Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Understanding and geo-referencing animal contacts: proximity sensor networks integrated with GPS-based telemetry

Understanding and geo-referencing animal contacts: proximity sensor networks integrated with... Background: In animal ecology, inter‑ individual encounters are often investigated using automated proximity log‑ gers. However, data acquired are typically spatially implicit, i.e. the question ‘Where did the contact occur?’ remains unanswered. To resolve this issue, recent advancements in Wireless Sensor Network technology have facilitated the geo‑ referencing of animal contacts. Among these, WildScope devices integrate GPS‑ based telemetry within fully distributed networks, allowing contact‑ triggered GPS location acquisition. In this way, the ecological context in which contacts occur can be assessed. We evaluated the performance of WildScope in close‑ to‑ real settings, whilst controlling for movement of loggers and obstacles, performing field trials that simulated: (1) different scenarios of encounters between individuals (mobile–mobile contacts) and (2) patterns of individual focal resource use (mobile– fixed contacts). Each scenario involved one to three mobile and two fixed loggers and was replicated at two differ ‑ ent radio transmission powers. For each scenario, we performed and repeated a script of actions that corresponded to expected contact events and contact‑ triggered GPS locations. By comparing expected and observed events, we obtained the success rate of: (1) contact detection and (2) contact‑ triggered GPS location acquisition. We modelled these in dependence on radio power and number of loggers by means of generalized linear mixed models. Results: Overall we found a high success rate of both contact detection (88–87%: power 3 and 7) and contact‑ triggered GPS location acquisition (85–97%: power 3 and 7). The majority of errors in contact detection were false negatives (66–69%: power 3 and 7). Number of loggers was positively correlated with contact success rate, whereas radio power had little effect on either variable. Conclusions: Our work provides an easily repeatable approach for exploring the potential and testing the perfor‑ mance of WildScope GPS‑ based geo‑ referencing proximity loggers, for studying both animal‑ to‑ animal encounters and animal use of focal resources. However, our finding that success rate did not equal 100%, and in particular that false negatives represent a non‑ negligible proportion, suggests that validation of proximity loggers should be under‑ taken in close‑ to‑ real settings prior to field deployment, as stochastic events affecting radio connectivity (e.g. obsta‑ cles, movement) can bias proximity patterns in real‑ life scenarios. Keywords: Contact‑ triggered GPS, Fully distributed Wireless Sensor Networks ( WSNs), Movement ecology, False positives and false negatives, Proximity loggers *Correspondence: francesca.cagnacci@fmach.it Biodiversity and Molecular Ecology Department, IASMA Research and Innovation Centre, Fondazione Edmund Mach, Via Mach 1, 38010 San Michele all’Adige, TN, Italy Full list of author information is available at the end of the article © The Author(s) 2016. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/ publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Ossi et al. Anim Biotelemetry (2016) 4:21 Page 2 of 14 date. One is a so-called master–slave configuration where Background only the ‘master’ logger has the capacity to communicate Proximity loggers are increasingly used animal-borne with other loggers (the ‘slaves’) and triggers GPS schedul- devices [1] (see also Table  1), with great potential for ing when contact occurs, but contacts and GPS locations behavioural ecology studies (e.g. predator–prey interac- between the ‘slaves’ are not recorded [26]. Another is a tions [22]; intra-specific relationships [23]; social systems fully distributed configuration where all loggers commu- [24]) and wildlife epidemiology (i.e. by evaluating contact nicate equally with each other, with both contacts and rate [10]). Proximity loggers contain a radio unit that is triggered GPS positions recorded by all loggers in the able to receive and transmit a low-power signal (ultra- network [19]. The fully distributed configuration char - high frequency, UHF, range 300  MHz–3  GHz) from/ acterizes the prototype WildScope [19] and has recently to other nearby loggers, thus forming a Wireless Sensor also been proposed for some commercial loggers [27], Network (WSN). In this way, loggers ‘detect’ the proxim- although so far empirical applications in animal ecology ity of one another. A micro-controller and a power unit remain limited. are essential complements to the radio sensor for deploy- Fully distributed configurations of WSNs somewhat ing proximity loggers on animals, so that radio transmis- parallel those of acoustic tags in the marine environ- sion and detection are self-powered, and recording and ment (e.g. detection of pelagic fish in harbours [28]), in storing of information automated. Proximity loggers that they combine a ‘mixed’ network of mobile and fixed can be used as both biologging units, i.e. deployed on loggers. Hence, they also provide new opportunities to animals for subsequent data retrieval via recapture [5], simultaneously investigate both geo-referenced encoun- and as biotelemetry devices, i.e. equipped with a remote ters between individuals (mobile–mobile contacts) data retrieval system, such as a GSM modem [9, 19], or and the use of specific habitat resources by individuals a WSN comprised of mobile proximity loggers and fixed (mobile–fixed contacts). The latter has long been recog - base stations (see Table 1 and [2, 9]). However, proximity nized as an important application of WSNs [5] but has loggers frequently provide only spatially implicit informa- thus far seldom been used in animal ecology studies (but tion, often processed with network-based analytical tools see [10, 11] and Table  1). In this work, we explore the (e.g. Social Network Analysis [6, 12]), where the sampling functionality and potential of WildScope WSN ([19] and units are the contacts between individuals (or ‘edges’) Additional file 1) for animal ecology investigation. rather than the individuals wearing the devices. The value Proximity loggers—and GPS-based geo-referencing of proximity information can be further enhanced by proximity loggers—should be carefully calibrated prior to making animal-to-animal contacts spatially explicit [25], use [13], as with all biologging devices. To this end, sev- providing ecologists with the added value of knowing eral authors have attempted to identify both the potential where animal encounters occur. Thus, the ecological con - and the limitations of WSNs, including contact reciproc- text of a given contact can be inferred (e.g. which habitat ity [15], effect of antenna orientation [5] and RSSI/dis - features characterize the contact site). Recent advance- tance relationship ([11, 13]; see also Table 1). However, no ments in proximity logging technology have incorporated published studies have been presented so far to investi- indirect geo-referencing of animal contacts (see Table 1). gate performance of GPS-based geo-referencing proxim- In particular, Encounternet [12] and BATS [20] are WSN ity loggers in animal ecology and in particular to describe systems