussion will focus mainly on the management decisions that derived from elk telemetry data collected during the study period. The five main objectives for the study were to determine elk: 1) seasonal distribution patterns, movements, and population size; 2) harvest levels and illegal kills; 3) age, sex composition, and productivity; 4) use areas, calving areas, and density by livestock allotments; and 5) fidelity to seasonal range (Beale 1994). We will discuss parts of objectives 1, 2, 4, and 5 in this paper.


Mature and calf elk were captured for radio tagging using a variety of methods. Round corral traps and aerial net-gunning methods were used for mature elk, and search-and-capture methods were used for elk calves. Traditional wintering areas for elk determined where corral trap sites were located. Aerial net-gun sites were determined from summer elk populations reflecting no marked animals. Mature cow elk were predominately the target animal of the main study; however, a few mature bulls and 7 month-old bull calves also were collared with transmitters. A total of 23 corral trap sites were established and 2 net-gun sites were used. During 3 years of the study, newborn calves (within 2 to 3 days of birth) were captured in June and radio tagged with breakaway collars.

Ground and aerial telemetry data were gathered. For ground locations, azimuth readings were taken from fixed towers and hand-held Yagi antennas. An average of two aerial survey flights per month were made, with telemetry data recorded and downloaded into the Geographic Information System (GIS) computer for analysis. Individual animal data entered consisted of: 1) date trapped, 2) location trapped, 3) sex, 4) age, 5) radio frequency, 6) ear tag number, and 7) individual animal number. After each flight was made, the coordinates and date were entered according to the individual animal number. The GIS computer program ArcInfo was used to enter the data while ArcView was used to query the data and print maps. Base maps (Figure 1) are generated with feature layers such as land ownership boundaries, roads, lakes, streams, etc. Telemetry data for elk as well as moose, cougar and deer are filed in their respective layers and are combined with the base layers to display collared animal patterns.


The Fishlake Elk Unit and the Monroe Elk Unit telemetry studies were a result of ongoing complaints from land owners and permittees concerning high elk numbers and forage competition. Once the elk were trapped and the collars in place, the telemetry data were recorded and maps printed. The maps were studied by a steering committee formed to oversee the project. The committee was made up of a combination of wildlife managers, livestock men, and sportsmen.

Figure 1. GIS map showing forest boundary, primary roads, and private land, combined with all elk telemetry points on Monroe Mountain.

Population objectives were a main focus of the elk study. Overall numbers within the respective herd units, mortality, recruitment, and hunter harvest success were integrated into the study design. Maps were generated to show concentrations for different seasons. An "all points" map (Figure 2) demonstrated elk concentration areas and migration routes. Telemetry data maps gave managers a good idea where elk were concentrating and sharing summer forage with livestock. Fall telemetry data allowed managers to target specific elk with hunts designed to reduce the population levels. Winter telemetry data revealed where habitat work could be done to improve winter range.

Figure 2. GIS map showing elk fidelity to winter range on Monroe Mountain.

Elk Calf Study

June telemetry data specific to cow elk helped managers understand the components required to sustain critical elk calving areas. A special study was conducted to determine elk calf-bed site selection in the aspen (Populus spp.) ecosystem. It was evident that, in both study areas, the aspen forest type on private land was an important factor in the reproductive success of the elk herd. Elk calves were caught and collared with minimal data taken at capture time. Bed location was recorded with a hand held Global Position System (GPS) receiver and within a 3-day time frame re-visited to gather detailed data about the bed site and surrounding forest components. Radio-tagged calves were then recaptured a few days later to determine growth factors and relocation selection.

Data summary. Flinders (1995), in his study on neo-natal elk, concluded the following points that have helped land managers understand the intricate components that comprise critical elk calving habitat. Of the 109 bed sites, 75% were in an aspen/snowberry (Symphoricarpos oreophilus) community, 9% were in mixed conifer-aspen (at least 5% confer) with either white fir (Abies concolor) or Douglas fir (Pseudotsuga menziesii), 6% in gambel oak (Quercus gambelii)-mixed brush, 3% aspen-chokecherry (Prunus virginianus), 2% in sagebrush (Artimesia tridentata and A. cana)-forb, 2% in snowberry/forb meadow, and 1% in a booth willow (Salix boothei) riparian community. It was also noted that calf elk loss to predators was only 6%. Thus it appears that under current land and predator management practices there are no significant threats to recruitment during the first few months of life. Forest canopy cover occurred at 97% of the bed sites and averaged 34+21% (range 2-83%). Deadfall (fallen trees or branches) was present in 96% of the bed sites, snags in 41 %, and boulders in 23 %. Thus, there was a preponderance of evidence suggesting that deadfall, specifically aspen deadfall, was a strong indicator of neo-natal bed site habitat. The average distance from bed sites to graded roads was 3,137 m + 1,325 m (range 1,037-7,051 m), while the average distance to other roads was 1,271 m + 689 m (range 180-3,704 m). The average distance from bed sites to Forest Service authorized All Terrain Vehicle trails was 1,534 m + 1,000 m (range 1-4,975 m).

Management implications. The data presented reflect habitat component and access management implications as related to critical elk calving areas for land managers on the Fishlake National Forest. With much of the aspen ecosystem in poor to average health on the Fishlake Forest, large-scale projects are being proposed to rejuvenate aspen stands. Data from the elk studies conducted on the forest allow district biologists to work with foresters to design treatments and timing to assure the existence of, and the future of, critical elk calving habitat.

Season Distribution, Productivity, and Mortality Study

Population estimation. In the summer of 1992, the number of elk on a large proportion of the unit was estimated using the Lincoln-Peterson method with radio-tagged elk for the index. Values such as immigration, emigration, and death of marked animals in the population were accurately known from previous radio monitoring (Beale 1995). Population estimates were also made mathematically using yearling bull/cow ratios and harvest data. Helicopter counts over the entire unit were made starting winter of 1992 and have been conducted on an every other year basis (funding permitting).

Concern about high numbers of elk (ranging from 6,000 to 10,000 on the Fishlake Unit and 1,000-2,000 on the Monroe Mountain Unit) was addressed by the population data gathered from the study on both herds. Using the objectives described, the Fishlake herd is estimated at 5,000 elk and the Monroe Mountain herd at 800 elk

Harvest levels and seasonal range fidelity. The Fishlake Elk Herd Unit hunt strategy is to harvest yearling bulls with limited permits offered on a random draw. Access to elk during the hunting season is very good on the Fishlake Forest. One concern of game managers was to understand how many yearling bulls would survive hunting and recruit into the mature bull population the following season. Harvest level of yearling bulls was determined by the percent of radio-tagged yearlings harvested and the changes in yearling-to-cow ratios between pre-season and post-season classifications (Beale 1995). Study results indicate that hunters were harvesting 85% of the yearling bulls.

Telonics (Mesa, Arizona, USA) and Lotek (Newmarket, Ontario, Canada) radios with an expected longevity of 4 years were used on the elk studies. With a full range of data point collection during the life of the transmitters, fidelity to seasonal ranges was easily mapped (Figure 2).

In conclusion, after years of combined telemetry data studies, many of the problems, based on speculation, were solved. Problems still exist in trying to keep elk populations under Herd Unit Elk Plan objectives. As future projects are proposed that require ground-disturbing activities, GIS maps will help managers arrange timing for the projects around elk calving or breeding seasons. In some cases, a project may not be allowed because of critical habitats. Management decisions that have been made with the help from telemetry and GIS include: 1) where to harvest elk during hunting season that are causing problems on agriculture land in the spring; 2) when to harvest elk on private land in the fall; 3) where to place vegetative treatments for better seasonal distribution; 4) juxtaposition of cougar, elk and deer; and 5) how livestock grazing redistributes elk patterns. Other maps of interest that have been generated show interaction of elk from different trap sites, interaction of elk from different herd units, elk calving range fidelity, seasonal distribution, migration patterns of bulls vs. cows, and many others.

Telemetry data combined with vegetation layers in the GIS computer environment allowed land managers to see the importance of the aspen habitat for elk calving and rearing. It is well known that aspen forest is an important habitat for elk over much of the West, but the intensive use of specific portions of the available forest for elk calving grounds, logically suggests these areas deserve special scientific and management attention (Boyce and Hayden-Wing 1979, Thomas and Toweill 1982). Forest managers must be able to identify and protect the critical components of calving habitats if elk management programs are to be successful (Flinders 1995).

Our data regarding roads and trails were interpreted as a good case for access management requiring careful monitoring of recreational use in calving areas during the critical reproductive period. That data should lead to seasonal closures of some roads and/or trails and involve intense scrutiny of any new access development (Flinders 1995).

The information obtained from this study will help biologists meet the objectives in their elk management plans. For example, seasonal distribution data and maps will aid in establishing hunting unit boundaries and season dates for antlerless hunts, so that harvest will remove those animals using the areas where the problem exist. Distribution and movement data can benefit range managers when allocating the available forage resource between livestock and elk on the different allotments during different months and seasons (Beale 1995).

Through the combination of telemetry and GIS, we have been able to make better, more sound judgments regarding big game on the Fishlake National Forest. However, we caution ourselves that the possibility of managing the land from a chair in front of a computer, with so much fascinating data, could get easier than going out in the field where things continue to change and wildlife constantly surprise us.


We would like to thank the Fishlake Elk Study Steering Committee; the Monroe Mountain Seeking Common Ground Steering Committee; Don Beale, Norm Bowden, and Ver Don Durfee, Utah Division of Wildlife Resources; Dr. Jerran Flinders, Brigham Young University; those from the Fishlake National Forest, Richfield District Bureau of Land Management; and volunteers who assisted on the project.


Beale, D. M. 1994, Seasonal distribution, productivity and mortality of elk on the Fishlake herd unit. Utah Division of Wildlife Resources Publication No. 97-16.

Boyce, M. S. and L. D. Hayden-Wing. 1979. North American elk: ecology, behavior, and management. University of Wyoming. 294pp.

Flinders, S. H., D. M. Beale, and J. T. Flinders. 1995. Bedsite selection by neo-natal elk in an aspen dominated ecosystem. 401 WIDB, BYU, Provo, UT.

Thomas, J. W., and D. E. Toweill, editors. 1982. Elk of North America, ecology and management. Stackpole Books Publ., Harrisburg, PA. 698pp.

Chronobiological Analysis of Animal Locations: Development of an Automatic Recording System and Principles of Data Processing.

K. M. Scheibe and K. Eichhorn

Cornelia Heinz Institute for Zoo Biology and Wildlife Research Berlin,
Pf 601103, D-10252 Berlin, Germany

T. Schleusener

IMF technology GmbH, Grosse Müllroser Strasse 46,
D-15232 Frankfurt (Oder), Germany

W. J. Streich and C. Heinz

Cornelia Heinz Institute for Zoo Biology and Wildlife Research Berlin,
Pf 601103, D-10252 Berlin, Germany


A telemetric system for location and for continuous recording of different types of behavior such as activity, resting and feeding has been developed and tested on Przewalski horses in a semi-reserve. The system combines a GPS-receiver with the features of the behavior recording system ETHOSYS. Data analysis starts with the identification of periods of activity, resting and feeding. Locations can be subsampled according to the different behavioral states.


Wildlife species such as large herbivores influence structure and composition of their habitats; they can be described as ecosystem engineers in the sense of Jones et al. (1994). The different types of herbivores select their home ranges and use the different structures inside these areas according to their evolutionary specialization. Factors such as predators, human disturbance, thermal exposure or insects influence the utilization of space. Telemetric methods may be used to understand the influence of animals on a habitat or to describe the anthropogenic factors that endanger individuals, populations, and species or that change the animals’ impact on the environment. At the last telemetry symposium in Marburg, Casaer et al. (1998) presented their approach to evaluate habitat utilization by definition of substructures of animal positions. As an additional approach, we present a strategy based on the combination of behavior and location.

Animal behavior is the basic interrelationship between the individual and its environment (Simpson 1969); and as the most effective adaptive mechanism, it has a regulative function (Hafez 1968). Behavior is organized in distinct functional circles as described by v. Uexküll (1973). Animal activities are directed to factors of the environment, the result of the action operating back on the animal. In this sense, animal behaviors as correcting elements are components of control systems. Feedback control systems are characterized by their temporal course. Typically, damped or sustained oscillations of the correcting element (behavior, e.g. feeding) and of the controlled member (internal state, motivation) result from regulation. The utilization of space depends on the actual behavioral state of an individual. This behavioral state oscillates as a result from regulative behavior, and, accordingly, the utilization of different places oscillates periodically. Animals do not disperse homogeneously over a home range or territory, but use different structures to satisfy special behavioral demands. The basic structure of a territory or home range for mammals was formulated by Hediger. He distinguished between homes of first or second order for resting, feeding places, drinking places, marking places, places for comfort behavior and animal tracks (Tembrock 1987). These different places are used by an individual following its regulative behavior.

Based on visual observations and activity records, we formulated the following demands for an automatic system for the analysis of the utilization of substructures in animal home ranges or territories:

The system should be able to record simultaneously the most important behaviors and locate the animals at fixed time intervals.

Resting, feeding and locomotion should be recorded continuously with a resolution in time sufficient for activity bouts of 1-2 h; this requires a sample interval of 15 min.

Locations should be taken at least in 1 h intervals to represent the different behavioral states. The precision should be in the range between 10 m to 50 m, to find correlations to larger vegetational units and to places of behavioral significance.

The recording unit should be a closed system for simple application and minimal impact on wild animals and able to work independently for up to 12 months to follow an individual during the different seasons.

Data should be stored and downloaded on request over distances of 100-1,000 m.

The resulting system should be tested and its performance (to analyse temporal and spatial interrelations between individuals and environments) investigated.


The daily time budget was observed by point sampling (observation interval 15 min) from two red deer (Cervus elaphus) and eight European bison (Bison bonasus) in a large, nature-like enclosure (110,000 m2). During 50 h of observation, 200 samples were taken for each species.

The behavior recording system ETHOSYS (Scheibe et al. 1995, Scheibe et al. 1998) has been used on roe deer, red deer, mouflon, Heck cattle and Przewalski horses in several large enclosures and on mouflon in the wild. Characteristic 24-h records were selected for comparison.

We followed the movements of one Przewalski horse living in a semi-reserve of 0.42 km2 in a group of eleven animals. Optical angle recording instruments were used to localize the animal from three fixed points at the border of the reserve. Fixes were taken every 5 minutes and analyzed using the software Locate II. The fixes were classified into locations during resting (lying, standing), locomotion, and grazing.

Following the demands for an automatic behavior recording and location system, a new system, called ETHOLOC was developed (Figure 1) and a prototype produced. The recording unit consists of two basic recording systems, a behavior recording system equivalent to ETHOSYS with four channels, and a Global Positioning System (GPS) location system using a receiver type GPSMS1 from m -blox (Switzerland) based on SIRF-technology and active antenna. Data are stored on board and transmitted via a two-way radio link on request.

Figure 1. ETHOLOC block diagram.

The recording unit is designed as a collar, with all elements arranged around the collar and the heaviest elements at the lowest point. The GPS aerial is fixed to the neck of the animal. The configuration for horses weighs 1 kg; it weighs 520 g without the waterproof-shielded mechanical cover. Two D-size LiNMH batteries with 26Ah provide the electric power for the system. The expected lifetime is 0.5-1 year depending on the duration of the location process. Power consumption is 180 m A without location or communication. During location, an additional 130 mA are required for a maximum of 2 min. Communication requires another 12 mA. The storage capacity is designed for a recording time without data transmission of 20 days (2046 behavior data sets and 512 GPS locations). A bi-directional radio link on the 70 cm band serves for data transmission. The receiving station is connected to a laptop computer. Communication worked over a distance of >150 m in our application using a l ¼ antenna on the receiving station. With a Yagi antenna, communication distances of more than 500 m can be achieved. A 1 mW radio beacon is attached for tracking and is cut off during communication. The data format of the positions enables a post-processing correction of the GPS fixes using a reference station, but until now, we worked with non-corrected values. The prototype was tested on a Przewalski horse in the semi-reserve mentioned previously.


For large herbivores, the daily cycle of behavior is mainly determined by feeding and resting as shown by the examples from the two species. Both behaviors occupied more than 80% of the time budget (Figure 2). Social behavior such as mutual grooming or locomotion without feeding was displayed, respectively, for 1% and 6% of the observation period.


Figure 2. Time budget of European bison (Bison bonasus) left, and red deer (Cervus elaphus), right.

Following the course of behavior over the day, we found a regular change between activity and resting in all herbivorous species observed (Figure 3). The ultradian rhythms differed from each other, as the basic frequency differed according to the different feeding types that the animals represent. In all large herbivores we found cycles of activity and resting with period length between 1 h and 6 h.



Figure 3. Typical activity patterns on single days, recorded by ETHOSYS from individuals in different large nature-like enclosures or in the wild (mouflon) during early summer. The different species represent different feeding types (roe deer are concentrate selector, red deer are intermediate type, mouflon and Heck-cattle are ruminant roughage eater, and Przewalski horses are non-ruminant roughage eater).

We followed the movements of one Przewalski horse, living in a semi-reserve in a group of eleven animals, to get a first impression of the interdependencies between behavior and location. During a 4 h period, we recorded a complete cycle of behavior, beginning with resting, followed by locomotion and then a longer feeding bout, which changed to resting after locomotion again (Figure 4). The different behaviors were spatial separated. Resting occurred primarily on one special sandy place while feeding occurred on a clear trace. Locomotion without feeding was seen between the resting place and the feeding area. The trace of feeding resulted from movements over several feeding grounds. The feeding grounds differed clearly from the surrounding vegetation in height and composition due to the long-time impact of the animals.

Figure 4. Visual locations of a Przewalski horse during a 4 h observation with behavior indications.

The automatic locations from ETHOLOC show a fairly random distribution over the whole area if all locations are displayed together (Figure 5). A behavior record from ETHOLOC is displayed in Figure 6 for a typical day. The ultradian rhythms of activity and feeding become clearly visible. Periods of resting and feeding were defined using an automatic peak detection procedure and the locations subsampled according to these behaviors. Results are displayed in Figure 7 and Figure 8 for resting and feeding separately. As non-corrected values are displayed, the GPS-typical error of up to 100 m influences the locations, so that some of them occurred even outside the enclosure. Resting is predominantly shown in the central southern part of the enclosure, near the preferred resting place. Three locations are near the wood at the border of the enclosure. These places are used by the horses as shelters against wind or rain. Feeding took place on locations around the resting place and on a central track near the northern part of the enclosure.

Figure 5. All automatic locations of a Przewalski horse by ETHOLOC.

Figure 6. Behavior record of a single day from a Przewalski horse, measured by ETHOLOC.

Figure 7. Locations of a Przewalski horse during resting periods.

Figure 8. Locations of a Przewalski horse during feeding periods.


Locations of large herbivores depend on their behavioral state, which varies more or less regularly. The most important behaviors that should be classified are feeding and resting. Locomotion may be a third behavior of importance that might be underrepresented in the restricted area of a semi-reserve. The parallel recording of behavior and location by means of ETHOLOC gives insight into the substructure of animal territories or home ranges. A system for recording grazing, resting, and location was developed as described by Rutter et al. (1977), but size and sensor attachment as well as lack of automatic data transmission restricted its application to tame and well-adapted domestic animals. ETOLOC, with a closed collar for the animal, avoids as far as possible influencing the animal and is applicable to wild animals. The strategy to analyze locations depending on at least the three basic behaviors mentioned will be essential for understanding not only interrelations like these between animals and vegetation, but also for the analysis of factors like anthropogenic disturbances or interspecific concurrence.


We thank the keepers of the Przewalski horses, Mrs. and Mr. Broja, for support and access to the animals. The presentation was supported by Deutsche Forschungsgemeinschaft.


Casaer, J., M. Hermy, R. Verhagen, P. Coppin 1998. Analysing habitat utilization by Thiessen polygon and triangulated irregular network interpolation. Pages 435-440 in T. Penzel, S. Salmons, M.R. Neumann (eds.), Biotelemetry XIV, Proceedings of the Fourteenth International Symposium on Biotelemetry, Marburg, Germany, April 6 - 11, 1997, Tectum Verl. Marburg.

Hafez, E.S.E. 1968. Adaptation of Domestic Animals. Philadelphia 1968.

Jones, C.G.; Lawton, J.H., Shachak, M. 1994. Organisms as ecosystem engineers. Oikos 69:373-386.

Rutter, S. M., N. A. Beresford, G. Roberts 1997. Use of GPS to identify the grazing areas of hill sheep. Comp. Electronics Agric. 17:177-188.

Scheibe, K.M., K. Eichhorn, T. Schleusner, A. Berger, J. Langbein 1995. Biorhythmic analysis of behavior of free ranging domestic and wild animals by means of a new storage-telemetry system. Pages 271-276 in Cristalli, C., C.J. Amlaner, M.R. Neumann (eds.), Biotelemetry XIII, Proceedings of the Thirteenth International Symposium on Biotelemetry, Williamsburg, Virginia U.S.A., March 26-31, 1995.

Scheibe, K.M., Th. Schleusner, A. Berger, K. Eichhorn, J. Langbein, L. Dal Zotto, W.J. Streich 1998. ETHOSYS – new system for recording and analysis of behavior of free-ranging domestic animals and wildlife. Appl. Anim. Behav. Sci. 55:195-211.

Simpson, G.G. 1969. Evolution: Methoden und derzeitiger Stand der Theorie. in Roe, A. and G.G. Simpson, Evolution und Verhalten. Suhrkamp Verl. Frankfurt a.M.

Tembrock, G. 1987. Gesetzmäßigkeiten tierischen Verhaltens. Pages 19-72 in Scheibe, K. (ed.), Nutztierverhalten. Fischer Verl. Jena.

Uexküll, J. v. 1973. Theoretische Biologie. Suhrkamp, Frankfurt.

Bird Altitude Radio Telemetry

Harlan D. Shannon

Center for Conservation Research and Technology, University of Maryland, Baltimore County, Baltimore, Maryland, USA

George S. Young

Department of Meteorology, Pennsylvania State University, University Park, Pennsylvania, USA

William S. Seegar

United States Army Edgewood Research, Development, and Engineering Center, Aberdeen Proving Ground, Maryland, USA

M. Blake Henke

Center for Conservation Research and Technology, University of Maryland, Baltimore County, Baltimore, Maryland, USA

Sheldon L. Struthers and Valerian B. Kuechle

Advanced Telemetry Systems, Inc., Isanti, Minnesota, USA


Radio telemetry transmitters equipped with pressure sensors were developed to measure bird flight altitudes. Transmitter design parameters and data recording methods are presented. Sensor accuracy was investigated by comparing transmitter-derived altitudes to those derived from a control altimeter. Results suggest that after correcting for various meteorological processes, an accurate measurement of bird flight altitudes can be obtained using these transmitters.


Various methods have been used to identify temporal and spatial variations in bird flight altitudes. Ground-based observations often provide generalized descriptions of bird flight altitudes based upon visual estimates. Such observations are typically subjective, providing little if any quantitative information (Kerlinger 1989). Furthermore, the volume of air space through which observations can be made is limited by the distance an observer can see birds, often from a fixed location. More precise measurements of bird flight altitudes have been obtained by following birds with aircraft, estimating bird flight altitudes using aircraft altimeter readings (Pennycuick 1972). These data are most accurate when bird flight altitudes change slowly over time, enabling the pilot to maintain an altitude comparable to that of the bird. Although this method enables observers to obtain more precise measurements over greater distances, avian flight behavior and/or inclement weather can limit the effectiveness of this technique. For example, soaring birds often make rapid ascents relative to aircraft, thereby limiting observer abilities to continually obtain accurate bird altitude data. Various types of radar have been used successfully to gather bird altitude data, including tracking (Able 1977), vertical pointing (Kerlinger and Gauthreaux 1985), and surveillance radars (Gauthreaux 1991). Radar power, range, and resolution limitations often limit the volume of air space over which precise bird flight altitude data can be acquired using these tools. In general, the closer the birds are to these radars, the more precise the radar-derived altitude measurements. Given the limitations of radar, ground- and aircraft-based observation methods, new technologies have been developed to obtain more precise bird flight altitude measurements over broader areas.

This paper describes a conventional radio telemetry transmitter designed to measure bird flight altitudes. These transmitters, like aircraft altimeters, provide altitude based upon atmospheric pressure measurements. This relationship between altitude and atmospheric pressure is briefly discussed. Transmitter specifications and data recording and manipulation methods are presented. A study was conducted to evaluate the accuracy of transmitter altitude measurements. Results from this study are provided, as is a discussion of various corrections that can be applied to improve the accuracy of these measurements.


Altitude is related to atmospheric pressure through the expression,



Z is the altitude (m) above sea level,

To is the surface temperature (K),

G is the lapse rate (K m-1) (i.e., negative of the change in temperature with height),

P and Po are the atmospheric pressures (mb) at altitude Z and sea level respectively,

R is the gas constant for dry air (287 m2 K-1 s-2), and

g is the acceleration due to gravity (9.81 m s-2).

Altimeters are often calibrated assuming a USA standard atmosphere (i.e., To = 288oK, Po = 1013.25 mb, and G = 6.5 K m-1) such that variations in altitude Z are a function of pressure P alone. The conventional radio telemetry transmitters discussed herein employ this methodology to measure bird flight altitudes.

Radio Transmitters

Thirty-one conventional radio telemetry transmitters, each containing a Motorola pressure sensor (MPX4115A), were designed and built by Advanced Telemetry Systems to measure bird flight altitudes. Each transmitter weighed between 34.7 grams and 37.3 grams, and transmitted on a frequency between 149.000 MHz and 151.999 MHz. Changes in pressure, hence altitude, were expressed through changes in transmitter pulse rates. An Advanced Telemetry System model R4000 receiver was used in conjunction with an Advanced Telemetry System Data Collection Computer (DCC II) to monitor and log these radio telemetry transmissions. The DCC II computes the time interval (in milliseconds) between successive transmitter pulses, and averages over six consecutive intervals (i.e., seven successive pulses) to calculate an average pulse rate. Averaged pulse rates are time tagged and stored in the memory of the DCC II during tracking. Once tracking is complete, these time-tagged data are downloaded into personal computers, imported into spreadsheets, and input into transmitter-specific equations that relate atmospheric pressure to transmitter pulse rates. These pressure data are then used to solve equation 1, and thereby to determine bird flight altitudes.

Transmitter Accuracy Evaluation

Transmitter performance was evaluated by comparing transmitter altitudes to those provided by a Pretel Altiplus A2 (5 m vertical resolution) altimeter. Measurements were taken from each transmitter and the control (Pretel) altimeter during level flight in a Cessna 172 aircraft. These flight levels were approximately 304.8 m (1,000 ft), 609.6 m (2,000 ft), 914.4 m (3,000 ft), 1371.6 m (4,500 ft), 1676.4 m (5,500 ft), and 1,981.2 m (6,500 ft). Note that the aircraft altimeter measures altitude based upon the atmospheric pressure outside the aircraft. The transmitters and control altimeter were housed within the aircraft cabin, where the atmospheric pressure was slightly less during the test flight. An analysis of transmitter altitude accuracy was completed using altitude data obtained from the Pretel altimeter because of these pressure differences.

The receiver scanned each frequency for 15 seconds, obtaining one altitude measurement from each transmitter during each period of level flight. These measurements were time tagged and logged in the DCC II. The control altimeter provided altitude data continuously; however, this instrument is incapable of logging these data. In addition, the pilot was unable to hold the plane strictly level during each period of approximately level flight. Therefore, an average aircraft altitude was computed for each period of level flight based upon the range of values observed on the control altimeter during each period. The difference between the transmitter derived altitude and the average control altimeter altitude was calculated for each transmitter for each period of level flight. The mean difference was then calculated for each level by averaging the differences for all 31 transmitters at that level.


Figure 1 illustrates the mean difference for each level. The uncertainty in these means was calculated assuming a 95% confidence interval (t-test), and is shown by the error bars in Figure 1. These results indicate that the altitude readings obtained from the transmitters were on average 7 m to 30 m less than those readings obtained from the control altimeter. Differences between transmitter-derived altitudes and the average assumed altitude at each level are primarily attributed to two factors. During each period of level flight, aircraft altitudes often varied by 30 m to 60 m. These deviations are largely responsible for the uncertainty in the average assumed altitude of each level. Although the uncertainty in the average assumed altitude is greater than the mean difference between this altitude and the transmitter-derived altitudes at each level, the consistent negative bias in transmitter-derived altitudes suggests that an additional factor is contributing to these differences. The most likely factor contributing to this bias is calibration differences between the transmitters and the control altimeter.


Although reasonable altitude estimates can be obtained by measuring the pressure at flight level and solving Equation 1 assuming a standard atmosphere, more precise altitude readings can be derived by solving this equation using observed values of the surface temperature To, surface pressure Po, and lapse rate G . Measured values for these variables improve altitude calculations because these values often deviate significantly from the constant values assumed in a standard atmosphere. Examples of how changes in these variables affect altitude readings follow.

Figure 1. Mean difference between transmitter-derived altitudes and the approximate altitude of each flight level.

In a standard atmosphere, the sea-level atmospheric pressure is assumed to be 1013.25 mb. It is common, however, for these pressures to vary by 20 mb or more from the standard value. At sea level, a 1 mb uncertainty in surface pressure is approximately equivalent to an 8.3 m uncertainty in altitude. This relationship was obtained by calculating the altitude at pressure levels 1 mb above and below the standard value assuming a standard atmosphere. Given this relationship, a 20 mb uncertainty in surface pressure translates into a 166.4 m uncertainty in altitude. Given this variability in sea-level pressure, corrections for changes in sea-level pressure are necessary to reduce the error in calculated altitudes.

Variability in the atmospheric temperature profile can also influence altitude measurements. In a standard atmosphere the surface air temperature is 288oK (15oC). In reality, temperatures often vary significantly on hourly and daily time scales. Table 1 shows the uncertainty in calculated altitudes assuming a 15oK uncertainty in surface temperature (i.e., between 0oC and 30oC). These values were obtained by solving equation 1 assuming To = 273oK and To = 303oK, where all other variables were defined by their standard values. Table 1 shows that the uncertainty in altitude measurements increases with height. When expressed as a fraction of the approximate altitude, however, the uncertainty in altitude measurements attributed to the variability in surface temperature is relatively constant, approximately 5% of the altitude.

Table 1. Uncertainty in altitude calculations attributed to variations in the surface temperature.

Pressure Level (mb)

Approximate Altitude (m)

Uncertainty (m)



















The lapse rate in a standard atmosphere is 6.5 K km-1. Many birds, specifically soaring species, confine the majority of their flight to the boundary layer. This layer is defined as that part of the lower atmosphere directly influenced by the earth’s surface, and is where thermal updrafts and downdrafts often predominate (Stull 1988). The lapse rate in the boundary layer is normally 9.8 K km-1. The error in this variable when assuming a standard atmosphere is therefore 3.3 K km-1. The resulting errors in altitude are shown in Table 2. Although these errors are not as great as those associated with uncertainties in the surface temperature, a more precise altitude measurement will be obtained by using measured values of the lapse rate.

Table 2. Uncertainty in altitude calculations attributed to variations in the lapse rate.

Pressure Level (mb)

Approximate Altitude (m)

Error (m)



















The combined affects of temperature and pressure uncertainties on altitude calculations can be significant. For example, assume the surface pressure is 1023.25 mb, the surface temperature is 295.5oK, and the lapse rate is 9.8 K km-1. For a bird soaring at a pressure level of 900 mb, the calculated altitude is 1089.4 m. In contrast, the calculated altitude assuming a standard atmosphere is 987.5 m. These differences in altitude calculations can be significantly greater depending upon atmospheric conditions. It is therefore important to use the most accurate data available when calculating altitude based upon atmospheric pressure measurements.


The authors thank Jim Dayton, Tom Maechtle, and Mike Yates for their assistance in obtaining and testing these transmitters.


Able, K. P. 1977. The flight behavior of individual passerine nocturnal migrants: A tracking radar study. Anim. Behav. 25:924-935.

Gauthreaux, S. A., Jr. 1991. The flight behavior of migrating birds in changing wind fields: Radar and visual analyses. Amer. Zool. 31:187-204.

Kerlinger, P. 1989. Flight strategies of migrating hawks. University of Chicago Press, 375 pp.

Kerlinger, P. and S. A. Gauthreaux, Jr. 1985. Seasonal timing, geographical distribution, and flight behavior of broad-winged hawks during spring migration in south Texas: A radar and visual study. Auk 102:735-743.

Pennycuick, C. J. 1972. Soaring behaviour and performance of some East African birds, observed from a motor-glider. Ibis 114:178-218.

Stull, R. B. 1988. An introduction to boundary-layer meteorology. Kluwer Academic Publications, 666 pp.

An Improved Radio Attachment Technique for a Variety of Marine Bird Species

Scott H. Newman

Oiled Wildlife Care Network, Wildlife Health Center,
School of Veterinary Medicine, University of California,
Davis, California, USA

S. Kim Nelson

Oregon Cooperative Fish and Wildlife Research Unit,
Oregon State University, Department of Fisheries and Wildlife,
Nash Hall 104, Corvallis, Oregon, USA

John Y. Takekawa

United States Geological Survey, Biological Resources Division,
San Francisco Bay Estuary Field Station, Post Office Box 2012,
Vallejo, California, USA

Esther E. Burkett

California Department of Fish and Game, Wildlife Management Division,
1416 Ninth Street, Sacramento, California, USA

Richard T. Golightly

Humboldt State University, Wildlife Department,
Arcata, California, USA


We modified a subcutaneous radio attachment technique for use on four marine bird and diving seabird species (American coots, Xantus’ murrelets, marbled murrelets, and western gulls) to study post-oil exposure and rehabilitation survival, breeding biology, habitat use, distribution, and movements. The modified subcutaneous anchor, suture, and epoxy attachment provided a secure attachment method for these birds without resulting in significant detrimental effects. Mean tracking time was 18-159 days with a maximum duration of 51-145 days.


Successful radio telemetry studies require a transmitter attachment method capable of providing adequate duration of attachment without causing major adverse effects or altered behavior in study subjects. When an appropriate attachment method is established for a given species, telemetry studies enable researchers to monitor survival (short- or long-term), individual movements, home range utilization, foraging ranges, and breeding or nesting behavior.

