Global positioning system (GPS) telemetry technology allows us to monitor and to map the details of animal movement, securing vast quantities of such data even for highly cryptic organisms. We envision an exciting synergy between animal ecology and GPS-based radiotelemetry, as for other examples of new technologies stimulating rapid conceptual advances, where research opportunities have been paralleled by technical and analytical challenges. Animal positions provide the elemental unit of movement paths and show where individuals interact with the ecosystems around them. We discuss how knowing where animals go can help scientists in their search for a mechanistic understanding of key concepts of animal ecology, including resource use, home range and dispersal, and population dynamics. It is probable that in the not-so-distant future, intense sampling of movements coupled with detailed information on habitat features at a variety of scales will allow us to represent an animal's cognitive map of its environment, and the intimate relationship between behaviour and fitness. An extended use of these data over long periods of time and over large spatial scales can provide robust inferences for complex, multi-factorial phenomena, such as meta-analyses of the effects of climate change on animal behaviour and distribution.
In the history of science, rapid conceptual advances have often been stimulated by technological innovations. Such technological milestones included Galileo and Kepler designing and looking into telescopes, finding evidence for Copernicanism (and falsifying the Tolemaic system); Hooke and van Leeuwenhoek peering into microscopes and describing cells, thus laying the basis for modern microbiology; Sanger and Maxam and Gilbert developing DNA sequencing, which spawned molecular biology; and high-speed computers fostering the emergence of nonlinear dynamics. In this Theme Issue we herald an exciting synergy between animal ecology and global positioning system (GPS)-based radiotelemetry. New telemetry technology allows us to monitor and to map the details of animal movement, securing vast quantities of such data even for highly cryptic organisms. Conceptual advances have been made with roots in Lagrangian characterization of the details of movement by individual animals, which ultimately leads to Eulerian characterization of population process. Recent theory uses stochastic movement models to challenge our mechanistic understanding of home range and dispersal. We are beginning to tackle the complexities of memory and perception, and how these behavioural processes yield the individual variation that is so pervasive in nature. It is probable that in the not-so-distant future, intense sampling of movements coupled with detailed information on habitat features at a variety of scales will allow us to represent an animal's cognitive map of its environment. Models of resource selection by animals give us fundamental insights into the mechanisms behind the distribution and abundance of organisms. Indeed, this powerful synergy between science and technology is rapidly shaping the very structure of the discipline of ecology.
In animal ecology, reality is observed at approximately the same space–time scale as an observer's experience. Direct human observations cannot provide thorough and standard data that will allow falsification of ecological hypotheses for all animals and all kinds of research questions. The most logical step is moving the point of observation from the observer to the observed (i.e. the studied animal). For example, to quantify accurately where animals go, using technology to let animals show us where they are has many advantages. Interestingly, marine biology, where boundaries to human senses are apparent, was the first field of animal ecology to use telemetry (Boyd et al. 2004). Biotelemetry (the remote measurement of state variables of individual, free-living animals) has been the technological revolution that has allowed the expansion of the mechanistic approach to the ecology of large animals (Cooke et al. 2004). From the first time-depth recorders attached to Weddell seals (Kooyman 1965) to the multi-sensor ‘daily diary’ (Wilson et al. 2008), biotelemetry devices have evolved, spawning increasing numbers of studies and peer-reviewed publications (Ropert-Coudert et al. 2010). Continued technological developments have broadened the potential (and definition) of bio-telemetry devices to allow sampling of environmental variables around tagged animals (Fedak 2004; Hooker et al. 2007).
Among the variables that can be measured for a free-living animal, its position in space allows intuitive and immediate ways to relate the animal to its environment. One could ask what phenotypes (adaptations) are required to be in a certain place at a certain time. What are the consequences for being there? Why is the animal there? Animal positions provide the elemental unit of movement paths and show where individuals interact with the ecosystems around them. Those interactions ultimately determine animal fitness (Nathan et al. 2008). In other words, animal locations provide the ‘live’ point of contact between ecology and evolution. Thus, knowing where animals go is of utmost importance as human activities affect climate and habitat, and thereby challenge the very persistence of species (Thomas et al. 2004).
Animal tracking (the monitoring and recording of animals' sequential positions) initially, and for a long time, relied on VHF (very high-frequency) technology; that is, animals have been equipped with transmitters emitting at radio frequencies that can be received by radio receivers (Craighead 1982). A few automated radiotelemetry systems have been developed (e.g. Cedar Creek, University of Minnesota; Rongstad & Tester 1969; ARTS, http://www.princeton.edu/~wikelski/research/index.htm; Croofoot et al. 2008), but, even for automated systems, VHF technology requires receivers to be close enough to the animals to triangulate animal positions. Therefore, animal tracking has traditionally relied on researchers to be in the field, with the potential to affect animal behaviour (Cooke et al. 2004; but see, for marine biology, Hill 1994 on geolocation, and Wilson & Wilson 1988 and Wilson et al. 2007 on dead-reckoning).