that infer spatial contextualization of proximity (1) how to design field calibration studies based on GPS- patterns by post-processing, and specifically by estimat - based geo-referencing proximity loggers, (2) what type of ing the distance between mobile loggers worn by individ- information GPS-based geo-referencing proximity log- uals, and fixed loggers deployed in the study area. In this gers collect and (3) how reliable they are at collecting it. scenario, the geo-referencing of proximity patterns essen- To address these issues, we developed and tested a pro- tially depends on the spatial overlap between a network tocol to derive data on the occurrence and location of of fixed loggers and animal movements, i.e. ultimately on animal encounters in semi-controlled conditions using the range of the target species. Where monitored animals WildScope GPS-based geo-referencing proximity log- move over wide areas (as is typical of medium to large gers. The work represents an extension of a preliminary mammals), coverage with fixed loggers becomes unfea - field calibration [19] that measured the average contact sible. In this case, geo-referencing of proximity patterns distance threshold of loggers at different radio transmis - between individuals is possible via integration of a GPS sion powers, in static and controlled open-air conditions sensor in the logger [13]. This capability is provided by [10]. Here, we introduced two well-recognized sources of those WSNs which integrate proximity detection with stochasticity in contact detection, i.e. the movement of GPS-based telemetry to simultaneously provide GPS loggers [11] and the presence of obstacles in the network location acquisition with animal-to-animal contacts [19]. [29]. Specifically, in our calibration exercise we mimicked Two different configurations have been developed to Ossi et al. Anim Biotelemetry (2016) 4:21 Page 3 of 14 ‑ ‑ ‑ ‑ ‑ ‑ ‑ Table 1 Review of proximity logger deployments Study Logger Logger Target species Mobile Static loggers Target Radio fre- Geo- Remote data down- Calibration scenario Power con- brand or weight (g) loggers detection quency referencing load sumption project name range (m) of encounters analysis [2] Zebranet Not Zebra (Equus Yes (only Yes (base sta Not reported 900 MHz No Yes Not reported Yes reported burchellii) for data tions ‘daily A GPS module Via radio between retriev drive’) is embedded mobile loggers and ing) in the logger base stations to track zebra move ments [3] Sirtrack Ltd. 30 Common Yes No 0.3–0.5 160 MHz No No (recapture neces Tests on captive Possums No brushtail sary) [4] possum (Trichosurus vulpecula) [5] Sirtrack Ltd. 125 Racoon (Pro- Yes No 1–1.5 916.5 MHz No No (recapture neces Laboratory test to assess Yes cyon lotor) sary) (1) threshold contact distance in dependence of antenna orientation; (2) contact duration and robustness Field tests on raccoons to evaluate (1) interlogger distance variation pre and post deployment; (2) contacts reciprocity between loggers [6] Sirtrack Ltd. 150 Eurasian badger Yes No 1.5–2.5 916.5 MHz No No (recapture neces Laboratory test to assess No (badger); (Meles meles)/ sary) threshold contact not domestic distance reported cattle (Bos the taurus) weight of the cattle logger [7] Sirtrack Ltd. 120 Tasmanian devil Yes No 0.3–0.5 916.5 MHz No No (recapture neces Laboratory test to assess No (Sarcophilus sary) threshold contact harrisii) distance Field test on captive Tas manian devils to assess contact distance, using video camera as ground truth [8] Sirtrack Ltd. Not White tailed Yes No 0–1.0 916.5 MHz No No (recapture neces Field test on captive white No reported deer sary) tailed deer to compare (Odocoileus observed and recorded virginianus) contacts Contacts reciprocity assessed Ossi et al. Anim Biotelemetry (2016) 4:21 Page 4 of 14 ‑ ‑ ‑ ‑ ‑ ‑ ‑ Table 1 continued Study Logger Logger Target species Mobile Static loggers Target Radio fre- Geo- Remote data down- Calibration scenario Power con- brand or weight (g) loggers detection quency referencing load sumption project name range (m) of encounters analysis [9] CraneTracker ~100 Whooping Yes (only Yes (base sta Not reported 2.4 GHz No Yes Field test on wild turkeys Yes crane (Grus for data tions) A GPS is Via radio between (Meleagris gallopavo) and americana) retriev embedded mobile loggers on crane species other ing) in the logger and base stations than Grus americana, to to track placed at wintering evaluate logger function longrange and breeding sites; ing prior to deployment migratory via GSM modem on target species movements during migration phase [10] Sirtrack Ltd. Not Eurasian badger Yes Yes 0.5–2 916.5 MHz No No (recapture neces Laboratory test to evaluate No reported (Meles meles)/ sary) (1) interlogger distances domestic at different heights cattle (Bos and combinations; (2) taurus) contacts reciprocity; (3) contacts duration Field test on cattle to compare observed and recorded contacts [11] Encounternet 1 Longtailed Yes Yes 5–30 433 MHz No Yes Field test to evaluate (1) No manakin Via radio between RSSI/distance relation (Chiroxiphia mobile loggers ship; (2) duration and linearis) and base stations, robustness of contacts; then via radio to a (3) effect of antenna ori PC by means of a entation on connectivity; masternode (4) effect of movement on connectivity (simula tion) [12] Encounternet 10 New Cal Yes Yes (base sta 20–40 for 433 MHz Yes Yes Field test on a fixed No edonian tions) mobile Via cross Via radio between network made up of crow (Corvus loggers triangulation mobile loggers dead quails, to model the monedu- 100 for static between and base stations, probabilistic relation loides) loggers base stations then via radio to a ship between RSSI and and mobile PC by means of a distance in function of loggers. RSSI masternode height from ground, values used antenna orientation, to estimate habitat type. Details of interlogger calibration model are distance provided in [13, 14] [15] Sirtrack Ltd. Not Cattle (Bos Yes Yes (base sta 2.0–3.5 916.5 MHz No No (recapture neces Laboratory test to assess No reported taurus) tions) sary) interlogger distance (mobile–base station) Field test on dairy cows to measure loggers reciprocity in measuring contact duration Ossi et al. Anim Biotelemetry (2016) 4:21 Page 5 of 14 ‑ ‑ ‑ ‑ ‑ Table 1 continued Study Logger Logger Target species Mobile Static loggers Target Radio fre- Geo- Remote data down- Calibration scenario Power con- brand or weight (g) loggers detection quency referencing load sumption project name range (m) of encounters analysis [16] Encounternet 65–70 Galapagos sea Yes Yes (base sta 10 433 MHz No Yes Laboratory test to assess No lion (Zalophus tions) Via radio between (1) RSSI/distance relation wollebaeki) mobile loggers ship; (2) effect of antenna and base stations, orientation on connectiv then via radio to a ity; (3) duration and PC by means of a robustness of contacts masternode Field test (1) to measure loggers reciprocity; (2) to compare observed and recorded contacts; (3) to assess RSSI/distance relationship in outdoor conditions [17] Bat Monitor 30–50 Flying fox fam Yes Yes (base sta Not reported < 1 GHz No Yes Not reported Yes ing Project ily (Pteropodi tions) A GPS is Via radio between dae) embedded mobile loggers and in the logger base stations, then to track bats via GSM modem movements to the central database [18] Encounternet 1.3 Barn swallow