Devising radio attachment techniques for the diverse array of avian species has posed great challenges to ornithologists and biologists alike. However, aquatic avian species have posed an even greater challenge because of their natural history and the environment they inhabit. Alcids, for example, can inhabit old growth forests ecosystems, coastal cliffs, caves, or marine islands; they can nest several hundred feet in the air or in burrows several meters underground; they can forage near shore or hundreds of kilometers out at sea; they can dive to depths of 180 m or catch prey at the sea surface; and they range in body size from less than 100 grams to more than a kilogram. Therefore, a telemetry attachment method for these species should be secure and able to tolerate harsh marine conditions encountered by wading and diving marine birds; and at the same time, not adversely affect the birds’ flight, diving capability, or waterproofing.

Multiple transmitter attachment methods have been developed, including harnesses, adhesives, neck collars, sutures, and surgical implants (Kenow et al. 1997). Radio telemetry studies of small alcids and waterfowl have been hampered by low transmitter retention times, radio failure, short transmission ranges, or adverse behavior (Duncan and Gaston 1990, Ralph et al. 1992, Kuletz et al. 1995). Stainless steel or surgical gut harnesses enclosed in plastic tubing appeared to disrupt the normal behavior of Xantus’ murrelets (Synthliboramphus hypoleucus) (Hunt et al. 1979). Marbled murrelets (Brachyramphus marmoratus) fitted with harnesses of polyethylene tubing died soon after release, and murrelets with surgically implanted transmitters responded poorly to the anesthetic and never behaved normally after release (Quinlan and Hughes 1992). Harnesses used to attach radios to mallards (Anas platyrhynchos) have resulted in both behavioral and reproductive problems (Pietz et al. 1995, Rotella et al. 1993). Suture, glue, or suture-and-glue attachments in both alcids and waterfowl resulted in short retention times and inadequate tracking duration (Wanless et al. 1985, 1988a, 1988b, Duncan and Gaston 1990, Quinlan and Hughes 1992, Houston and Greenwood 1993. Rotella et al. 1993, Burns et al. 1994, Kuletz et al. 1995, Kuletz and Marks 1997, Newman et al. 1999a).

Increased radio retention time was accomplished for mallard ducklings (Mauser and Jarvis 1991) and adults (Pietz et al. 1995) using a subcutaneous anchor and 3 sutures. This attachment method had a longer retention time than previous attachment techniques and also had minimal adverse effects on birds. We modified this subcutaneous anchor attachment to study breeding season distribution and movements of Xantus’ murrelets in California, marbled murrelets in California and Alaska, and to study post-oil spill rehabilitation survival of American coots (Fulica americana) and western gulls (Larus occidentalis). In this paper, we describe the attachment modifications, and the effectiveness of a modified subcutaneous anchor attachment for the species studied, and discuss the advantages and potential disadvantages of this radio-marking technique.


We used 3.5 g radio transmitters (Model PD-2, Holohil Systems Limited, Ontario, Canada) with 15-cm external whip antennas and front and rear suture channels for Xantus’ murrelets. We used 2.0 g transmitters (Model BD-2G, Holohil Systems Limited) with 15-cm whip antennas and front and rear suture channels for marbled murrelets. Stainless-steel anchors (Advanced Telemetry Systems, Isanti, Minnesota) were slightly flattened and attached to the bottom surface of the front end of the transmitter with Superglue and finished with epoxy (Hysol Epoxi-Patch, Dexter Corporation, Seabrook New Hampshire). The bottom of the radio was roughened for better epoxy adhesion using 200-grain sandpaper. We used 7.5 g transmitters (Model B2040, Advanced Telemetry System, Isanti, Minnesota) with front and rear suture channels, with a 15 cm whip antenna, and stainless-steel anchors already attached for American coots and western gulls. These radios were programmed with a duty cycle of 8 hours on and 16 hours off, and were equipped with a mortality sensor.

All transmitters were attached to birds using the modified subcutaneous anchor technique (Newman et al. 1999a). Modifications to the traditional anchor attachment included: 1) use of an inhalant anesthetic, isoflurane, to facilitate radio attachment, decrease pain and alleviate stress; 2) placement of the subcutaneous anchor farther cranial on the bird’s dorsal side (back) compared to the more caudal location of the anchor on ducklings and ducks; 3) using a 16 g hypodermic needle to create an entry port for the anchor placement instead of making an incision using a scalpel; 4) not trimming any feathers at the attachment site to decrease the possibility of waterproofing problems or thermal burns secondary to the exothermic epoxy; 5) application of marine epoxy to initially hold the radio in place until a healthy fibrous- tissue reaction firmly attached the anchor (and transmitter) to the birds; and 6) placement of a single subcutaneous non-absorbable suture at the caudal suture port of the transmitter for further stability. In addition, the actual transmitter attachment time was 5-10 minutes compared with 15 minutes reported for the traditional anchor attachment method (Pietz et al.1995).

We administered an inhalation anesthetic, isoflurane, to facilitate handling and minimize the pain and stress associated with transmitter attachment. Isoflurane is safer and has multiple advantages over methoxyflurane, which has previously been used for field anesthesia on birds (Rotella and Ratti 1990, Pietz et al. 1995). Xantus’ and marbled murrelets were sedated using a prototype field anesthesia system (Newman et al. 1999a). Coots and gulls were sedated using a standard vaporizer and anesthetic machine. Birds were exposed to anesthetic until they no longer had a palpebral reflex, showed signs of decreased muscle tone (no leg or wing withdrawal reflex), and developed slow, deep respiratory patterns. Murrelets required anesthetic exposure for < 30 s and were then removed from the anesthetic mask. Coots and gulls were maintained on isoflurane (2-3% flow rate) for the duration of the attachment procedure.

We radio tagged 113 Xantus’ murrelets during April and May in 1996 and 1997 at Santa Barbara Island, California (Whitworth et al. 1997); 28 marbled murrelets in May, June and August 1997 at Año Nuevo Bay, California (Burkett et al. 1998); and 9 marbled murrelets in Auke Bay, Juneau, Alaska, during June 1998 (Nelson et al. 1999). In April 1995, 75 American coots (37 oiled and rehabilitated; 38 not oiled, not rehabilitated) were radio tagged, wing-clipped and relocated to an outdoor enclosure (0.5 ha land and 0.5 ha pond) at the University of California, Davis (Anderson et al. 1999, Newman et al. 1999b). During October and November 1997, 27 western gulls (7 oiled and rehabilitated; 10 not oiled, but rehabilitated; 10 not oiled, not rehabilitated) were radio tagged at either Vandenberg Air Force Base, California or International Bird Rescue Research Center, Berkeley, California (Golightly et al. 1999).


Xantus’ Murrelets

We located 92 of 113 (81%) radio-tagged, free ranging Xantus’ murrelets during breeding and post-breeding dispersal periods in the Southern California Bight. Mean tracking duration was 18 ± 15 (mean ± SD) days with the longest tracking duration of 51 days. On average, murrelets were relocated 40-60 km from Santa Barbara Island with the farthest location being 567 km (Whitworth et al. 1997), The total distances traveled by individual murrelets was quite large, with one bird moving more than 1,700 km in 47 days (Whitworth et al. 1997). Foraging distribution and habitat use during different months of the breeding season were also documented for Xantus’ murrelets. One Xantus’ murrelet, which had been radio tagged in 1996, was recaptured in 1997, in healthy condition. The radio was not present. One bird, which had been radio tagged in 1996, was sighted again in 1997 incubating eggs with the dead transmitter still attached.

Marbled Murrelets in California

In 1997, 28 (100%) marbled murrelets were relocated and tracked for 45 ± 15 days in a breeding biology and habitat-use study in central California. The longest tracking duration was 78 days. Following marking, marbled murrelets dispersed far to the north (224 km) and south (181 km) from Pt. Año Nuevo (Burkett et al. 1998, 1999) and birds continued to navigate inland to forest nest locations (Big Basin Redwoods State Park and surrounding habitat). These observations suggest that radio tagging had minimal, if any, detrimental effects on the ability of marked individuals to fly.

One marbled murrelet radio-tagged in 1997 was recaptured 4 months later with the radio still attached, but attached only by part of the subcutaneous anchor. The suture was still tied to the radio but had pulled out from the skin. There was no evidence of infection or inflammation at the anchor attachment site. Another marbled murrelet had completely lost its radio and the attachment site was healthy and completely healed.

Marbled Murrelets in Alaska

In a similar study on marbled murrelet breeding biology and habitat use in Juneau, Alaska, in 1998, 9 of 9 (100%) murrelets were relocated after marking. Mean tracking time was 67 ± 30 days and the longest duration was 84 days (Nelson et al. 1999). Following capture and marking in late May, birds primarily remained in Auke Bay and Fritz Cove. However, beginning in mid-June, 6 of the 9 birds were detected flying inland during the early morning hours to potential nest sites on the slopes and ridges along the Mendenhall River Valley. At the same time, all nine murrelets left the Auke Bay area and began flying up to 124 km from inland sites to western Icy Strait, near the mouth of Glacier Bay, on a daily basis. Average flight distances were 85 ± 37 km from inland sites. These observations again suggest that radio tagging had minimal, if any, detrimental effects on the ability of marked individuals to fly.

Despite consistent inland attendance patterns by 6 marked individuals to inland nest sites, no nest sites were located. This was primarily related to the fact that inland visits were short in duration and incubation was not documented (24-hour shifts on alternate days by both adults; Nelson 1997). We suspected that these murrelets were failed breeders, post-breeders, or sub-adults. However, we do not know whether the transmitters had an effect on breeding, or whether other factors, such as El Niño, reduced breeding effort and success. Few observations of adults with fish in 1998 and complete breeding failure by several seabird species in the Gulf of Alaska suggest the latter.

American Coots

In 1995, 75 radio-tagged American coots were wing-clipped and relocated to an outdoor enclosure. Birds were monitored for 140 days following oil exposure and rehabilitation (Anderson et al. 1999, Newman et al. 1999b). At this time, remaining birds (those that had not molted and emigrated, n=22, or died, n=11) were recaptured and 21 transmitters were removed. Before transmitters were removed, the following was observed at the attachment sites: feathers had pulled out from the attachment site and were attached to the epoxy on the bottom of the radio; the suture was still tied to the radio but had pulled out from the skin; refeathering had either started or was completed beneath the radio; the anchor was still in place and fibrous tissue could be palpated at the anchor attachment site. No attachment sites were inflamed or infected based on visual inspection. Necropsies and histological examination of 11 coots that died from other causes (Newman et al. 1999b) revealed that healthy fibrous tissue surrounded the anchors and no infections had occurred. Of birds that molted and emigrated from the enclosure, the longest tracking duration was 145 days when no further tracking flights were performed. Birds were tracked as far as 160 km from Davis.

Western Gulls

In 1998, 27 (100%) of radio-tagged western gulls were relocated in a southern California post-oil exposure and rehabilitation release study. Mean tracking time was 159 ± 51 days with the longest tracking duration being 235 days. Gulls were tracked as far north as San Luis Obispo and as far south as Long Beach (approximately 350 km). Repeated telemetry locations and minimum convex polygons and adaptive kernels were used to determine that spatial habitat use during the fall, winter and spring were similar for oiled and rehabilitated gulls, and control gulls (Golightly et al. 1999).


The broad spectrum of data collected from this variety of studies demonstrated that the modified subcutaneous anchor attachment was a minimally invasive, durable attachment technique with little affect on the species studies. Using isoflurane to sedate birds facilitated rapid transmitter attachment and had no effect on survival or fitness of released birds. The suture and glue held these transmitters in place immediately following attachment, whereas the subcutaneous anchor held these radios in place over the long term. Re-inspection of the attachment site also indicated that infection was not a concern; this should be the case if appropriate anchor sterilization and clean attachment techniques are implemented. Feather loss beneath the attachment site may pose a slight increase in thermoregulatory demand for radio-tagged birds, but did not result in severe compromise or mortality of these aquatic species, including the diving seabirds we studied. Furthermore, in the case of the coots, refeathering occurred within several months. In addition, data from one recaptured Xantus’ murrelet and one recaptured California marbled murrelet indicated that the anchor can be expelled from under the skin, causing the entire transmitter to detach. Furthermore, these birds were not physically harmed by the detachment process. Similar findings were documented by Pietz et al. (1995), suggesting that transmitter loss was relatively benign and did not affect the health of birds.

The large flight distances documented in these studies also indicated that the attachment location or technique does not impede wing movements or flight in the species we studied. Furthermore, documented survival duration and specific flights to and from known foraging areas for these species indicated that foraging behavior was not affected. Intermittent recapture of Xantus’ murrelets and marbled murrelets, and monthly recapture of coots indicated that body mass was maintained following transmitter attachment.

In previous studies on productivity, the subcutaneous anchor attachment did not alter survival rates of ducklings from radio-tagged adult mallards (Pietz et al. 1995). Western gulls also demonstrated visitation of known breeding areas in the spring; however, limited transmitter life prevented confirming nesting of gulls from this study. Tracking results suggested that breeding behavior and nest initiation were not significantly disrupted by the anchor attachment method for at least 5 of 18 California marbled murrelets (28%) known to be reproductively active; however, nest success may have been affected (Burkett et al. 1998). By contrast, nest success was documented in 23 of 40 radio-tagged marbled murrelets in Desolation Sound, British Columbia (Cooke et al. 1999). Overall fecundity (29%) was greater in marked than unmarked marbled murrelets. Marbled murrelets monitored in Alaska demonstrated repeated flights from known foraging areas to inland old-growth forest regions, which were likely the locations of nest sites (Nelson et al. 1999). Unfortunately, limited tracking flights early in the study prevented confirmation of nest sites at these inland areas. This compilation of data suggested that while possible effects of this marking technique on reproductive success were possible, aquatic species – and, in particular, diving seabirds – can successfully breed and fledge offspring despite handling and radio tagging. Further studies are needed to determine the extent of effects on reproductive success and to study possible effects on productivity in other species of marine birds.

In all of these studies, the broad range of habitat utilized by these species (especially Xantus’ murrelets) and our inability to track hundreds of kilometers on every telemetry flight often limited the mean tracking duration. In addition, we believe that limitations in tracking duration were related to the transmitter life span rather than the detachment of the transmitter from birds, or from bird mortality. These studies determined that the modified subcutaneous anchor attachment technique provided a secure and reliable attachment method for multiple marine and aquatic bird species. This attachment does not appear to have significant detrimental effects on flight capabilities, foraging behavior, or survival, and may have minimal affects on nest success.


We are grateful for funding and support provided by multiple federal and state agencies (California Department of Fish and Game (CDFG), CDFG - Office of Oil Spill Prevention and Response, Oiled Wildlife Care Network, United States Navy, United States Geological Survey - Biological Resources Division, United States Fish and Wildlife Service, United States Forest Service (USFS), USFS Pacific Northwest Research Station, Alaska Department of Fish and Game, Channel Islands National Marine Sanctuary, Channel Islands National Park, Año Nuevo State Reserve, and Big Basin Redwoods State Park); universities (Humboldt State University, Oregon State University, and University of California, Davis); and private companies (Big Creek Lumber Company and Ecoscan Resource Data). We are also grateful to the hundreds of personnel and volunteers who helped with these projects.


Anderson, D. W., S. H. Newman, P. R. Kelly, S. K. Herzog, and K. P. Lewis. In press. Experimental release of oil-spill rehabilitated coots: Lingering effects on survival and behavior. Environmental Pollution.

Bight: 1995-1997. Final Report, U. S. Geological Survey, Biological Resources Division, California Science Center, Vallejo and Dixon, California; Wildlife Health Center, University of California, Davis, California; Naval Air Weapons Station, Pt. Mugu, California; and California Department of Fish and Game, Office of Oil Spill Prevention and Response, Sacramento, California.

Burkett, E. E., H. R. Carter, J. Y. Takekawa, S. H. Newman, and R. T, Golightly. 1998. Movement patterns and habitat preferences of Marbled Murrelets in central California: A radio telemetry study. In Proceedings of the 25th Annual Pacific Seabird Group Meeting, January 21-25. Monterey, California.

Burkett, E. E., H. R. Carter, J. Y. Takekawa, S. H. Newman, R. T. Golightly, and J. L. Lanser. 1999. Marbled Murrelet radio telemetry in central California in 1997 and 1998: A reduced breeding effort in an El Nino year. In Proceedings of the 26th Annual Pacific Seabird Group Meeting, February 24-28, Blame, Washington.

Burns, R. A., L. M. Prestash, and K. J. Kuletz. 1994. Pilot study on the capture and radio tagging of Murrelets in Prince William Sound, Alaska, July and August, 1993. Final Report, U. S. Fish and Wildlife Service, Anchorage, Alaska. Project 1193051 B- Exxon Valdez Restoration Project.

Cooke, F., L. W. Lougheed, G. Kaiser, and S. Boyd. 1999. Survival and fecundity of Marbled Murrelets at Desolation Sound, B. C. Page 28 in Proceedings of the 26thAnnual Pacific Seabird Group Meeting, February 24-28, Blaine, Washington.

Duncan, D. C., and A. J. Gaston. 1990. Movements of Ancient Murrelet broods away from a colony. Pages 109-113 in S. G. Scaly ed. Auks at Sea, Studies in Avian Biology 14.

Golightly, R. T., Newman, S. H., Carter, H. R., Craig, E. N., Van Wagenen, B., Mazet, J. K., 1999. Survival and behavior of western gulls following exposure to oil and rehabilitation. Abstract in Proceedings of the Western Section of the Wildlife Society, Monterey California, January 1999.

Houston, R. A. and R. J. Greenwood. 1993. Effects of radiotransmitters on nesting captive mallards. Journal of Wildlife Management 5 7:703-709.

Hunt, G. L., Jr., R. L. Pitman, M. Naughton, K. Winnett, A. Newman, P. R. Kelly, and K. T. Briggs. 1979. Summary of marine mammal and seabird surveys of the Southern California Bight. Volume III - Investigator’s Reports. Part III. Seabirds - Book II Reproductive ecology and foraging habits of breeding seabirds. Report for the U. S. Department of Interior, Bureau of Land Management by the Center for Coastal Marine Studies, University of California at Santa Cruz.

Kenow, K. P., C. E. Korschgen, F. J. Dein, A. P. Gendron-Fitzpatrick, and, E. F. Zuelke. 1997. Evaluating the effects of telemetry transmitter attachment techniques on waterfowl: A review and recommendations. Page 41 in J. E. Austin and P. J. Pietz (eds). Forum on Wildlife Telemetry: Innovations, evaluations and research needs. Snowmass, Colorado.

Kuletz, K. J., D. K. Marks, R. A. Burns, L. M. Pretash, and D. A. Flint. 1995. Marbled murrelet foraging patterns and a pilot productivity index for murrelets in Prince William Sound, Alaska. Exxon Valdez oil spill restoration project final report, Project 94102. U. S. Fish and Wildlife Service. Anchorage, Alaska.

Kuletz, K. J., and D. K. Marks. 1997. Post-fledging behavior of a radio-tagged juvenile Marbled Murrelet. Journal of Field Ornithology 68(3):421-425.

Mauser, D. M., and R. L. Jarvis. 1991. Attaching radio transmitters to one-day-old Mallard ducklings. Journal of Wildlife Management 55(3):488-491.

Nelson, S. K. 1997. Marbled murrelet (Brachyramphus marmoratus). in A. Poole and F. Gill eds. The Birds of North America, No. 276. The Academy of Natural Sciences, Philadelphia, Pennsylvania and The American Ornithologists’ Union, Washington, D. C.

Nelson, S. K., D. L. Whitworth, and S. H. Newman. 1999. Pilot study of the breeding and foraging distribution of marbled murrelets in northern Southeast Alaska, May-July 1998. Oregon Cooperative Wildlife Research Unit, Corvallis, OR; Wildlife Health Center, School of Veterinary Medicine, University of California, Davis.

Newman, S. H., J. Y. Takekawa, D. L. Whitworth and E. E. Burkett. In press. Subcutaneous anchor attachment increases retention of radio transmitters on seabirds: Xantus’ and marbled murrelets. Journal of Field Ornithology.

Newman, S. H., D. W. Anderson, M. H. Ziccardi, J. G. Trupkiewicz. F. S. Tseng, M. M. Christopher and J. G. Zinkl. 1999. Experimental release of oil-spill rehabilitated American coots (Fulica americana): Effects on health and blood parameters. In press. Environmental Pollution 107: 295-304.

Pietz, P. J., D. A. Brandt, G. L. Krapu, and D. A. Buhl. 1995. Modified transmitter attachment method for adult ducks. Journal of Field Ornithology 66 (3): 408-417.

Quinlan, S. E., and J. H. Hughes. 1992. Techniques for radio tagging of marbled murrelets. Pages 117-121 in H. R. Carter and M. L. Morrison (eds). Status and conservation of the Marbled Murrelet in North America. Proceedings of the Western Foundation of Vertebrate Zoology 5.

Ralph, C. J., S. L. Miller, and B. O’Donnell. 1992. Capture and monitoring of foraging and breeding of the Marbled Murrelet in California during 1990. Interim Report: Technical Report 92.07, California Department of Fish and Game, Sacramento, California. 16 pp.

Rotella, J. J. and J. T. Ratti. 1990. Use of methoxyflurane to reduce nest abandonment of mallards. Journal of Wildlife Management 54:627-628.

Rotella, J. J., D. W. Howerter, T. P. Sankowski and J. H. Devries. 1993. Nesting effort by wild mallards with 3 types of radio transmitters. Journal of Wildlife Management 5 7:690-695.

Wanless, S. M., M. P. Harris, and J. A. Morris. 1985. Radio-monitoring as a method for estimating time budgets of Guillemots Uria aalge. Bird Study 32: 170-175.

Wanless, S. M., M. P. Harris, and J. A. Morris. 1988a. The effect of radio transmitters on the behavior of Common Murres and Razorbills during chick rearing. Condor 90: 8 16-823.

Wanless, S. M., J. A. Morris, and M. P. Harris. 1988b. Diving behavior of guillemot Uria aalge, puffin Fratercula arctica and razorbill A/ca torda as shown by radio telemetry. Journal Zoological Society of London 216: 73-8 1.

Whitworth, D. L., J. Y. Takekawa, H. R. Carter, S. H. Newman, T. W. Keeney, and P. R. Kelly. 1997. Foraging distribution and dispersal of Xantus’ murrelets in the Southern California

Progress with the Development of the
GPS-Based Flight Recorder for Homing Pigeons

Karen von Hünerbein, Hajo Hamann, and Wolfgang Wiltschko

Institute of Zoology, J.W.-Goethe Universität Frankfurt, Siesmayerstr. 70,
60323 Frankfurt am Main, Germany


In order to track the flight paths of homing pigeons, we are developing a Global Positioning System (GPS) flight recorder. The device consists of a GPS receiver board, a datalogger, an antenna, the power supply, and a casing. The first prototype (100 g) was tested and worked well 50% of the time. There were problems with passive antennas and timing. A lighter device is being built on the basis of a hybrid GPS board.


Pigeons are able to return home from places where they have never been before. This homing and orientation behavior has already been intensively studied by Keeton (1974), Schmidt-Koenig (1991), Papi (1991), and Wiltschko and Wiltschko (1994). The navigation system is thought to consist of map and compass mechanisms. This model was developed by Kramer (1953) and is still considered to be valid. It has been shown that pigeons have two types of compasses, a sun compass (Schmidt-Koenig 1991) and a magnetic inclination compass (Wiltschko and Wiltschko 1988). The map is thought to consist of gradients of physical factors that gradually change over geographic distances.

The traditional method of studying the orientation behavior of pigeons has been by watching them with binoculars and by determining their direction when they disappear (vanishing bearing). With this method, only the first few miles of the flight path of a pigeon can be observed. While it was possible to establish the two compasses with this method, it has brought us little insight into the nature of the map. Questions still to be answered are, at which point of their flight do the pigeons correct initial deviations from the homing direction, and, which factors can they use to determine where they are compared to their home loft? Such questions could be answered with a method of tracking the complete flight path from the release site to the home loft.

There have been several attempts to track pigeons: by airplane (Michener and Walcott 1967, Wagner 1974), by conventional radio tracking (Wolff 1997) and by an Italian compass route recorder (Bramanti et al. 1988). The automatic route recorder stores the bearings of flight only, but allows a reconstruction of a flight path with the assumption of a constant speed. All these systems have major drawbacks, such as high cost, limited range, and/or lack of precision (Hünerbein et al. 1997, Hünerbein and Müller 1998).

We strongly felt that a flight recorder with a high precision of position and timing was needed. It would have the following advantages:

It would be possible to observe the flight over longer periods of time, not just for a few minutes after takeoff.

Data analysis with self-made measures would be possible (e.g. index for directness of homing, and accuracy of homing direction).

Complete tracks or parts of tracks could be compared.

The data from flight recorders would be on a ratio scale (absolute numerical values), so that experimental groups of animals could be compared with statistically standard methods.

The flight recorder should measure the position of the pigeon during its flight and record these positions on a datalogger. At the end of the pigeon’s flight, there should be a download of the position data to a computer. Then the flight track or parts of it should be calculated or displayed on a map. The data should be in a convertible format allowing processing by a PC.


Why Did We Choose GPS?

After comparing all major technical navigational system (Hünerbein et al. 1997), we decided that the Global Positioning System (GPS) (Schrödter 1994) would best fit the following requirements: 1) recording of position, speed, and heading with good accuracy (100-300 m); 2) small dimensions (70 mm x 40 mm x 30 mm); 3) low weight, about 30 g (with antenna and power supply); 4) sampling rate 5 sec to 5 min; 5) operation time 3-12 h; and 6) data protection in case of power failure. GPS offers high accuracy of position, possible sampling rate of once per second, no transmissions in flight, and worldwide availability. Major drawbacks of GPS for this application are size and weight of components, power consumption, high cost of devices, possible loss of the device and of all data if the animal does not return to the loft, and influence of the military on precision.

System Components

Our device consists of a GPS antenna, GPS receiver board, datalogger, power supply, casing, and possibly some part that would allow recovery of a lost device (Figure 1).

Figure 1. System components of GPS-based flight recorder.

There are two basic types of GPS antennas: patch antennas (which look like a metal square with cut corners) and quadrifilar helix antennas (with four wires wound around an axis). One can also distinguish between passive antennas (without an amplifying circuit) and active antennas (with an amplifying circuit). The active ones have better signal reception; the passive ones use less power and save the weight of the circuit.

Currently we are working with the following antennas: WiSi Patch antenna, passive, 50 mm x 50 mm, 13.5 g (used for the first prototype); Murata patch antenna, active, 25 mm x 25 mm, 11.8 g (including amplifier); and Navsymm Quadrifilar helix antenna, active or passive, 30 mm x 5 mm + size of amplifier, 9.5 g (without amplifier).

We have tried to find the smallest and lightest GPS boards available. In the past few years, several boards with a weight of less than an ounce have come to the market. We now are working with the following boards: Jupiter OEM Board by Rockwell, 43 mm x 75 mm, 25 g, 5V, 200 mA; and, a hybrid board by a Swiss company, 30 mm x 30 mm, 6.2 g, 3.3 V, 150 mA.

The datalogger has been designed by Ralf D. Mueller, Department of Technical Computer Sciences, University of Frankfurt. It basically consists of an EEPROM and a PIC chip. The datalogger stores part of the NMEA messages; it can store up to 8,000 (32,000) positions depending on the size of the EEPROM. The sampling rate can be chosen by the user and is variable from 5 sec to once a day. The type of field to be stored can be chosen by the user. The datalogger is very light (3g +/- 1 g depending on the packaging) and uses very little power, 2 mA in active read/write cycles.

Mueller also developed an interface (datalogger to PC and vice versa) and a Windows front end for reconverting the data and for producing it in several formats so that it can be used with Fugawi map software.

The power supply has been the hardest problem to solve. The GPS boards used need a very high continuous current to be able to operate. At an operation time of three hours, the theoretical characteristics of the ideal battery would be: 5V, 750 mAh capacity, 10 g, and 250 mA of maximum continuous current. The battery closest to this requirement is the Panasonic CR2, with 3V, 1,000 mAh capacity, 11 g, 250 mA of maximum continuous current.

We intend to use primary lithium cells with a voltage of 3.6 V (probably AA cells by SAFT), because the voltage suits the 3.3 V GPS board we have. There is no alternative to lithium cells because they offer the highest capacity in relation to the weight. We do have the advantage of not needing a long operating time, as the pigeon experiments only take a few hours.

We have been using a homemade protective casing of epoxy (3 mm thick). We are considering embedding the next prototype into thin plastic material, either by fitting a flexible foil precisely to the surface of the electronics or by using fluid material that hardens.

First Prototype

Last year, a prototype recorder system was developed consisting of a WiSi patch antenna, Jupiter board, datalogger, and epoxy casing (F. Jöst and S. Wolff, University of Darmstadt, Germany, personal communication). This prototype yielded valid data (Table 1) on several outdoor tests, including several car drives around Darmstadt.

We did 76 tests with the prototype (with unlimited power) and found that it worked well, storing a large amount of valid data in 50% of the cases. However, in several instances no new position data was stored. We found several possible causes for failure: lack of antenna strength, either hardware error or programming error in the datalogger resulting in bad timing, and a short circuit caused by a wire close to a pin of the chip in the datalogger.

Second Prototype

As a consequence of the experience with the first prototype, we are now using an active antenna in the second prototype. We are also reducing the amount of wires in the datalogger. The second prototype consists of the Swiss hybrid board, a Murata patch antenna, the datalogger by Mueller with some modifications, and a suitable lithium battery. Testing is planned for summer of 1999.

Table 1. Example of good data stored by the datalogger and reconverted by the Windows front end in a static measurement on a parked car.

Type of

Message time latitude longitude







Any device used should have as small an impact as possible on the animal’s natural behavior; otherwise, one measures the animal’s reaction to some property of the device. The first major problem is weight. It is very exhausting for pigeons to carry a load of 30 g, which is between 6% and 10% of their body weight. Any device carried on the back of a bird will also disturb its aerodynamic flight properties. Studies by Gessaman and Nagy (1988) have shown that pigeons can be greatly influenced by transmitter loads with a weight of 2.5% and 5% of their body weight. They have shown that the birds slow down by 15-28% on 90 km flights, and that their CO2 production increases by 41-50%. Since we have no choice but to work at the upper limit of the pigeon’s load carrying capacity, we shall start by measuring short flight paths of 10-20 km. Still, there needs to be further improvement at this point.

The second problem is the magnetic field created by the devices during operation. There is evidence that the pigeons use the earth’s magnetic field for navigation. All electrical devices using power create a magnetic field during operation. We will measure the strength of this field and try to compensate for it by rearranging the components to minimize this effect and shield the worst parts with thin aluminium foil.

GPS is also an expensive system. Some pigeons get lost during homing, and some biologists cannot afford to buy new devices every year. We still need to find some way of recovering lost devices.

In conclusion, the system we are presenting for homing pigeons solves many of the problems of tracking systems currently in use. It yields sufficiently accurate position fixes with a variable sampling rate, stores data in a way that it is not lost when the power supply fails, and the expenditure for the experiment itself is quite low. Tracking will be limited to a few hours, though, and 30 g is still heavy enough to influence the flight behavior of the birds. Our device will greatly increase the number of species that can be tracked with GPS. For the first time, it will be possible to use GPS on birds.


This project has been financially supported by the DFG. The German Society of Telemetry (Arbeitskreis Telemetrie) enabled the first author to come to the 15th International Symposium on Biotelemetry. Partners in building and testing the hardware and software were Frank Joest and Dr. Stefan Wolff of the University of Darmstadt, Germany; Rainer Hartmann, Prof. Klinke, Ralf D. Müller, and B. Klauer from the University of Frankfurt, Germany; and Dr. Wolfgang Lechner, Rueter Nachrichtentechnik, Minden, Germany. For supplying samples of components at no cost and for being very helpful with information we thank the following companies: WiSi, Rockwell, Murata, Bosch, and Varta. For encouragement, interest, and cooperation we thank God, Dr. Sandra Woolley, University of Birmingham; Willi Schmid, Prof. Lipp, University Irchel, Zürich, Switzerland; and Mohamed al Bowardi, NARC, United Arabis Emirates.


Bramanti, M., L. Dall' Antonia and F. Papi. 1988. A new technique to monitor the tracks of birds in flight. Monitore zool. ital. (N.S.)21:177-178.

Gessaman, J., and K Nagy. 1988. Transmitter loads affect the flight speed and metabolism of homing pigeons. The Condor 90:662-668.

Hünerbein, K. and R. Müller. 1998. Development of a GPS flight recorder for birds. Pages 86-98 in Conference Proceedings of the European Telemetry Conference 98 (ed: Arbeitskreis Telemetrie e.V.) May 4th-7th. Braunschweig.

Hünerbein, K., W. Wiltschko, R. Müller and B. Klauer. 1997. A GPS-based flight recorder for homing pigeons. Volume II, pages 561-570 in Conference Proceedings of the GNSS 97 (ed.: Deutsche Gesellschaft für Ortung und Navigation) April 21st-25th,1997. Bonn.

Keeton, W. T. 1974. The orientational and navigational basis of homing in birds. Pages 47-132 in D. Lehrmann, R Hinde and E. Shaw (eds.), Advances in the Study of Behavior. Academic Press, New York, San Francisco, London.

Kramer, G. 1953. Wird die sonne bei der heimfindeorientierung verwertet? J. Orn. 94:201-219.

Michener, M., and C.Walcott. 1967. Homing of single pigeons - Analysis of Tracks. J. Exp. Biol. 47:99-131.

Papi, F. 1991. Olfactory navigation. Pages 52-85 in P. Berthold (ed.), Orientation in Birds. Birkhäuser Verlag, Basel.

Schmidt-Koenig, K., and J.U. Ganzhorn. 1991. On the problem of bird navigation. Pages 261-283 in P.P.G. Bateson and P.H. Klopfer (eds.), Perspectives in Ethology.

Schrödter, F. 1994. GPS-Satelliten-Navigation: Technik, System, Geräte, Funktionen und praktischer Einsatz. Poing; Franzis-Verlag.