The advent of satellite telemetry allows remote tracking of animal positions and movements. Telemetry using GPS, in particular, has several technical advantages, including the ability to determine position on the surface of the Earth (or in the air) with high precision and accuracy 24 h a day, with position updates available in rapid succession. Argos satellites using Doppler-based positioning (Tomkiewicz et al. 2010) are less precise but provide the only automated reporting of data available in North America. We now can obtain nearly continuous, systematically scheduled datasets of animal positions, though the technology is subject to some serious technical limitations (Tomkiewicz et al. 2010).
Cooke et al. (2004) listed several limitations for monitoring animals remotely with biotelemetry. The most apparent is the cost of units to be fitted to individual animals. Unfortunately, funding often dictates that few individuals in a study can be equipped with the very advanced, but very costly, units (Rodgers 2001; Hebblewhite & Haydon 2010). We expect prices to drop as more researchers use GPS technology, because off-the-shelf devices are subject to free-market forces. Standardization of device functions and data types will help prices to fall as well, and can also be an advantage for comparing among studies and for data sharing in large-scale studies (Tomkiewicz et al. 2010; Urbano et al. 2010). In addition, GPS wildlife telemetry is based on a widespread technology and can take advantage of components developed for other sectors of business, with a huge market. This should assure constant technological improvements and decreased costs, and stimulate attention to a second large limitation: at present, GPS telemetry is confined to large animals or those capable of carrying solar chargers for batteries (Tomkiewicz et al. 2010). GPS technology was pioneered on large vertebrates, such as elephants (Douglas-Hamilton 1998), moose (Rodgers et al. 1996; Edenius 1997) and bears (Schwartz & Arthur 1999). The size of the devices is decreasing, however, and each decrease in size increases the range of animal species for which they are available (e.g. foxes, tortoises, pigeons). Recent technological advances have also allowed GPS applications in marine environments (Schofield et al. 2007; Tomkiewicz et al. 2010).
Clearly, the choice of a technical tool for a specific study requires critical evaluation in light of the goals and scope of the study. Hebblewhite & Haydon (2010) compare advantages and disadvantages of GPS telemetry in ecology, and offer examples of areas of study that have benefited from this technology. Ecologists and managers must evaluate, during the design of a study, the potential advantages and challenges, from data handling to analyses, of using GPS-based technology. Providing a guide to the accomplishment of this task is the reason for this Theme Issue.
To make good use of GPS-based location data one must have a measure of device error. The most difficult aspect in testing GPS devices and building a valid theory of error is the change in performance of devices from the laboratory to free-ranging animals (Tomkiewicz et al. 2010). System function is influenced by environmental conditions (e.g. climatic factors, habitat types, terrain roughness) and animal behaviour (e.g. movement, orientation of the collar). The effect of the latter is still largely unexplored because of the difficulties of setting up appropriate field tests (Heard et al. 2008). The two types of error that plague GPS telemetry are spatial inaccuracy of the locations acquired and missing data in the form of failed location attempts (Frair et al. 2004). Their combined effect can lead to mistaken inferences on animal spatial behaviour, especially those involving movement paths and habitat selection (Frair et al. 2010). When planning a field study with GPS technology, users should evaluate the potential effect of errors on their analyses within the context of their ecological questions. Pragmatically, they should assess the potential sources of error in the field, test the sensitivity of GPS units accordingly and consider potential solutions, if necessary. Frair et al. (2010) provide a comprehensive synthesis of predominant causes of errors, their implications for ecological analyses and tools for correcting analyses.