Yes Yes (base sta 0–40 for 433 MHz No Yes Field tests to evaluate (1) No (Hirundo tions) mobile Via radio between body effect, antenna rustica) loggers mobile loggers orientation and environ 100 for static and base stations, ment effect on RSSI/ loggers then via radio to a distance relationship; (2) PC by means of a interlogger variability; masternode (3) contacts reciprocity [19] WildScope 440 Roe deer Yes Yes 5–30 2.4 GHz Yes Yes Field test ‘in vitro’ to evalu Yes (Capreolus A GPS is trig Via GSM modem ate effect of height from capreolus)/ gered in case from mobile log ground, casing and domestic of contact gers to the central radio power on contact horse (Equus between database distance caballus) proximity Field test on horses to loggers compare observed ver sus recorded contacts in relation to distance and radio power Field test on roe deer to evaluate functioning of contacttriggered GPS acquisition in case of contact Ossi et al. Anim Biotelemetry (2016) 4:21 Page 6 of 14 ‑ ‑ Table 1 continued Study Logger Logger Target species Mobile Static loggers Target Radio fre- Geo- Remote data down- Calibration scenario Power con- brand or weight (g) loggers detection quency referencing load sumption project name range (m) of encounters analysis [20] BATS 2 Mouse eared Yes Yes (base sta 50 868 MHz Yes Yes Not reported Yes bat (Myotis tions) Via cross Via radio between myotis) triangulation mobile loggers and between base stations base stations and mobile loggers. RSSI values used to estimate interlogger distance [21] BATS 2 Fringelipped Yes Yes (base sta 10 868 MHz No Yes Field test ‘in vitro’ to model No bat (Trachops tions) Via radio between relationship RSSI/dis cirrhosis) mobile loggers and tance, also in function of base stations antenna orientation Field test ‘in vivo’ on mouse eared bats to link RSSI variation to individual movement, using video camera as ground truth The table collects information on some of the most well-known proximity logger deployments. Details on target species, proximity logging features, calibration attempts and geo-referencing functionalities are provided We report the detection range of interest for the specific study, which does not necessarily pair the maximum distance potentially covered by the loggers of any given brand Ossi et al. Anim Biotelemetry (2016) 4:21 Page 7 of 14 stereotyped individual encounters and individual use of download. The fixed loggers are equipped solely with a focal resources [11], designing and performing field tri - radio unit, to allow contact detection with mobile log- als in semi-controlled conditions. During the trials we gers (see Additional file  1 and Picco et al. [19] for further described and predicted logger performance (in terms of details). contact detection and contact-triggered GPS locations WildScope adopts a fully distributed configuration of acquisition) based on the results provided in Picco et  al. the WSN: in contrast to the master–slave configuration, [19] (objective 1). Eventually, we compared the predicted all loggers transmit and receive a radio signal on the same and observed data to derive estimates of contact success channel, so that connectivity is allowed throughout the rate and contact-triggered GPS location success rate, network (see Additional file  1). A contact between two thus testing the reliability of WildScope GPS-based geo- mobile loggers (i.e. a dyad) denotes an event of animal referencing proximity loggers (objective 2). proximity, representing an animal encounter [13]. When a contact occurs between two mobile loggers, WildScope Methods triggers the acquisition of a GPS location (i.e. ‘contact- WildScope proximity sensor network triggered GPS location acquisition’) that is independent WildScope GPS-based geo-referencing proximity log- from the scheduling of periodic GPS acquisition, which gers are designed to be deployed in a network of mobile happens in parallel (see Additional file  1 and Picco et al. and fixed sensors (Fig.  1). Mobile loggers are commonly [19] for details). Finally, a contact occurring between a deployed on animals, whilst fixed loggers are deployed at mobile and a fixed logger represents the interaction of an static stations. The mobile devices include, among other individual with a static feature, e.g. use of a focal resource components, a radio unit for contact detection that can [19]. be set at different transmission powers, a Global Posi - Lifetime properties, details on power settings, data tioning System component for individual localization storage and data processing of WildScope are described and a quadriband GSM/GPRS modem for remote data extensively in Picco et al. [19] (see also Additional file 1). Fig. 1 Conceptual scheme of a mixed Wireless Sensor Network. The network is composed of mobile loggers deployed on animals (circles), equipped with GPS unit for localization, fixed loggers deployed on focal resources (squares) and a base station (triangle). Grey-filled forms denote the presence of a modem on the unit for remote data retrieval. Mobile loggers can be in contact via radio with each other (dash‑dot line). Data download can also happen via radio through contact detection (dotted lines), either between mobile and fixed loggers (and from here to the base station) or directly from the mobile loggers to the base station Ossi et al. Anim Biotelemetry (2016) 4:21 Page 8 of 14 Wireless Sensor Networks at work: network design and data recording (objective 1) The primary goal of our experiments was to design field trials in close-to-real settings that simulated stereotyped encounter patterns between individuals and between an individual and a focal resource [11], and to describe logger performance. In particular, we designed three dif- ferent behavioural scenarios (Fig.  2). The first simulated use of focal resources by one individual (mobile–fixed loggers) and the others additionally mimicked behav- ioural modes that imply proximity between individuals (mobile–mobile loggers; Fig.  3), specifically: encounter and avoidance (e.g. territory defence); paired movement in parallel directions (e.g. mother–calves); and random, Fig. 3 Photograph of deployment scenarios. Two people move the mobile loggers on an established path to simulate a pair of individu‑ als moving side by side independent movement of individuals in the same sur- roundings (e.g. for feeding purposes). The experiment took place in San Michele all’Adige, Italy, from 14 May 2014 to 20 May 2014, in dry and windless conditions, for a total of 7  days. The study site encompassed a regular and flat terrain with some trees and bushes (Fig.  3). We intentionally included these environmental obstacles in our trials to account for the generic noise of connectivity typical of the environments in which loggers are deployed on animals [10, 29, 30]. We set the loggers to two radio transmission powers: 3 and 7, since they corresponded to average contact distances appropriate to studying the type of stereotypical encoun- ters described above (see Additional file  1 for details on the power settings; Picco et  al. [19] for average contact distances). We set up the experimental network of fixed and mobile loggers as follows. Two fixed loggers were fitted on trees at 1 m height, within plastic water-proof boxes; mobile loggers (one to three, according to the scenario) Fig. 2 Graphical description of the three deployed scenarios to test were