Wagner, G. 1970. Verfolgung von Brieftauben im Helikopter. Revue suisse Zool. 77(2):39-60.

Wolff, C. 1997. Influence of artificial magnetic information on the correction of false-oriented clockshifted pigeons. In Conference Proceedings of RIN 97, Orientation and Navigation — Birds, Humans and Other Animals April 21st-23rd, 1997 (ed: Royal Institute of Navigation): 8-1-8-10.

Wiltschko, W., and R.Wiltschko. 1988. Magnetic orientation in birds. Pages 67-121 in Richard F. Johnston (ed.), Current Ornithology. New York, Plenum Publishing Corporation.

Wiltschko, W., and R. Wiltschko. 1994. Avian orientation: Multiple sensory cues and the advantage of redundancy. Pages 95-119 in M.N.O. Davies and P.R. Green (eds.), Perception and Motor Control in Birds. Berlin, Heidelberg, Springer Verlag, 1994.

Improved Automatic Telemetry Receiving System for Data Acquisition of Radio Locations and Pulse-Coded Signals Based on Doppler Effect

Franz Schober

Research Institute of Wildlife Ecology, University of Veterinary Medicine, Vienna, Austria

Ralf Boegel and Manfred Eschbaumer

Berchtesgaden National Park, Berchtesgaden, Germany


In 1992, a Doppler-based automatic direction finding and location system for wildlife applications was introduced. Recent improvements focused on antenna and circuit design and on new control software. The capacity of the system covers up to 30 receivers (remote-controlled by a central PC) and 100 RF channels. Location, pulse width, and interval are collected every 12 minutes. Bearing accuracy varies with topography and reaches up to 1 degree.


While various telemetry systems with automatic data collection of pulse-coded parameters are commercially available, only a few attempts have been made to design automatic direction finding (ADF) and location systems. Satellite telemetry may be adequate for migrating animals; however, data quality and temporal resolution are not sufficient for most territorial animals. Global Positioning System (GPS) telemetry systems offer superior data quality but are restricted to large animals and short battery life or low data rates. Several ground-based ADF systems have been developed to guarantee continuous and unbiased data sampling. Former designs include: 1) mechanically rotating Yagi antennas with maximum signal strength or 0-peak detection (Cochran et al. 1965, Boegel 1988), 2) time-of-arrival (TOA) systems (Lemnell et al. 1983), and 3) Doppler direction finders (Den Boer 1991, Angerbjörn and Becker 1992, Olesen and Kristiansen 2000). Advantages and drawbacks of various designs have been discussed (Schober et al. 1993). While the early designs of Doppler systems required long pulses (>100 msec) and/or high transmitter output power (>1 W), the most important design requirement of the system described in this paper was full compatibility with commonly used, short-pulsed, low-power wildlife transmitters in order to avoid increased transmitter weight and/or reduced lifespan.


Doppler direction finders operate with circular arrays of omnidirectional antennas. A rotating antenna dipole is simulated by switching from one element to the other electronically. When receiving carrier frequency pulses, the resulting phase modulation in the received signal is used to determine the direction of the incident wave front (Burchard 1989, 1992). To eliminate phase errors caused by the Doppler receiver and achieve valid bearings, left and right turn readings have to be averaged. Basic specifications of the prototype design introduced in 1992 (Schober et al. 1993) have not been changed: 2m-band (130-170 MHz), channel separation 20 kHz, sensitivity -130 dBm, dynamic range 100 dB, selectivity "3 kHz (-6 dB), Doppler antenna array (Ø 1.5 m) with 8 dipoles, rotation frequency 500 Hz (1 bearing result within 8 ms), measurements of bearing angle, pulse width, and interval. The Doppler receivers are not only capable of direction finding, but also of decoding pulse duration and pulse interval-modulated signals (PDM/PIM) from transmitters with various sensors. Recent modifications concern mechanical improvements of the antenna array, the implementation of a microcontroller into the receiver, and a pulse filter that greatly improves the robustness of pulse measurements against noise. Furthermore, completely new control software has been developed.

A complete system consists of 3-30 Doppler receivers, which can be placed at optimal locations due to solar power supply, radio data links, and an optional repeater (Figure 1). The system control is handled by a central PC or a notebook computer, which can also be used as mobile data readout unit offering access to logged data during fieldwork.

All receiver settings can be controlled via a RS-232 interface from the central PC and address: channel selection (0-99), frequency fine tune (­3.75 kHz to +3.75 kHz in 1.25 kHz steps), trigger level (2 / 5 / 7 dB and adjustable), input attenuator setting (-30 / -20 / -10 / 0 dB), measurement mode (continuous rotation / pulse-triggered bearing / pulse measurements with selectable dipoles), averaging of bearing results (2 / 4 / 8), pulse filter (enabled / disabled). Measurement capabilities including reliability parameters were extended in the present design and the following details sent back to the control unit: bearing angle (0.1 deg resolution), pulse width (1 msec resolution), pulse interval (1 msec resolution), signal-to-noise ratio (0.5 dB resolution), frequency offset (40 Hz resolution), battery voltage (0.1 V resolution), antenna check (o.k. or failure), receiver status (data new / data not actual / pulse failure / receiver time-out).

Figure 1. Block diagram of the Doppler direction finding and location system.

The software package is a Windows 95/98 application and is divided into several modules and three basic operation modes: 1) "manual" mode (see receiver control window), 2) automatic data sampling, 3) playback mode (visualization of recorded data).

System Configuration and Initialization

System configuration and initialization is divided according to the specific function and stored in different files:

All receivers with coordinates, antenna zero directions, and colors for plotting (adfsys.ini-file).

All beacons with coordinates and colors for plotting (adfsys.ini-file).

All channels with alias names, colors for plotting, and target areas (adfsys.ini-file).

Back-environment map definitions for graphic display and plotting options (adfsys.ini-file).

Default settings for statistic data aggregation (adfsys.ini-file).

Sensor types for each transmitter with modulation type, regression parameters, sensor units, definitions of valid pulse durations and intervals for all channels and sensors (sensors.ini-file).

Receiver Control and Automatic Data Sampling

A receiver control window is assigned to each specific receiver. This window summarizes all actual receiver settings and measurement results (as previously described). In this way, manual control of all receiver settings is possible as well as the viewing of actual receiver settings during automatic data sampling. Automatic data collection is controlled by a special routine. Through this routine, the user specifies all channels to be scanned by specific or all receivers and the sampling rates (up to 12 data sets per min). This module also handles a dynamic adjustment of all receiver settings according to the signal quality. Due to the enormous influence of wave propagation conditions on signal quality, these adjustments can be both channel and receiver specific. The characteristics of the procedure for the dynamic adjustment of receiver settings are stored in a separate file (scanning.ini) and can be modified by the user according to the specific requirements of the study. A protocol of the data communication (commands to and answers from the receivers) is listed in a receiver diagnostics window.

Data Checks, Conversion, and Display

Data plausibility checks are of utmost importance in automatic data sampling devices and are realized on various levels: 1) signals that violate channel specific pulse duration and interval criteria are ignored, 2) fixes that violate target areas are excluded from further analysis, 3) data are aggregated in order to minimize corresponding errors and to avoid autocorrelated results. This aggregation of data occurs according to fixed (within the same intervals as selected for data sampling) or dynamic time intervals, which are derived from the statistic variance of the data. Thus, the dynamic aggregation procedure is self-adaptable and adjusts the time window according to the replicability of the results and thus to the activities of the tracked animal. Within the aggregation time window, animal locations are calculated as medians of X/Y coordinates of all pairwise bearing intersections and are plotted continuously in channel-specific colors on a back-environment map (Figure 2).

An indication of the location accuracy is derived from the standard deviation of all intersections and is expressed as a (user-defined) multiple of SDX and SDY. All results derived from PDM/PIM-coded signals are listed in real units in a receiver overview window together with raw measurements, medians, and standard deviations (Figure 2). A data conversion routine handles the transformation of PDM/PIM signals to real units according to classifications or regression functions specified in the sensors.ini-file. This data conversion routine covers most patterns of data encoding from simple classifications like "heads up" or "heads down", over linear regression functions up to the third order such as a temperature transformation to complex transformations like the calculation of air pressure levels which require temperature as an additional variable (two-dimensional regression function). Using reference transmitters at known altitudes with air pressure and temperature sensors, a barometric calculation of altitudes from air pressure differences is also possible and extends the capabilities of the ADF system to three-dimensional fixes.

Figure 2. Automatic data collection with graphic data display and receiver diagnostics window.

Playback Module

While all features described above refer to data acquisition, aggregation, viewing, and plotting of results in real time (online operation), this module handles data viewing and plotting of stored files in various scales (playback mode). For this purpose, two scroll bars offer a stepwise access to individual data sets, bearings, or locations (Figure 2). While one scroll bar skips stepwise from one location to the next (plotting and summarizing of aggregated data), the other bar scrolls stepwise through the aggregation window, providing access to all measurements of the individual receivers. All characteristics of data plausibility checks, aggregation, and display are identical to the online mode and also apply to the playback mode. However, editing of individual measurements may influence the variance of the data and thus can result in changes of the aggregation time window and the aggregated results. Editing options cover a deletion of data sets, an editing of pulse duration and interval results, or enable a shift of animal location by mouse click. An import of back-environment maps is possible from a DXF- (AutoCAD) or Bitmap-file. Thus, existing data sets from GIS layers can be used and imported into the software package. For further analysis, logged telemetry data can also be exported to other software packages in various formats (home range analysis, statistics, databases or spread sheets). While a GIS-export aggregates data in the same way as previously described, other export options cover the raw data of individual receivers and may include transformation of PDM/PIM results to real units. To reject implausible data from the exported file, several filter options are available. An extended ASCII option covers the implementation of individual receiver settings to the exported data file. An option for creating a summary statistic of a logged data file is also available.

Study Area

The system was established at the northern slope of the European Alps in Berchtesgaden Biosphere Reserve. The study area is a complex, alpine landscape covering an altitude range of 500 m to 2700 m a. SL.


Figure 3 shows bearing performance for a 5 mW transmitter, 4 different receiver gain settings (signal attenuation) and short (0.4 km) and long (9.5 km) transmitter distances. Bearing precision under optimal conditions is better than 0.5 degrees (SD = 0.34 deg). However, maximum bearing accuracy is limited by symmetry failures of the antenna array and lies within 2.5 degrees at maximum (Schober et al. 1993). While bearing precision decreases with distance (SD for optimum gain setting: 2.3 deg), bearing accuracy remains in the same magnitude (bias of the median for optimum gain setting: < 2.5 deg). As long as signal-to-noise ratios are not critical, signal attenuation does not have a strong influence on bearing performance. In contrast to these rather robust findings, bearing performance is highly influenced by topography and transmitter antenna polarization (Figure 4). While bearing precision remains in the same magnitude (2.8 < SD < 5.6 deg; the high SD for the 90° radial antenna orientation is an outlier caused by critical signal-to-noise ratios), accuracy can be extremely biased by multipath effects (simultaneous reception of the direct and one or more reflected signals). If antenna polarization is constant, multipath can be detected by an increasing shift in bearing angles that is correlated to signal attenuation. If the signal-to-noise ratio of the direct or reflected wave becomes too small, bearing angles skip exclusively to the direction of the direct or reflected (-30 db) wave front (Figure 4a). The polarization of the transmitting antenna greatly influences the chance for multipath effects (see Figure 4b). In the tested configuration, radial antenna polarizations had a very small influence on bearing performance while tangential polarizations resulted in bearing errors of # 30 degrees. However, this may vary greatly with the geometric configuration between transmitter, Doppler receiver and topography.

Figure 3. Bearing performance for optimal transmitter locations (line of sight, transmitter height 5 m), a: 0.4 km distance, b: 9.5 km distance.


The presented system has some striking advantages, including very high data rates on a 24-hours-a-day basis, potential for studying multiple species, data collection of locations and ecophysiologic data with a high degree of standardization, optimum bearing synchronization, expandability of the system to a network of tracking stations, and low costs for data collection and system maintenance. However, there are some obvious drawbacks, including the high cost of system set-up and highly variable data quality. While bearing performance under optimal conditions is superior, the influence of the topography on the accuracy of bearings and locations is enormous. Therefore, multipath effects can result in complete outliers. As antenna orientation of radio-tagged animals determines the chance for multipath effects in complex terrain, average data quality in wildlife applications is hardly predictable. Such effects cannot be detected on the level of bearing angles, and data plausibility checks are therefore restricted to the location-finding routine of the software. Furthermore, the results show that standard deviation is not a useful indicator for bearing accuracy, and thus location errors cannot be calculated accurately from bearing precision. In general, best bearing accuracies are achieved at medium gain settings where signal-to-noise ratios are maximum. This can be used to optimize receiver settings and data quality. As location errors increase with distance from the bearing stations, average data quality can be greatly improved by establishing a dense network of Doppler receivers and, if possible, by omitting all collected data which are not plausible results of simultaneous bearings from at least 3 bearing stations. In general, average performance of the system will be a function of landscape complexity and cover as well the behavior and morphology of the tagged animals (e.g. preferred habitats, possibilities of vertical antenna fixation). In this context, the suitability of the system will be best for big, soaring birds; terrestrial animals will deliver intermediate results and subterranean or aquatic species will be most critical.

Figure 4. Bearing performance for a multipath situation (a) and various transmitter antenna orientations (b) (a: distance 5.1 km, transmitter height 15 m, b: distance 5.1 km, transmitter height 3 m).


Angerbjörn, A. and D. Becker. 1992. An automatic location system. Pages 68-75 in Priede, I. G., and S. M. Swift (eds.), Wildlife Telemetry, Remote Monitoring and Tracking of Animals. Proc. 4th Europ. Conf. on Wildl. Telemetry. Ellis Horwood, Chichester, U.K.

Boegel, R. 1988. Automatic radio tracking. Pages 115-124 in Acte du Colloque International: Suivi des Vertèbres Terrestres par Radiotélémétrie. Monaco, France, 12.-13. Dezember 1988. Eds.: Parc National du Mercantour / CLS-ARGOS / IBM France.

Burchard, D. 1989. Direction finding in wildlife research by Doppler effect. Pages 169-177 in Amlaner, C. J. Jr. (ed.), Biotelemetry X. University of Arkansas Press, Fayetteville, U.S.A.

Burchard, D. 1992: Doppler-Peiler mit verbesserten Eigenschaften. UKW-Berichte 2/92:66-80.

Cochran, W. W., D. W. Warner, J. R. Tester and V. B. Kuechle. 1965. Automatic radio tracking system for monitoring animal movements. Bioscience 15:98-100.

Den Boer, M. H. 1991. An automatic Doppler radio tracking system: application, performance, and terrain features. Pages 167-171 in Uchiyama, A., and C. J. Amlaner Jr. (eds.), Biotelemetry XI. Waseda University Press, Tokyo, Japan.

Lemnell, P. A., G. Johnsson, H. Helmersson, O. Holmstrand and L. Norling. 1983. An automatic radio-telemetry system for position determination and data acquisition. Pages 31-46 in Pincock, D. G. (ed.), Proc. 4th Int. Conf. on Wildl. Biotel., Halifax, Nova Scotia, Canada.

Olesen, C. R. and N. U. Kristiansen. 2000. Design and evaluation of an automatic localisation system based on Doppler antennas used in wildlife studies. In Eiler, J. (ed.), Biotelemetry XV, Proceedings of the 15th International Symposium on Biotelemetry (this volume).

Schober, F., R. Boegel, W. M. Bugnar, D. Burchard, G. Fluch and N. Rhode. 1993. Automatic direction finding and location system based on Doppler effect. Pages 327-336 in Mancini, P., S. Fioretti, C. Cristalli, and R. Bedini (eds.), Biotelemetry XII. Editrice Universitaria Litografia Felici, Pisa, Italy.

Testing Performance of the Birth Archive-Tag on Black Bears: An Increase or a Decrease?

Merav Ben-David

Institute of Arctic Biology, University of Alaska Fairbanks,
Fairbanks, Alaska 99775, USA

Sheldon L. Struthers

Advanced Telemetry Systems, Isanti, Minnesota 55040, USA

H. David Moll

College of Veterinary Medicine, Virginia Polytechnic Institute and
State University, Blacksburg, Virginia 24061, USA

Michael R. Vaughan

Virginia Cooperative Fish and Wildlife Research Unit, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061, USA


To determine timing of parturition for free-ranging mammals, we developed a birth archive-tag that records changes in pressure in the vaginal wall during delivery. We tested the performance of the tag on two wild-caught black bears. Results indicated that the tags registered a significant decrease in pressure within 48 hours of appearance of cubs. Thus, the tag can serve as a useful tool for determining timing of parturition in mammals.


Timing of breeding is one of several adaptive properties evolved to maximize reproductive output (Williams 1966, Lack 1968). Variation in timing of breeding between individuals, and within the same individuals between breeding attempts in different years, reflect adaptive responses to varying individual circumstances (Drent and Daan 1980, Meijer et al. 1990). Determining timing of parturition for most species of free-ranging mammals poses a major difficulty to wildlife researchers because parturition usually occurs in a concealed location.

Studies on humans indicated that increases in levels of vaginal pressure occurred during childbirth as well as in relation to stress levels (Samples et al. 1988, Dougherty et al. 1991). Similar changes in pressure are likely to occur during parturition in other mammals, thus creating a reliable marker for timing of parturition even in elusive species. Therefore, we developed an implantable, birth archive-tag that is designed to record changes in pressure in the vaginal wall during delivery. In this study we report results from testing the performance of the tag on two wild-caught female black bears (Ursus americanus) held in captivity at Virginia Polytechnic Institute and State University. We hypothesized that pressure will significantly increase during parturition and will provide a reliable measure of timing of delivery.


The Tag

The birth archive-tag is composed of a non-volatile memory chip (8,160 byte capacity), a pressure transducer (BT-300, 0-25psi, 25mV sensitivity), an analog to digital converter, and a microcontroller. All these components with a battery (3.6 volts) are encapsulated in electrical waterproof, biologically inert resin (specific gravity 1.12). The sensor measures pressure in dimensionless units, which are calibrated in an environmental chamber to psi (pounds per square inch) or Torr (mm Hg) (Table 1). The sampling frequency of pressure and data storage is programmable by the operator prior to the sterilization or implantation of the tag, using an external-reader interface-circuit and communication software between a personal computer and the tag. Communication from tag to interface is 4096 Baud, interface to tag 1024 is Baud, and from interface reader to PC is 9600 Baud. Two stainless steel communication contacts (ground and I/O) are used to send information between the tag and the reader and are embedded in the resin encapsulating the tag. External tag dimensions are 16 mm x 15 mm x 42 mm (Figure 1).

Once sampling schedule is set and the tag begins logging, the sampled pressure is stored by the microprocessor and maintained in memory even in case of battery failure. The tag recognizes the shift from recording data to offloading data by detecting the high voltage level from the reader through the I/O pin. Further details regarding the technical specification of the tag can be obtained from S. L. Struthers at

For testing the tag performance in our experiment, sampling frequency was set to 5 sec because deliveries in bears can be in the order of a few seconds (M. R. Vaughan, personal observation). Data storage was set as the highest pressure recorded every 600 sec (10 min).

Table 1. Readings of the 2 prototype tags used in this experiment in relation to pressure (in pounds per square inch) as measured in an environmental chamber.

Pressure (psi)

Readings of Tag 100

Readings of Tag 101*































*Readings of Tag 101 drifted during the test.

@ 256 +35 = 291; 256+26 = 282.


Figure 1. The birth archive-tag during the removal surgery from a captive black bear at Virginia Polytechnic Institute and State University, on April 1, 1996.

The Bears

Two adult, wild-caught female black bears from an ongoing research on the reproductive biology of bears by the Virginia Cooperative Fish and Wildlife Research Unit (Bears 46 and Bear 47) were selected for the experiment. Bears were brought into captivity in late summer and were held individually in circular cages of 2.7-m radius. Each cage contained a wooden den and was lined with straw. Bears were taken off their feed in late December and entered hibernation shortly thereafter. Both bears were confirmed pregnant through repeated progesterone and ultrasound analyses, prior to tag implantation. Tags were surgically implanted into the vaginal wall of each bear on January 11, 1996, and were surgically removed on April 1, 1996. Dens were checked for the occurrence of cubs every other day beginning January 15, 1996. All methods used in this experiment were approved by Independent Animal Care and Use Committees at the University of Alaska Fairbanks, and Virginia Polytechnic Institute and State University.

Surgical Procedure

For the implanting surgery, animals were sedated to a surgical plane of anesthesia with a 2:1 mixture of ketamine:xylazine at a dosage of 4mg/kg ketamine and 2mg/kg xylazine administered using Telinject darts and a blowgun. An 8-cm by 5-cm area lateral to the vulva was clipped and routinely prepared for aseptic surgery. A fenestrated sterile surgery drape was placed on the planned incision site. A 3-cm skin incision was made just lateral to the vulva using a # 10 scalpel blade. Using scissors and blunt dissection, a tunnel was created 5-cm deep and just lateral to the vaginal mucosa. Care was taken not to penetrate the mucosa. The birth-archive tag (previously gas sterilized using Amprolene) was inserted into the tunnel. The skin incision was closed using 2-0 absorbable suture material in an interrupted pattern. A prophylactic dose of 4mg/kg-body weight LA200 (Oxytetracycline) was administered IM. The removal surgery followed a similar protocol (Figure 1).


Three cubs were observed with Bear 46 on January 21, 1996, and two cubs were observed with Bear 47 on February 11, 1996. The removal surgery was performed on April 1, 1996, to ensure that the mothers would not abandon the cubs. A veterinary inspection a week prior to tag removal has shown that the tags shifted outward from their original implantation site towards the vaginal lobes. Following the removal of the tags, data was downloaded to a PC and analyzed. Data was smoothed using a default kernel parameter from the point of stabilization post surgery to 2 days past known parturition day. Probability of change in pressure was calculated for a moving window of 60 minutes (S-Plus for Windows). Pressure readings are shown in Figures 2–4; both tags registered a significant decrease in pressure (P = 0.01) shortly before cubs were observed (Figure 3).

Figure 2. Pressure (dimensionless values) plotted against time sequence (intervals of 10 minutes) for the period surrounding implantation of Tag 100 in Bear 46. Note the large increase in pressure during surgery. Similar results were observed in Tag 101 that was implanted in Bear 47. Readings of pressure in Tag 101 drifted during the period between surgery and several days before parturition, similarly to the readings obtained in the environmental chamber.

These results indicated that within the 48 hours before the cubs were first observed, the tags registered a significant decrease in pressure, probably in response to vaginal dilation. Compared to other mammals, bear cubs are considerably smaller in size at birth (Robbins 1993) and therefore may not register an increase in pressure. While this observation is contrary to our hypothesis, the significant reduction in pressure in correspondence with the appearance of cubs suggests that the birth archive-tag can serve as a useful tool for determining timing of parturition in mammals. The technological advances since 1996 in increased efficiency of data storage and improvements in controlling microprocessor operations (thus increasing battery life) will assist in improving the performance of the birth archive-tag. Furthermore, a modification of the shape of the tag may assist in reducing the movement of the tag inside the vaginal wall and prevent the inconsistency in tag measurements. Further experimentation with other species of wildlife will also be necessary for refining the performance of the tag.


Figure 3. Pressure (dimensionless values) plotted against time sequence (10 minutes intervals) during the period surrounding delivery in Bear 46 (above), and Bear 47 (below).

Figure 4. Pressure (dimensionless values) plotted against time sequence (10 minutes intervals) during the period following delivery in Bear 46. The tag shifted outward and pressure changes were more pronounced thereafter. Similar phenomenon was observed in Bear 47, although the tag shifted outward immediately after delivery.


We wish to thank the numerous students who assisted in care and maintenance of the bears and helped in the surgeries. Our thanks are extended to E. and O. Dromi for help in programming the sampling frequency and data storage. The Alaska Cooperative Fish and Wildlife Research Unit provided travel funds for Merav Ben-David.


Dougherty, M. C., K. R. Bishop, R. A. Mooney, P. A. Gimotty and L. B. Landy. 1991. Variation in intravaginal pressure measurements. Nursing Research 40:282-285.

Drent, R. H. and S. Daan. 1980. The prudent parent: energetic adjustments in avian breeding. Ardea 68:225-252.

Lack, D. 1968. Natural selection and family size in the starling. Evolution 2:95-110.

Meijer, T., S. Daan and M. Hall. 1990. Family planning in the kestrel (falco tinnuculus): the proximate control of covariation of laying date and clutch size. Behaviour 114:83-116.

Robbins, C. T. 1993. Wildlife Feeding and Nutrition (4th ed). Academic Press, New York.

Samples, J. T., M. C. Dougherty, R. M. Abrams and C. D. Batich. 1988. The dynamic characteristics of the circumvaginal muscles. J. Obstetetric Gynecologic and Neonatal Nursing 17:194-201.

Williams, G. C. 1966. Adaptation and natural selection: a critique of some evolutionary thought. Princeton University Press. Princeton, New Jersey.

Use of Transmitter Darts for Immobilizing and Locating Wild Boars in Dense Forest Stands

Gunter Sodeikat and Klaus Pohlmeyer

Institute of Wildlife Research at the School of Veterinary Medicine Hannover, Bischofsholer Damm 15, D-30173 Hannover, Germany

Ernst Finkelmann

AWEK fL-electronic, Ackerstr. 21, 38179 Klein Schwülper, Germany


The Institute of Wildlife Research in cooperation with AWEK fL-electronic has developed an immobilization dart with an integrated radio transmitter. The dart transmitter is a modification of a Dist-Inject-Aluminum-Dart screwed to the back of a regular Dist-Inject dart. The transmitter is produced by using SWOs and other parts with high loadability. The dart’s flight is not significantly affected by adding the transmitter. The transmitter dart (weight 18 g, 150 MHz with whip antenna, or 433 MHz with an integrated antenna) has a range of at least 3/4 of a mile in dense vegetation with any radio-tracking receiver, and battery life is about 12 hours. This modified dart has many valuable applications in wildlife research.


In their fieldwork, biologists and veterinarians often have problems capturing wild animals for marking, treatment, or health-control. The use of immobilization darts (syringes) in forested areas is difficult, because the animals darted often escape into dense stands where it is difficult to find them. A barbed dart transmitter stays with the animal, but often biologists are unable to find the darted animal before the immobilization wears off.


The Institute of Wildlife Research in cooperation with AWEK fL-electronic has developed an immobilization dart with an integrated radio transmitter. The dart transmitter (150 MHz with whip antenna, or 433 MHz with integrated antenna) is a modification of a Dist-Inject-Aluminum-Dart (like a Cap Chur dart) screwed to the back of a regular Dist-Inject dart. The transmitter is produced by using SWOs and other parts with high loadability. Despite the weight of the transmitter (18 g), we have a high target-precision up to 30 m (CO2-rifle). Adding the transmitter does not significantly influence the dart’s flight, based on the result of tests of target accuracy and flight movements filmed with a high-speed video camera (1,000 frames/second) using a 3 ml Dist-Inject-Aluminum-Dart (40 g), at air pressure of 10 bar, distance 15 m, V target of 42 m/s, and E target of 35.3 Joule. The range of the transmitter dart with our receiver system (or with any radio tracking receiver) is at least 3/4 of a mile in dense vegetation, and battery life is about 12 hours. The transmitter batteries are easy to change. The transmitter darts are ready for use in Dist-Inject Aluminum-Darts (13 mm diameter). They are distributed by Peter Ott Ltd., Dist-Inject Equipments (P.O. Box 16, CH-Basel/Switzerland).


Presently biologists and veterinarians do not try to immobilize large species of wildlife that cannot be visually followed (by helicopter) after they are darted, for example, if they are in dense brushwood. This transmitter dart will now allow them to do this. These darts can be used in a variety of circumstances such as for problem wildlife in urban areas where darting from the ground is possible.

It is very necessary that researchers or anyone using immobilization darts recover all the darts that they shoot. When a dart misses the animal or falls out after injection, the transmitter dart is easy to recover. In such cases this modified dart is very valuable. The Institute of Wildlife Research has successfully used these darts to immobilize wild boars (Sus scrofa) in dense forest stands.

The Use of a Telemetric Egg for
Monitoring Bird Incubation

Karen Bauman, Tim Snyder, Cheryl Asa, and Mike Macek

Saint Louis Zoo, Forest Park, Saint Louis, Missouri 63110, USA


We sought to better understand incubation in waterfowl and eagles using a telemetric egg. While not a new idea, incubation has been studied in only six of 160 species of waterfowl and four of 286 species of raptors. The collection at the Saint Louis Zoo provides a unique opportunity to gain further knowledge about these taxa. Our results indicate that temperature and position data can be successfully collected using this method.


The reproductive success of any species depends on its ability to produce and raise young. The avian strategy of external incubation of embryos within eggs is a biologically risky one, since the eggs are subjected to environmental variability and are completely dependent on the adults making the proper behavioral choices. Many basic biological questions remain about the incubation microclimate, such as which components of the incubation process are vital to the successful hatching of young. Of the 42 species studied to date, temperature was the component of this microclimate that was investigated in all the studies, with findings suggesting the average incubation temperature is approximately 34 °C. There are many other aspects of the microclimate that have been less thoroughly studied such as the effect of humidity, affect of position in the nest, how the behavior choices by the incubating adult affect incubation, and even how often the eggs are turned. For example, it is accepted that eggs are turned in the nest, yet studies quantifying this phenomenon are few (Howey 1984, Koontz 1991, Gee 1995), and it is not known whether turning rate differs among species or how it might relate to hatchability. Many aviculturists believe that the size of the egg relative to the size of the yolk may affect how often the eggs must be turned. It has been suggested that the turning rate may not be as critical in eggs with larger yolks (Harvey 1990).

Eggs from non-domestic species artificially incubated in the later stages of development have a significantly better hatching rate than those artificially incubated for the complete period. It is unclear what occurs within this critical first third of incubation that affects the success rate. It is possible the female incubates her clutch differently during this time than in the later phases; whereas, current artificial incubation techniques hold temperature, humidity, and turning rate constant throughout incubation. A better understanding of the biological factors that occur during this phase of incubation are essential to improving the current protocols and would significantly contribute to the propagation of endangered birds.

The earliest study of nest incubation dynamics was in 1907 when Eycleshymer published his paper entitled "Some observations and experiments on the natural and artificial incubation of the egg of the common fowl" in the Biological Bulletin. To date, incubation temperature has been studied in 42 avian species, with the majority of this work focused on passerine and Arctic birds. Huggins (1941) surveyed 37 species in 11 orders using the methodology of inserting a thermister wire in the space between the shell and the embryo through a small hole in the eggshell, the wire then extended from the nest into a blind. Data were collected using a potentiometer, galvanometer, and a strip chart recorder. Many other investigators (Drent 1970, Norton 1972, Derksen 1977, Walsberg 1978, Burger 1979, Snelling 1972, and Spellerburg 1969) also worked during this time using the thermister design described in Huggins (1941). In 1974, Varney designed the first telemetric egg, built within the hollowed shell of a real egg. It was equipped with two temperature sensors and a light sensor. Data was transmitted to an antenna at a receiving station approximately 40 m from the nest, pulses converted and recorded on a strip chart. The development of the telemetric egg was an important step in studying incubation, as it is less invasive; because it is used remotely, it reduces possible effects of disturbance from researchers near the nest area. It is also not possible to collect data on parameters other than temperature using a thermister since the wire lead potentially interferes with natural turning of the egg and other phenomena. In subsequent years, the technology has improved and has been incorporated into recent research. Schwartz (1977) added surface temperature sensors and a subcarrier oscillator to reduce effects of noise. He also filled the egg with paraffin to better emulate heat transfer from the surface to the core. In 1977, Howey greatly improved the design of the telemetric egg, adding temperature probes at six points, relative humidity, light, and position sensors, all packaged inside a fiberglass shell with the shape and weight of the real egg. The transmitters could send signals up to two miles and data were recorded directly into a decoder and saved to a computer-readable magnetic tape. Koontz (1991) and Gee (1995) had a similar egg design to that of Howey, but were able transfer signal from the receiver directly into a computer. Both also developed programs for objectively eliminating "bad" or erroneous data points.

Incubation has been studied in only six of 160 species of waterfowl and 4 of 286 species of raptors. The size and the depth of the bird collection at the Saint Louis Zoo provides a unique opportunity to further the knowledge of both the basic and applied biology of these taxa.


A pilot study to evaluate the ability of a telemetric egg to record core temperature and position was conducted at the Saint Louis Zoo between July and November of 1998. Two species, the spotted whistling duck (Dendrocygna guttata) and the Bateleur eagle (Terathopius ecadatus), with dissimilar incubation patterns, were used. The duck produces large clutches of medium-sized eggs laid in a nest box located outdoors in mid summer. In contrast, the eagle lays only one large egg and nests on an open, indoor platform in the fall.

The telemetric egg (Advanced Telemetry Systems, Isanti, Minnesota) consisted of a transmitter oriented along the long axis affixed to the inside surface of a commercially available plastic Easter egg. The transmitter contained a core temperature sensor located in the center of the transmitter. Temperature data were transmitted by varying the pulse rate of the main pulse (20 ms) in the manner standard to all temperature transmitters. Position of the egg was indicated by the location of two mercury switches, with the switches oriented in a cross formation so that each switch position represented an axis. Position data indicated what quadrant (Quadrant 1 - 4) of the egg was in the down position by the orientation of each of the switches, i.e., for Quadrant 1, the first switch was in the open position, while the second switch was closed. Quadrant 4 was indicated when both Switch and Switch 2 were closed and, therefore, it was the default quadrant when no signal was detected. Telemetric eggs were evaluated and calibration checked in an artificial incubator for one week prior to use by comparing results to the incubator thermometer and to temperature dataloggers (Onset Computers, Bourne, Massachusetts) also placed in the incubator as controls. Although the telemetric egg was considerably smaller than the real eagle egg and slightly larger than the real duck egg, all three females accepted and incubated it.