Large datasets, made available by advanced technology, present serious problems of data management, as underlined for microarrays from genomic techniques (Hess et al. 2001), physiological and behavioural data from sensors (Cooke et al. 2004; Rutz & Hays 2009), animal-borne video imagery (Moll et al. 2007), and high-resolution spatio-temporal movement data (Nathan et al. 2008). Challenges include preservation of data integrity and consistency, avoidance of data redundancy, automation of data download, filtering and storage, management of specific data types, and definition of standards for objects and formats (Urbano et al. 2010). For GPS-based locations (and other animal-borne data from biotelemetry), flexibility in integrating data from different devices, the ability to manage spatial time series and the possibility to integrate tools for data analysis and visualization in a single software environment are also desirable characteristics (Urbano et al. 2010). Tools to meet these requirements already exist (Shekhar & Chawla 2003; Hall & Leahy 2008) but have rarely been used to develop ad hoc solutions for management of biotelemetry data (but see Cagnacci & Urbano 2008; Hartog et al. 2009). Better systems to support these data ensure better quality and efficient allocation of resources (i.e. funds are used for developing persistent data management systems instead of for manual data handling), while better systems also feed back to ways of conducting research. The choice to tag individual animals, the potential consequences on the welfare and behaviour of those animals (Tomkiewicz et al. 2010), and large costs argue that the data produced should persist in science beyond the initial study. After being used for its original research purposes, location data should be used to answer other ecological questions. Those questions can extend over long periods of time and over large spatial scales beyond the limits of the original research (and funding). Such extended use can provide robust inferences for complex, multi-factorial phenomena, such as meta-analyses of movement (Hebblewhite & Haydon 2010) or of the effects of climate change on animal behaviour and distribution. Despite several initiatives currently being developed (e.g. http://www.movebank.org and http://www.topp.org, accessed 28 February 2010; see Urbano et al. 2010 for other examples), persistence of animal ecological data is lower than that for other fields of biology. In the future, GPS-based location data and other biotelemetry datasets will hopefully converge towards initiatives similar to GenBank for genetic sequences (http://www.ncbi.nlm.nih.gov/GenBank, accessed 28 February 2010).
Datasets of GPS-based locations are, essentially, time series of spatial data, usually collected with high-frequency, systematic schedules. This fact has two important consequences. First, position time series represent movement paths (Nathan et al. 2008), and the higher the frequency of positions, the more trustworthy the movement paths. Trustworthiness also depends on the physiological and behavioural characteristics of the animals that affect step length (Calenge et al. 2009; Owen-Smith et al. 2010). Second, position time series are highly correlated datasets, both in time and space (Boyce et al. 2010; Fieberg et al. 2010); the higher the frequency of positions, the stronger the correlation (or autocorrelation, i.e. correlation that changes as a function of temporal or spatial distance between observations; sensu Fieberg et al. 2010). These two consequences provide opportunities for GPS-based research. Datasets of high-frequency positions present opportunities to study how animals interact with their environments, while also offering opportunities and analytical challenges for extracting behavioural information and dealing with correlated data in model. On one side, these opportunities and challenges should promote the adoption of analytical methods that: (i) model correlation explicitly as an intrinsic and relevant property of data (Wittemyer et al. 2008; Boyce et al. 2010; Fieberg et al. 2010), with particular emphasis on individual-level variability and processes (Gillies et al. 2006; Forester et al. 2007); (ii) disentangle patterns of trajectories that correspond to behavioural phases or to specific behaviours (Morales et al. 2004; Fryxell et al. 2008; Boyce et al. 2010; Merrill et al. 2010); and (iii) investigate the mechanisms underlying autocorrelation, such as internal state and memory (Dalziel et al. 2008; Van Moorter et al. 2009). On the other side, the abundance of data from few individuals decreases the quality of inferences (due to small sample sizes; Rodgers 2001; Fieberg et al. 2010), including population-level inferences (Morales et al. 2010).
When a new technology with the potential to improve empirical research emerges, enthusiastic adoption may lead to reassessment of classic questions. While returning to old questions is insightful, it should not restrict the scope of research. Science should be enhanced by, not limited to, technology. In other words, more robust raw data should underpin more comprehensive theoretical frameworks.
Understanding how and why animals move are the goals of mechanistic approaches to animal ecology, including how and why animals use specific resources, how and why animals interact with conspecifics, and how and why they compete and reproduce. Ultimately, we gain an understanding of fitness. Therefore, modelling animal movement is basic for all questions in animal ecology. Ecology is fundamentally spatial, with ecological processes occurring on heterogeneous landscapes. Movement is the glue that ties ecological processes together. This is a dynamic and growing area of research that includes mechanistic approaches to understanding home range formation (Mitchell & Powell 2004; Moorcroft & Lewis 2006; Van Moorter et al. 2009; Kie et al. 2010; Smouse et al. 2010), assessing the effect of memory on orienting or limiting movement (Gautestad & Mysterud 2006; Dalziel et al. 2008; Van Moorter et al. 2009; Wolf et al. 2009), estimating movement and habitat preference simultaneously (Rhodes et al. 2005; Johnson et al. 2008; Fieberg et al. 2010), incorporating individual stochasticity in dynamic movement models (Jonsen et al. 2003; Morales et al. 2004; Patterson et al. 2008) and upscaling from individual movement to population dynamics (Morales et al. 2010). Critical evaluation of the predictions of these models requires dense location data and, thus, thrives on large, high-frequency GPS-based datasets (Smouse et al. 2010).