fitted onto two-litre bottles filled with a saline solu - WildScope. The path followed by each mobile logger is described by tion to mimic the body effect of an animal [10, 19] and a line of a different colour. The sequence of contemporary actions is then fixed to rigid wooden supports, with the antenna represented by letters, whereas numbers indicate different loggers. oriented upwards (Fig.  3). For consistency, we chose the Scenario 1: one mobile logger moving towards two fixed loggers same site and logger support used for the field calibration (mimic of focal resource use) (box 1, upper left); scenario 2: two mobile loggers (mimic of encounter and avoidance movements) and two described in Picco et al. [19] that we expanded to ‘noisier’ fixed loggers (box 2, upper right); scenario 3: three mobile loggers settings. (mimic of parallel and random movements) and two fixed loggers In each of the three scenarios described above, one or (box 3, bottom left). The shaded circles represent the expected range more experimenters moved the mobile logger(s) on a of contact detection for the fixed loggers (stars) at the two power predetermined path for a period of 17 min, which repre- settings (dark-shaded small circle: power 3; light-shaded large circle: power 7). The radii of these circles, which are not reported here on sented the trial (Fig. 3). Each trial comprised a ‘script’ of the proper scale, were derived from the outcome of previous tests [8]. actions controlling the position of every logger in space The dark green circles represent trees and bushes of various sizes in and time (Table  2), at 1-min intervals (i.e. the temporal the arena where we performed the experiments resolution of contact detection chosen or ‘epoch’; see Ossi et al. Anim Biotelemetry (2016) 4:21 Page 9 of 14 Table 2 Example of  scenario for  testing contact detection Table 3 Example of  match between  expectations and contact-triggered GPS location and observations Minute Action Expected event Reason for expectation Minute Expected event Observed event Typology event 1 Stop No contact Distance M1–L1 and M1– 1 No contact No contact True positive ( TP) L2 > 2.38 m 2 No contact No contact True positive ( TP) 2 Stop No contact Distance M1–L1 and M1– 3 No contact No contact True positive ( TP) L2 > 2.38 m 4 No contact No contact True positive ( TP) 3 Stop No contact Distance M1–L1 and M1– 5 Start contact with L1 Contact with L1 True positive ( TP) L2 > 2.38 m GPS triggered GPS location True positive ( TP) 4 Stop No contact Distance M1–L1 and M1– L2 > 2.38 m 6 In contact with L1 Contact with L1 True positive ( TP) 5 Movement Start contact with L1 Distance M1–L1 < 2.38 m 7 In contact with L1 No contact with L1 False negative (FN) GPS triggered Detection of a contact 8 End of contact with No contact with L1 True positive ( TP) L1 6 Movement In contact with L1 Distance M1–L1 < 2.38 m 9 Start contact with L1 Contact with L1 True positive ( TP) 7 Movement In contact with L1 Distance M1–L1 < 2.38 m GPS triggered GPS location True positive ( TP) 8 Stop End of contact with L1 Distance M1–L1 > 2.38 m 10 End of contact with No contact with L1 True positive ( TP) 9 Movement Start contact with L1 Distance M1–L1 < 2.38 m L1 GPS triggered Detection of a contact 11 Start contact with L2 Contact with L2 True positive ( TP) 10 Movement End of contact with L1 Distance M1–L1 > 2.38 m GPS triggered GPS location True positive ( TP) 11 Movement Start contact with L2 Distance M1–L2 < 2.38 m 12 End of contact with Contact with L2 False positive (FP) GPS triggered Detection of a contact L2 12 Movement End of contact with L2 Distance M1–L2 > 2.38 m 13 No contact Contact with L2 False positive (FP) 13 Movement No contact Distance M1–L2 > 2.38 m 14 Start contact with L2 Contact with L2 True positive ( TP) 14 Movement Start contact with L2 Distance M1–L2 < 2.38 m GPS triggered No GPS location True positive ( TP) GPS triggered Detection of a contact 15 In contact with L2 Contact with L2 True positive ( TP) 15 Stop In contact with L2 Distance M1–L2 < 2.38 m 16 In contact with L2 Contact with L2 True positive ( TP) 16 Stop In contact with L2 Distance M1–L2 < 2.38 m Example of matching the expected and observed data in terms of contact Example of a 17-min trial ‘script’ for testing contact detection and contact- detection and contact-triggered GPS location acquisition. The results refer to the first scenario, as for Table 2 (one mobile logger M1 and two fixed loggers L1 triggered GPS location acquisition with WildScope loggers, deployed in a mixed network (mobile and fixed loggers). The script refers to the first scenario (Fig. 2, and L2, power 3, see also Fig. 2), and specifically to the results obtained for the mobile logger. The recorded events depend on the comparison of expected and one mobile logger M1 and two fixed ones, L1 and L2), with loggers set at power 3 where the expected contact detection threshold is 2.38 m [19]. The table refers observed events only to the actions taken by the mobile logger. Both expected contact detection and contact-triggered GPS location expected events are reported, with the associated reason for expectation We performed three replicates of the trials for each experimental setting and power, for a total of 18 tri- als. We used three to five loggers to perform the trials, Additional file  1). u Th s, for each minute, we knew the depending on the scenario mimicked, and reused the respective position (and distance) of loggers from each same loggers across replicates to avoid biases arising other. We were able to express the ‘expected’ connectivity from a single-logger malfunctioning. between loggers based on the findings of Picco et al. [19], who determined the average contact distance in base- Estimation of contact detection and contact-triggered GPS line conditions (static loggers, open-air, no obstacles; see location success rate (objective 2) also [10]). Notably, Picco et  al. [19] determined an aver- We used the proximity and contact-triggered GPS loca- age contact distance of 2.38  ±  1.65  m for power 3 and tion data collected during the trials to measure the 7.31 ± 2.97 m for power 7. Thus, we expected loggers to reliability of contact detection and acquisition of con- be in contact if they were at or below the linear distance tact-triggered GPS locations by WildScope in close-to- of 2.38  m for power 3 and 7.31  m for power 7 (see also real settings. We did this by comparing expected and Fig.  2: such threshold distances are represented as buff - observed contact events and acquisitions of contact- ers around the loggers). Similarly, every observed contact triggered GPS locations during the trials, to eventually leads to the acquisition of an expected contact-triggered determine the success rate both for contact detection GPS location. For each trial, we were thus able to express and for acquisition of contact-triggered GPS locations. all expected events (Table  2), that we then cross-refer- Specifically, we transposed contacts and contact-trig - enced with the observed ones, based on the timestamp gered GPS locations into a list of true positives (TP; (Table 3). expected and occurred contact detection events or Ossi et al. Anim Biotelemetry (2016) 4:21 Page 10 of 14 expected and occurred contact-triggered GPS loca- Results tions), false negatives (FN; expected events which were Contact detection and contact-triggered GPS location not