Study One

A pair of spotted whistling ducks was allowed to mate and lay two clutches naturally. Once the female began pulling down, the clutch was considered complete and one egg was removed from the nest and replaced with the telemetric egg, leaving a clutch of 9 real eggs and the telemetric egg. The removed egg was measured, weighed whole, then the yolk and white portions were weighed separately so that yolk mass could be compared to turning rate. The hen was allowed to incubate her clutch for the first eleven days. Then all of the eggs, including the telemetric egg, were removed from the nest and relocated to the artificial incubator (Humidaire Incubator Co., New Madison, Ohio) until 2 days prior to hatching, when they were moved into a still air hatcher (Brinsea Products Inc., Titusville, Florida). Removing the eggs from the hen stimulated her to lay another clutch and provided the opportunity to collect data on temperature and turning rate from the artificial incubator. The same protocol was followed for the second clutch (consisting of 10 real eggs and the telemetric egg), except that the eggs were left with the hen throughout the entire period of incubation. On the day of pipping, all eggs, including the telemetric egg, were moved to a still air hatcher (one chick had hatched immediately prior to this move) and data recorded until all eggs had hatched.

Study Two

Two pairs of Bateleur eagles mated naturally and each laid one egg. For management reasons, neither female was allowed to naturally incubate the egg for fear of breakage. Data on egg yolk mass were not collected due to the single-egg clutches. Within 24 hrs of laying, the egg was moved to the artificial incubator and replaced with the telemetric egg. Pair 1 was allowed to incubate the telemetric egg for 14 days and Pair 2 for 10 days before the egg was removed to stimulate the female to lay again. Both females laid a second clutch; however, the telemetric egg batteries had weakened by this time and could not be used. Three days of daytime time-lapse videography was recorded and used to validate egg-turning telemetry data.

In both studies, temperature and humidity dataloggers were placed within 5 feet of the nest to measure ambient conditions. Data from the telemetric eggs were automatically collected four times per minute and transferred continuously to a Data Collection Computer (Advanced Telemetry Systems, Isanti, Minnesota). The dataloggers recorded temperature and relative humidity values at five-minute intervals and data were stored in the logger. Loggers were downloaded into a computer at the end of each incubation period to avoid disturbing the nesting birds.


Study One

During Week 1, mean nest temperature (Tn) for Clutch 1 was 35.4°C, and 33.4°C for Clutch 2 (Table 1). During Week 2, mean Tn was 35.6°C for Clutch 1 and 32.6°C for Clutch 2. In Weeks 3 and 4, when eggs from Clutch 1 were moved into the artificial incubator, mean temperatures were 38.7°C and 36.9°C, respectively. Clutch 2 was naturally incubated until pipping. During Week 3, mean Tn was 35°C. Week 4 had the broadest temperature range, with a 15.3°C spread, mean Tn was 33.0°C. In Clutch 2, the temperature appeared to increase on Day 10 of incubation, which coincided with when Clutch 1 was moved to the artificial incubator for management protocols (Figure 1). There was no temperature trend noted across the incubation period for Clutch 2, nor was there any apparent daily or weekly pattern of time spent incubating (Figure 2). Ambient temperatures were not collected for Clutch 1, but dataloggers were added to give a more complete picture for Clutch 2. Temperature patterns for the telemetric egg and the ambient datalogger were very similar during the first 2 weeks of incubation. However, in Weeks 3 and 4 the patterns diverged. In Week 3, ambient temperature seemed to decrease while the telemetric egg temperature remained fairly constant; then in Week 4, ambient temperature increased while the telemetric egg temperature was variable before finally decreasing (Figure 2). In Clutch 1, all eggs were fertile, and 8 of the 9 eggs hatched. In Clutch 2, 8 of 10 were fertile, and 4 of the fertile eggs hatched. This demonstrates that while embryos can survive within this range of recorded temperatures, it seems hatchability for Clutch 2 was lower.

Study Two

The mean temperature during Week 1, for Pair 1 was 34°C and for Pair 2 was 34.1°C (Table 1). During Week 2, mean Tn was 35°C for Pair 1 and 34.6°C for Pair 2. Mean egg temperature seemed to increase slightly over the period of incubation in both pairs, although, as with the ducks, no apparent pattern of time spent incubating either within a day or week was observed (Figure 3 and Figure 4). Analysis of time-lapse videography data collected for three days for Pair 1 to validate the position sensor suggested that our assumption that the bird was present on the nest when the egg temperature was similar to body temperature and absent from the nest when the egg temperature dropped to or near ambient was correct. Video data of Pair 1 turning the egg confirmed telemetry results indicating a change in egg position at those times. Position data from the telemetric egg was clearer and more consistent with Pair 1 than with Pair 2. However, the results were difficult to interpret due to the high number of readings from Quadrant 4, which served as a default quadrant as well as a position indicator.

Figure 1. Spotted whistling duck, Clutch 1 vs. Clutch 2. Note that in Clutch 1, all eggs were moved to artificial incubator on Day 11.

Figure 2. Telemetric egg vs. nest box ambient temperature for spotted tree duck, Clutch 2.

Table 1. Weekly nest temperatures.


Min. EC

Max. EC

Mean EC



Week 1




  Week 2




  Week 3 a




  Week 4 a





Week 1




  Week 2




  Week 3




  Week 4






Week 1




  Week 2





Week 1




  Week 2




a indicates artificial incubator


Figure 3. Telemetric egg vs. nest ambient temperature for Bateleur eagle, Pair 1.

Figure 4. Telemetric egg vs. nest ambient temperature for Bateleur eagle, Pair 2.



Study One. During Week 2 when Clutch 2 had a mean temperature of 32.6°C, it seems likely that the pair was not incubating Clutch 2 reliably because the decrease in egg temperature mirrors the decrease in ambient temperature (Figure 2). The decrease in temperature at this point in incubation may explain why the hatchability of Clutch 2 seemed poor (50% hatched vs. 90% in Clutch 1). It is likely the embryo’s chorioallantois system is not yet attached at this time since attachment in the chicken is on about Day 9 (Freeman and Vince 1974) and in cranes on about Day 12 (Gee 1995), so the need for parental protection from environmental conditions is reported to be the greatest (Drent 1975). Temperature patterns for the telemetric egg and the ambient datalogger were very similar during the first two weeks of incubation (Figure 2), which is unlike a previous report that ambient temperature and egg temperature patterns did not correlate (Spellerberg 1969). More repetitions are needed to further elucidate the pattern for this species in the first two weeks of incubation. In general, the data from both clutches agree with the findings for waterfowl (Huggins 1941, Howey 1985) and other birds that the mean nest temperature under natural conditions is approximately 34°C. The higher hatching rate for Clutch 1 was unanticipated because incubator temperatures were much higher (mean T Week 3: 38.7°C, Week 4: 36.9°C) than 34°C and our findings from Clutch 2.

It is interesting to note that there was an apparent increase in the Tn of Clutch 2 on Day 10, which coincides with the range (Day 10-14) when waterfowl managers generally move the eggs to an artificial incubator to increase survivorship. We did not observe an increasing temperature trend across the incubation period as noted by Burger (1979) and Spellerburg (1969), although this may be a result of individual differences and the small numbers in this pilot study. There were no apparent daily or weekly incubation patterns noted, not surprisingly, as in whistling ducks, unlike most waterfowl species, both male and female share in the task of egg incubation. Although ambient temperature data were not collected for Clutch 1, it is not significantly cooler in August than July in Saint Louis, so it is unclear why the temperatures for Clutch 1 during Days 4-10 were higher than for Clutch 2.

Study Two. Both pairs of Bateleur eagles incubated their eggs at very similar temperatures, which agrees with findings for raptors (Huggins 1941, Schwartz 1977, Varney 1974, Snelling 1972) and other birds where mean nest temperature is approximately 34°C. Since there were no other eggs in with the telemetric egg, due to the biology of this species, it was not possible to know if the recorded temperatures would have resulted in successful hatching. It may be that although both females appeared to accept the smaller sized telemetric egg and incubated it properly, they perceived a difference and the pattern for incubation might have differed had it been a real egg. There was a slight increase in mean egg temperature during the incubation period in both pairs as seen in Spellerburg (1969) and Burger (1979). No apparent pattern of incubation either within a day or week was observed since in Bateleur eagles, both male and female share in the task of egg incubation; therefore, the nest is rarely unattended.

We feel that some modifications to the egg design would strengthen the information gathered in subsequent years. The addition of temperature sensors placed on the shell so that there would be four such sensors evenly spaced across the largest circumference of the egg as seen in Schwartz 1977; Howey 1977, 1984; Koontz 1991, and Gee 1995). The core temperature sensor would remain. Data from the surface temperature could then be compared to the core temperature providing further information about the nest environment. Howey (1984) found that temperature sensors on the nest substrate of a goose nest were 8° cooler than the sensor closest to the brood patch and 4° cooler than the core sensor. The surface sensors could also confirm egg-turning data from the position sensor if such a temperature variance exists in other species.

Egg Position

The position sensor to measure egg turning did not work as well as expected, although there are indications that the egg was turned before and/or after an absence from the nest, indicated by a drop in temperature at that time. This was confirmed in the analysis of the videography data for Bateleur eagle Pair 1. Unfortunately, video data could not quantify the number of turns or verify which quadrant the egg was in. The results, therefore, were difficult to interpret due to the high number of readings from Quadrant 4, which served as a default quadrant. The lack of clear, consistent data is puzzling as the position sensor worked properly during testing in the incubator. Position data from Study 2 seemed cleaner and more consistent than data from study 1. Within Study 2, Pair 1, which was closer to the antenna than Pair 2, seemed to have less noise in the data. It is possible that the range of the position sensor was shorter than expected or that noise interfered with signal decoding. The design could be improved with a modification so that spurious noise is not defaulted to the fourth quadrant. This would be possible with the addition of a non-existent fifth quadrant to serve as the default.

General Observations

Although the telemetric egg was considerably smaller than the real eagle egg and slightly larger than the real duck egg, all three females accepted and incubated it. However, modifying the external shape from the commercially available plastic Easter egg to an egg that is the proper shape and size of the birds should be investigated. It would be best to have eggs of multiple sizes to accommodate the variation in egg sizes of birds. Changing the external shape and size of the telemetric egg would also decrease the risk of rejection by the female.

In this pilot study, presence of a bird on the nest was assumed when the egg temperature was similar to body temperature, whereas absence from the nest was assumed when the egg temperature dropped to or near ambient. Unfortunately, it is not possible in many situations to use videotape to verify presence or absence on the nest as we did with the Bateleur eagles. Howey (1977, 1984), Koontz (1991) and Gee (1995) addressed this issue by incorporating a photocell into the egg design. Although, the photocell appeared to work well in the abovementioned studies, it is not clear that the photocell would function with cavity-nesting birds, such as the whistling ducks. The photocell was practical for geese, cranes, and swans, which have nests that are out in the open. However, the inside of the nest box may be too dark for the photocell to differentiate when the female in the nest box is on the nest and when she is off. If this is indeed the case, one solution would be to develop a technique for remote monitoring of activities of the parent bird around the nest. This would also facilitate a study of the role each parent plays in the maintenance of the incubation conditions.

We believe that these results indicate that our telemetric egg can accurately measure elements of the incubation environment, such as temperature and turning rate. However, as this pilot study shows, further modification of the technology is necessary to obtain more meaningful results. Although this study had a relatively small number of individuals, data from this pilot study agree with previous findings. More repetitions are needed to further elucidate the patterns for these species and to increase knowledge relating to the basic and applied biology of incubation for these taxa. In further studies, developing and testing other sensors to collect data on aspects of incubation, such as humidity and behavior of the parent bird relating to presence on the nest, should be emphasized.

Literature Cited

Burger, A. E. and A. J. Williams. 1979. Egg temperatures of the Rockhopper Penguin and some other Penguins. The Auk 96:100-105.

Derksen, D. V. 1977. A quantitative analysis of the incubation behavior of the Adelie penguin. The Auk 94:552-66.

Drent, R. 1975. Incubation. Pp. 333-420 in Avian Biology, vol. V, Editors D. S. Farner, J. R. King, and K. C. Parkes. New York. Academic Press.

Drent, R. H. 1970. Functional aspects of incubation in the herring gull. Behavior Supl. 17:1-132.

Eycleshymer, A. C. 1907. Some observations and experiments on the natural and the artificial incubation of the egg of the common fowl. Biol. Bull. 12:360-374.

Freeman, B. M. and M. A. Vince. 1974. Development of the Avian Embryo. New York. John Wiley & Sons.

Gee, G. F., J. S. Hatfield, and P. W. Howey. 1995. Remote monitoring of parental incubation conditions in the greater sandhill crane. Zoo Biology 14:159-172.

Harvey, R. L. 1990. Practical Incubation. Great Britain. Payn Essex Publishers.

Howey, P. B., R. G. Board, D. H. Davis, and J. Kear. 1984. The microclimate of the nests of waterfowl. IBIS 126:16-32.

Howey, P. W., R. G. Board, and J. Kear. 1977. A pulse-position-modulated multichannel radio telemetry system for the study of the avian nest microclimate. Biotelemetry 4:169-80.

Huggins, R. A. 1941. Egg temperatures of wild birds under natural conditions. Ecology 22(2): 148-157.

Koontz, F. W. 1991. Integrating biotelemetric, computer and video technologies to monitor physiological temperatures of zoo birds and reptiles. Pages 7-18 in Biotelemetry Applications for Captive Animal Care and Research, Cheryl Asa (editor). Silver Springs, MD. American Association of Zoological Parks and Aquaria.

Norton, D. W. 1972. Incubation schedules of four species of Calidridine sandpipers at Barrow, Alaska. The Condor 74:164-176.

Schwartz, A., J. D. Weaver, N. R. Scott, and T. J. Cade. 1977. Measuring the temperature of eggs during incubation under captive falcons. Journal of Wildlife Management 41(1):12-17.

Snelling, J. C. 1972. Artificial incubation of sparrow hawk eggs. Journal of Wildlife Management 36(4):1299-1304.

Spellerberg, I. F. 1969. Incubation temperatures and thermoregulation in the McCormick Skua. The Condor 71:59-67.

Varney, J. R. and D. H. Ellis. 1974. Telemetering egg for use in incubation and nesting studies. Journal of Wildlife Management 38(1):142-148.

Walsberg, G. E. and J. R. King. 1978. The energetic consequences of incubation for two Passerine species. The Auk 95:644-655.


GPS Radio Collar 3D Performance as Influenced by Forest Structure and Topography

R. Scott Gamo and Mark A. Rumble

Center for Great Plains Ecosystem Research, Rocky Mountain Research Station, 501 East St. Joe, Rapid City, South Dakota 57701, USA

Fred Lindzey and Matt Stefanich

Wyoming Cooperative Fish and Wildlife Research Unit,
Department of Zoology and Physiology, University of Wyoming,
Laramie, Wyoming 82071, USA


Global Positioning System (GPS) telemetry enables biologists to obtain accurate and systematic locations of animals. Vegetation can block signals from satellites to GPS radio collars. Therefore, a vegetation-dependent bias to telemetry data may occur which if quantified, could be accounted for. We evaluated the performance of GPS collars in 6 structural stage categories of 3 forest vegetation types to attempt to quantify the relation between GPS locations and forest vegetation.


Radio telemetry has enabled wildlife biologists to remotely monitor the movements and activities of free-ranging animals in their natural habitats since the 1960s. Locations of animals can be obtained from aircraft, triangulation of radio signals from the ground, or from satellite positioning systems. Triangulation of telemetry signals from the ground and use of aircraft facilitate visual observations of animals. If animals are not seen, their locations from either the ground or aircraft are subject to error and bias (Springer 1979, Lee et al. 1985, Saltz and Alkon 1985, White and Garrott 1990). Bias in locations of animals also occurs if relocation schedules are not random, or if animals are not located because of their activity or the habitat they occupy. Nonetheless, radio telemetry techniques over the past 30 years have proven especially useful in studying the habitat use and selection patterns of wildlife. Thus, experimental designs have been developed that minimize the bias in this application of radio telemetry.

Transmitter collars adapted to animals that measure Doppler shift from satellites (Craighead and Craighead 1987) or Loran-C (Dana et al. 1989) enabled researchers to locate animals in remote locations and during any time period, which enabled greater flexibility in experimental designs than conventional telemetry. Satellite and Loran-C collars also offered solutions to problems associated with locating animals due to vegetation density, topography, darkness, weather, and remoteness (Mech 1983, Rodgers and Anson 1994). Locations obtained with these collars varied from 100 m to 1000 m from the actual location of the animal (Rodgers and Anson 1994), and thus are inadequate for fine-scale studies of animal habitat (Rempel et al. 1995).

The most recent advance in telemetry techniques is the incorporation of a Global Positioning System (GPS) unit in a radio telemetry collar for animals (Rodgers and Anson 1994, Rempel et al. 1995). The Navigation Satellite Time and Ranging (NAVSTAR) GPS is a satellite-based navigation system consisting of 24 high earth-orbit satellites located in 4 orbital planes. GPS radio telemetry incorporates the advantages of Loran-C and Doppler shift satellite telemetry systems and is more accurate.

GPS receivers operate on a line-of-sight principle. Successful locations depend on satellite visibility and may be affected by animal activity, topography, and vegetation characteristics (Gerlach and Jasumback 1989, Rempel et al. 1995, Moen et al. 1996, Rumble and Lindzey 1997). Delusion of precision (DOP), a measure of the quality of satellite geometry with smaller values indicating lower triangulation error, can also affect the ability of the GPS collar to obtain a location. Blocking the satellites can: 1) cause positional delusion of precision to exceed the threshold of three to five, resulting in a 2-dimensional (2D) location; 2) reduce visible satellites to three, resulting in a 2D location; or 3) reduce the number of visible satellites to under three, making location impossible. If four satellites are visible to the GPS receiver, then latitude, longitude, and elevation can be calculated resulting in a 3-dimensional (3D) location, which, if differentially corrected, can be within 10 m of the true location (Wells et al. 1986). If three satellites are visible to the GPS receiver, the elevation value obtained from the last 3D position is substituted to solve the equations resulting in a 2D location. The error of this 2D location is directly related to the vertical distance that an animal has moved since its last 3D location, which, in areas of high relief, could be significant. Finally, when less than three satellites are visible, no location can be obtained.

Visibility of satellites in various vegetation or topographic types and the tendency of secretive animals to use areas of dense canopy cover may influence the accuracy of GPS locations and whether the locations are representative of the proportion of time an animal spends within a habitat. Inferences of animal habitat selection directly drawn from GPS telemetry data are biased toward areas of open canopy (Rempel et al. 1995, Moen et al. 1996). However, if the magnitude and character of this bias can be determined, GPS telemetry can be critically evaluated against conventional telemetry techniques for studies of animal movements and habitats. In addition, biologists can then evaluate data relative to biases and possibly correct for some biases in GPS locations by employing techniques similar to those used in animal sightability models (Samuel et al. 1987, Steinhorst and Samuel 1989).

The effects of forest vegetation and topographic relief on obtaining accurate and representative locations of GPS receivers are unclear. GPS telemetry collars for large ungulates are expensive, thus, the benefit of using these collars depends on obtaining many accurate locations of animals. When describing habitat data, particularly at the microsite level, we believe 3D locations are the most useful. We examined the effects of forest structure and topography on obtaining 3D locations from GPS large-mammal radio collars in the Black Hills of Wyoming and South Dakota.

The objectives of this study were to quantify the effects of forest vegetation and topography on 3D locations obtained from GPS elk collars, and, if possible, to model the effects to correct bias associated with open canopy. A preliminary evaluation suggested that it might be possible to model bias of GPS locations in open canopy areas (Rumble and Lindzey 1997). The hypotheses we tested were: 1) forest vegetation characteristics will not affect the number of 3D locations obtained by GPS collars; and 2) topography will not affect the number of 3D locations obtained from GPS collars.


Study Area

The study area is in the Black Hills National Forest (BHNF). Annual precipitation ranges from an average of approximately 46 cm to 66 cm (Orr 1959). Most precipitation occurs between April and July. January is typically the coldest month with mean temperature extremes of -11oC to 1.8oC. July and August are the warmest months with mean temperature extremes of 15oC to 29oC (Miller 1986). Elevations range from approximately 915 m to 2,207 m. Ponderosa pine (Pinus ponderosa) occurs throughout the Black Hills and is a dominant tree species. Other important tree species include white spruce (Picea glauca) occurring on north facing slopes, wetter sites, and at higher elevations, and quaking aspen (Populus tremuloides) (Hoffman and Alexander 1987).

Field Methods

We used three GPS radio telemetry collars (Version 2.10; Lotek Engineering, Inc.) designed for application on elk (Cervus elaphus). Each GPS unit was placed 2 m from the nearest tree on a tripod 0.5 m above the ground to approximate the height of a GPS collar on a bedded elk. Each individual GPS unit was programmed to collect locations at 30-min intervals and was placed at its location for approximately 24 h. If all satellites were operational, the current constellation of satellites guaranteed that six satellites were always visible (User's Manual 1994, Lotek Engineering, Inc., Newmarket, Ontario, Canada). Thus, 100% coverage in open environments was expected. Each GPS unit searched for satellites for up to 90 s; the first 25 s were spent searching for four satellites (3D location). If four satellites were not visible, the GPS unit searched for three satellites (2D location) for the remaining time with one caveat: a portion of the time remaining from locations obtained in <90 s was available to search >90 s for satellites on subsequent locations.

We stratified ponderosa pine, white spruce, and quaking aspen vegetation types into structural stages that included sapling-pole (2.5 cm to 25 cm diameter at breast height [DBH]) and mature timber (>25 cm DBH) with three overstory cover categories (OCC) of 0-40%, 41-70%, and >71% in each DBH category (Buttery and Gillam 1983). In each of these structural stages, we stratified slope into three categories of 0-30%, 31-60%, and >60%. In aspen stands, we tested GPS collars at the same location during summer when leaves were present and during winter when leaves were absent. We used Geological Information System (GIS) coverage of forest vegetation in the BHNF and digitized elevation models to find prospective stands of each vegetation structural stage-slope category. We verified the vegetation structural stage-slope category of each stand before placing the GPS collars for testing. We also placed a GPS collar in a meadow with no overstory component and negligible slope for comparison of forest locations to a benchmark location.

At each site, we characterized vegetation and topographic features. Measurements included OCC, percent slope, basal area (BA), DBH, average tree height, and average percent of the visible horizon. OCC was visually estimated with a spherical densiometer (Griffing 1985) positioned directly over the GPS collar. We used a 10-factor prism to determine trees >15.2 cm DBH to be measured in a variable radius circular plot. Trees <15.2 cm DBH were measured in a 4.9 m-radius fixed plot. We recorded the DBH of all trees using tree calipers. Stand density index (SDI), a useful, relative measure of tree density, was calculated using the equation in McTague and Patton (1989) using average tree DBH, tree density, and BA. We also calculated an index of DBH x tree density for each site. We measured percent slope as the uphill reading of a clinometer for the first 50 m from the GPS unit at a height of 1 m. The angle of the visible horizon (lowest angle that sky was visible) was measured using a clinometer at 45o increments (n=8). The average of these angles was subtracted from 180° . The difference was then divided by 180 to estimate the percentage of visible horizon at each site.

Data Analysis

The hypothesis of no effect of vegetation on percent 3D locations was tested using 2-way analysis of variance between DBH categories and among OCC categories. There were no differences (P > 0.3) between DBH categories of ponderosa pine, white spruce, or aspen without leaves. Therefore, we pooled data for DBH categories and tested for differences among OCC categories using one-way analysis of variance for each forest type. We used Tukey’s multiple range test if Levene’s test (Milliken and Johnson 1984) for homogeneity of variance was non-significant and Dunnett’s T-3 test (Dunnett 1980) if Levene’s test was significant for post hoc comparisons of DBH categories.

Within each vegetation type, we used linear regression to model the average percent 3D locations with OCC, percent slope, timber density, DBH, tree height, DBH x tree density index, SDI, and percent of visible horizon. Multiple linear regression was used to model the percent of 3D locations using the better predictors obtained from linear regression. Variables were entered if they significantly (P = enter < 0.1) contributed to the model.


We measured 36 sites each in ponderosa pine and in quaking aspen without leaves. Due to malfunction in the GPS units, only 19 sites each were measured in spruce and quaking aspen with leaves. Two meadow sites were measured for a relative measure of optimum performance of GPS collars.

As predicted, meadow sites had the highest percent of 3D locations. The percent of 3D locations declined from the 0-40% OCC to 41-70% OCC and >71% in ponderosa pine and white spruce (Figures 1a, 1b). In ponderosa pine stands with 0-40%, OCC had more 3D locations than 41-70% OCC (P < 0.03) and >71% OCC (P < 0.02). White spruce stands 0-40% OCC had more (P = 0.002) 3D locations than stands > 71% OCC. Percent 3D locations in white spruce 41-70% OCC did not differ from 0-40 or >71% OCC. Percent 3D locations were the same (P > 0.99) among OCC categories of quaking aspen without leaves (Figure 1c). Percent of 3D locations in sapling-pole size quaking aspen stands with leaves differed from mature (P < 0.09). Failure of one GPS collar prevented quantitative comparisons among vegetation structural stages of quaking aspen with leaves.


a) Ponderosa pine b) White spruce

c) Aspen without leaves

Figure 1. Graphs a, b, and, c are resulting percentages of 3D locations in each OCC structural stage.

Single regression analyses showed a decrease in OCC as the best predictor (P = 0.01, adj. R2 = 0.27) of percent of 3D GPS locations in ponderosa pine. In white spruce, OCC also was the most highly correlated variable (P < 0.01, adj. R2 = 0.5) in predicting percent 3D GPS locations. Average percent of visible horizon was the most highly correlated (P < 0.01, adj. R2 = 0.37) single variable in obtaining 3D locations in aspen sites without leaves. In aspen with leaves, no significant correlations between percent 3D GPS locations and vegetation or topography were found.

Multiple regression improved the amount of variation in percent 3D GPS locations that could be accounted for by vegetation and topography in ponderosa pine and aspen without leaves stands. Multiple regression added slope to OCC as the most highly correlated (P < 0.01, adj. R2 = 0.39) variable in predicting 3D locations in ponderosa pine. Positively correlated average percent of visible horizon and negatively correlated DBH resulted in an improved (P < 0.01, adj. R2 = 0.40) predictive model in aspen without leaves. Multiple regression analysis of white spruce resulted in no improvement on the single regression model. No significant multiple variable model could be derived for aspen with leaves (P > 0.1).


Dense vegetation decreases the probability of obtaining 3D locations from GPS collars. We found that dense vegetation and steeper topography decrease the ability of GPS collars to obtain locations. Tree boles are physical barriers that can block GPS signals from satellites (Rempel et al. 1995, Moen et al. 1996). We anticipated that the index of tree density X DBH would reflect the blockage of satellite signals by tree boles. However, this variable was not as highly correlated with percent 3D GPS fixes as OCC. Theoretically, satellite signals can penetrate the needles of conifers (Rempel et al. 1995). In white spruce and ponderosa pine >71% OCC, 25% and 35% of location attempts, respectively, were 3D fixes. Therefore, some GPS signals are penetrating the conifer needles in white spruce and ponderosa pine. However, because OCC had the highest correlation with percent 3D fixes, our data suggest that increased canopy cover of conifers reduces the visibility of satellites to the GPS receiver on the collar.

We expected that the increased canopy coverage due to leaves (in deciduous aspen) would reduce the percent of 3D GPS locations compared to leafless (or winter) stands. In seven of 19 of our aspen stands with leaves, the percent of 3D locations increased from the number of 3D locations obtained at the same site without leaves. However, we are unable to explain these increases in 3D locations when leaves were present.

In a preliminary study by Rumble and Lindzey (1997), percent 3D GPS fixes showed similar declines with increasing OCC categories of ponderosa pine. The collars we used included updated communications software from the unit used by Rumble and Lindzey (1997). This software enabled the GPS unit to track 8 satellites when searching for satellite links to calculate the position, which resulted in reducing the time needed to establish a communication link with the satellite. The shorter time used to establish communication with satellites should increase the chance of obtaining a 3D location since more opportunities become available to establish links with satellites.

Regression analyses showed that the structural components of forest stands, particularly OCC, and the physical features of the landscape, such as percent available horizon or slope, can partially block or reduce the view of satellites from GPS collars. However, none of the OCC levels completely blocked all signals as evidenced by 3D locations obtained at all sites. The satellites used to acquire fixes are not geostatic; they orbit around the earth. Therefore, satellites are sometimes in a good configuration to obtain a 3D fix, while minutes or hours later they are in a configuration that only allows for a 2D fix. Furthermore, when compared to meadow sites that obtained about 72% 3D locations, the success of obtaining 3D locations in forest stands was good.

Either 2D or 3D locations are adequate for studies that describe general movements and home ranges. However, delineation of habitats used by animals within a home range is best addressed with the accuracy obtained with 3D locations. Researchers will need to account for bias in GPS locations due to vegetation and topography by testing the performance of GPS collars in forest vegetation and slope categories in their study area. Because GPS technology is changing, tests must be completed for each software or component upgrade before collars are placed on animals.

GPS technology in radio telemetry collars enables researchers to determine animal locations systematically at night, during poor weather conditions, or during other situations that are impractical for using conventional telemetry methods. Animal locations will be biased toward open habitats, but that bias can be considered and the tradeoffs of obtaining data from animals in remote areas, at night, or during winter storms must be evaluated. Continued advances in equipment will lead to further improvements in collar performance. At this time it would be inappropriate to extend these data to other forest types, topography, or GPS collars.


The study plan for this research was designated as Forest Service RC-927. Financial support for this study was provided by the United States Department of Agriculture Forest Service, Rocky Mountain Research Station, Black Hills National Forest, and the Wyoming Game and Fish Department.


Buttery, R.F., and B.C. Gillam. 1983. Forest ecosystems. Pages 43-71 in R.L. Hoover and D.L. Wills, (eds.), Managing forested lands for wildlife. Colorado Division of Wildlife in coop. with USDA Forest Service, Rocky Mountain Region, Denver, CO. 459p.

Craighead, D.J., and J.J. Craighead. 1987. Tracking caribou using satellite telemetry. Natl. Geogr. Res. 3:462-479.

Dana, P.H., W. Fowler and D. Hindman. 1989. Automated animal-tracking system: tracking elk with Loran-C. Proceedings Western Goose Association.

Dunnett, C. W. 1980. Pairwise multiple comparisons in the unequal variance case. Journal of the American Statistical Association. 75:796-800.

Gerlach, F.L., and A.E. Jasumback. 1989. Global positioning system canopy effects study. USDA Forest Service, Technology and Development Center, Missoula, MT. 18p.

Griffing, James P. 1985. The spherical densiometer revisited. Southwest Habitater 6(2). USDA, Forest Service, Southwestern Region. Albuquerque, NM.

Hoffman, G. R., and R. R. Alexander. 1987. Forest vegetation of the Black Hills National Forest of South Dakota and Wyoming: a habitat type classification. Res. Pap. RM-276. U.S. Dept. Agric., For. Serv., Rocky Mtn. For. Range Exp. Stn. Fort Collins, CO. 48p.

Lee, J.E., G.C. White, R.A. Garrott, R.M. Bartmann and A.W. Allredge. 1985. Assessing accuracy of a radiotelemetry system for estimating mule deer locations. Journal of Wildlife Management. 49(3):658-663.

McTague, J. P., and D. R. Patton. 1989. Stand density index and its application in describing wildlife habitat. Wildlife Society Bulletin. 17:58-62.

Mech, L.D. 1983. Handbook of animal radio-tracking. Univ. of Minn. Press. Minneapolis. 107p.

Miller, J. R. 1986. Rapid City climate. S. Dak. School of Mines Tech. Found. Rapid City, SD. 66p.

Milliken, G. A., and D. E. Johnson. 1984. Analysis of messy data, Vol 1: designed experiments. Van Nostrand Reinhold Co., New York: 473p.

Moen, R., J. Pastor, Y. Cohen and C.C. Schwartz. 1996. Effects of moose movement and habitat use on GPS collar performance. Journal of Wildlife Management 60(3):659-668.

Orr, H. K. 1959. Precipitation and streamflow in the Black Hills. Stn. Pap. RM-44. U.S. Dept. Agric., For. Serv., Rocky Mtn. For. and Range Exp. Sta. Fort Collins, CO. 25p.

Rempel, R.S., A.R. Rodgers and K.F. Abraham. 1995. Performance of a GPS animal location system under boreal forest canopy. Journal of Wildlife Management 59:543-551.

Rodgers, A.R., and P. Anson. 1994. News and applications of the Global Positioning System. GPS World. July:20-32.

Rumble, M. A., and F. Lindzey. 1997. Effects of forest vegetation and topography on global positioning system collars for elk. Pages 492-501 in: Proc. ACSM/ASPRS Res. Tech. Inst. Vol. 4.

Saltz, D., and P.U. Alkon. 1985. A simple computer aided method for estimating radio-location error. Journal of Wildlife Management 49(3):664-668.

Samuel, M. D., E. O. Garton, M. W. Schlegel and R. G. Carson. 1987. Visibility bias during aerial surveys of elk in northcentral Idaho. Journal of Wildlife Management 51:622-630.

Springer, J.T. 1979. Some sources of bias and sampling error in radio triangulation. Journal of Wildlife Management 43:926-935.

Steinhorst. R. K., and M. D. Samuel. 1989. Sightability adjustment methods for aerial surveys of wildlife populations Biometrics. 45:415-425.

Wells, D.E., N. Beck, D. Delikaraoglou, A. Kleusberg, E.J. Krakiwsky, G. Lachapelle, R.B. Langley, M. Nakiboglu, K.P. Schwarz, J.M. Tranquilla and P. Vanicek. 1986. Guide to GPS Positioning. Canadian GPS Associates, Fredericton, N.B., Canada.

White, G.C., and R.A. Garrott. 1990. Analysis of wildlife radio-tracking data. Academic Press, Inc. San Diego. 383p.

Design and Evaluation of an
Automatic Localization System Based on
Doppler Antennas Used in Wildlife Studies

Carsten Riis Olesen

National Environmental Research Institute, Department of Landscape Ecology, Kaloe, Grenaavej 14, DK-8410, Roende, Denmark

Niels Udengaard Kristiansen

Aarhus University, Department of Zoofysiology, C.F. Mollersalle, Building 131, DK-8000, Aarhus C, Denmark


An accurate, labor-saving automatic wildlife location system, based on Doppler direction-finders at three locations and 1-1½ watt transmitters, was developed to analyze dispersal and habitat choice in roe and red deer. Dense Danish forests made Global Positioning System alternatives impractical. Calibrations of the automatic system are stable for long periods with diurnal maximal standard deviations of 3 degrees. Mean error distances and error ellipses in the best and worst case range from 63 m and 0.4 ha to 162 m and 9.2 ha, respectively. Horizontal antenna positions induce polarization problems and significantly affect signal strength and precision of estimated locations. System hardware and software and examples of animal movement patterns are described.