Perhaps the most crucial limit to understanding why animals move is understanding what resources they use at specific places and times, and at different scales (Fryxell et al. 2008; Wittemyer et al. 2008; Beyer et al. 2010; Gaillard et al. 2010; Morales et al. 2010; Owen-Smith et al. 2010; Smouse et al. 2010). Despite the availability of robust theoretical frameworks, such as optimality (Owen-Smith et al. 2010) and functional responses (Merrill et al. 2010), and despite the development of well-tested and innovative analytical approaches (Dalziel et al. 2008; Wiens et al. 2008), comparing animal choices with availability of resources remains difficult (Beyer et al. 2010). GPS-based locations provide robust spatio-temporal datasets with the potential for fine-scale associations between animals and habitat features. Nonetheless, GPS-based locations do not provide information on the actual resources that animals use or information on activities of animals. Obtaining data on resources and behaviours will be possible only with integration of data from other bio-sensors (e.g. accelerometers, magnetic switches; Cooke et al. 2004; Rutz & Hays 2009) and high-precision remote sensing of climatic and environmental variables (Hebblewhite & Haydon 2010; Urbano et al. 2010). If both animal-borne information and environmental information reach the same level of detail and reliability, the assessment of the relationship between habitat use and performance of animals will be a much more affordable task (Gaillard et al. 2010).
We should note at this point that researchers who observed their animals directly 50 years ago, before the advent of telemetry (and those who still observe their animals today), could see what resources their animals used and could watch and note their behaviours. Regular VHF telemetry, used to find the animals and observe them more regularly or to obtain data on key parameters such as mortality, is still the technique of choice for several research questions. New GPS-based telemetry is a formidable technological advance but it is not the best technology for all purposes. Perhaps more researchers should observe their research animals directly today. Moreover, GPS-based location datasets are not yet what they can potentially be. The technology is limited by battery life and by archival memory (Hebblewhite & Haydon 2010; Tomkiewicz et al. 2010). Analogously, several advanced analytical techniques suggested in this Theme Issue are not yet developed fully, or still require exceptional computational power and highly specific skills. Conscious, pragmatic revisits to well-tested methods could be the solution of choice for an appropriate use of GPS-based datasets (Fieberg et al. 2010; Frair et al. 2010; Kie et al. 2010).
In this Theme Issue, we confront both pragmatic and theoretical issues of using GPS-based location data and derive examples from the existing deployments. We thus lay a platform to orient scientists in a critical choice of the right tool for their studies, offering an overview of the technology currently available and discussing its advantages and limitations (Hebblewhite & Haydon 2010; Tomkiewicz et al. 2010). We advise on appropriate handling and use of GPS location data (Frair et al. 2010; Urbano et al. 2010) and analysis. We focus on peculiarities of these datasets, such as correlation and recognition of spatio-temporal patterns that are common to other ecological and biological data (Boyce et al. 2010; Fieberg et al. 2010; Merrill et al. 2010). We promote the need to lay strong theoretical foundations before taking up a study, and offer syntheses and perspectives of the approaches used in several theoretical frameworks, such as movement (Smouse et al. 2010), resource and space use (Beyer et al. 2010; Kie et al. 2010; Smouse et al. 2010), optimality (Owen-Smith et al. 2010), functional responses (Merrill et al. 2010), population dynamics (Morales et al. 2010) and habitat–performance relationships (Gaillard et al. 2010).
Scientists are compelled to make the best use of their data, to account for research efforts and economical investments, and to enhance integrated and comprehensive theoretical frameworks. We have discussed how better datasets can help ecologists in their search for a mechanistic understanding of key traditional concepts and new frameworks of animal ecology. Only integration of analytical and technical advancements within solid theoretical frameworks will allow us to tackle the intimate complexity of ecosystems, and their sensitivity to a changing planet.
The idea of this Theme Issue was stimulated by discussions at the GPS-Telemetry Data: Challenges and Opportunities for Behavioural Ecology Studies workshop organized by the Edmund Mach Foundation (FEM) in September 2008 and held in Viote del Monte Bondone, Trento, Italy. Funding of the workshop by the Autonomous Province of Trento to F.C. is gratefully acknowledged.
M.S.B. was supported by the Alberta Conservation Association and the Natural Sciences and Engineering Research Council of Canada. F.C. was partly supported by grant N. 3479 BECOCERWI, Autonomous Province of Trento, Italy. F.C. greatly thanks all those who strongly supported the idea of a discussion meeting on GPS-telemetry data from its very beginning—particularly C. Chemini, A. Rizzoli, R. Viola—and those who supported her interest in GPS-telemetry, including A. Brugnoli, C. Furlanello, R. Giovannini, S. Nicoloso, U. Zamboni, M. Zanin and many others. Finally, she warmly thanks A. Parasporo, A. Cipriano, S. Colzada, G. Cagnacci and T. Cipriano for the personal support.
One contribution of 15 to a Theme Issue ‘Challenges and opportunities of using GPS-based location data in animal ecology’.
- © 2010 The Royal Society