recorded by the loggers) and false positives (FP; acquisition in simulated scenarios of encounters unexpected but observed events). In this way we derived and movement (objective 1) estimates of contact success rate (RTP ) by comput- For each combination of number of loggers per power, we contact ing the ratio of true positives to the total expected and matched the observed and expected events for both con- unexpected events (RTP   =  TP/[TP  +  FP  +  FN]). tact detection and contact-triggered GPS location acqui- contact We also computed the false-negative rate of contact sition. Hereafter we provide a description of the mobile detection (RFN ) as the ratio of false negatives to logger performance, as a guideline for users who wish contact total unexpected events (RFN   =  FN/[FN  +  FP]). to repeat this validation exercise (see Tables  2, 3 for the contact For the analysis on success rate of contact-triggered described example and Fig.  2 for the corresponding sce- GPS locations, we did not include false-positive events, nario). From the beginning until minute 6 we classified i.e. records acquired out of the expected contact-trig- only true-positive events, i.e. M1 performed as expected gered GPS locations, because these represented spuri- (no contact from minute 1 to minute 4; contact detec- ous information that would not bias the analyses. Thus, tion with L1 at minutes 5 and 6, with consequent acqui- we computed the success rate of contact-triggered sition of contact-triggered GPS location). At minute 7, GPS locations (RTP ) as the ratio of true positives M1 was moved away, but still within the expected range GPS to the sum of true-positives and false-negative events of contact detection with L1. Contrary to expectations, (RTP  = TP/[TP + FN]). M1 broke the contact with L1. We classified this unex - GPS We used RTP , RFN and RTP to evalu- pected end of contact as a false-negative event. From contact contact GPS ate the effect of radio transmission power on the perfor - minute 8 until minute 11, the expected and observed mance of WildScope whilst controlling for the number events matched. At minute 12 and 13, contrary to our of loggers in the network. Specifically, we fitted a gen - expectations, M1 kept the contact with L2. We classified eralized linear mixed model (GLMM) to evaluate the this unexpected duration of the contact as a false-positive dependence of RTP , RTP and RFN on event. At minute 14, the scenario predicted a new con- contact GPS contact power level (two levels, 3 and 7) and number of loggers tact between M1 and L2, with the consequent acquisition (three levels: 1, 2 and 3). We used the 3 replicates within of a contact-triggered GPS location. Since the contact each combination of power per number of loggers in between M1 and L2 had not been interrupted during the network to estimate the variance. Since the depend- minutes 12–13, M1 kept the contact with L2 at minute 14 ent variable is binary (i.e. contact recorded or not, 1 and (true-positive event), but did not acquire a contact-trig- 0, respectively), we fitted a binomial distribution with a gered GPS location. This was correct, since the contact logit link function. For all the three analyses, we prepared had been never interrupted. We thus classified the lack of a list of candidate models including all the potential acquisition of a contact-triggered GPS location as a true- combinations of the two covariates, fitted either as fixed positive event. In the last 2 min of the trial M1 performed or random effects. We adopted this procedure to evalu - as expected. ate whether the fitted covariates contributed to explain We adopted a similar approach for each combination of the variance of the intercept only, or whether they were power and number of loggers involved in the scenarios. significant predictors of the model (see also Additional file 2). Estimation of contact detection and contact-triggered GPS We then applied a model selection procedure based on location success rate (objective 2) AIC scores to determine the model of best fit [31]. We We derived estimates of contact success rate based on further evaluated the importance of each of the terms 3095 total expected contact events, of which 2757 were retained in the best-fit model to contribute to the good - true-positive events. We computed the false-negative rate ness-of-fit of the model by means of an ANOVA based from 231 false-negative and 108 false-positive events. We on deviance procedure (see Additional file  2). We applied based our assessment of contact-triggered GPS location an F test to account for overall differences within the success rate on 104 realized contact-triggered GPS loca- variances of the covariates levels retained in the best-fit tions out of 113 expected locations (Table 4). model. Lastly, we tested the significance of β-coefficients We found a high success rate of contact detection with respect to the reference level by means of a Student’s RTP with respect to the expected events that we contact t test. described for the simulated scenarios, for both powers All the statistical analyses were programmed in SAS tested (89% for power 3; 89% for power 7) and across software 9.3 [32]. the number of loggers involved in the trials (86% for one logger; 87% for two loggers; 91% for three loggers). Ossi et al. Anim Biotelemetry (2016) 4:21 Page 11 of 14 Table 4 Contingencies to  estimate contact success rate, false-negative rate and  contact-triggered GPS location success rate Trial Contact success rate False-negative rate Contact-triggered GPS location success rate Tot. events True pos. False neg. False pos. Tot. events True pos. 1L‑P3 204 184 20 0 14 14 1L‑P7 204 167 29 8 8 8 2L‑P3 480 414 46 20 17 11 2L‑P7 480 423 40 17 15 13 3L‑P3 864 774 52 38 29 28 3L‑P7 864 795 44 25 30 30 The table summarizes the contingency data used to estimate contact success rate, false-negative rate and contact-triggered GPS location success rate. ‘Tot. events’, ‘True pos.’, ‘False neg.’ and ‘False pos.’ denote, respectively, the total events, true positives, false negatives and false positives used to estimate the rates. 1L-P3 = trials with one logger set at power 3; 1L-P7 = trials with one logger set at power 7; 2L-P3 = trials with two loggers set at power 3; 2L-P7 = trials with two loggers set at power 7; 3L-P3 = trials with three loggers set at power 3; 3L-P7 = trials with three logger set at power 7 The model selection and the ANOVA based on devi - of loggers in the trials (86% of total failures for one log- ance procedure indicated that the number of loggers in ger; 70% of total failures for two loggers; 60% of total the network positively affected the contact success rate failures for three loggers). The false-negative rate was (Table  5a), whilst the effect of power was negligible (see negatively correlated with the number of loggers in the Additional file  2). The F test confirmed the significance of network, and this relationship approached significance the number of loggers in predicting the pattern of contact (F   =  3.57; P  =  0.0538; Table  5b, see also Additional 2,15 success rate (F  = 6.11; P < 0.05). file 2). 