The conventional hand-held triangulation approach is commonly used in wildlife studies (Kenward 1987, White and Garrott 1990), although a number of drawbacks are known (Angerbjorn and Becker 1992).

One of the most limiting factors in conventional triangulation is the considerable need for manpower to be able to gather more than only a very limited number of animal locations per day. Another important limitation is the need for the single observer to move to a minimum of three separate positions around the animal, making triangulation impossible when animals are on the move, either freely or as a result of observer-induced disturbance. An automatic location system offers the potential of higher precision of estimated locations in a fixed area, multiple location estimates usable for time step-analyses and analyses of habitat choice. Global Positioning System (GPS) based locations offer all these possibilities without being limited to a particular study area, and even have the potential of being quite precise. Regrettably, communication between GPS receivers and satellites works at frequencies that have a very low degree of penetration of any kind of obstruction. Rodgers et al. 1996 report that observation rates in a GPS-based moose tracking system declined with increasing tree density, height, and canopy closure. The lowest observation rate was 0.10 in a red pine stand with a tree spacing density of 1.8 meter.

Most forests in Denmark are planted and very dense. In Danish coniferous plantations, the canopy is closed and practically no light passes through to the ground, which is therefore devoid of ground cover. Deciduous forests are rather open during winter, but the canopy closes to near 100 % after bud break. Accordingly, our pilot study using GPS receivers (Trimble Pathfinder Pro-XL) proved that at most forest positions in the study area, no GPS fixes were possible. We developed this automatic wildlife location system (AWLS) to study the dispersal, movement, and land use pattern of two species of deer known to spend most of their time in dense forest. This AWLS uses the VHF band and correct bearings are, therefore, based on line-of-sight, but signals transmitted through forest are still received and the degree of reflection influencing precision can be evaluated.


Study Area

The study area is in central Denmark, approximately 50 km due east of Århus. The core study area is approximately 50 km2. It is bordered at the western and southwestern side by Lake Tange, through which the Gudenå River flows. The elevation of the water surface in Lake Tange is 14 m above sea level (a.s.l.). In the northern part, the city of Bjerringbro (8,000 inhabitants), makes a man-made border to animal dispersal. Eastward, the landscape is an open plateau of intensively managed arable farmland with an average altitude of 50 m a.s.l. About a third of the area is forest, mainly covering the far west and southernmost areas. The forest is mainly dense coniferous forest of spruce and pine although the southernmost forest is mixed forest with broadleafed plantations of beech and oak. From the rather flat plateau of arable land to the river valley, some slopes with height differences of 35 m occur. The study area and the locations of the three fixed radio receiver towers are shown in Figure 1.

Figure 1. Study area with location of the three fixed Doppler receiver towers. Bearing lines as the result of signals from a reference transmitter in open landscape are shown.


The animal-worn collar electronics consists of a transmitter, controller, and power source (Figure 2). For improving real-time characteristics, we have added a micro-power Radio Clock Receiver (DCF77, Physikalisch-Technische Bundesanstalt, Braunschweig, Germany). The FM transmitter is produced by IBN-DLO (Institute for Forestry and Nature Research, Wageningen, The Netherlands) and delivers 1-1½ W into the whip antenna. The uplink timing logic is generated by a Microchip (Microchip Technology Inc., Arizona, USA) PIC12C508 micro controller (Figure 2). All collar electronics are programmed to transmit five 150 ms pulses with a repetition rate of 500-2,500 ms, every half hour (one uplink). The pulse interval and the uplink real time serve as identification. Transmitting five signals in an uplink increases the statistical possibility of calculating location estimates of high precision. All transmitters are tuned to the same frequency for practical reasons, although the receiver and computer system of the fixed direction-finder towers allows for the use of multiple frequencies. Power source is three SAFT (Saft, Romainville, France) LSH14 lithium batteries, with a maximum life span of 1½ years.

At the side of the plastic housing of the red deer collar, a l /2 stiff antenna is mounted, pointing a little backwards when the animal keeps its head high, but retaining a more vertical position when the animals are foraging. The collar band is a 5 cm wide, soft nylon band. Total weight of the red deer collar is 750 g. In the roe deer collars, a soft l /4 antenna is mounted in the neck position. The antenna is pointing a little to the rear when the animal is holding its head high, but otherwise hold a quite vertical position, as the collar glides down the neck of the foraging animal. The housing is compact and the collar band is a 4 cm wide, flexible plastic band. Total weight of the roe deer collar is 475 g.

Figure 2. Schematically representation of the electronics of the animal transmitter collars.

Automatic Direction-Finding System

Three fixed receiver towers are located on some of the highest elevations in the landscape (Figure 1). Two of the towers (T1, T2) are placed in the open landscape, distant from human settlements, with a 220 V power line supplying power. A wooden enclosure (1.5 m x 3.0 m) contains and protects the technical equipment. A 10 m high wooden antenna mast is adjacent to the enclosure, bringing the center of the Doppler antenna to a total height of 11 m above ground. The third Doppler antenna (T3) is mounted at the top of a farm silo 25 m above ground, with the connected technical equipment in the adjacent farm buildings. Resulting antenna altitudes are 61 m, 56 m, and 78 m above sea level. Mean distance between towers is 3,336 m. The receiving system consists of several integrated electronic components (Figure 3) including:

A multi-channel receiver, ICOM IC-R10 (ICOM Inc., Osaka, Japan), with a computer interface, in the 150 MHz frequency range in the narrow band FM mode.

Doppler Systems Inc. (Carefree, Arizona, USA) DDF6000 Radio Direction-Finder using the Simulated Doppler System (also known as the Pseudo- or Quasi-Doppler System), consisting of an 8-element circular antenna (actually 8 individual antennas) plus the controller and bearing-computing box, which includes a computer interface.

IBM-style PC (with MS-DOS operating system).

Data Acquisition Program.

Radio Clock Receiver (HKW-Elektronik, Thüringen, Germany) that will keep the PC Real Time Clock at correct time.

Figure 3. Schematic representation of the technical equipment of the automatic location system towers.

Data Acquisition and Processing

The Doppler System computes bearings from pulsed signals and data output is given on the serial data line, with processing time up to 900 ms. This calls for asynchronous data logging. Our Data Acquisition Program first identifies the telemetry pulse and logs the time. Then it waits for Doppler System bearing output and saves it all to a text file.

The PC sends and receives both data and control commands through its serial COM ports. After appropriate setup commands for the ICOM FM receiver and the Doppler System, the program will ask the FM receiver for the S-meter value (antenna signal strength). If this value rises above a defined threshold, a telemetry pulse is expected. An input buffer is opened for the Doppler System bearing data and the Real Time Watch is logged. If the above threshold value lasts for 150ms (± a fixed time tolerance), the bearing data, the S-meter value and the Real Time Stamp are appended to the data text file. If the pulse is shorter or longer, it is not considered a telemetry pulse. Finally, the input buffer is closed and the program resumes the "ask for S-meter" loop.

Individual transmitters are identified on the basis of pulse repetition rate and the uplink time. If all five pulses are present in the data file, the ID job is easily done with a frequency cross correlation of the time stamps. After adding a Radio Clock Receiver to the transmitters, the ID job is much more reliable, so even an uncompleted uplink with one pulse can be identified.

Apparently due to weak signals or background noise, the number of bearings from the Doppler System may differ from the five signals transmitted. If the Doppler System data output holds fewer bearings than signals received by the ICOM receiver, the resulting extra time stamps are deleted. If the Doppler System data output holds more than five bearings, the mean angle (Zar 1999) of all possible combinations are computed and the bearing raising the resulting angular deviation the most is deleted. This is repeated until the number of bearings and pulses matches. Our experience shows that the extra bearings are either similar to the first five or they are spread out randomly. Therefore, we find this reducing method acceptable.

It is useful to appoint a quality index to an uplink. A first class uplink we define as 5 equal bearings. Missing bearings or variation in the uplink bearings will lower the class. In some uplinks the majority of bearings are equal and one or two are obvious outliers. To identify the outliers, a fit product of the angular deviation (Zar 1999) and the number of bearings are computed for all possible combinations of the bearings in the uplink. The unwanted bearings are marked and the result is written to the data file. In addition, the resulting mean bearing and the bearing standard deviation are recorded for any later statistical analyses.

In a post-processing program, the bearings from the three towers are combined through the transmitter identification and the time criteria to form sets of uplinks. Offset errors of the direction-finder may be entered separately as an initial tower-specific value, and all bearings are corrected accordingly. If the operator wants, he can omit the previously mentioned outliers and omit low-quality uplinks. After adding the tower positions (UTM), the data are ready for processing in a triangulation program such as LOAS (Ecological Software Solutions, Sacramento, California, USA).

In this study all triangulations were made using the LOAS program. To calculate estimated locations (accuracy = 0.000001) and error ellipses, the Maximum Likelihood Estimator (95% confidence, Chi square) (Lenth 1981, White and Garrott 1990) was used. To improve the display, all resulting data were imported into ArcView Geographical Information System (GIS) (Environmental Systems Research Institute, Inc., California, USA). All other statistics were handled in the SAS software (Version 6.12) (SAS Institute, Inc., Cary, North Carolina, USA).

For evaluating the precision of the AWLS we used a reference transmitter placed in a good position in open landscape. This transmitter was either switched to continuous-pulse mode, used for tower calibration, or to 2-minute transmissions every third hour throughout the day as a solid reference. Reference bearings were then used for correcting offset errors at the individual towers. In this study, data were only corrected if a daily mean value exceeded ± 0.5° from the true bearing. Test transmitters were placed at different positions representing the range of variation in landscape topography and vegetation cover. Test transmitters were placed approximately 1½ m over ground surface. All positions were verified with the D-GPS (Trimble Pathfinder Pro-XL) system, having a mean location error of less than 10 m.

Animal Capture

The total population of free-ranging red deer in Denmark is only approximately 10,000 animals. In the study area, the estimated population is 125 animals. Red deer are exposed to heavy hunting pressure and are accordingly very shy animals, rarely seen moving in the open landscape in daylight. During the winter season, often with snow cover, the red deer were fed with sugar beet at a distance of 30 m from a pre-prepared hide. Deer were immobilized at a distance of 25-40 m using a Dan Inject CO2 pressure gun (Dan Inject Aps, Børkop, Denmark). As tranquilizer, a mixture of 0.7 ml Etorfin vet. (4 mg/ml) and 0.8 ml Rompun (50mg/ml) was used for a 100 kg animal (Eriksen 1978). To be certain of recovery of the darted animal in dense forest, we developed a 1.5 g VHF tracking transmitter, which was incorporated in the dart. A barbed syringe prevented the dart transmitter from dropping off the animal. After locating the immobilized animal by hand-held triangulation, the radio collar was attached and the animal marked with reflecting earmarks and given an appropriate antidote after approximately 10 minutes of handling.

Roe deer are numerous and widespread in Denmark. Apparently they adapt very quickly to human made changes in the landscape. Although heavily hunted, they are quite visible, even during daytime. Roe deer were not darted but driven into a soft net and handled without being immobilized. Handling time was reduced to approximately 5 minutes before the animal was released.


AWLS Performance

Calibration of the three Doppler antenna system towers forms the basis of precise location estimates. Although system calibrations were done carefully, daily averages may differ from the true value (Table 1.). The stability of the tower calibration is typically characterized by standard deviations in the range of 1-3 degrees with maximum deviations of 6 degrees. Changing weather conditions such as rain seems to have some influence on the stability of the calibration, even under line-of-sight conditions. Within these limits, the tower calibration holds for weeks. The stability of multiple location estimates on the reference transmitter is shown in Figure 4. Estimated locations are scattered around the true position with a mean error distance of 63 m and mean error ellipse of 0.40 ha.

Table 1. Mean and standard deviation of recorded reference bearings during 24 hours on three receiving towers.


Number of bearings

Mean bearing

Standard deviation

True bearing



94.3 °

2.9 °

93.7 °



38.8 °

1.9 °

42.2 °



326.4 °

1.3 °

326.2 °


Figure 4. Graphical representation of location estimates and error ellipses of the system reference transmitter recording every third hour during a 24-hour period. The true location of the transmitter is on a farm silo, just between the farm buildings.

Signals from the animal transmitters are received at distances up to 15 km or even more in good conditions. This does not mean that acceptable location estimates can be calculated, but it offers the opportunity to stay in contact with the marked animals when they move outside the core study area. The S-meter value drops when distance to the transmitter is increased. At greater than 7 km, the S-meter value is down to approximately 80 (Figure 5.). Bearing error seem to be unaffected by long distances (Figure 6.), although weak signals with a S-meter value under 70 tend to give large bearing deviations and rather imprecise estimates of location. Due to the effect of landscape topography on the precision, it is difficult to make a clear evaluation of system range. In the chosen study area, we are confident of having precision as shown in Table 2 within approximately 50 km2.

Figure 5. Receiver S-meter values as a function of distance to a true transmitter location (vertical and 45 degree antenna position).

Distance, Tower – true location (m)

Figure 6. Bearing errors (numerical values) as a function of distance to a true transmitter location (vertical and 45 degree antenna position).

Test Transmitter Locations

Tests of precision and stability of precision over 24 hours with transmission each half hour from 4 locations typical of the variation in landscape topography and vegetation cover result in mean error distances (estimated location - true location) of 63 m in the open landscape to 162 m in one of the most difficult locations for triangulation found in the study area (Table 2). At this location, the transmitter was situated in dense coniferous forest at the slope side of a narrow, eroded valley. Estimated locations of this test transmitter were not evenly scattered around the true location, but felt within a circle with a radius of 81 m, which was biased 162 m in a westerly direction toward the position of the nearest receiving tower. Mean error ellipses ranged from 0.4 ha to 9.2 ha (best to worst case). Radius from true location covering 50% and 90% of the estimated locations ranged from 57 m to 159 m and 106 m to 216 m, respectively (Table 2.).

Table 2. Precision of the Automatic Wildlife Location System. Transmitters were placed at different fixed locations, which reflects the differences in landscape topography and vegetation cover of the study area. Tests are run simultaneously through a 24-hour period. Figures given as xx ± xx are mean ± STD. N is number of estimated locations each based on sets of 15 bearings.


Open landscape

Alt.=57 m


Dense coniferous forest

Alt.=45 m


Mixed forest

Alt.=45 m


Dense coniferous forest:

hills and slopes

Alt.=29 m


Mean error distance

Estimated – true

63.4 ± 37.2

118.0 ± 31.4

138.8 ± 49.0

161.9 ± 42.3

Mean error ellipse

(MLE) (ha)

0.40 ± 0.39

0.92 ± 0.73

1.02 ± 0.83

9.18 ± 1.78

Radius from true location covering 50 % of the estimated locations (m)





Radius from true location covering 90 % of the estimated locations (m)





Transmitter-receiver distance (m)

Tower 1

Tower 2

Tower 3

















Due to the fixed vertical position of the Doppler antennas, polarization problems may occur when the antenna of the animal collar is not in the vertical position. To test this we took 15 bearings from a range of known locations each with either a vertical, 45-degree angle, or horizontal antenna position. Mean bearing error, relative values of S-meter, error distance, and error ellipse area do not differ for the vertical and the 45- degree antenna position, but horizontal antenna positions markedly reduce signal strength (S-meter value), and increase mean bearing error and relative values of error distance and error ellipse area (Table 3).

Table 3. Results of test trials with different transmitter - receiver antenna polarization.


Mean bearing error (numerical)

mean ± std (n)

S-meter recordings

Relative error distance.

Estimated - true location

Relative error ellipse area (MLE)

Vertical antenna position

2.8 ± 2.1 (80)




45 degree antenna position

2.6 ± 1.8 (81)




Horizontal antenna position

5.6 ± 3.8 (69)





Animal Movements

Using ArcView, estimated locations and error ellipses give a clear picture of animal habitat choice (Figure 7). In this case the position of female red deer No. 8 during the period of March 10-12 was estimated. Time- step analyses of the data show a distinct pattern of using arable crop fields (carrots and winter wheat) in the open landscape during darkness (1900 hours to 0700 hours) and resuming cover in the adjacent dense forest during daylight (0700 hours to 1900 hours). This movement pattern was repeated each of the days of March 10-12. Positive identification of the marked animal during night surveys proved that the AWLS estimated locations were exact within the limits of our capabilities of far distance visual mapping of the animal location.


Many kinds of VHF telemetry systems ranging from the hand-held antenna systems to mobile or aircraft systems and manually operated fixed antenna towers have been used in wildlife studies (Kenward 1987, White and Garrot 1990). Recently GPS has been given more attention. Still, conventional VHF triangulation is often the only possible way to track smaller animals or animals living in dense forest areas. Few attempts of developing an automatic VHF triangulation system for wildlife studies have been conducted. Angerbjorn and Becker (1992) described an automatic location system based on the Doppler direction-finder system, but did not report any statistics on precision of their system. The technical advantages of our AWLS is the use of eight-element Doppler antennas and the use of a radio clock system in both the receiver and the transmitter system along with logging and post-processing software.

Figure 7. Estimated locations of female red deer (Deer No. 8), March 10-12, 1999.

Average bearing error and standard deviation of the AWLS is <3 ° , which is similar to those reported by Lee et al. 1985, Garrott et al. 1986 and Kufeld et al. 1987, when dealing with signals that are not always of line-of-sight. Only a few authors present statistics of error distance from estimated location to true location and error ellipse area. Lee et al. (1985) had error polygons < 5 ha from the best tower positions and < 20 ha from poorer good tower positions. Compared to these results our AWLS is more precise, ranging from mean error ellipse areas of 0.4-9 ha for the best and the worst case. Despite the size of error, conclusions about animal movement pattern using estimated locations must be evaluated relative to landscape mosaic or patch size. The precision of our AWLS still has the potential for improvement as we develop a more sophisticated software application, dealing with the short-term variations in bearing accuracy on the permanent reference transmitters. Overall precision would also be improved by establishing more towers, but this calls for additional funding. The range of a location system is not easy to define, as it is very much dependent on landscape characteristics. In this way, effective range may partly be monitored through the size of the error ellipses.

Polarization problems in wildlife telemetry have been addressed by Beaty and Tomkiewicz (1990), who state that when two antennas are completely cross-polarized, the signal loss can be as large as 20 dB. In reality, complete cross-polarization will rarely occur because of animal-induced antenna movements and signal reflections in the landscape. However, even a small degree of polarization means some loss of signal and the potential for reduced precision. The polarization problem is a disadvantage of fixed antennas and, therefore, care must be taken to ensure the best possible antenna position on the animal. In hand-held triangulation, polarization problems are easily corrected by rotating the antenna to obtain the best signal.

Our animal-collar transmitters are prepared for FSK modulation, making it possible to identify animals by digital coding or transmitting any kind of data taken from the animal, e.g. activity, heart rate or temperature, in future system developments. Making an on-line connection (mobile phone and modem) to all towers would facilitate systems control and data collection.

Developing and testing the AWLS and capturing the very shy Danish red deer was very labor intensive, but the established system offers us the advantage of being able to devote a greater proportion of time to the ecological research into the dispersal and habitat choice of the marked animals. Data from the AWLS made it possible to detect a distinct pattern of time-specific habitat choice in red deer. All estimated locations during daylight were in areas of dense forest. During darkness, all estimated locations were in open, arable land.


The authors want to thank Kees Van’t Hoff for valuable technical advises, Dave Cunningham (Doppler Systems Inc.) for advice and repairing Doppler antennas. Robert Thrane and associates, Mogens Rosengaard and Kim Bengtsson for their everlasting endurance at late and cold hours of capturing the animals. Niels Due Jensen (Ormstrup Estate), Ulrik Kristensen, Gudenå Centralen and other private landowners for giving their permission for establishing receiver towers, working in the area, and marking the animals. Dr. Chris J. Topping for comments on and improvement of the language. The work was supported by the Danish Research Council program: Humans, Landscape and Biodiversity.


Angerbjorn, A. and D. Becker. 1992. An automatic location system for wildlife telemetry. In Priede, G., and Swift, S.M. (eds.) Wildlife Telemetry, Remote Monitoring and Tracking of Animals. (Based on the Proceedings of the fourth in a series of European meetings on wildlife telemetry). Ellis Horwood Limited.

Beaty, D., and S. Tomkiewicz. (1990). Polarisation: The effect on range performance. Telonics quarterly Vol 3. No.3.

Eriksen, E. Medical immobilisation and capture of wild cows and deer. Ph.D. thesis. Royal Veterinarian and Agricultural University, Copenhagen (In Danish) P.198.

Garrot, R.A., G.C. White, R.M. Bartmann, and D.L.Weybreight. 1986. Reflected signal bias in biotelemetry triangulation systems. J. Wildlife Management 50(4):747-752.

Kenward, R. 1987. Wildlife Radio Tagging — Equipment, Field Techniques and Data Analysis. Academic Press. London. P. 222.

Kufeld, R.C., D. Bowden, and J.M. Siperek. 1987. Evaluation of a telemetry system for measuring habitat usage in mountainous terrain. Northwest Science, Vol. 61, No. 4: 249-256.

Lee, J. E., G.C. White, R.A. Garrot, R.M. Bartmann, and A.W. Alldredge. 1985. Accessing accuracy of a radiotelemetry system for estimating animal locations. J. Wildlife Manage. 49(3): 658-663.

Lenth, R. V. 1981. On Finding the Source of a Signal. Technometrics. Vol. 23. No. 2: 149-154.

Rodgers, A.R., R.S. Rempel, and K.F. Abraham. 1996. A GPS-based telemetry system. Wildlife Society Bulletin, 24(3):559-566.

White, G. C. and R.A. Garrot. 1990. Analysis of Wildlife Radio-Tracking Data. Academic Press, Inc. San Diego, California. P.391.

Zar, J.H. 1999. Biostatistical Analysis (Fourth Edition). Prentice-Hall, Inc. New Jersey.

Radio Tracking via Triangulation: Are Small Sample von Mises Maximum Likelihood Estimates Normal?

Richard M. Pace, III

United States Geological Survey, Louisiana Cooperative Fish and Wildlife Research Unit, 309 Forestry, Wildlife, and Fisheries Building,
Louisiana State University, Baton Rouge, Louisiana, USA


I conducted simulation experiments to examine distributional properties of von Mises Maximum Likelihood Estimates (VMMLE) under conditions frequently found in wildlife radio-tracking studies, that is 3 or 4 azimuths per location, standard deviations of 2° and 5° , under 3 tower configurations. Surprisingly, even when using only 3 azimuths per location estimate, VMMLE were relatively unbiased and sometimes bivariate normal. Location estimates were frequently skewed and often more peaked than a normal distribution. Average estimated variance-covariance (VC) components of VMMLE matched empirically generated (VC), but confidence ellipses did poorly at capturing true locations.


Implementation of Lenth's (1981) estimators transformed the art of radio tracking wildlife via triangulation into a well-defined estimation process. In particular, the von Mises maximum likelihood estimates (VMMLE) and associated confidence ellipse have become the standard location point and aerial estimators used in wildlife research when >2 bearings are collected per location. The attractiveness of VMMLE, aside from the fact that software is readily available for its use, is no doubt due to the well-known large sample of properties of maximum likelihood estimates such as consistency and normality. However, the typical wildlife radio-tracking operation employs very small sample sizes (e.g. n=3 or 4 azimuths per location). The properties of VMMLE under these small sample-size conditions have not been previously reported. I conducted a set of simulation experiments designed to examine the distributional properties of VMMLE under various conditions found in wildlife radio-tracking studies. In particular, I show that even though repeated, sample VMMLE are usually skewed from the bivariate normal, and resultant confidence ellipses do a poor job capturing the true location.


I conducted simulation experiments to examine distributional properties of VMMLE under conditions frequently found in wildlife radio-tracking studies. I used 3 tower configurations in a hypothetical 1 km2 study area: 1) 3 towers optimally located to minimize average confidence ellipse (White 1985), 2) 4 towers optimally located to minimize average confidence ellipse (White 1985), and 3) 3 towers arranged outside but along the southern edge of the study area (Figure 1). During each simulation experiment, I generated multiple sets of error-prone bearings from one of the 3 receiving tower configurations to fixed points within the study area. I treated bearing errors as being normally distributed with mean 0 and standard deviation of either 2º or 5º. I placed a 100 m grid over the study area and generated 1000 sets of bearings for all non-redundant grid points as judged from configuration symmetry (66 for 3-tower configurations and 36 for 4-tower configurations). For each of the 336 combinations of grid point, bearing error standard deviation, and tower configuration, I examined univariate normality of X and Y, Mardia’s tests of multivariate normality (skew and kurtosis), plots of estimated locations, the shape of 2-dimensional kernel density estimate (bandwidth 2), and actual coverage rates of 95% confidence ellipses. I also estimated location bias and compared VC components of the generated location estimates to means of estimated VC for individual locations.


I found abundant evidence for lack of bivariate normality for small sample VMMLE of location (Table 1). Although fitted 2 dimensional kernels revealed most tower configuration by beacon location by bearing precision combinations produce estimates with "normal-like" (i.e., smooth, monotonic, and unimodal) distributions, VMMLE were frequently skewed and commonly more peaked (large kurtosis) than the bivariate normal (Figure 2). Averaged over the entire region studied for the each of the various configurations, VMMLE were relatively unbiased (Table 1). Mean variance-covariance (VC) components of VMMLE matched empirically generated (VC) fairly well, but confidence ellipses did poorly at capturing true locations (Table 1). Although on average VMMLE characteristics were desirable, some combinations of transmitter locations and tower configurations, appeared were biased along one ore more axes and may have VC components different from those predicted by VMMLE.

Figure 1. Tower configurations used in a simulation study of radio tracking designed to examine distributions of small-sample von Mises maximum likelihood estimates of location. The three configurations were a linear arrangement of 3 towers (® ), 3 towers arranged to minimize the size of an average confidence ellipse (1 ), and 4 towers arranged to minimize the size of an average confidence ellipse (Ú ) (White 1985). Open circles were grid points for simulation of repeated radio tracking.


Von Mises maximum likelihood estimates with sample sizes of 3 and 4 perform surprisingly close to what is expected of maximum likelihood estimates asymptotically. However, a general assumption that VMMLE estimates are normally distributed for all locations within a study area is not warranted. Furthermore, calculated confidence ellipses poorly represent estimates relative to true locations. This suggests that we not abandon VMMLE, for it clearly performs well in many situations (Pace and Weeks 1990), but that we reconsider measures of quality about each estimate. For example, M. D. Samuel (personal communication, Madison, Wisconsin, USA) has developed a set of computer routines that attempt to adjust habitat use data by accounting for radio tracking precision (Samuel and Kenow 1992). Currently it generates subsamples about each estimated location based on a bivariate normal random number generator and the estimated VC components of VMMLE. I suggest that this and similar efforts would be improved if the investigator’s computer driven subsample procedures simulated actual tower configurations and bearing error structures rather than relying on estimated confidence ellipse to describe VMMLE distributions.

Table 1. Statistics used to assess distributional properties of VMMLE location estimates produced by simulating repeated (n=1000) radio tracking of beacons at known grid points in a hypothetical 1 km2 square study area under varying tower configurations and bearing error standard deviations (SD).


3 Towers

4 Towers








Grid Points















% Failure Rates for Certain Normality Tests

K-S (x)2







K-S (y)






















Mean Sample Attributes








VAR(X) 6







VAR(Y) 6







Bias of X (m)







Bias of Y (m)







1 Optimal for minimizing average ellipse size across a square study area (White 1985).

2 Kolmogrov-Smirnov univariate test on normality (P<0.001).

3 Mardia’s multivariate normal test of skewness (P<0.0001).

4 Mardia’s multivariate normal test of Kurtosis (P<0.0001).

5 Percent of true locations actually captured by 95% confidence ellipse.

6 Difference between empirical variance of X or Y based on simulation and mean estimated variance of X or Y from VMMLEs.

Figure 2. Example contours produced by 2 dimensional kernel fit to VMMLE estimates at 2 hypothetical fixed locations. Note the modest amount of skewness present in both distributions.


Lenth, R. V. 1981. On finding the source of a signal. Technometrics 23:149-154.

Pace, R. M., III and H. P. Weeks, Jr. 1990. A nonlinear weighted least-squares estimator for radiotracking via triangulation. J. Wildl. Manage. 54:304-310.

Samuel, M. D. and K. P. Kenow. 1992. Evaluating habitat selection with biotelemetry triangulation error. Journal Wildlife Management 56:725-734.

White, G. C. 1985. Optimal locations of towers for triangulation studies using biotelemetry. J. Wildl. Manage. 49:190-196.

Radio Telemetry of Echidnas and Platypus

Lyn A. Beard and Gordon C. Grigg

Physiological Ecology Laboratory, Department of Zoology and Entomology, University of Queensland, Brisbane, Australia 4072


We used implanted transmitters to track echidnas and record body temperatures year round. The resultant discovery of hibernation in these non-placental mammals (but not in their close relatives, platypus), sometimes in quite benign climates, poses questions about the origin of the ability to hibernate and the origin of endothermy itself. To help answer some of these, we developed a configuration for implantable transmitters to monitor heart rate in the field as an index of metabolic rate.


Australia is home to two out of the three surviving monotremes in the world, the short-beaked echidna (Tachyglossus aculeatus), and the platypus (Ornithorhyncus anatinus). The third, the long-beaked echidna (Zaglossus bruijnii), is found in New Guinea to Australia’s north. Monotremes are the only egg-laying mammals and have traditionally been considered among the most primitive of mammals. They are also poor homeotherms. However, Australia’s echidna is successful enough to be the only native mammal to enjoy an Australia-wide distribution, from hot, dry deserts to wet tropics to cold alpine regions. It was the occurrence of echidnas in the latter, even atop Australia’s highest peak, Mount Kosciusko (2,230 m) in winter that first prompted our study. If these poorly homeothermic mammals existed in Australia’s coldest region, were they living and breeding there by choice or were they merely lost?

When we began our study 13 years ago, a good deal of basic biological knowledge about this species had been gathered from specimens brought into captivity, serendipitous observations in the field, and dissection of dead specimens (Griffiths 1968, 1978). However, there was little known about their behavior in the wild. Although reasonably common, echidnas are extremely cryptic, rarely seen even by avid bushwalkers, and there is no way to attract them into a trap. Thus, radio telemetry offers the only real opportunity to follow and study these animals in the wild and there had been only a few short-term studies of this type (Augee et al. 1975, Griffiths et al. 1988) at the time we began ours.



We needed to be able to track our study animals and, as we were interested in the homeothermic capabilities of echidnas, particularly as they related to a cold climate, we required implantable temperature sensitive radio transmitters as well. Early attempts to attach external tracking transmitters to these animals for extended periods had failed. Echidnas have no neck for a collar and external gluing was short-lived due to their habit of pushing through tight holes and vegetation. So we learned, by default, to rely on the internal temperature-sensitive transmitter for both tracking and body temperature measurements.

We used a 2-stage transmitter (originally Austec Enterprises, Canada; more recently Sirtrack Limited, New Zealand) with internal loop antenna around an AA lithium battery operating on the 150-151 MHz band. To economize on battery life, our transmitters had a slower pulse rate than commonly supplied: about 1,500-1,600 ms pulse interval at 30o C. This gave the original transmitters a working life of more than one year with the lithium batteries available at the time and 2 ½ to 3 years, depending on temperature, with the current transmitters and 2,400 mAh batteries now available.

Each transmitter was coated in an elvax mixture (Grigg and Beard, these Proceedings) to render it biologically inert. The smooth, ovoid package weighed approximately 33 g (most of our experimental animals were between 2 kg and 4 kg with the largest over 6 kg) and was allowed to float freely in the body cavity. Transmitters were implanted under halothane anaesthesia (induced in a flow-through box at 5% and maintained on a mask at 2-4%) in the peritoneal cavity via a mid-ventral incision along the linea alba. The incision was closed with an internal layer of dissolving sutures (vicryl, Ethicon) followed by an external layer of vertical mattress sutures in silk. This type of suture was necessary to maintain the integrity of the healing wound, as echidnas otherwise could split their sutures by flexing very strong abdominal muscles when curling up into a defensive posture or digging. Silk was chosen because it is easy to tie, less prone to unravel than a monofilament type, and wears out more easily in the field. We released the animals as soon as possible following the operation, believing that the stress associated with holding a wild animal in captivity is a hindrance to successful recovery. All procedures were carried out under as sterile conditions as possible, following a protocol approved by the University of Queensland’s animal experimentation and ethics committee. Animals healed well. There was no infection and transmitters could be recovered cleanly and replaced in subsequent operations using the same incision site.

Animals were tracked using a Telonics (Mesa, Arizona, USA) TR2 receiver-scanner and RA2A hand held "H" antenna. As they are low to the ground and are often down burrows, our minimum range was 200 m to 300 m routinely in lightly timbered, undulating country, often up to 600 m, and over 1 km with good line of sight. As echidnas have a home range averaging 0.5 km2, this was usually sufficient. The record was 11 km from across a lake to an echidna in a log on a hill on the opposite shore. We also used the sometimes-disadvantageous side effect of uneven signal radiation from the internal loop antenna to deduce movement when the animal was out of sight, such as in a nursery burrow.

There were no great surprises at first when the animals were released in late summer. Body temperatures were quite heterothermic but with a consistent daily pattern (Grigg et al. 1989, 1992) ranging from 32o C to 33o C when active to 27 o C or 28o C during overnight inactivity (Figure 1). However, within two months, the power of telemetry was demonstrated when one of the echidnas was found under a rock pile with a transmitter pulse rate indicative of a body temperature of 9o C. The next day, however, it was found to have warmed up and moved. So we had evidence that echidnas can become torpid for short periods, which had been suspected from previous anecdotal reports.

Figure 1. Annual pattern of body temperature in a male echidna from Australia’s alpine region.