2,15 The analysis of false-negative rate RFN indicated The success rate of contact-triggered GPS location contact that the false negatives constituted the majority of total RTP was high at both powers (power 3: 88%; power GPS failures for both powers (67% of total failures for power 3; 7: 96%) and across the number of loggers in the trials 69% of total failures for power 7) and across the number (100% for one logger; 75% for two loggers; 98% for three loggers). The F test indicated a marginal significance of Table 5 Summary of  the best model accounting for  the the number of loggers in predicting the pattern of con- observed patterns of  contact success rate (a), false-neg- tact-triggered GPS location success rate (F   =  3.71; 2,15 ative rate (b) and  contact-triggered GPS location success P = 0.0492). In particular, RTP was significantly lower rate (c) GPS when two loggers were deployed, whilst there were no β-Coefficient SE df t P differences between trials with one and three loggers (Table 5c and Additional file 2). (a) Model contact success rate Intercept 2.2893 0.08323 15 27.51 <.0001 Discussion 1 logger −0.4716 0.1653 15 −2.85 0.0121 In this paper we describe contact detection by Wild- 2 loggers −0.3635 0.1277 15 −2.85 0.0123 Scope, a fully distributed proximity logging Wireless 3 loggers – – – – – Sensor Network that allows GPS-based geo-referencing (b) Model false-negative rate proximity detection, which we tested within a semi-con- Intercept 0.6931 0.1768 15 3.92 0.0014 trolled set-up. Previous studies have performed valida- 1 logger 1.1192 0.4203 15 2.66 0.0177 tion exercises of proximity loggers in order to evaluate 2 loggers 0.1386 0.2647 15 0.52 0.6082 the effect of body obstruction on radio connectivity (e.g. 3 loggers – – – – – [9]) or to measure the relationship between logger dis- (c) Model contact‑triggered GPS location success rate tance and radio connectivity (e.g. [13]). The final goal Intercept 4.0604 1.0086 15 4.03 0.0011 of the majority of these calibrations was to measure the 1 logger 11.2386 4.4765 15 0.03 0.9803 average threshold distances of contact detection (e.g. [6]) 2 loggers −2.9618 1.0881 15 −2.72 0.0157 in order to choose the correct transmission setting that 3 loggers – – – – – corresponds to the distance of biological interest (e.g. β-Coefficients, standard errors and significance for each level of the number of 1.5–2.0  m for perpetuation of bovine tuberculosis (bTB) loggers in the network, for the model of contact success rate (a), false-negative rate (b) and contact-triggered GPS location success rate. Since the number of in cattle herds [10]). Here, we also used average contact loggers was fitted as a categorical variable in the models, the β-coefficients refer threshold distance, as derived from Picco et  al. [19] and to the difference with respect to the reference level (three loggers) Ossi et al. Anim Biotelemetry (2016) 4:21 Page 12 of 14 made the assumption that contacts should occur at such Although mixed networks of fixed and mobile loggers a distance, within defined error bounds, i.e. we calculated represent the typical case for testing contact detection, expected events of contact detection. However, Picco they have rarely been included in animal-borne stud- et  al. [19] did not take into account the influence of fac - ies (but see [10, 11]). In this paper we propose a simple tors such as the presence of obstacles [29] and the move- mixed network and provide a description of its function- ment of loggers [11] on contact detection based on radio ing in the presence of moving subjects that mimic com- connectivity. Here, we built our experiments to control mon animal behaviours. We propose this exercise as a for movement and obstacles and demonstrated that, convenient way for users to understand how proximity despite the loggers being generally reliable, there was a loggers might work in typical deployment scenarios and noticeable rate of failures in contact detection. to easily test the reliability of a given proximity detection Furthermore, in our study the majority of failures were system whilst accounting for stochastic bias. Moreover, caused by false-negative events, i.e. expected contacts mixed networks have a much greater potential than being that did not occur. This is consistent with the increased ‘just’ a technical tool to remotely download data acquired noise arising from movement of loggers (as proposed by by mobile devices (e.g. [2, 9]). The integration of mobile Rutz et  al. [13]) and presence of obstacles [29]. Had the and fixed loggers in a fully distributed WSN can help effect of movement and obstacles been negligible, we address a variety of important themes in animal ecology, would have detected similar rates of false negatives and including focal resource use in a given habitat by an indi- false positives arising as a consequence of equal distri- vidual [11]; identification of crucial corridors for animal bution of errors within the variance of contact distance movement or cross-road sites [33]; and consequences of threshold determined by Picco et al. [19]. In contrast, the resource distribution on animal social networks [14]. majority of failures due to false negatives might be due In this work, we also demonstrate the reliability of a rel- to a reduced connectivity compared with controlled set- atively new function of GPS sensors—that is, contact-trig- tings, caused by logger movement and presence of obsta- gered GPS location acquisition, in application to animal cles in the network, i.e. an actual ‘lower’ threshold value ecology. Aside from our anomalous finding that contact- in the ‘noisier’ conditions compared with the baseline triggered GPS location success rate was reduced using ones of Picco et al. [19]. We also found that contact suc- two loggers, the overall performance of this GPS-based cess rate increased with the number of loggers, and false- geo-referencing proximity logger system was highly satis- negative rate decreased. This can be explained by the factory. Furthermore, the fully distributed (peer-to-peer) hypothesized stochasticity of the ‘noise’ of the system, capability of WildScope and other logging systems with which is characteristic of radio transmission and thus a fully distributed configuration considerably extends of WSN deployments [29, 30]. In particular, the noise the type of ecological question that can be addressed in caused by the environmental conditions [30] in addition comparison with master–slave configurations. All log - to the presence of obstacles and fine movements [29] is gers in the network can exchange proximity information likely to exert stochastic variation in connectivity that with all others, whilst contacts and contact-triggered GPS leads to the occurrence of errors (either false positives or locations are recorded in parallel by all loggers involved false negatives) in contact detection. Given the stochas- in the contact, thus permitting ‘encounter direct mapping’ tic nature of this ‘noise’, its overall effect on connectivity (sensu [23]). This has the potential to elucidate previously should be less than proportional to the number of loggers little-known ecological and behavioural processes. For included in the network. In other words, it should not example, social interactions in a group of individuals can strictly depend on the number of loggers deployed in the be detected and geo-referenced, as well as movement pat- network. Conversely, the number of contacts between terns of a group of individuals approaching or abandoning sensors (i.e. the true positives) increases proportionally a point resource (e.g. feeding stations). with the number of loggers deployed in the network. Finally, contact-triggered GPS locations are addi- Therefore, if the number of contact events increases tional to periodic GPS acquisition for all individuals beyond the number of associated errors, so too does the equipped with loggers. In this way, studies on spatially success rate. Further research in this direction should explicit contact detection can be combined with ‘tradi- disentangle the overall stochasticity of the system with tional’ movement research [25]. The flexible scheduling the variability linked to specific deployment features (but of both contact-triggered and independent (periodic) see [29, 30]). In particular, future studies should address GPS location acquisition allows the user to adjust the the physics of transmission of WSN (but see [13]). Radio tool in accordance with study-specific requirements. For transmission power had a negligible effect on contact instance, mother–calf interactions can last a long time, success rate and false-negative rate, demonstrating that so the investigator of parental care might decrease the observed patterns are robust to the power adopted. scheduling frequency of the GPS device after the first Ossi et al. Anim Biotelemetry (2016) 4:21 Page 13 of 14 30  min of contact, to save battery lifetime. Conversely, Additional files the spatial contextualization of aggressive encoun- ters between individuals, which can be instantaneous, Additional file 1. WildScope proximity loggers. This additional file might require an intense but brief GPS sampling regime. describes in detail the technical components (hardware and software) of Indeed, we should note that when animal GPS detection WildScope GPS‑based geo ‑referencing proximity loggers. is possible at a very high rate, i.e. when battery consump- Additional file 2. Model selection procedure. This additional file tion constraints are irrelevant (e.g. large marine birds describes in detail the procedure to select the best model describing contact detection success rate, false negative rate and contact‑triggered provided with solar-powered GPS units) high-frequency GPS location success rate). animal trajectories can provide identical information on animal-to-animal contacts as proximity detection sen- sors [34]. However, GPS frequency trades-off with bat - Abbreviations tery lifetime (e.g. [19, 35]), with constraints including the TP: true positive; FP: false positive; FN: false negative; RTP : ratio of true‑ contact length of the monitoring period for each individual, and positive events of contact detection; RTP : ratio of true‑positive events of GPS contact‑triggered GPS location acquisition. recapture feasibility. If recapture is a viable option and/ or the monitoring period required is short, then GPS Authors’ contributions frequency can be kept at a maximum, thus guaranteeing FO, SF, GP, AM, DM, FC designed the experiments. FO, SF, DM, FC designed and carried out the analysis. FO, DM, NG carried out the experiments. FO, FC wrote semi-continuous monitoring from which it is possible to the paper, with all other authors’ contribution. All authors read and approved extrapolate proximity patterns between individuals. the final manuscript. Notwithstanding study-specific trade-offs, the spa - Authors’ information tial information associated with contacts can remarkably Our team is mainly constituted by academic researchers belonging to differ ‑ increase the robustness of inference on animal encounters ent fields. FO, SF, FC, NG, JG and BT are animal ecologists, whilst GP, AM and [36], enhancing a process-based interpretation of these DM are information engineers with long experience in Wireless Sensor Net‑ work applications. These two components of the team worked side by side observations [37]. This may prompt a completely new set to assess the potential of WSN for animal ecology investigation and wildlife of questions in behavioural ecology (e.g. do aggregation management. patterns in winter depend on resource distribution? Is pro- Author details vision of maternal care altered by habitat composition?), Biodiversity and Molecular Ecology Department, IASMA Research and Inno‑ which could not be addressed with previous technology. vation Centre, Fondazione Edmund Mach, Via Mach 1, 38010 San Michele all’Adige, TN, Italy. UMR CNRS 5558 “Biometrie et Biologie Evolutive”, Université Claude Bernard Lyon 1, Bat G. Mendel 43 Bd du 11 Novembre Conclusions 1918, 69622 Villeurbanne Cedex, France. Istituto dei Sistemi Complessi, CNR, We tested a novel biotelemetry tool that integrates prox- 4 Via Madonna del Piano 10, 50019 Sesto Fiorentino, FI, Italy. Department imity detection and contact-triggered GPS location of Information Engineering and Computer Science (DISI), University of Trento, Via Sommarive 9, 38123 Povo, TN, Italy. Center for Scientific and Technologi‑ acquisition, in a fully distributed mixed wireless net- cal Research, Bruno Kessler Foundation, Via Sommarive 18, 38123 Povo, TN, work. This technological advancement has the potential 6 Italy. Ecosystems and Environmental Management Group, The University to bring together two branches of ecology thus far kept of Brighton, Huxley Building, Lewes Road, Brighton, East Sussex BN2 4GJ, UK. Organismic and Evolutionary Biology Department, Harvard University, 26 relatively distinct, namely movement ecology [37] and Oxford St, Cambridge, MA 02138, USA. animal encounters [23]. Moreover, we suggest further enhancements of the application of mixed proximity Acknowledgements We acknowledge the work of Sandro Nicoloso and Michele Corrà in manufac‑ logger networks to more comprehensively understand turing WildScope. The research was conducted with the support of Forestry animal use of specific focal resources within a habitat. and Wildlife Service of the Autonomous Province of Trento. We are also grate‑ However, these wide-reaching and innovative applica- ful to three anonymous reviewers for insightful comments on previous drafts. tions of a novel GPS-based geo-referencing proximity Competing interests logger must take into account the limitations of their use. The authors declare that they have no competing interests. We present a simple and repeatable series of scenarios to Availability of data and materials test the functioning and reliability of GPS-based geo-ref- The data set supporting the conclusions of this article is available in the Dryad erencing proximity loggers, prior to deployment in field repository. conditions. We foresee further steps for calibrating these Consent for publication tools, including probabilistic modelling of error rates The subjects represented in Fig. 3 give their consent for publication of the associated with contact detection as a function of the dis- image. tance between loggers, in conditions as close as possible to final deployment [13], i.e. on wild animals. Ossi et al. Anim Biotelemetry (2016) 4:21 Page 14 of 14 Funding 17. Sommer P, Kusy B, McKeown A, Jurdak R. The big night out: experiences This project was partly funded by Autonomous Province of Trento (PAT ), under from tracking flying foxes with delay‑tolerant wireless networking. In: grant BEARNET, Protocol No. S044‑5/2012/226570. F.O. was granted three ‑ Real‑ world wireless sensor networks; 2014. p. 15–27. yearly scholarships financed by the European Union (European Social Funds), 18. Levin II, Zonana DM, Burt JM, Safran RJ. Performance of encounternet Aosta Valley Autonomous Region and the Italian Ministry for Work and Social tags: field tests of miniaturized proximity loggers for use on small birds. Politics. PLoS ONE. 2015;10(9):e0137242. 