A bigger surprise was revealed after several more months tracking through winter when a classic pattern of true hibernation emerged, virtually indistinguishable from that of advanced eutherian mammals such as the arctic ground squirrel (Figure 1). Note the drop in body temperatures to almost ambient for extended periods, interrupted by periodic arousals to normal operating temperatures, as seen in northern hemisphere hibernators.


This discovery stimulated us to wonder about those other monotremes, platypus. The following winter we applied the same techniques to a population of these animals in the Thredbo River. This is one of the coldest rivers inhabited by platypus, being fed by meltwater from snow and with a mean winter minimum temperature of 2.8 oC. Platypus are smaller than echidnas (approximately 600-1,100 g), so the transmitters were fitted with ½ AA batteries and the AA-sized loop antenna was convoluted to fit the smaller package. Even with the resultant range attenuation, we could still get 100-400 m from an animal in the water, quite adequate for our purposes. We never found any evidence of hibernation in these animals (Grigg et al. 1992) but telemetry again revealed, within a couple of days, interesting behaviors such as burrow sharing by males, which has not been previously reported (T. Grant, personal communication).

Automatic Monitoring

The body temperature data for both echidnas and platypus were originally gathered manually, often camping within range of the animal to get hourly readings. However, this was time-consuming and tedious, so we invented a simple automatic recording system. This consisted of a specially constructed electronic timer that switched a tape recorder and radio receiver on and off at user-determined intervals, with the antenna mounted in a tree to maximize the receiving range (Grigg et al. 1990, Grigg and Beard, these Proceedings). The resultant tape needed to be read later in the lab, measuring the intervals between recorded pulses for each reading to determine temperature, but the system compressed continuous monitoring over days or even weeks into the length of a cassette tape.


These days, more complete body-temperature profiles may be gained by the implantation of a miniaturized, temperature-sensitive datalogger (e.g., Tidbit, Onset Corp., USA). The disadvantage of this equipment is that it is dependant on recovering the unit from the animal to download the data. If the animal is lost or the logger fails before the end of the observation period (we usually leave our animals in the field for over a year), all data is lost. But this is not the case if the data is gathered progressively, by monitoring the radio transmitter.

We have now recorded year-round body temperature patterns for echidnas from several different areas and are discovering that echidnas hibernate (albeit facultatively) even in the mild climate of southeastern Queensland, where frosts are comparatively uncommon and it rarely snows. Conventional thinking holds that hibernation is a means of overwintering in extreme conditions such as cold and/or lack of food. Neither of these scenarios holds true for echidnas. In the alpine areas, echidnas enter hibernation as early as February (late summer) when their food source of ants and termites is still plentiful and there is no snow cover. Moreover, they emerge from hibernation at the coldest time of the year (July-August) to breed. Echidnas resident in southeastern Queensland enjoy even more benign conditions. The challenge, then, is to explain why echidnas hibernate. Perhaps trying to understand this could hold the key to understanding the origin of the ability to hibernate, or even the origin of endothermy itself.

Heart Rate Telemetry

To this end, we were interested to learn more about how hibernation might figure in an echidna’s yearly energy budget. Pilot experiments indicated that heart rate could be used as a reasonably satisfactory predictor of oxygen consumption and, hence, metabolic rate. For this we needed to be able to monitor heart rate under natural conditions in the field throughout the whole hibernation season.

We had attempted heart-rate monitoring early in the study using heart-rate transmitters (Stuart Enterprises) configured, like the temperature transmitters, as smooth, ovoid, wax-coated bodies with the electrodes as stainless steel buttons embedded in the wax at each end. These transmitters were triggered by the QRS complex of the wave of electrical depolarization associated with each heartbeat and, like the temperature transmitters, were allowed to float freely in the body cavity.

Unfortunately, at least half of the time, the transmitters were silent or returned scrambled signals from random firing or triggering by muscular activity. We assumed this to be because of the poor coupling of the electrodes to the animal and inappropriate sensitivity. Nonetheless, we obtained some good data, including some from two animals in hibernation, that registered heart rates as low as 3-4 per minute at body temperatures around 10oC. We also obtained some tantalizing support for the current controversial theory that metabolic rate (as measured by heart rate in this case) is depressed during hibernation further than it would be simply as a consequence of the Q10 effect of lower body temperature (Grigg and Beard, 1996).

Recently, with Louise Kuchel, we sought to improve the methodology, this time using transmitters developed by Sirtrack Limited. Like the earlier transmitters, these new heart-rate transmitters are triggered by the difference in the electrical potential between two electrodes detected as the wave of depolarization associated with each heart beat spreads out through the body. This is more economical on battery power than continuous-wave models, which require the transmitter to be "awake" at all times and therefore allow for potentially long-term monitoring (up to a year). There are two important considerations. One is to have the electrodes from the transmitter far enough apart to detect the potential difference. The other is to have the electrodes well enough "coupled" to the animal to pick up what can be quite a weak signal in smaller animals.

At first we tried implanting the transmitter body subcutaneously on the ventral flank, with two longish electrode leads terminating in wire loop electrodes tunnelled under the skin to attachment points on the ventral surface. The best placement for the electrodes was found, using a Maclab physiograph (AD Instruments), to be with one near the posterior end of the sternum and the other on the lateral abdomen just anterior to the hind leg. However, the electrode leads broke, the electrodes themselves broke off, and the transmitter body migrated under the influence of gravity to a position uncomfortable for the echidna.

We decided, therefore, to try again with an intraperitoneal implant, and, due to the echidna’s extremely strong body musculature and habit of curling into a tight ball when disturbed, to leave it unattached to any body parts. We are now having good success with a stainless steel, 10 mm diameter button electrode on the body of the transmitter and a 10 x .6 mm "bullet" electrode held distant at the end of a 6 cm, sturdy but flexible whip made of stainless steel fishing trace inside silastic tubing.

The new transmitters also have increased sensitivity, maximum between 0.10 mv and 0.15 mv, and a small, user-adjustable sensitivity screw. Measurements with the Maclab, and experience, have shown that a sensitivity setting of about 0.3 mv is suitable for this transmitter configuration in echidnas. However, random firing, often triggered by muscular activity, can still be a problem, as can double firing if the transmitter picks up the "T" wave of the EKG as well as the targeted QRS complex.

To minimize these effects, the transmitter has a delay circuit so that, once triggered, it will not fire again within a factory-set time interval after the first stimulus. For echidnas, this latent period is set at 300 ms, which is just a bit shorter than the time between heartbeats at the maximum heart rate observed for these animals in the wild. In addition, use of a square, flat "Keeper" brand battery gives the whole transmitter a flattened body shape that encourages it to stay in the best orientation for picking up heart rate, with the button electrode facing toward the skin to minimize interference from muscular activity.

Our study animals now routinely carry a temperature-sensitive transmitter, heart-rate transmitter, and temperature datalogger, all in the peritoneal cavity. With this combination, we can achieve almost a full year’s monitoring of heart rate and up to three years’ body temperature data both in real time and electronically archived. The combined weight of the hardware is approximately 85 g, which is less than 0.5% of the body mass of the 2.5-4.5 kg animals in our study. As far as can be ascertained, there seems to be no interference with the animals’ normal behavior. Telemetered individuals breed successfully and exhibit spectacular pre-hibernation weight gains.

Future Plans

As to the future, we would like to see further miniaturizations in dataloggers, especially for implantation; increased battery energy densities; and, hopefully, commercial realization of heart-rate loggers, all of which would greatly facilitate studies of this type. Also, we would like to see the development of implantable, long-lived loggers or archival tags that can be downloaded and reprogrammed remotely by radio while within the animal.


We are grateful to the transmitter manufacturers, Austec and Sirtrack, who have always been ready and willing to help with technical problems and provision of equipment at (inevitably) short notice. Telonics have been similarly helpful with maintenance of receiving equipment. Hastings Data Loggers (Port Macquarie, Australia) was extremely prompt with supply and advice on dataloggers. We would also like to thank Joerg Suchau, electronics technician at the University of Sydney, for designing and building the timers crucial to the automatic recording system described in this paper. A. Woakes and P. Butler at the University of Birmingham were responsible for the helpful suggestion of using a "bullet" electrode for our latest configuration of the heart rate transmitter. Thanks are due also to T. Grant, who joined us in the field; to the many willing (often volunteer) assistants including J. Fletcher, who helped gather data in the field; and to our colleague in the early work, M. Augee, who read more tapes than he cares to remember.


Augee, M.L., E.H.M. Ealey and I.P. Price. 1975. Movements of echidnas, Tachyglossus aculeatus by marking-recapture and radio-tracking. Australian Wildlife Research 2:93-101.

Grant T.R., Beard L.A. and G.C. Grigg. 1992. Movement and home range of platypus in the Thredbo River, NSW. Pages 263-267 in M.L. Augee (Ed.) Platypus and Echidnas, Sydney. Roy. Zool. Soc. NSW.

Griffiths, M. 1968. Echidnas. Pergamon Press, London.

Griffiths, M. 1978. The Biology of the Monotremes. Academic Press, New York.

Griffiths, M., F Kristo, B. Green, A.C. Fogerty and K. Newgrain. 1988. Observations on Free-living, lactating echidnas, Tachyglossus aculeatus, (Monotremata:Tachyglossidae), and sucklings. Australian Mammalogy 11, 135-144.

Grigg, G.C., L.A. Beard and M.L. Augee. 1989. Hibernation in a monotreme, the echidna Tachyglossus aculeatus. Comp. Biochem. Physiol. 92A:609-612.

Grigg G.C., Beard L.A. and M.L. Augee. 1990. Echidnas in the high country. Aust. Nat. Hist. 23(7):528-537.

Grigg G.C., Augee M.L. and L.A. Beard. 1992. Thermal relations of free-living echidnas during activity and in hibernation in a cold climate. Pages 160-173 in M.L. Augee (Ed.) Platypus and Echidnas, Sydney. Roy. Zool. Soc. NSW.

Grigg, G.C. and L.A.Beard. 1996. Heart rates and respiratory rates of free-ranging echidnas – evidence for metabolic inhibition during hibernation? Pages 13-21 in Geiser, F., A.J. Hulbert and S.C. Nicol (eds) Adaptations to the Cold: Tenth International Hibernation Symposium. University of New England Press, Armidale.

Grigg, G.C. and L.A. Beard. Application of radiotelemetry to studies of the physiological ecology of terrestrial vertebrates. This volume.

Long-Time Recording of Behavior Rhythms in
Red Deer and Przewalski Horse:
Methods, Results and Importance

Anne Berger and Klaus M. Scheibe

Institute for Zoo Biology and Wildlife Research Berlin, Pf 601103,
10252 Berlin, Germany

Alain Brelurut

Institut National de la Recherche Agronomique, Theix,
63122 St-Genes-Champanelle, France

Knut Eichhorn, Annemarie Scheibe and W. Jürgen Streich

Institute for Zoo Biology and Wildlife Research Berlin, Pf 601103,
10252 Berlin, Germany


The storage-telemetry-system ETHOSYS was used on Przewalski horse under semi-reserve conditions and on enclosure-kept red deer in a one-year investigation of the time pattern of activity and feeding. A specific time pattern of behavior parameters was obtained from either species for one year. Various stress conditions of either species could be identified and evaluated by biorhythmic analysis of these continuous behavior data.


Behavior as an organism-environment relationship on the basis of exchange of information is a source of knowledge about the relationship between the organism and the environment. Behavior can also be used as an indicator of internal state of the individual. The internal state of an individual may be judged from time budgets as well as from circadian rhythms and shorter rhythms (known as ultradian rhythms). Synchronization of internal periodicities with environmental periodicities is an elementary mechanism of adaptation. If the individual comes under stress conditions, these complex rhythmic couplings will be disturbed, and the population-specific pattern of daytime activity will be modified. Synchronization between the organism and the environment can be measured by biorhythmic analysis. We used biorhythmic analysis to identify environmental conditions, which the animals experience as stressful.

Biorhythmic analysis depends on the capability to measure and recognize biological rhythms. This calls for long-term, continuous recording of parameters, which are characterized by standardized measurement intervals and constant measurement accuracy.


The storage-telemetry-system ETHOSYS (Scheibe et al. 1998) allows for non-invasive, continuous, and long-term registration of activity and feeding in large, grazing animals. A motion sensor with an adjustable time window and a tilt switch in a collar scans movements and positions of the head. Any movement which is transmitted to the collar is defined as "activity" and stored in channel one. Downward bending of the head together with movement characteristic of food intake is evaluated as "feeding" and stored in a second channel. All data measured are stored in the collar at pre-selected intervals. When the animal is within the reception range of a central station, these data are radioed to that station from which they can be picked up independent of the animal's current position. The telemetry system is available as a commercial set under the name ETHOSYS.

Daily phases of activity and feeding were selected. Phases from which no activity was recorded were defined as rest phases. Daily totals were calculated for activity and feeding and the percentage of activity in daylight hours was related to that in dark hours for each day.

Those initial macroscopic analyses were followed by closer structural investigation of each of the time series (Scheibe 1999). The original data series were subdivided into data sets for seven consecutive days, with a delay by one day between subsequent data sets. From these data sets, autocorrelation functions were calculated. From the autocorrelation functions, power spectra were derived. The periods of these power spectra were tested for significance, which finally gave the significant periodic components of the original data series.

The values of these significant periods are used to calculate the "Degree of Functional Coupling" (DFC), which expresses the percentage of the total intensities of significant and harmonic periods to the intensities of all significant periods. Harmonic periods, in this context, are defined as periods that are synchronized by an integer-number relationship with the circadian rhythm (this means 24 hours divided by 1, 2, 3 and so on gives harmonic periods). DFCs can vary between 100% (which implies total synchronization between organism and environment) and 0% (which implies total desynchronization caused by disturbance of the internal state of the organism).

We studied four Przewalski horses (Equus ferus) and four red deer (Cervus elaphus) in semi-free ranging groups under more or less stress-free conditions. The horses belonged to a group of twelve mares kept in the 44-ha enclosure of semi-reserve Schorfheide/Liebenthal. Two of the deer lived in an enclosure in Slovakia of about 40 ha; the other two came from enclosures of INRA in France of about 2 ha.


Figure 1 shows the general distribution of activity and feeding for horses and deer. For the entire year, the two species showed a polyphasic daily pattern with activity peaks at sunrise and sunset. The intensity of feeding was generally lower than the intensity of activity in both species.

In Przewalski horses, some of the seasonal variations in time patterns occurred in a leapwise manner within the same week. For instance, in summer, there was a sudden change from predominant daytime activity to predominant overnight activity. Horses drastically increased activity and feeding as soon as fresh plants began to emerge and air temperature went up in spring. High feeding activity occurred also in autumn before food became rare in wintertime. In red deer, activity and feeding decreased in wintertime. All seasonal variations in time patterns occurred as gradual transitions.

Figure 2 presents the monthly variation in the daily totals of activity and feeding for horses and red deer. In Przewalski horses, annual activity and feeding levels vary independently from one another. Activity was at its highest in summer and lowest in winter, but feeding was at its highest in autumn and spring, with minimum values being recorded in summer. Annual average of daily activity was 64%. In red deer, activity was reduced in the winter season and was high in summer, with floating transitional periods in between. Maximum and minimum values of the two behavioral parameters did not deviate from one another over the course of the year. The annual average of daily activity (33%) was generally lower as in Przewalski horses.

Figure 1. General distribution of activity and feeding for Przewalski horses (above) and red deer (below) in the course of one year. Each day is represented by a horizontal line. Days follow each other from top to bottom. The behavior intensity increases from white over gray to black.

Figure 2. Monthly variation in the daily totals of activity (above) and feeding (below) for Przewalski horses and red deer.

Figure 3 shows monthly variation in the ratio of daytime and night behavior. In horses, July was the only month with higher activity and feeding at night than during the day. In all other months, daytime activity and feeding were higher than those at night. Strong overnight activity in summer can be attributed to high air temperatures and disturbances by flying insects during the day. In red deer, daytime activity and feeding were higher than overnight activity and feeding from February to April. A trend to higher overnight activity was observed from June through September. Transitions, however, were fluent between those phases.

Figure 4 presents results of monthly power spectra of activity and feeding for both species. Weighted means of intensities of significant and harmonic periods are depicted. In Przewalski horses, time pattern of activity and feeding was moved to lower frequencies (up to 18-hour period length) at times of good food supply (during autumn and spring). In red deer, periods were moved to higher frequencies up to 3 hours in period length at times of high food supply. These variations in the ultradian structure of activity and feeding were due to seasonal variations of nutritional conditions and can be an indicator of the different species-specific feeding strategies.

Figure 3. Monthly variation in the ratio of day- and night-activity (above) and day- and night-feeding (below) for Przewalski horses and red deer.

Figure 4. Weighted means of intensities of all significant and harmonic periods of monthly power spectra for activity and feeding for both species.

DFCs were used to identify and evaluate irritative stimuli under usually nature-like conditions. Various stress conditions could be recognized in both species. DFCs over one year from one Przewalski horse are presented in Figure 5. DFCs varied around a medium level (about 65%) but, on three occasions, they were reduced to or close to zero. The first two phases were related to hunting activity in the surrounding area. The third phase of reduced DFCs was in June and coincided with the start of shooting at a firing range that opened 1 km away.

Figure 5. Degree of Functional Coupling (DFC) for activity over one year of one Przewalski horse under semi-reserve conditions.

The DFCs for activity and feeding of one red deer for the entire period of measurement are given in Figure 6. Mean DFC values were 82% for activity and 83% for feeding. Sudden drop of DFCs to minimum values was in all three cases attributed to sudden human interference with the environment of the animals. The DFC for activity dropped to 0% on August 21, the day that the animal's antlers were cut off. The DFC for feeding dropped to 0% on September 15 and May 1, days that the feeding regime was changed. DFCs declined within behavior that was directly affected by the disturbance. So a change of food led to minimum values of feeding DFCs, but was of no impact on activity DFCs.


Biorhythmic analysis based on long-term records of continuous time series can give on the one hand information on general population-specific way of life (such as activity levels, level in light and dark time or

Figure 6. Degree of Functional Coupling (DFC) for activity and feeding over one year of one red deer kept in an enclosure of about 40 ha.

predominant periods in behaviors) and on the other hand information on the internal state of animals by harmonic analysis. This internal state is influenced by physiological processes (for instance the rut, birth or illness), as well as by human disturbances and other forms of stress. Assessment of the internal state is important for species-specific keeping methods, breeding, identification and evaluation of stress in wildlife and the possible successful renaturalization of different animals.


Scheibe, K. M., T. Schleusner, A. Berger, K. Eichhorn, J. Langbein, L. Dal Zotto and J. Streich. 1998. ETHOSYS - New System for Recording and Analysis of Behaviour of Free-Ranging Domestic Animals and Wildlife. Appl. Anim. Behav. Sc. 55:195-211.

Scheibe, K. M., A. Berger, J. Langbein, J. Streich and K. Eichhorn. 1999. Comparative Analysis of Ultradian and Circadian Behavioural Rhythms for Diagnosis of Biorhythmic State of Animals. Biol. Rhythm Res. 30(1):1-18.

Passive Integrated Transponders as a Method for Relocating Legless Lizards in Underground Habitats

Linda A. Kuhnz

Moss Landing Marine Laboratories, Post Office Box 450,
Moss Landing, California 95039, USA


Passive Integrated Transponders tags and a mobile scanner are used to relocate legless lizards underground. Lizards can be found in various microhabitats and as deep as 11.5 cm in the soil (0 = 10.6), within the depth where they are presumed to reside most of the time. Due to the small size of adult animals (5-8 mm diameter), new methods were developed to avoid injury and mortality due to tagging.


The California legless lizard (Anniella pulchra) is a rare, burrowing animal that lives underground, primarily in sandy soil. They are difficult animals to study because of their subterranean life history. Populations of these lizards, found mostly along the coast of California, have been severely reduced because of anthropogenic disturbances. They are protected as a Species of Special Concern in the state of California (California Department of Fish and Game, 1999). Legless lizards are small and snake-like in appearance, averaging 2.9 g in weight, 179 mm in length and 6 mm in diameter (Figure 1). To date, the accepted, non-destructive method for locating legless lizards consisted of placing wood cover-boards on the soil surface. These boards are surveyed periodically by digging under them to check for the presence of lizards. Due to the cryptic nature of the lizards, no previous studies have produced information regarding their movement patterns or microhabitats.

Traditionally, the use of Passive Integrated Transponder (PIT) tags in field biology has been primarily limited to the identification of manually recaptured animals or those that must pass near a stationary tag reader (Camper and Dixon 1988, Germano and Williams 1993, Jemison et al. 1995, Parmenter 1993). The methods developed for this study allow us to track the activity of over 600 individual animals in their underground environment without recapturing or disturbing them, potentially with less bias toward slow or easy to capture lizards.

Figure 1. The California legless lizard, a cryptic animal that lives in small, underground burrows.

This study is being conducted at Moss Landing on the coast of central California. The Moss Landing Marine Laboratories, located on the shores of Monterey Bay, were destroyed in the Loma Prieta earthquake of 1989. Surveys of the rebuilding location revealed a large local population of legless lizards. In 1997-1998, the site was searched to depletion and 3500 Anniella were moved a short distance to an area of restored sand dune habitat.



Six hundred sixty-six lizards were randomly chosen for tagging from among animals with a snout-vent length (SVL) greater than 110 mm and weight greater than 2.5 g recovered from the Moss Landing Marine Laboratories rebuilding site. This size class was selected based on an evaluation of the internal anatomy and concern that tagging smaller lizards could be detrimental. Some animals fitting these criteria were later rejected if their diameter appeared too small to accommodate the tag without risk. A small group of lizards (n=25) was tagged in October 1997, six months before the remaining lizards were marked. The general health and survival of this group was used to determine the feasibility of mass lizard tagging. This group of lizards is being held in the lab for long-term monitoring.

Six hundred forty-one lizards were tagged between May 22 and June 30, 1998. All lizards were implanted intra-abdominally with AVID #2023 PIT-tags, 11.5 mm x 2.1 mm in size and weighing 0.06 g. Each glass-encased unit was encoded with a unique 10-digit alphanumeric sequence. Six biologists conducted the tagging. Initially, tags were injected using a 12-gauge needle, coated in antibiotic ointment, introduced into the abdomen slightly to the right of the ventral midline, anterior to the gonads and posterior to the lungs and liver. To facilitate tissue closure, the incision was held together for 10-20 s. Refinements in procedures were made and subsequent tagging was accomplished by inserting the needle only far enough to create an opening in the peritoneum and by massaging the tag through the incision. During their recovery, lizards were held in separate plastic boxes (35 cm length, 20 cm width, 10 cm height) with modified lids for air circulation and containing sand about 15 cm deep.

Lizards exhibiting visceral protrusion or signs of infection were injected subcutaneously with 1 mg/ml Amikacin (5 mg/kg wt.) using a 28 gauge sterile needle. Maintenance doses were half the initial treatment. Betadine was used as a topical antiseptic. Lizards that were moribund were euthanized with chloroform.

Tag Detection

The external scanning device was an AVID Power Tracker II (Figure 2), normally used as a hand-held detector. The reader emits an electromagnetic field that activates the tag so that the unique code is transmitted back to the scanner. Modifications for use in the field included the addition of a 1.5 m extendable aluminum pole, the addition of a reset switch on the pole, and a revision to the power source to accommodate a double external Nicad battery. A pivot was placed at the base of the pole so that the orientation of the reader could be changed on slopes. To search for lizards, the reader was swept as close to the ground as possible without causing disturbance. Circular movements were used to minimize the effects of tag orientation on sensitivity.

To evaluate the effects of tag orientation on tag sensitivity and to ensure that all tags could be sensed from the same distance, loose, unburied tags were tested with the scanner from a constant distance. PIT tags were placed in various positions during testing (horizontal, vertical and diagonal).

The depth at which loose, buried tags could be detected underground was assessed by burying PIT tags in dry sand and recording the greatest depth at which the tag could be read. The effect of moisture was tested in the same manner by placing tags in wet sand. Tags also were tested for sensitivity variability in duff-covered sand and duff-covered wet sand. PIT-tagged lizards were placed in a plastic container that prevented them from moving. These containers were buried under progressively deeper sand (dry and wet) until the tag could no longer be detected. All detection depths were recorded to the nearest 0.5 cm.

Figure 2. The PIT tag detection device was an AVID Powertracker II, which was modified for use in the field.


Marked lizards were released in a series of experiments (i.e., short-term dispersal studies) into large field enclosures, and randomly throughout a 5.8-acre area. Approximately 41% of this habitat was searched biweekly. Large areas were covered in dense bushes and berry thickets that are inaccessible to searchers. Other lizards were being held in the lab for long-term observation and for microhabitat selection studies.


Lizard Tagging

A total of 666 lizards were PIT-tagged (weight range = 2.6 - 7.3 g, 0 = 4.7 g; SVL range = 112 - 172 mm, 0 = 133 mm). Lizards were held in the lab for 15-91 days before release in the field. Thirty-three lizards (4.95%) died, possibly due to tagging. Lethal injuries consisted of visceral protrusions, probable bowel obstruction, and wound infection (Table 1). The mortality rate was 6.15% for lizards tagged during the first four tagging sessions. This rate dropped to 3.75% for sessions 5-6, and the mortality rate for lizards tagged thereafter dropped to 0%. Twenty other lizards had minor visceral protrusions, wound infections, or subcutaneous tag placement. These animals were successfully treated with minor surgical procedures and antibiotics.

Table 1. Possible causes of mortality due to PIT tagging. The number of lizards affected and the percentage of the total tagged are shown.

PIT Tag-Related Injuries

Number of Lizards Affected

Percent of Total

Intestinal protrusion



Postcaval vein protrusion



Wound infection



Bowel Obstruction






Twenty-four of the 25 original tagged lizards were held successfully in the lab for 18 months with no apparent detrimental effects. We observed that movement and speed did not appear to be affected in the PIT-tagged animals. To date, tag loss has been <1% and none of the tags appear to have malfunctioned or broken. Thirty-five tagged lizards were recaptured 9-10 months after they were released in the field. They were very active and well fed. PIT-tag wounds were completely healed, in most cases with little or no scarring. We will continue to periodically recapture lizards for assessment over the coming years.

Tag Detection

There was little variability in the sensitivity range of loose, unburied PIT tags when tags were in the horizontal or diagonal position (Table 2). Buried, loose tags in the horizontal position could be read to nearly the same depth under dry sand, wet sand, duff-covered dry sand, and duff-covered wet sand. Readings for buried tags in the vertical position were highly variable with a mean depth range of about half that of those in a horizontal position.

Tagged Animal Studies

In short-term dispersal studies conducted in the field, 87-100% of the lizards were detected within 10 days of release, many of them more than once. During the period from August 1998 to May 1999, all of the lizards released into large field enclosures were relocated and 51% of the free-roaming, randomly released lizards were detected at least once; 42% of those detected have been relocated more than 2 times. Some lizards have been found on a regular basis, up to nine times. Preliminary results show that although legless lizards are capable of moving 7 m or more in a 4-hour period, most remain within a few meters of their release site over long periods. Closely spaced sampling intervals (i.e., every 15 minutes) showed that the animals move frequently.

Table 2. The results of testing for PIT-tag readability under varied conditions. Tags were implanted in lizards two weeks prior to testing.

Tag-Sand Condition



Depth Range (cm), 0


Loose, Unburied



10.0-10.5, 10.5


Loose, Unburied



10.5-10.5, 10.5


Loose, Buried, Dry



9.5-12.0, 10.4


Loose, Buried, Dry



2.0-10.5, 5.2


Loose, Buried, Wet



10.0-12.0, 10.7


Loose, Buried, Duff-Covered, Dry



9.5-11.0, 10.5


Loose, Buried, Duff-Covered, Wet



10.0-10.5, 10.5


Implanted in Lizard, Dry



9.0-11.5, 10.5


Implanted in Lizard, Wet



10.0-11.5, 10.7



Comprehensive studies of legless lizards can only be carried out through the PIT-tagging of animals. External methods of tagging legless lizards (Higgins eternal ink, permanent ink marker) are short-term or unreliable (personal observation) and due to morphological differences, other common methods of marking lizards and snakes (e.g., toe or scale clipping and freeze branding) are not possible with Anniella.

The intra-abdominal placement of PIT tags for long-term marking is essential for animals that live in a fossorial environment. Tag placement in the peritoneum slightly increases the probability of injury to the animal; however, the loss of tags placed subcutaneously from friction is unacceptably high (Jemison et al. 1995, Germano and Williams 1993). Subcutaneously placed tags can be broken or easily lost through skin lesions (Camper and Dixon 1988, Germano and Williams 1993, Keck 1994). The small diameter of legless lizards (5-8 mm) is a challenge; protruding viscera is the major problem to overcome. The incidence of protrusion was higher in lizards that were in an early gravid state and whose abdominal turgor pressure may have been higher than normal. It is clear that mortality can be avoided or reduced by experienced workers and by using the modified tagging technique we developed.

Implanted PIT tags can be expected to continue to function appropriately for the life of the lizards. Failure of the tags is rare and they are maintenance free. They can be expected to be readable for 15-20 years (Camper and Dixon 1988, Germano and Williams 1993, Jemison et al. 1995, Parmenter 1993). Field tagging of lizards is not only possible, but also desirable. The application of a bio-compatible glue over the injection site may decrease the incidence of visceral protrusion and offer additional protection against infection.

The methods developed for this study can be used in other investigations of fossorial or cryptic animals. Studies of manually recaptured, PIT-tagged animals have been compromised when the subjects become either skittish or used to being handled. Modifying the tag reader so that these animals can be detected from a distance may be an improvement.

As data from our studies accumulate, we find that the relocation rate of tagged lizards is very high, considering our limited access into impenetrable vegetation. Legless lizards appear to burrow deeper and become more dormant during winter and, as the warmer weather returns, we have detected many "missing" lizards. We would welcome the development of better equipment with longer distance tag detection ranges. Faber (1997) has obtained a 15 cm range using similarly sized transponders. This study will continue for five years or more, and we look forward to gaining new insights into legless lizard population redistribution, the effects of habitat heterogeneity on movement, and the longevity of these lizards.


Procedure development was coordinated with Drs. Norman Scott, James R. Dixon and Stephen B. Ruth. Methods were approved and conducted under a Memorandum of Understanding with the California Department of Fish and Game. Special thanks to Thomas Kieckhefer for his technical expertise and to Peter Slattery, James Oakden, Jo Guerrero, Skyli McAfee, Kristy Uschyk, Daniel Grout and other members of the Benthic Ecology Laboratory for their support and assistance. Major funding was furnished by the California State University system. Facilities and materials were provided by the Benthic Ecology Laboratory, Moss Landing Marine Laboratories and ABA Consultants.


California Department of Fish and Game, 1999. Amphibian and reptile species of special concern in California.

Camper, J.D. and J.R. Dixon. 1988. Evaluation of a microchip marking system for amphibians and reptiles. Texas Parks and Wildlife Department, Research publication, 7100-159:1-22.

Faber, H. 1997. Der einsatz von passiven integrierten Transpondern zur individuellen markierung von bergolchen (Triturus alpestris) in Freiland. Naturschutzrelevante Methoden der Feldherpetologie 7:121-132.

Germano, D.J. and D.F. Williams. 1993. Field evaluation of using passive integrated transponder (PIT) tags to permanently mark lizards. Herpetological Review 24(2):54-56.

Jemison, S.C., L.A. Bishop, P.G. May and T.M. Farrell. 1995. The impact of PIT-tags on growth and movement of the rattlesnake Sistrurus miliarius. Journal of herpetology. 29(1):129-132.

Keck, M.B. 1994. Test for detrimental effects of pit tags in neonatal snakes. Copeia 1:226-228.

Parmenter, C. J. 1993. A preliminary evaluation of the performance of passive integrated transponders and metal tags in a population of the flatback sea turtle (Nataor epressus). Wildl. Res. 20:375-381.

Radio Tracking via Triangulation: A Simple Moving Window Estimator for Describing Movement Paths

Richard M. Pace, III

United States Geological Survey, Louisiana Cooperative Fish and Wildlife Research Unit, 309 Forestry, Wildlife, and Fisheries Building,
Louisiana State University, Baton Rouge, Louisiana, USA


For many vertebrates, radio tracking is the only practical method to gather movement path data; but, because individual location estimates have low precision, they may poorly represent a travel path. I used a moving window to pool bearings collected during brief intervals and estimated locations along travel paths in 2 field trials. The increased sample sizes improved estimated location precision and more closely tracked the true path compared with raw estimates. Estimates were easily produced with existing software.


Insight into hierarchical scaling in ecology has stimulated interest about the use of animal movement paths to measure response to spatial scale (Turchin 1998). Testing hypotheses in this area of interest requires that movement path data be acquired and compared under differing experimental conditions (e.g. 2 levels of environmental heterogeneity). Radio tracking seems to be a reasonable means by which to acquire these data, and may be the only practical means to gather movement path data on some vertebrates. Unfortunately, radio-tracking data—if gathered via triangulation—can be imprecise relative to the scale at which one wishes to study movement phenomena (White and Garrott 1990).

Location data useful in describing movement paths are necessarily spatially correlated: they represent sets of points (states) along a continuous course of movement. Anderson-Sprecher (1994) used state-space, time-series modeling estimates that incorporated dependence among successive observations to build robust estimates of wildlife location. Using his estimation procedures to develop movement paths from radio tracking data encounters two constraints. First, his iterated, extended Kalmann-filter-smoothing estimators require a moderately complex set of computations not readily accessible to wildlife ecologists. Second, the state-space model incorporated in these estimators assumes a random walk by the animal, and a random walk means no periodically directional movement. However, I have frequently observed animals tracked during half-day tracking sessions showing periodic directional behavior.

To overcome potential deficiencies in the state-space approach, I investigated the usefulness of moving windows for estimating travel paths. Herein, I describe the technique, demonstrate its use in 2 field trials, and compare performance against time-independent location estimates.


Radio Tracking

A radio tracking system established to collect travel path data might consist of 2 or more receiving stations (towers) of known location from which sets of bearings to the target animal are measured periodically. For example, in southern Louisiana, my students and I have used 2 or 3 people with truck- or boat-mounted antennas (mobile stations) simultaneously measuring bearings every 5 minutes to describe the movement paths of coyotes (Canis latrans) and black bears (Ursus americanus). Using 3 people is advantageous because if 1 tracker has to relocate his mobile antenna in order to improve his reception or improve the overall antenna configuration, the time series of location data is not interrupted. Hence, the result of a tracking session is a multivariate time series of sets of bearing-station pairs.

Standard procedure for estimating travel paths from these data would be to use an estimator such as von Mises maximum likelihood (VMMLE) or Andrews (Lenth 1981) estimates to combine sets of bearing-station pairs for a given time point to produce an estimated location for that time point (time-independent estimate). With 2 or 3 bearings per location estimate, these estimated locations are generally of low precision and quite often skewed (White and Garrott 1990; Pace, these Proceedings). One alternative is to use an iterated, extended Kalman-filter smoother to estimate travel path in a state-space analysis (Anderson-Sprecher and Ledolter 1991).

Moving Windows

A simple modification of these time-independent estimates would be to use all bearings measured during a reasonably short interval of time (a window) in the VMMLE or Andrews procedures to produce each estimated location, then moving the window 1 time step to estimate the subsequent location. That estimate should be considered a smoothed estimated location for the average time of the combined bearings. For tracking coyotes and bears, we used 10 minutes as the interval over which we combined bearings. Thus, if 3 observers each recorded a bearing on the target animal at 5-minute intervals during a 6-hour tracking session (e.g., 1200 hours, 1205 hours, …, 1800 hours), then the total of 9 bearings recorded at 1215 hours, 1220 hours, and 1225 hours would be combined to estimate location at 1220 hours. Similarly, bearings recorded at 1220 hours, 1225 hours, and 1230 hours would be pooled to estimate location at 1225 hours, and so on.

Travel Paths

To examine performance of moving-window estimation at describing travel paths, I used 4 simulated tracking sessions and 1 actual tracking session of 2 field-trial travel paths. Travel paths resulted from my attempt to mimic the travels of a black bear over a 6-hour period by meandering through mostly forested bear habitats, wearing a radio collar, and traveling approximately the same speed and total distance of bears observed in my study. Locations along a path were recorded and stored every 1.25 minutes using a military-grade Global Positioning System (GPS) receiver (± 3 m). The radio tracking system consisted of 3 operators using 2 truck-mounted and 1 stationary null-peak antenna arrays. Operators were instructed to record bearings to a primary target (me) and a stationary beacon every 5 minutes. Operators used 2-way radios to coordinate activities and mobile operators were free to move so as to improve reception or configuration. Repeated bearings on the stationary beacon were used to calibrate antennas and estimate bearing precision.

I used the 3 most frequently used receiving stations, the estimated standard deviation (3.0° ) for the radio tracking system, and the parameters governing actual tracking to simulate 2 radio tracking sessions on each of the 2 observed travel paths. That is, I selected a location at each 5- minute interval as determined by GPS, calculated true directions to the 3 receiving stations, and added a normal (m =0, s =3) random error to each true bearing to simulate measurement errors in radio tracking (Pace 1988). Simulated sessions used all 3 receiving stations or the 2 stations never split by the paths. For both observed and simulated tracking sessions, I calculated VMMLE estimates using both single-time (n=2 or 3) and 10-minute moving window (n=6 or 9) bearing sets. I calculated distances between estimated locations and their concomitant true location in time along the travel path and plotted empirical distribution functions for both single-time and moving window-based estimates.


Using a moving window through time to pool sets of azimuths was clearly superior to time-independent location estimation (Figure 1). Median distances between estimated locations and their same-time-equivalent true locations were 79 m, 64 m, 48 m, and 42 m for window-based estimates compared with 108 m, 114 m, 58 m, and 55 m for time-independent estimates of travel paths for Path 1 with 2 towers, Path 2 with 2 towers, Path 1 with 3 towers, and Path 2 with 3 towers, respectively. Pooling azimuths through time acted to smooth perceived paths by removing some of the abundant "white noise" that often occurs at the same or larger scale than behavioral phenomena such as movement paths (Figure 2). Smoothing had a profound effect on perceived travel path statistics calculated for radio tracking location estimates. For example, the time-independent estimate for Path 1 when tracking was accomplished with 2 towers, had angular concentration of 0.088, angular deviation 126, linearity of 0.065, and mean distance between locations of 150 m. These compared unfavorably to the window–based estimator that had angular concentration of 0.11, angular deviation 120, linearity of 0.096, and mean distance between locations of 93 m, which were quite close to the actual path statistics.

Figure 1. Empirical distributions of distance errors incurred in the estimation of 2 travel paths from simulated radio tracking based on 2 or 3 towers with bearing SD=3° .

Figure 2. Predicted travel paths (diamonds) based on time-independent VMMLEs (A) or ~10-minute moving window VMMLEs (B) overlaid onto true paths of a meandering biologist. Radio tracking data were simulated based 2 towers with bearing SD=3° .


Based on simulation experiments, performance gains of the moving-window approach over time-independent estimates were more dramatic for 2-tower tracking trials (Figure 2) than for 3-tower tracking trials (Figure 3). This was important because movement scale of the meandering biologist was large in relation to tower arrangement and equipment sensitivity. Hence, during actual tracking sessions, it was necessary for one or the other of the mobile stations to relocate several times during the tracking session which would cause wide fluctuation of location precision using time-independent estimates. During actual tracking sessions, I observed several spurious (>15° error) bearings occur. Increased sample sizes greatly increased the precision of individual estimated locations and afforded the opportunity to remove spurious bearings directly or to reduce their influence through the use of a "robust" estimator such as Andrews (Lenth 1981), rather than allowing them to distort the character of the movement path.

Figure 3. Predicted travel paths (diamonds) based on time-independent VMMLEs (A and C) or 10-minute moving window VMMLEs (B and D) overlaid onto true paths of a meandering biologist. Radio tracking data were simulated based 3 towers with bearing SD=3° .

The moving window approach had 2 advantages over the state-space analysis approach of Anderson-Sprecher and Ledolter (1991). Estimates were easily produced with existing software. They require no assumed movement model, so requirements on the form of a random walk (nondirectional and isotropic) are not needed. The ease of implementation and good performance of the moving window approach makes it a potentially valuable tool in the estimation of travel paths.


This work was sponsored by the Louisiana Cooperative Fish and Wildlife Research Unit: Louisiana Department of Wildlife and Fisheries, Louisiana State University Agricultural Center, United States Geological Survey, Biological Resources Division and the Wildlife Management Institute cooperating. I am grateful for field assistance given by my students, D. Hightower, M. Giordano, and J. Pike.


Anderson-Sprecher, R. 1994. Robust estimates of wildlife location using telemetry data. Biometrics 50:406-416.

Anderson-Sprecher, R. and J. Ledolter. 1991. State-space analysis of wildlife telemetry data. J. Am. Stat. Assoc. 86:596-602.

Lenth, R. V. 1981. On finding the source of a signal. Technometrics 23:149-154.

Pace, R. M., III. 1988. Measurement error models for common wildlife radio-tracking systems. Minn. Wildl. Rep. 5. 19 p.

Pace, R. M., III. Radio tracking via triangulation: Are small sample size von Mises maximum likelihood estimates normal? These Proceedings.

Turchin, P. 1998. Quantitative analysis of animal movement: measuring and modeling population redistribution in animals and plants. Sinauer Associcate, Inc., Sunderland, Massachusetts, USA. 396p.

White, G. C., and R. A. Garrott. 1990. Analysis of wildlife radio-tracking data. Academic Press, Inc., San Diego, California, USA. 383p.

Radio Telemetry Error and Wildlife Habitat Analyses

Karen M. Goh

Wildlife Research Group, Faculty of Agricultural Sciences,
University of British Columbia, 2357, Main Mall,
Vancouver, British Columbia, Canada V6T 1Z4

Kevin Sallee

Ecological Software Solutions, 3154 53rd Street,
Sacramento CA 95820, USA

Aaron D. Gladders and David M. Shackleton

Wildlife Research Group, Faculty of Agricultural Sciences,
University of British Columbia, 2357, Main Mall,
Vancouver, British Columbia, Canada V6T 1Z4


We compared three methods to classify locations in radio telemetry studies: circles, ellipses and points. Data from a 1997-1998 cougar study were used to determine the precision of the methods relative to one another; they were found to differ. Simulations to assess the accuracy in high-, mid- and low-complexity habitats were performed using real-life error parameters. All estimators performed worst in the high-complexity habitat, although shapes were better than points.


Forest harvesting and human activity in North America have caused some concern for the viability of wildlife populations and their habitats. The spotted owl Strix occidentalis (Lamberson et al. 1994, Lehmkuhl and Raphael 1993, Andersen and Mahato 1995), marbled murrelet Branchyramphus marmoratus (Rodway et al. 1993a, 1993b), grizzly bear Ursus arctos (Mattson et al. 1996, Wielgus et al. 1994), and northern goshawk Accipiter gentilis (Squires and Ruggiero 1996) all face population declines due to habitat loss. High elevation, logged areas act as a mortality sink for the endangered Vancouver Island Marmot (Marmota vancouverensis) when they disperse to such areas instead of to their natural habitat of subalpine meadows (Bryant 1996, Bryant and Janz 1996). Research in high snowfall, forested areas (McNay and Bunnell 1995) revealed that black-tailed deer (Odocioleus hemionus columbianus) require mature forests during deep-snow winters, fueling conflict between the forest industry and conservation. The ability to properly assess the habitat requirements of a species is becoming increasingly important for biologists and resource managers.

Conventionally, radio telemetry studies use a maximum likelihood (ML) point estimate to describe an animal’s location, and its use in Geological Information Systems (GIS) and habitat analyses is straightforward. However, ML estimates are not exact; each point has an associated error ellipse. Ellipses are based on the orientation of tracking towers around the location, the bearing error, observer error, and distance of the towers to the true location. However, a circular shape is easier to use with GIS than an ellipse. To address the uncertainty in radio telemetry locations while maintaining simplicity of analyses, we addressed the precision and accuracy of three methods of analyzing locations: circles, points, and ellipses.


Using the LOAS program (Sallee 1999a), the Andrews-M or ML estimates, as well as the 95% error ellipse, were generated for 700 cougar (Puma concolor vancouverensis) locations collected during our study on Vancouver Island, British Columbia. For each location, a circular buffer was created based on the average of the ellipse major and minor axes. The habitat at each location was queried with the GIS for each location type and compared across types. To address the complication of more than one habitat falling within a circle or an ellipse, we used a compositional analysis (Aebischer et al. 1993) and permutated to test for significance.

To determine which method most accurately described the habitat at the true location, simulations were conducted. A total of 1000 random points were created using Biotas (Sallee 1999b) to represent the true locations. The corresponding point, circle, and ellipse estimates were created, using error parameters from 95 test collars collected while conducting the cougar research. The ability of each estimate to describe the true location across 3 habitat complexities (low, medium, high) was tested using a compositional analysis for each estimate.

The three levels of habitat complexity were generated from an initial real landscape of forest cover polygons. Adding 40 m riparian buffers and tessellating the landscape to a minimum polygon size of 9 ha, created the high complexity landscape. The medium complexity landscape corresponded to the raw forest cover data. Dissolving the boundaries of the forest cover polygons created the low complexity landscape.


The precision compositional analysis showed that circles and ellipses performed with the same level of precision (p > 0.10). However, points were significantly different from both ellipses and circles (p < 0.001). Results of the compositional analysis to test accuracy showed that all three estimators classified the true habitat more than 85% of the time. However, the circles and ellipses were consistently able to classify the true habitat at high complexity, while the point estimator began to misclassify habitats (Figure 1).

Figure 1. Percent of the 3 estimates that included the true habitat in varying habitat complexity. Points misclassified more locations in the high complexity, while circles and ellipses still included the true habitat.

The ability of points, circles or ellipses to properly classify the true habitat was dependent on the landscape complexity (Figure 1) and observer error (low error=increased accuracy and precision). We suggest that by using a pilot study to estimate observer error, researchers could determine which of the three location descriptors works best for their study area.


We thank Forest Renewal B.C., the British Columbia Ministry of Environment, Lands, and Parks, MacMillan Bloedel Ltd., and Mountain Equipment Co-op for funding the cougar field project.


Aebischer, N.J., P.A. Robertson, and R.E. Kenward. 1993. Compositional analysis of habitat use from animal radio-tracking data. Ecology 74. 1313-1325.

Andersen, M.C. and D. Mahato. 1995. Demographic models and reserve designs for the California spotted owl. Eco. Applic. 5: 639-647.

Bryant, A.A. 1996. Reproduction and persistence of Vancouver Island marmots (Marmota vancouverensis) in natural and logged habitats. C. J. Zool. 74: 678-687.

Bryant, A.A., and D.W. Janz. 1996. Distribution and abundance of Vancouver Island marmots (Marmota vancouverensis). C. J. Zool. 74: 667-677.

Lamberson, R., B.R. Noon, C. Voss, and K.S. Mckelvey. 1994. Reserve design for territorial species: The effects of patch size and spacing on the viability of the northern spotted owl. Cons. Bio. 8:185-195.

Lehmkuhl, J.F. and M.G. Raphael. 1993. Habitat pattern around northern spotted owl locations on the Olympic peninsula Washington. J.Wildl. Manage. 57:302-315.

Mattson, D.J., S. Herrero, R.G. Wright, and C.M. Pease. 1996. Science and management of Rocky Mountain grizzly bears. Cons. Bio. 10: 1013-1025.

McNay, R.S. and F.L. Bunnell. 1995. Behavioural limits to movement: The effect on habitat choices for Columbian black-tailed deer. Trans. Congr. Int. Union Game Biol. 21: 295-303.

Rodway, M.S., H.M. Regehr, and J-P. L. Savard. 1993a. Activity levels of marbled murrelets in different inland habitats in the Queen Charlotte Islands, British Columbia. C. J. Zool. 71: 977-984.

Rodway, M.S., H.M. Regehr, and J-P. L. Savard. 1993b. Activity patterns of marbled murrelets in old-growth forest in the Queen Charlotte Islands, British Columbia. Condor. 95: 831-848.

Sallee, K. 1999a. Location Of A Signal (LOAS program). Ecological Software Solutions.

Sallee, K. 1999b. Biotas program. Ecological Software Solutions.

Squires, J.R., and L.F. Ruggiero. 1996. Nest-site preference of northern goshawks in southcentral Wyoming. J. Wildl. Manage. 60: 170-177.

Wielgus, R.B., F.L. Bunnell, W.L. Wakkinen and P.E. Zager. 1994. Population dynamics of Selkirk Mountain grizzly bears. J. Wildl. Manage. 58: 266-272.

Radio Tracking via Triangulation: The Folly of Censoring Locations on the Basis of Error Ellipse Size

Richard M. Pace, III

United States Geological Survey, Louisiana Cooperative Fish and Wildlife Research Unit, 309 Forestry, Wildlife, and Fisheries Building,
Louisiana State University, Baton Rouge, Louisiana, USA


I simulated radio tracking that used von Mises maximum likelihood estimate to calculate location estimates and error ellipses. I examined the effect of removing 1%, 5%, or 10% of locations with the largest confidence ellipses on location precision and overall geographic distribution of these data. Censored data sets produced distorted views of utilization distributions and only slightly improved mean accuracy. I demonstrated that a previously proposed method of detecting "bad" location data caused no obvious distortions in clean data and did well detecting spurious azimuths.


As described by Lenth (1981), von Mises maximum likelihood estimate (VMMLE) and its associated confidence ellipse have become the standard point and aerial location estimators used in wildlife research. An apparently common extension of VMMLE has been to use error ellipse size as a criterion for the exclusion of observed location data (data censoring). Most researchers are apparently unaware that such censoring, unless carefully applied, is not non-informative with respect to location. White (1985) clearly demonstrated that, given a fixed level of equipment precision, a strong association exists between error ellipse size and beacon location relative to receiving stations. Hence, the likelihood of censoring an observation can depend heavily on the animal’s location relative to receiving stations. If receiving stations are few or fixed, zones within a study area exist for which the censoring rate will be high. The subset of non-censored data may present a biased view of an animal's home range. I conducted a set of simulations to examine the effect of simple ellipse-size-based censoring on the perceptions of animal movement data gathered using typical wildlife radio tracking systems.


I simulated radio-tracking experiments with 2 tower configurations and 2 levels of bearing precision and used VMMLE to calculate location estimates. I conducted all simulation trials in a hypothetical 1 km2 square study area and used either 3 towers or 4 towers arranged in accordance with the recommendations of White (1985) for minimizing the size of an average confidence ellipse. Each trial consisted of establishing (N=1000) of true locations within the study area bounds by way of a uniform random sample, simulating the radio tracking of those points by adding a normal (mean=0, SD) error (appropriately adjusted for circular data) to each true bearing from all towers to each true location and calculating VMMLE location estimate and associated error ellipse size. By consistently using either SD = 2° or SD = 5° , I created 4 sets of what I term "clean" data reference as T3S2, T3S5, T4S2 and T4S5 depending on the number of towers and the bearing error standard deviation. I also created 2 "dirty" data sets: 3- and 4-tower configuration data sets for which each bearing set had a 10% chance of containing 1 spurious bearing. Spuriousness was defined as adding a uniform (5, 45) random error in addition to the normal (0, 2) error already added to the bearing. For these 6 data sets I examined the results of censoring all locations for which error ellipse size could not be calculated (convergence failure or a negative variance component) plus 1%, 5% or 10% of locations having the largest ellipses.

As an alternative to error-ellipse-size-based censoring, I examined the value of a calculated goodness-of-fit (GOF) statistic for identifying poor location estimates. The GOF statistic is the sum of squared deviations between the observed tracking bearing from each tower and the bearing from that tower to the estimated location. This statistic is compared to a ?2(number of towers -2) value at whatever Type I error rate (a) the investigator chooses. I examined the distribution of locations censored on the basis of GOF and compared the performance of GOF against ellipse-size-based censoring. For dirty data, I also examined the performance censoring on the basis of convergence of the Andrews estimate (Lenth 1981).


Error ellipse size was only weakly related to actual location error (Figure 1). Hence, the removal of all locations with calculated error ellipses above some pre-specified size, while producing very small gains in average accuracy (Table 1) would result in considerable data of equal quality of those data not censored. A much worse result was that the dominant factor determining error ellipse size in clean data sets was location (Figure 2 and Figure 3). In comparison, locations censored due to GOF test failure were randomly distributed and, in the clean data, occurred at a rate close to a (Table 1).

Figure 1. Graphs depicting the poor relationship of location error to ellipse area resulting from simulated radio tracking and von Mises maximum likelihood estimates of location.

Figure 2. Using simulations based on 3 receiving towers (1), A and C show locations which had the largest confidence ellipses (10%) or had no associated confidence ellipse, whereas B and D show the distribution of the data remaining of locations with large ellipses are removed, for SD=2 (A and B) and SD=5 (C and D).


Results based on the examination of dirty data indicated that error-ellipse-size-based censoring was biased by location (Figure 4). If censoring was based on deleting location estimates with the top 10% of ellipse sizes, only 27% and 24% of location estimates derived from data containing a spurious bearing would be removed from 3- and 4-tower configurations, respectively. By contrast, GOF-based censoring with a = 0.01 resulted in removal of 92% of location estimates derived from data containing a spurious bearing would be removed from both 3 and 4-tower configurations. Andrews estimation was effective at removing only about 20% of location data containing spurious bearings. Accuracy statistics for Andrews were N of 965, mean error of 30.7 m, and median error of 17.0 m for 3 towers and N of 978, mean error of 17.8 m, and median error of 12.8 m for 4 towers.

Figure 3. Using simulations based on 4 receiving towers (10, A and C show locations which had the largest confidence ellipses (10%) or had no associated confidence ellipse, whereas B and D show the distribution of the data remaining of locations with large ellipses are removed, for SD=2 (A and B) and SD=5 (C and D).

Figure 4. Comparison of locations elected for censor (0) and those remaining (Ê ) when censoring is performed by (A) goodness-of-fit (a =0.01) or (B) removal of 10% of locations with the largest confidence ellipses for a 4 tower tracking system with 10% of locations containing a spurious bearing.

Table 1. Results of simulations used to examine effect of ellipse-size-based censoring on radio tracking data gathered under various conditions and using von Mises maximum likelihood estimated locations from bearings taken from 3 (T3) or 4 (T4) stationary towers and standard deviation of bearing error was 2° (S2), 5° (S5), or 2° with a 10% chance of a spurious (5-45° ) additional error to one bearing.



1% Removal by ellipse size1

(a = 0.01)

Configuration N Mean E2 Median E3 N Mean E2 Median E3 N Mean E2 Median E3

T3 S2










T3 S5










T4 S2










T4 S5











10% Removal by ellipse size


T3 Spurr










T4 Spurr










1. Also removed were all locations which produced variance of X or Y < 0 and therefore the confidence ellipse was not defined.

2. Mean location error (m).

3. Median location error (m).


A commonly held belief among investigators using radio tracking via triangulation is that one can improve the overall precision of a set of location data if suspect locations are removed. Censored locations should be non-informative with respect to location. That is, the likelihood of an observation being censored should not depend on an animal’s location relative to receiving stations. If it does, the data remaining after censoring may present a biased view of an animal's home range. Therefore, ellipse-size-based censoring is not a valid method of improving the overall accuracy of a radio tracking data set. Using ellipse-size-based censoring may bias location, causes unnecessary loss of relatively good data, and produces little gain in overall data precision. Its use should be strongly discouraged. By contrast, GOF testing appeared to perform fairly well on two data sets known to contain spurious bearings. Further evaluation of GOF testing seems warranted, especially given the inability of "robust" estimation procedures to adequately improve data under sample sizes typical of most radio tracking triangulation efforts.


Lenth, R. V. 1981. On finding the source of a signal. Technometrics 23:149-154.

White, G. C. 1985. Optimal locations of towers for triangulation studies using biotelemetry. J. Wildl. Manage. 49:190-196.

Application of Radio Telemetry to Studies of the Physiological Ecology of Vertebrates

Gordon C. Grigg and Lyn A. Beard

Physiological Ecology Laboratory, Department of Zoology and Entomology, University of Queensland, Brisbane, Australia 4072


We present some generalizations about the use of radio telemetry in physiological ecology, particularly to monitor body temperature and heart rate, drawn from work on free-ranging camels, echidnas, platypus, Tasmanian devils, quolls, crocodiles, bearded dragon lizards, goannas, freshwater turtles, and Queensland lungfish. We identify some potential pitfalls and describe some case studies, including the questions that drove the research and a brief review of some of the results.


On modest budgets, we have been using radio telemetry extensively since 1986 to study various ecological and physiological questions in a range of Australian vertebrates. Our most common application has been to use temperature-sensitive radio transmitters implanted surgically into the body cavity (intraperitoneal implants) for long term monitoring of body temperature and for locating animals to periodically determine body mass, reproductive status, etc. We have also used transmitters for gaining long term data on heart rate as an index of short term and seasonal changes in metabolic rate. The following species have been the focus of attention by ourselves, associates, or students in our laboratory: Queensland lungfish, cane toads, turtles, lizards, crocodiles, platypus, echidna, bandicoots, quolls, Tasmanian devils, and camels. We have also used dataloggers in echidnas, camels, donkeys and cattle.

These studies represent use of telemetry in a range of habitats (open plains, closed forest, freshwater streams) and over of a range of body sizes from 50 g to about 1,000 kg. Monitoring signals has been undertaken on foot; from a vehicle, boat, or aircraft; or remotely from a fixed recording site. In the course of these studies we have had to overcome many problems and have developed some solutions which may be helpful to others. Hence, the purpose of this paper is not to provide a primer on radio telemetry but, rather, to focus on particular case studies in the hope that others planning similar projects may be able to benefit from our experience and save some time.


We use Telonics TR2 scanner/receivers with extended range, 150-154 MHz, in almost all of our work; and we are conservative in our choice of transmitters, tending to stay with the same brand(s) for a long time and getting to know the suppliers. Initially we used the heart rate and body temperature transmitters made by Jerry Stuart from Grass Valley, California. Then we moved on to Austec Transmitters (Edmonton, Canada), and we are now using Sirtrack transmitters (Wellington, New Zealand) almost exclusively. All of these manufacturers have been very helpful; and, in the course of our long association with Sirtrack, we have often developed technical solutions to particular problems collaboratively.

Considerations When Choosing a Transmitter

The choice of transmitter will be determined by the biological questions being asked, how long the study will run, and the size and habits of the study animal. There will almost always be a compromise between transmitter size and longevity, as size is determined primarily by the size of the battery, and battery size determines the life and power output of the transmitter. Antenna configuration will also affect the size and shape, and implanted transmitters rarely allow use of an efficient antenna length. Waterproofing is also a major consideration, particularly for implants.

Because we undertake mostly long-term studies and usually implant the transmitters, we almost always seek to maximize longevity of the transmitter and minimize its size. One simple trick is to choose a slow pulse rate to minimize power consumption. We like to use a pulse rate of about 40 beats per minute (bpm), that is 1,500 milliseconds between pulses or, when a temperature-sensitive transmitter is being used, 40 bpm at the modal body temperature of the animal being studied. This is slower than that chosen by many people, but it does not take much practice to become just as efficient at locating an animal with a slower pulse rate, with a big saving in power consumption. Additionally, many manufacturers now offer factory-set daily duty cycles that also can significantly increase the life of the transmitter. If, for example, you will not be tracking at night, a period of 8 hours each day can be specified when the transmitter would be turned off.

There is also an option of shortening the actual pulse width from a fairly standard 18 ms to 12 ms. While this is not recommended where picking up a signal may be difficult, it may be an option to increase longevity in situations where the study animal can be found easily within a small area. We have employed this option and not found it too limiting.

While waterproofing is necessary for externally mounted transmitters, it is paramount for an implant, as the transmitter will be bathed constantly in body fluids. We remain suspicious of materials used to "pot" radio transmitters because of uncertainties about both waterproofing and biological inertness. Earlier epoxies were known to allow water through, a fact which stimulated the development of a hybrid wax coating made from elvax (polyvinyl beads) and paraffin wax. The newer epoxies are much less permeable to water, but, even so, we have been caught a couple of times recently with moisture ingress. Accordingly, we choose to coat our transmitters wherever possible in the elvax mixture (80% paraffin wax, 20% elvax by weight), which we know from long experience is biologically inert and very waterproof. We like to have a complete coating 3 mm thick where the wax is the only barrier to the outside, or thinner when transmitters or dataloggers are already coated with epoxy.

Calibrating Temperature-Sensitive Transmitters and Retrieving Data

If really precise and accurate results are required, particular care needs to be taken in calibration, and some pitfalls need to be avoided. Calibration should not be carried out until the transmitter has had time to settle down after having been powered up. Typically, a week is required for this: a day for the battery voltage to decrease to a stable plateau and a week or so for the capacitor components to stabilize. The stabilization process can be confirmed by monitoring the transmitter in a constant temperature environment and waiting for the pulse interval to stabilize.

Calibration needs to be carried out in a temperature-controlled water bath against a certified thermometer, allowing sufficient time for the transmitter to come into equilibrium at each step of the process. Ideally, calibration should be undertaken by stepping the temperature up and then back down to make sure there is no hysteresis. The usual practice is to calculate a mathematical representation of the curvilinear relationship between pulse interval and temperature. This allows mathematical interpolation between the measured points, and mathematical translation of a measured pulse interval to temperature. It is worth checking, however, that the maximum error which can result from the limitations of goodness of fit of the line is within acceptable limits. We have found that, even though simple equations generate very high r2 values, errors can still be quite large because a calculated equation will always depart from the real line to some extent. We have improved the goodness of fit by resorting to fitting a fifth order polynomial. The curve of best fit for a particular transmitter will depend upon the characteristics of the thermistor and, indeed, the temperature-sensitive characteristics of the overall unit. The important thing is to check the relationship properly over the whole range of interest.

Finally, it is highly desirable to retrieve the transmitter at the end of the study in order to check that there has been no drift in the characteristics of the components, which will result in a shift of the calibration curve. A small amount of drift can sometimes be accommodated, if it is assumed to have been linear over the period between the two calibrations. How much error or uncertainty is acceptable will depend upon the particular circumstances and the nature of the questions being asked. The important thing is to know and understand the properties and limitations of the measurement system being employed.

Surgical Implantation

We favor implantation as the means of attaching the transmitter wherever possible. Apart from the obvious need to implant when deep body temperature is to be measured, we also favor implantation for tracking purposes, for many animals do not immediately lend themselves to the use of collars or back mounting by gluing. The advantage of implantation for tracking transmitters is that it eliminates problems in attachment and the likelihood of interference with the animal’s movements and normal behavior.

The main disadvantage of implanting is that there may be a severe constraint on the type of antenna employed and, thus, the range of the transmitter may be severely reduced. Ideally, a 150 MHz transmitter should have a whip antenna of 42 cm in length. It is uncommon to be able to deploy a transmitter with a 42 cm whip intraperitoneally. In the case of heart rate telemetry, an increase in range can be gained by using one of the electrode wires as an antenna. Our standard transmitter format for implantation currently is a Sirtrack two-stage transmitter and an AA-sized lithium battery with a loop antenna fitted snugly around the battery, potted in epoxy and coated in an Elvax mixture (Beard and Grigg, these Proceedings). This gives adequate range for most of our needs and makes a smooth, biologically inert, ovoid package with no edges, which can be allowed to float freely in the body cavity.

Smaller animals such as quolls and lizards require a smaller package, so a ½ AA battery is used. However, a loop antenna around an AA battery is smaller and results in less range. In this case, we have arranged for the loop antenna to be external to the body of the transmitter. Alternatively, we have used a ½ AA battery with the same sized loop antenna as we would have had to fit around an AA battery, but folded to fit snugly around the smaller battery. Although reduced, the range of such a unit was still adequate to track platypus in fresh water up to 300 m, and cane toads and bandicoots on land. Of course, it is nearly certain that platypus will be in the river or in a burrow into the bank, so a walk along the watercourse will usually detect the animals even if the transmitter range is reduced.

In the case of Tasmanian devils and quolls in the southern temperate rainforests of Tasmania, the combination of thick vegetation and frequent rainfall wetting the vegetation attenuated the signal so that ranges of less than 100 m were the norm. In that case, range was not a problem because the animals were also wearing tracking collars.

For lungfish, although only tracking was required, implantation was still employed as the most long-lived method of attachment. As these fish were to be tracked in water of varying conductivities and, therefore, variable degrees of attenuation, Peter Kind chose to employ a whip antenna trailing out through the body wall. Tracking was from aircraft flying at 150 feet, and using a scanner-receiver, or from a boat, canoe or on foot. It is, of course, useful to know that the fish can only be in the river. Resultant ranges of these 150 MHz transmitters were comparable to those being used by colleagues working on Mary River Cod in the same river, even though they were using transmitters in the 20-30 MHz band usually prescribed for freshwater work. A further reminder that telemetry still has its mysteries was provided by a lungfish transmitter which continued to be detected, apparently as easily as ever, even though the whip antenna broken had off at the base.

It is important to ensure that the transmitter is sealed against water, coated smoothly with a biologically inert material, sterilized properly prior to implantation, and implanted under sterile conditions. We usually do the implantation in the field. An indication of the success of implantation is that we have had echidnas breed successfully when they are carrying implanted temperature and heart rate transmitters plus a "Tidbit" (Onset Corp., USA) datalogger.

Interestingly, implantation into a large animal poses a special case because the signal becomes very significantly attenuated. Radio telemetry does not work well in sea water because the signal becomes so easily absorbed, and implanting into a large animal is, effectively, putting the transmitter into a large pool of dilute sea water. In large animals, therefore, a loop antenna will return only short ranges, about 30-50 m. We solved this problem in one of our studies by using a 42 cm whip and to a certain degree in another study by increasing the power output of the transmitters.

All of our implantation techniques have been subject to animal ethics committee approval. For most mammals and some reptiles we use gaseous halothane delivered in regulated concentration by a portable anesthetic machine. Gas is instantaneously variable in response to any differences in the animal’s condition occurring during the surgery. Mammals recover quickly once the halothane is turned off. Reptiles, however, take longer to be anesthetized and to recover after the operation. This is partly because many reptiles breath-hold for long periods and because they have a lower metabolic rate, so metabolism of the anesthetic agent occurs more slowly. Injectable anesthetics suffer from not being flexible once the dose is given and also may have extremely long recovery times, sometimes days. A less time-consuming and probably safer alternative for reptiles is the use of cold immobilization plus local anesthetic; used successfully by us in goannas, bearded dragons and small crocodiles. Cane toads and lungfish were anesthetized with MS222.

For most animals we have found that a mid-ventral incision into the body cavity is the most convenient way to introduce transmitters. The smoothly coated transmitter is then allowed to float freely in the peritoneal cavity, finding its own resting place among the viscera. The linea alba in mammals provides a convenient muscle break along which to cut and is also where the layers are thinnest. A mid-ventral incision is also convenient for lizards, goannas, water dragons and crocodiles. Toads require an incision slightly off-center to avoid the large anterior abdominal blood vessel, and turtles are best accessed via the soft skin in the axillary region of the hind leg, with the turtle on its side so the viscera fall away from the incision site.

The method of closure can also be a consideration. While single, interrupted sutures are adequate for many animals, we found at the first attempt that echidnas require stronger sutures because they are basically a "ball of muscle" and can split their sutures in the process of curling up into their defensive posture. Subsequently we have used vertical mattress sutures and silk thread, which holds well while the wound is healing but wears out over a period of time in the field, eliminating the need to recapture the animal for removal of stitches.

Large mammals such as camels, cows, and donkeys cannot be laid on their backs, so the incision is made on the flank in a similar position to that required for a Caesarian section on a cow. We also tried tying transmitters and dataloggers to the last rib of camels to facilitate retrieval at the end of the study.

We sterilize the transmitters with a glutaraldehyde-based instrument sterilant ("Aidel," Whitely Industries, Australia) and rinse them thoroughly with sterile water before implantation. In some difficult field situations and particularly for reptiles, we use povoiodine (a bacteriostat) that does not need to be rinsed off. However, this is potentially less effective. Many of the implantations are carried out in the field so that an animal is kept in captivity for minimum time. We prefer to release animals as soon as practical once the immediate effects of the procedure are over, believing that a rapid return to freedom and a routine lifestyle best promote a rapid recovery by wild animals.

Signal Reception and Data Gathering

We use the Telonics TR2 Scanner-Receivers and, mostly, their RA-2AK H-antenna. Their roof-mounted whip antenna was used from a vehicle. This is such a standard procedure that we have only a couple of points to make. The first is to emphasize what we said about how, with a bit of practice, quite slow pulse rates can be monitored and tracked very effectively. If transmitter longevity is a consideration, it is useful to choose a pulse rate as slow as practical. The second point is that, in determining pulse interval by timing ten time intervals using a stopwatch, count zero for the first beat, not one.

Collecting data by hand is very time consuming. In order to monitor body temperature data from echidnas in a more continuous format, and well before implantable loggers became available, we devised a simple automated monitoring system. Body temperature studies usually require long runs of data, sampled against a known time base, so the capacity to do so automatically while the investigator is working elsewhere or sleeping is very valuable. Even with small, implantable, temperature-sensitive dataloggers, there is often an advantage in having data available during the study, without having to wait until the animal is retrieved and the logger removed one to two years later. We invented this comparatively simple system in the early days of the work on echidnas in Australia’s Alps (Grigg et al. 1990). We incorporated a simple computer system which would time the interval between adjacent pulses and store that automatically. Such systems work well if there is a very good signal-to-noise ratio. With a free ranging animal, however, a good signal-to-noise ratio cannot be relied upon because when the animal is well away from the receiving station, electronic ears are not nearly as good as human ears at detecting the signal amid the noise. Hence, a human with a stopwatch will be able to obtain data from a lot of transmissions to which electronic ears would be deaf. Accordingly, we resorted quickly to a much simpler system, which does not suffer from this drawback.

The core of this system is a timer that can be set to switch the pre-programmed receiver-scanner and the tape recorder on and off for a chosen length of time at predetermined intervals. For example, if we wish to sample body temperature from three individuals, with a dwell time on each individual of 15 seconds, the receiver would be set to listen on those three programmed frequencies for 15 seconds each, and the timer would be set to come on for 45 seconds. If we wanted to sample hourly, the receiver and tape recorder would then be turned off for 3,555 seconds. The tape recorder is turned on and off simultaneously with the receiver and, at that sampling rate, a C.120 cassette recording tape can record 80 hourly samples (3.3 days) before being turned over. Telonics scanners, as supplied, do not have appropriate dwell times for this purpose. We had our electronics technician, with the approval of Telonics, replace resistors inside the scanner to alter the options for the medium and slow dwell times to approximately 15 seconds and 30 seconds respectively, which allows both a short enough time to economize on tape space when pulse rates are relatively fast, and a long enough dwell time to reliably measure pulse interval when pulse rates are slow (e.g. when the animal is in hibernation). Another necessary modification is the addition of a small socket in the panel of the receiver which, when plugged by the timer lead, interrupts the power supply to the receiver. The power is restored when, and for as long as, the timer switches on, thus saving battery power which would otherwise be exhausted in a couple of days if the receiver were on constantly. Similarly the tape recorder uses a lot of power in "pause" mode, so it, too, is switched on and off by interrupting the power supply. The timer lead plugs into a socket in the power lead to the tape recorder. The whole system can be housed in a waterproof plastic or polypropylene box. We often camouflaged it with brush.

To increase reception range of the data-logging system, we usually employ an omni-directional whip antenna hauled high into a gum tree, using a weighted string thrown over a branch. The system is powered by rechargeable lead-acid "gel" cells, which are replaced periodically with fresh batteries during regular visits to turn or change the recording tape. Later, the investigator "reads" the tape using a device such as a Telonics TDP-2 Digital Processor for those sections of the tape in which there is a good signal-to-noise ratio, or ears and a stopwatch for those sections where the signal is lost in the background noise beyond the capacity of the TDP to discriminate. This system was used to collect data on echidnas (Grigg et al. 1989, Grigg et al. 1992), platypus (Grant et al. 1992), crocodiles (Seebacher et al. 1998), and pygmy possums (Broome and Geiser 1995).

This same system can be used to monitor visits by a telemetered individual to a particular location of interest. Using a shortened antenna or laying the antenna on the ground to reduce the sensitivity, the signal will be received only when an animal visits a particular area of interest such as a burrow or nest. We used it to monitor visits by a lactating echidna to feed its young in a nursing burrow (L. Beard and G. Grigg, personal observation). This application, of course, is limited to situations where visits, when they occur, are for a reasonable period of time. To pick up brief visits to an area, some other methodology is required, using a system which responds actively to the arrival of the subject.


Freshwater Turtles

It seems to be an almost invariable rule that the use of telemetry leads to a changed perception of the subject animal and how it makes a living. A recent example is work done on the Pleurodiran freshwater turtle (Emydura signata, Chelidae) (Manning and Grigg 1997). These turtles are commonly seen basking on logs and rocks in the rivers and creeks of southeastern Queensland. Such behavior was assumed to be thermoregulatory and the study aimed to describe the patterns. To monitor daily and seasonal cycles in body temperature (Tb), temperature-sensitive transmitters were implanted into the peritoneum under sterile conditions, using halothane-in-oxygen anesthesia, via an incision made in the left flank anterior to the left hind leg. Animals were released at their capture site and Tb monitored through winter, spring and summer. Some animals were monitored manually, but most were monitored with the automatic recording system. Environmental data was also collected. The outcome was a surprise; of 1,402 observations made during daylight hours on eight adult individuals, only ten showed Tb warmer than water temperature. There was no data to show that turtles routinely elevate their body temperature by basking. The turtles are thermoconformers. Presumably, turtles seen "basking" are not remaining out of the water long enough to warm up, and the biological significance of their emergent behavior remains unexplained. Favorite hypotheses might be that they dry off to retard algal or fungal growth on the skin, or that they bask for vitamin D synthesis. The study highlights the risk of drawing what appear to be obvious conclusions about the significance of a particular observation, and identify the need for direct measurements on other species of freshwater turtle.

Crocodiles as Thermal Models for Dinosaurs

Telemetry from large saltwater crocodiles poses particular difficulties. For one thing, saltwater attenuates a radio signal very effectively so, although a study the thermal relations of very large estuarine crocodiles (Crocodylus porosus) might have been more gratifying if conducted on free-ranging animals, we were lucky to be able to work at the Edward River Crocodile Farm on Cape York, where a large number of adult crocodiles live in naturalistic circumstances in water that is almost fresh. An additional problem is posed by the size of the animals, because body fluids, about one third the concentration of sea water, also attenuate the signal significantly. Other problems related to the large body size are those of capture and handling. Large crocodiles are fragile to handling, sustaining damage easily and being vulnerable to drug overdose. They are also dangerous. Whereas implantation would have been desirable, handling the animals was not really a practical option, especially as crocodiles of the size in which we were interested, up to a ton or more, are worth thousands of dollars each.

To address these difficulties, we trebled the power output of the transmitters, encased them in dental acrylic, inserted them into dead chickens, and fed the chickens to the crocodiles from the back of a truck driven into the large enclosure. Because crocodiles often retain stones within their stomach, perhaps as ballast, the sealed radio transmitters became pseudo-gastroliths. Range was less than ideal, and we were constrained in our capacity to get data automatically.

Nevertheless, enough data was gained from 12 individuals over a size range from 32 to 1,000 kg in both summer and winter to allow a good description of the pattern of daily and seasonal changes in body temperature and some speculations about dinosaurs (Grigg et al. 1998, Seebacher et al.1999). As crocodiles get larger, their daily variation in body temperature becomes less because of increased thermal inertia, and they become warmer. Very large crocodiles can enjoy a high and stable body temperature. We also found that their daily movements between water and land, differing in summer and winter, were very important influences over the measured body temperature. From heat exchange equations and knowledge of the ambient thermal conditions, it was possible to predict the empirical observations mathematically, and then extrapolate those analyses to larger reptiles. Thus we were able to show that large dinosaurs, without needing to be endotherms, and showing thermoregulatory behavior like that of crocodiles, could have had high and stable body temperatures even at high latitudes, throughout the year (Seebacher et al. 1999). At low latitudes, however, avoiding overheating may have been a potential problem and the large, tropical dinosaurs must have had access to either shade or water to escape radiant energy.

Telemetry of Heart Rate and Body Temperature in Lizards

In the early 1960s, it was determined under laboratory conditions that the Bearded Dragon (Pogona barbata), a local Agamid lizard, heated faster than it cooled (Bartholomew and Tucker 1963). This was counterintuitive. An ectothermic reptile would have been assumed to come into equilibrium with a new ambient temperature at a rate determined by its thermal conductance, and rates of heating and cooling could be expected to be the same during heating and cooling, as they were, in fact, in dead lizards they used as controls. Bartholomew and Tucker found that heart rates, too, were higher during heating than cooling, perhaps providing an explanation for the apparent change in thermal conductance. It has since been learned that, during heating, skin perfusion increases along with increased heart rate, enabling a more rapid uptake of heat. The phenomenon has been demonstrated in many reptile species. Its biological significance has been assumed to be that a basking reptile can gain heat rapidly, and lose heat more slowly when it leaves the sunshine for other behavior such as feeding or reproduction. Thus, behavioral thermoregulation is being augmented by a physiological control over thermal conductance in a way that minimizes the amount of time that an individual needs to spend basking. This has become one of the central paradigms in our understanding of reptile thermal relations, even though it is now known to be applicable only to those lizards that are greater than about 20 g in body weight (Fraser and Grigg 1984), which is actually the minority of reptile species. Although it has been studied extensively in the laboratory, it had not been examined under field conditions.

To rectify this lack of information, a radio-telemetric approach was used to monitor both heart rate and body temperature in free ranging bearded dragons (Grigg and Seebacher 1999). Epoxy-encased heart rate transmitters (30 mm x 25 mm x 10 mm, 10 g, Sirtrack) with a screw for sensitivity adjustment and a maximum sensitivity of 100 µV and two points for electrode attachment, were taped to the side of a lizards’ tail, just posterior to the vent. Using a straight surgical needle, one electrode was passed under the loose skin from the transmitter to the underside of the animal, ventral to the heart, and sutured in place. The second electrode was sutured under the skin at the base of the tail. Electrode leads were of plastic-coated surgical steel wire, with 10 mm of the active end stripped of insulation to form the contact. The longest lead also acted as the antenna. It was not possible to measure heart rate while the animals were moving due to interference from muscle activity. To verify that transmitter signals matched heart rate, ECG was monitored during transmitter installation using a MacLab physiograph (AD Instruments). Body temperature (Tb) was monitored with a Sirtrack 2-stage transmitter of the type described above, made small enough (10-15 g) by using a ½ AA-sized lithium battery, with adequate range ensured by having a 2 cm diameter loop antenna free of the transmitter and battery. We found that morning basking led to pronounced increases in Tb and in heart rate. When Tb reached between 35oC and 40oC, lizards started shuttling between sun and shade, and Tb remained relatively stable in the mid to high 30soC, rarely over 40oC, until late afternoon. Most significantly, however, as predicted by the paradigm, heart rate during heating usually exceeded heart rate during cooling, at any Tb. Further we found that, on those occasions when Tb reached very high levels (>40oC) and there was a risk of overheating, heart rates during cooling exceeded those during heating, apparently as the lizard sought to dump heat in the shade. This, together with the confirmation of higher heart rates during heating at Tb below preferred levels, seem to demonstrate convincingly that the exercise of physiological control of thermoregulation in reptiles is a reality. Work now in progress seeks to extend the study to varanids (Varanus varius) and crocodiles (C. johnstoni).

Echidnas, Platypus, Tasmanian Devils, and Eastern Quolls

The short-beaked echidna (Tachyglossus aculeatus), a Monotreme, hibernates in the winter, while the platypus (Ornithorhynchus anatinus), another Monotreme, apparently does not (Beard and Grigg, these Proceedings). The data on echidnas and platypus provide an interesting contrast. Unequivocally, echidnas are hibernators. But from five platypus monitored throughout winter in one of the coldest streams in Australia, we found no evidence that they have the capacity to show torpor, let alone hibernate. We could not, however, conclude that platypus never show torpor or hibernation, only that the ones we studied did not, and the results suggest that neither hibernation nor torpor is a routine occurrence in platypus.

A similar outcome was found after tracking and monitoring Tb throughout the winter in Tasmanian Devils and Eastern Quolls, two dasyurid marsupials, in the central highlands of Tasmania (Jones et al. 1997). Both species proved harder to trap in winter, suggesting that they may have the capacity for torpor or hibernation, a phenomenon not unknown in marsupials, having been described in Mountain Pygmy Possums, Burramys parvus, in the field by Broome and Gieser (1995). However, we found that both Eastern Quolls (about 1kg) and Tasmanian Devils (5-9 kg) maintained normothermia throughout the winter in a field study conducted near the Cradle Mountain National Park, despite their nocturnal habits and the extensive and prolonged snow cover. This represents an impressive energetic performance, and seems to put a final nail in the coffin of the now abandoned concept of marsupials as being metabolically inferior to eutherian mammals. However, as with platypus, all we could say is that hibernation and torpor are unlikely to be a regular occurrence in these species. It is still possible that, under certain circumstances, they may occur.

Despite this lingering uncertainty, it is far more likely that a representative result will have been obtained by long term telemetry from the animals free-ranging in their natural, familiar surroundings, than from a study of captive individuals. Indeed, captive echidnas are notoriously reluctant to enter hibernation, even when living outdoors in cold winters. This misled Augee (1978) to pronounce that they are not hibernators, although it was later determined that echidnas do hibernate (Grigg et al. 1989). It is worth noting that studies such as these could now be undertaken with considerably less effort because an implanted datalogger could be employed, instead of having to maintain contact with the animals throughout the winter and record data both manually and using an automatic monitoring system. However, a transmitter is still desirable in addition to a logger because it increases the chance of relocating and capturing the animal in order to retrieve the logger. It is also sensible for the transmitter to be temperature-sensitive because it affords the opportunity for Tb data to be gained as a backup, in case the logger cannot be retrieved, as well as to monitor Tb along with direct observations of behavior.

Camels, Donkeys, and Cattle

Schmidt-Nielsen et al. (1957) found that camels restricted from water access had the capacity to save water by avoiding sweating and tolerating a rise in body temperature as the day advanced. The excess heat could then be dumped at night, by radiation, and the camel could start the next day cooler than usual in anticipation of heat storage in the hot, dry day ahead. Thus camels could show daily ranges in body temperature of up to almost 7oC, much larger than those seen in most mammals. It was these observations that helped establish camels as the most celebrated example of desert adaptation among mammals. While not doubting their capacity, or the value of Schmidt-Nielsen’s seminal research, we were aware that his camels had been captive in a pen and were, presumably, prevented from pursuing thermoregulatory behaviors available to free-ranging camels. The possibility was there that such abandonment of thermal homeostasis may occur only very rarely in camels under natural circumstances, or that they may even be a captivity artifact.

To check, and in collaboration with Birgit Dorges and Jurgen Heuke, camel biologists living and working in the desert west of Alice Springs in Central Australia, and with much help from Jocelyn Coventry, a veterinarian, and Alex Coppock, land owner and grazier, we set about monitoring body temperatures in free-ranging camels. In our earliest attempts, using echidna-style transmitters with loop antennas, we found we had very little range because the large body size and its salt content attenuated the signal so effectively. However, a small loop antenna is actually a very inefficient radiator. To operate efficiently, a 150 MHz transmitter should, theoretically, have a 42 cm whip antenna. This is normally impossible in an implanted device, but quite possible in a camel.

Accordingly, in collaboration with Sirtrack, an implantable transmitter was devised that featured a 42 cm whip at one end, with a similar whip at the other end as a ground plane equivalent. There was plenty of room within the body cavity for this device to be implanted, via an incision just behind the last rib. In trials we found that the metal antennas, lying within the electrolyte-rich body fluids, acted like the electrodes of a battery and caused the active antenna to corrode. In the final version, the antennas were isolated electronically with a diode from the rest of the circuit and coated with silastic tubing. Because we wanted to retrieve the transmitters at the end of the study, we tied one end of the silastic tubing to the last rib, thus tethering the transmitter. Dataloggers (TidBit, Onset Corp. USA) were implanted in each camel at the same time, also tethered to the rib.

We found that this transmitter-antenna configuration provides a very strong signal that could be picked up from a kilometer or so away, sufficient for our purposes because all of the camels also wear a standard radio tracking collar. Another limitation of the Schmidt-Nielsen study (1957) was that the observations on camels were not compared with observations on other animals under the same circumstances; that is, there were no controls. We therefore implanted dataloggers in feral donkeys and cattle in the same area.

However, this project has been plagued by technical difficulties, despite much planning and careful trials, and we have had to repeat much of it. All of the loggers and transmitters failed in the first deployments. At first we thought there must have been some sort of interference between the loggers and the transmitters, but we use the same loggers and similar transmitters in echidnas routinely, so there is no a priori reason why the two cannot be combined. Then we speculated that the higher power of the transmitter, lying in an electrolyte bath of body fluids, was somehow damaging the circuits of both. But Sirtrack was unable to duplicate the failure in bench trials of the transmitters. Communication with other users of loggers threw suspicion onto their waterproofing and, by association, that of the transmitters as well, because both relied on the epoxy coating. Whereas our habit for implants has always been to coat the transmitter completely in Elvax to ensure biological inertness (Beard and Grigg, these Proceedings), we could not coat the camel transmitters completely because of the flexible antenna-to-body joins. Also, the otherwise complete wax coatings on the loggers were breached at the hole through the attachment tab, near the thermistor. Is epoxy more permeable to water at higher temperatures? We have been unable to prove conclusively the cause of the failures, but are blaming it on water ingress through the epoxy coatings of both the loggers and the transmitters. Sirtrack has replaced all of the transmitters, and Hastings Data Loggers, Australian agents for Onset, have stood by their guarantee. The financial losses associated with travel and management of large animals under field conditions are, of course, not recoverable.

We have encountered two more problems. The first is that we lost a logger, completely, from one of the camels. We presume that the gut enclosed and ingested it and expelled it through the anus. Migration of implanted hardware into the gut has been reported in humans (Singh and Godec 1992, Gomez et al. 1995, Majeski 1998) and this is a potential hazard with implanted devices, albeit a rare one. The second is that the summer of 1998-99 was very wet. Grass grew in profusion and the camels had no shortage of either good forage or water. We have our fingers crossed for 1999-2000. Perhaps it is appropriate to finish with this cautionary tale, a reminder that radio telemetry is an art as well as a science, and that Murphy’s Law can prevail.


The study of whole animal physiology in the laboratory is limited to finding out what they can do, whereas in the field it is more likely that research will uncover what they actually do. Data monitoring by radio telemetry offers the possibility of taking physiological studies out of the laboratory into the field. With an unlimited budget, almost anything is possible, such as studies of seals diving in Antarctica with the data being relayed to the researcher by satellite. However, even small-budget studies can yield valuable results, as we hope we have shown. Studies on thermal relations and energetics are obvious candidates for studies in which there is telemetry of physiological data, sometimes in concert with direct behavioral observations. Even simple tracking telemetry, however, gives the patient researcher the opportunity to keep tabs on an animal living under field conditions for many months or even years. Tracking may be necessary, too, to keep in touch with an implanted datalogger. Combined with periodic measurements of body length and weight and blood sampling for electrolyte or hormone status, reproductive status, and so on; long-term tracking applications make it possible to build up a comprehensive picture of the growth and lifestyle of a subject species in a way that would be impossible otherwise.


We thank Dave Ward, Kevin Lay and Darryl Olsen and staff at Sirtrack New Zealand (Havelock North) for their helpful and continuing assistance in the development and supply of transmitters; Dave Beaty and Stanley Tomkiewicz and staff at Telonics (Mesa, Arizona) for advice and assistance with receiving systems; and our previous telemetry equipment suppliers Jerry Stuart and Keith Brockelsby. All have been ready to listen and to help when we wanted small numbers (always), of specially customized variations (inevitably) of their standard products for particular applications. We are also grateful to the Australian Research Council and University of Queensland Research Grants which have funded much of the work, and we also wish to acknowledge the contributions of our numerous colleagues and collaborators in these studies.


Augee, M.L. 1978. Monotremes and the evolution of homeothermy. Pages 111-120 in M.L. Augee (ed.) Monotreme Biology, Sydney. Royal Zoological Society of New South Wales.

Bartholomew G.A. and V.A. Tucker. 1963. Control of changes in body temperature, metabolism and circulation by the agamid lizard, Amphibolorus barbatus. Physiol. Zool. 36:199-218.

Beard L.A. and G.C. Grigg. Radio telemetry of echidnas and platypus. This volume.

Broome L.S. and F. Gieser. 1995. Hibernation in the free-living pygmy possum, Burramys parvus (Marsupialia: Burramyidae). Aust. J. Zool. 43:373-379.

Fraser S. and G.C. Grigg. 1984. Control of thermal conductance is not significant to thermoregulation in most reptiles. Physiol. Zool. 57(4):392-400.

Gomez C., Dick M., R. Hernandez, A.G. Coran, D. Crowley, and G.A. Serwer. 1995. Peritoneal migration of an abdominally implanted epicardial pacemaker: a cause of intestinal obstruction. Pacing Clin. Electrophysiol.18:2231-2232.

Grant T.R., L.A. Beard, and G.C. Grigg. 1992. Movement and home range of platypus in the Thredbo River, NSW. Pages 263-267 in M.L. Augee (Ed.) Platypus and Echidnas, Sydney. Roy. Zool. Soc. NSW.

Grigg G.C., M.L. Augee, and L.A. Beard. 1992. Thermal relations of free-living echidnas during activity and in hibernation in a cold climate. Pages 160-173 in M.L. Augee (Ed.) Platypus and Echidnas, Sydney. Roy. Zool. Soc. NSW.

Grigg G.C., L.A. Beard, and M.L. Augee. 1989. Hibernation in a monotreme, the echidna Tachyglossus aculeatus. Comp. Biochem. Physiol. 92A:609-612.

Grigg G.C., L.A. Beard, and M.L. Augee. 1990. Echidnas in the high country. Aust. Nat. Hist. 23(7):528-537.

Grigg G.C., F. Seebacher, L.A. Beard, and D. Morris. 1998. Thermal relations of large crocodiles, Crocodylus porosus, free-ranging in a naturalistic situation. Proc. Roy. Soc. Lond. B 265:1793-1799.

Grigg G.C. and F. Seebacher. 1999. Field test of a paradigm: hysteresis of heart rate in thermoregulation by a free-ranging lizard, Pogona barbata. Proc. Roy. Soc. B (In press).

Jones M., G.C. Grigg, and L.A. Beard. 1997. Body temperatures of Tasmanian devils and quolls. Physiol. Zool. 70(1):53-60.

Majeski J. 1998. Migration of wire mesh into the intestinal lumen causing an internal obstruction 30 years after repair of ventral hernia. South. Med. J. 91(5):496-498.

Manning B. and G.C. Grigg. 1997. Basking behaviour is not of thermoregulatory significance to the "basking" freshwater turtle Emydura signata. Copeia 1997(3):579-584.

Schmidt-Nielsen K.B., S.A. Jarnam, and T.R. Haupt. 1957. Body temperature of the camel and its relation to water economy. Am. J. Physiol. 212:341-346.

Seebacher F., G.C. Grigg, and L.A. Beard. 1999. Crocodiles as dinosaurs: behavioural thermoregulation in very large ectotherms leads to high and stable body temperatures. J. Exp. Biol. 202:77-86.

Singh I., and C.J. Godec. 1992. Asynchronous erosion of inflatable penile prosthesis into small and large bowel. J. Urol. 147(3):709-710.

Telemetric Observation of Heart Rate, ECG, and Body Temperature in Deep Hibernation of Edible Dormice

Ralf Elvert and Gerhard Heldmaier

Department of Biology, Philipps-University, Karl von Frisch Strasse,
35032 Marburg, Germany


Heart rate, ECG and body temperature (Tb) were studied in hibernating dormice. Data were recorded by a modified physiological implantation system from April 1998 to January 1999. A custom-made software allowed continuous recording of Tb, heart rate, and metabolic rate without disturbing the dormice for several months. Telemetric monitoring was made in a temperature range of 30°C to 0°C where dormice showed normothermia (Tb of 36°C to 38°C) as well as daily torpor and hibernation (Tb of 2°C - 20°C).


Dormice are true hibernators and survive the cold and food-restricted season with a hibernation period of up to seven months. Edible dormice (Glis glis) are furthermore able to reduce their metabolic rate at all seasons of the year (Wilz and Heldmaier 1997). Hibernators lower their body temperature close to ambient temperature due to reduced energy requirements, only a fraction of what is needed during activity. During hibernation most of the energy required is spent for the repeated arousals that occur at distinct intervals in all hibernators.

This study focused on the development of a telemetric system for monitoring long-term changes of heart rate (HR), ECG, and body temperature (Tb) at different ambient temperatures. The changes in these parameters could also be analyzed during entrance into and arousal from torpor, and were investigated in different lethargic stages.


Housing Conditions

The dormice were housed in wire mesh cages (80 cm x 50 cm x 42 cm) inside a climate chamber with free access to water and food, consisting of sunflower seeds, nuts, rodent breeding chow (Altromin, Germany), and apples. For monitoring daily torpor and hibernation periods, food was removed. Ambient temperatures varied between 30°C and 0°C, and humidity was maintained constant at 80%. The dormice had free access through a revolving door to a sleep and nesting box placed outside the cage (Figure 1).

Figure 1. Schematic description of the entire setup for measuring body temperature, heart rate, and gas exchange

Telemetry System and Transmitter Implantation

The telemetry system consisted of implantable transmitters (ETA-F20), a telemetry receiver (RPC-1), an ECG and body temperature adapter (UA10), and an interface (Dacpad-71 B) converting analog signals for storing on a computer setup, which was controlled by self- developed software (QB45). This software included filter algorithms for suppression of noise and interferences. In each dormouse an ETA-F20 biopotential transmitter, weighing approximately 4.0 g (Data Sciences, DSI, St. Paul, USA), was implanted. A magnetically activated switch allowed the device to be turned on and off. Dormice were anesthetized with a mixture of Ketamine and Xylazine. The transmitter body was implanted into the abdominal cavity and fixed with the peritoneum. A pair of flexible leads extending from the housing were sutured subcutaneously. The negative lead was positioned at the right shoulder, the positive at the lower left chest (Kramer et al. 1993). The peritoneum was sutured with resorbable catgut (1.5 metric), and the skin with sewing silk (1.5 metric). After a recovery period of 3 to 6 weeks with free access to food and water, animals were prepared for long-term studies.

Prior to implantation the transmitters were calibrated in a temperature-regulated water bath in a range of 0.5°C to 41°C. The transmitters sent signal pulses in intervals correlating with voltage output. Body temperature was calculated from voltage output following calibration of the transmitters, using the equation: (modified from Ruf and Heldmaier, 1987). The radio signals were received by the RPC-1 (DSI, St. Paul, USA), located beneath the animal nesting box, and transformed into digital pulses. The signal output of the receiver was duplicated by a splitter and transmitted to a universal adapter (UA10, DSI, St. Paul, USA) which converted digital to analog signals. Analog outputs of the UA10 were connected to another interface (Dacpad-71 B, Datalog, Germany) which prepared transmitter signals for computer analysis (Figure 1). While one channel was used for calculating Tb, the other enabled calculation of HR from voltage peaks. Time intervals between these peaks correlated with HR. From scanning the signal input for 10 sec, several peaks were detected and a mean value was calculated. Ta was measured with a thermocouple placed inside the nest box. All parameters were monitored in 60 sec intervals and stored on the computer. An integrated snapshot mode enabled storage of raw mVolt-output values from heart rate channel in 1 msec intervals for analysis of ECG and PT-period, respectively.


Tb and HR were continuously recorded in two edible dormice from April 1998 to February 1999. Both parameters showed a nocturnal rhythm when dormice were normothermic (Figure 2).

Mean value of Tb of a single dormouse (Dormouse R4S) was 37.1°C during the night and 36.4°C during the day at an ambient temperature (Ta) of 15°C. HR varied between 367.05 beats per minute (BPM) during the night and 327.35 BPM during the day (mean values from dark and light phases). Gaps in data recording are caused by the dormouse leaving the nest box and being active in the wire mesh cage where transmitter signals could not be received.

A continuous recording of Tb and HR from 6 days shows almost daily drops in Tb and HR (Figure 3) at a Ta of 15°C. The first 24 h are extracted in Figure 2 and show the normothermic dormouse before entering the torpor period. On the second day, a decrease in Tb and HR could be observed, whereas this torpor event was interrupted at a Tb of 25.4°C. During the following days, Tb values approached Ta and decreased to 16.5°C. HR decreased to 20-30 BPM.


Figure 2. Body temperature (Tb), heart rate (HR) and ambient temperature (Ta) of a single day of Dormouse R4S in normothermia. Shaded areas indicate dark period.


Figure 3. Continuous 6-day recording of body temperature (Tb), heart rate (HR), and ambient temperature (Ta) from a single dormouse (Dormouse R4S).

When entering hibernation, Tb decreased close to the level of Ta (Figure 4). Ta was lowered during the hibernation period from 10°C to 4°C. After 19.5 h, Tb decreased from 38.1°C to 7.4°C after 48.5 h, while Ta was maintained at 7°C. HR decreased after an initial peak of 511 BPM to a stable level varying between 7 BPM and 20 BPM. The HR showed regular peaks every 20 min in deep hibernation (the period between Hour 48 and Hour 72 in Figure 4). At the end of this hibernation period, an arousal occurred (Hour 163.5). Dormouse R4S returned to normothermia just to enter another torpor bout with Tb and HR reduction to almost hibernation levels. Snapshots taken from ECG recording showed a duration of PT period in normothermia of 60 ms to 75 ms (II in Figure 4 and Figure 5). The numbers I, III and IV show ECG records from different lethargic phases. No. I was taken at Ta= 10°C and Tb was at 10.78°C. The PT period at this temperature was at 324 ms and was extended with decreasing Ta and Tb, respectively. At a Tb of 5.78°C, and Ta= 5°C a PT period of 531 ms could be observed (No. IV, Figure 5). In another dormouse (Dormouse R13G), a PT period of even 1,123 ms at Ta=0.5°C was measured.


Figure 4. Continuous 8-day recording of body temperature (Tb), heart rate (HR), and ambient temperature (Ta) from a single dormouse (Dormouse R4S). Numbers I – IV indicate time of snapshots taken for ECG analysis.

Figure 5. Snapshot recordings of ECG in normothermia, during daily torpor and in deep hibernation from a common dormouse (Dormouse R4S). Numbers I-IV indicate time of snapshots taken during recording from Figure 4.


In former studies involving monitoring hibernation periods for ECG, HR, and Tb analysis, most measurements included connecting sensors, lead wires, and catheters to restrained or anesthetized animals (Suomalainen and Sarajas 1951, Kulzer 1967, Pajunen 1992). The most important problem in dormice is that they are very sensitive to mechanical disturbances when hypothermic and immediately enter alarm arousals when disturbed (Pajunen 1992). Also, they do not enter lethargic phases when restrained. We applied a telemetric system for measuring Tb, HR, and ECG in unrestrained, freely moving dormice. Because dormice spent most of the time of the diel cycle in their nest and sleeping box, we were able to record physiological parameters in a long-term study including hibernation and daily torpor even at very low temperatures. Since animals had free access to a bigger cage, it was possible to house them for months in the climate chamber without disturbance except for weighing, feeding, and refilling water. HR varied between 230 BPM and 511 BPM in normothermia at Ta= 15°C. It was decreased during torpor and remained stable at a low level between 20 BPM and 30 BPM (Figure 3). Limiting food availability caused an increase in frequency of daily torpor until the dormice entered hibernation. In deep hibernation, which is defined as the maintenance of hypothermia for more than 24 h (Wilz 1999), a periodical pattern was developed showing regular fluctuations between 7–13 BPM as basic HR and 18–25 BPM peaks in 20 min intervals. A similar variability could be observed in PT duration, which increased from 60-75 ms in normothermia to more than 1100 ms in deep hibernation. In lethargic phases the duration of P to T wave showed a Ta dependency. It was decelerated with decreasing Ta. When food is less abundant, dormice minimize energy expenditure by reducing metabolic rate and heart rate. The most effective reduction of energy requirements is achieved by hibernation when hibernators adjust their metabolic rate to a fraction of euthermic level. The reduction of HR and thus energy expenditure of the heart muscle to a minimum completes energy savings. These mechanisms allow all hibernators to survive with the energy reserves accumulated in body fat.


I thank Dr. Michael Wilz for assistance in developing the experimental setup. Furthermore, I thank the Flora-Immerschitt Stiftung, which supported my participation at the 15th Symposium of Biotelemetry in Juneau, Alaska. The scientific project was supported by DFG.


Kramer, K., S.A.B.E. van Acker, H.P. Voss, J.A. Grimbergen, W.J.F. van der Vijgh and A. Bast. 1993. Use of telemetry to record electrocardiogram and heart rate in freely moving mice. J. Pharmacol. Toxicol. Meth. 30(4):209-215.

Kulzer, E. 1967. Heart rate of bats during lethargy and hibernation. Zeitschrift fuer vergleichende Physiologie (56):63-94. (in German, English summary).

Pajunen, I. 1992. A comparison of the heart rate at different ambient temperatures during long-term hibernation in the garden dormouse, Eliomys quercinus L. Cryobiology 29:414-421.

Ruf, T., and G. Heldmaier. 1987. Computerized body temperature telemetry in small animals: use of simple equipment and advanced noise suppression. Comput. Biol. Med. 17(5):331-340.

Suomalainen, P., and S. Sarajas. 1951. The heart rate in the hibernating hedgehog. Ann. Zool. Soc. Zool. – bot. Fenn. Vanamo, Tom. 14(2):1-8.

Wilz, M., and G. Heldmaier. 1997. Hibernation in edible dormice: influence of ambient temperature on breathing frequency and oxygen consumption. Verh. Dtsch. Zool. Ges. 90:1, 156.

Wilz, M. 1999. Hibernation, aestivation und taeglicher Torpor beim Siebenschlaefer (Glis glis, L.) PhD Thesis, Marburg University, Germany.