19. Picco GP, Molteni D, Murphy AL, Ossi F, Cagnacci F, Corrà M, Nicoloso S. Received: 8 August 2016 Accepted: 17 October 2016 Geo‑referenced proximity detection of wildlife with WildScope: design and characterization. In: Proceedings of the 14th international conference on information processing in sensor networks. 2015; p. 238–49. 20. Dressler F, Ripperger S, Hierold M, Nowak T, Eibel C, Cassens B, Mayer F, Meyer‑ Wegener K, Kolpin A. From radio telemetry to ultra‑low‑ power sensor networks: tracking bats in the wild. IEEE Commun Mag. References 2016;54(1):129–35. 1. Wilmers CC, Nickel B, Bryce CM, Smith JA, Wheat RE, Yovovich V. The 21. Ripperger S, Josic D, Hierold M, Koelpin A, Weigel R, Hartmann M, Page golden age of bio‑logging: how animal‑borne sensors are advancing the R, Mayer F. Automated proximity sensing in small vertebrates: design frontiers of ecology. Ecology. 2015;96(7):1741–53. of miniaturized sensor nodes and first field tests in bats. Ecol Evol. 2. Zhang P, Sadler CM, Liu T, Fischhoff I, Martonosi M, Lyons SA, Rubenstein 2016;6(7):2179–89. DI. Habitat monitoring with ZebraNet: design and experiences. In: Bulusu 22. Tambling CJ, Belton LE. Feasibility of using proximity tags to locate female N, Jha S, editors. Wireless sensor networks: a systems perspective. Nor‑ lion Panthera leo kills. Wildl Biol. 2009;15(4):435–41. wood: Artech House; 2005. p. 235–57. 23. Krause J, Krause S, Arlinghaus R, Psorakis I, Roberts S, Rutz C. Reality min‑ 3. Ji W, White PC, Clout MN. Contact rates between possums revealed by ing of animal social systems. Trends Ecol Evol. 2013;28(9):541–51. proximity data loggers. J Appl Ecol. 2005;42(3):595–604. 24. Prange S, Gehrt SD, Hauver S. Frequency and duration of contacts 4. Ji W, Clout MN, Douglas M, Day T, Hendra R. Mate ID: first trial of a novel between free‑ranging raccoons: uncovering a hidden social system. J device for measuring possum contacts. In: Sutherland G, editor. Advances Mammal. 2011;92(6):1331–42. in the biological control of possums: Report of a workshop sponsored 25. Bettaney EM, James R, St Clair JJH, Rutz C. Processing and visualizing by the National Science Strategy Committee for Possum & Bovine Tb association data from animal‑borne proximity loggers. Anim Bioteleme. Control. Wellington: Royal Society of New Zealand; 1999. p. 92–5. 2015;3(1):1–11. 5. Prange S, Jordan T, Hunter C, Gehrt SD. New radiocollars for the detection 26. Vectronic Aerospace GmbH. http://www.vectronic‑aerospace.com/ of proximity among individuals. Wildl Soc B. 2006;34(5):1333–44. wildlife‑monitoring/sensors/uhf‑id‑tags/. Accessed 29 Jul 2016. 6. Böhm M, Hutchings MR, White PC. Contact networks in a wildlife‑live ‑ 27. Zoological Society of London – Institute of Zoology. https://www.zsl.org/ stock host community: identifying high‑risk individuals in the transmis‑ science/research/badger‑ cattle‑ contact‑project. Accessed 29 Jul 2016. sion of bovine TB among badgers and cattle. PLoS ONE. 2009;4(4):e5016. 28. Klimley AP, Le Boeuf BJ, Cantara KM, Richert JE, Davis SF, Van Som‑ 7. Hamede RK, Bashford J, McCallum H, Jones M. Contact networks in meran S. Radio‑acoustic positioning as a tool for studying site ‑specific a wild Tasmanian devil (Sarcophilus harrisii) population: using social behaviour of the white shark and other large marine species. Mar Biol. network analysis to reveal seasonal variability in social behaviour and 2001;138(2):429–46. its implications for transmission of devil facial tumour disease. Ecol Lett. 29. Ceriotti M, Chini M, Murphy AL, Picco GP, Cagnacci F, Tolhurst B. Motes in 2009;12(11):1147–57. the jungle: lessons learned from a short‑term WSN deployment in the 8. Walrath R, Van Deelen TR, VerCauteren KC. Efficacy of proximity loggers Ecuador cloud forest. In: Real‑ world wireless sensor networks. Springer, for detection of contacts between maternal pairs of white‑tailed deer. Berlin; 2010. p. 25–36. Wildl Soc B. 2011;35(4):452–60. 30. Marfievici R, Murphy AL, Picco GP, Ossi F, Cagnacci F. How environmental 9. Anthony D, Bennett WP, Vuran MC, Dwyer MB, Elbaum S, Lacy A, Engels factors impact outdoor wireless sensor networks: a case study. In: 2013 M, Wehtje W. Sensing through the continent: towards monitoring migra‑ IEEE 10th international conference on mobile ad‑hoc and sensor systems; tory birds using cellular sensor networks. In: Proceedings of the 11th 2013. p. 565–73. international conference on information processing in sensor networks. 31. Burnham KP, Anderson DR. Information and likelihood theory: a basis for ACM; 2012. p. 329–40. model selection and inference. In: Burnham KP, Anderson DR, editors. 10. Drewe JA, Weber N, Carter SP, Bearhop S, Harrison XA, Dall SR, McDonald Model selection and multimodel inference: a practical information‑theo ‑ RA, Delahay RJ. Performance of proximity loggers in recording intra‑and retic approach. New York: Springer‑ Verlag; 2002. p. 49–97. inter‑species interactions: a laboratory and field‑based validation study. 32. SAS/STAT Software, Version 9.3 (2011) by SAS Institute Inc. PLoS ONE. 2012;7(6):e39068. 33. Garcia‑Sanchez AJ, Garcia‑Sanchez F, Losilla F, Kulakowski P, Garcia‑Haro J, 11. Mennill DJ, Doucet SM, Ward KAA, Maynard DF, Otis B, Burt JM. A Rodríguez A, Lopez‑Bao JV, Palomares F. Wireless sensor network deploy‑ novel digital telemetry system for tracking wild animals: a field test ment for monitoring wildlife passages. Sensor. 2010;10(8):7236–62. for studying mate choice in a lekking tropical bird. Method Ecol Evol. 34. Strandburg‑Peshkin A, Farine DR, Couzin ID, Crofoot MC. Shared 2012;3(4):663–72. decision‑making drives collective movement in wild baboons. Science. 12. Rutz C, Burns ZT, James R, Ismar SM, Burt J, Otis B, Bowen J, St Clair 2015;348(6241):1358–61. JJ. Automated mapping of social networks in wild birds. Curr Biol. 35. Tomkiewicz SM, Fuller MR, Kie JG, Bates KK. Global positioning system 2012;22(17):669–71. and associated technologies in animal behaviour and ecological 13. Rutz C, Morrissey MB, Burns ZT, Burt J, Otis B, St Clair JJ, James R. Calibrat‑ research. Philos Trans R Soc B. 2010;365(1550):2163–76. ing animal‑borne proximity loggers. Method Ecol Evol. 2015;6(6):656–67. 36. Tosa MI, Schauber EM, Nielsen CK. Familiarity breeds contempt: 14. St Clair JJ, Burns ZT, Bettaney EM, Morrissey MB, Otis B, Ryder TB, Fleischer combining proximity loggers and GPS reveals female white‑tailed deer RC, James R, Rutz C. Experimental resource pulses influence social‑ (Odocoileus virgilianus) avoiding close contact with neighbours. J Wildl network dynamics and the potential for information flow in tool‑using Dis. 2015;51(1):79–88. crows. Nat Commun 2015; 6: article number 7197. 37. Nathan R, Getz WM, Revilla E, Holyoak M, Kadmon R, Saltz D, Smouse 15. Boyland NK, James R, Mlynski DT, Madden JR, Croft DP. Spatial proximity PE. A movement ecology paradigm for unifying organismal movement loggers for recording animal social networks: consequences of inter‑log‑ research. Proc Natl Acad Sci USA. 2008;105(49):19052–9. ger variation in performance. Behav Ecol Sociobiol. 2013;67(11):1877–90. 16. Meise K, Krüger O, Piedrahita P, Mueller A, Trillmich F. Proximity loggers on amphibious mammals: a new method to study social relations in their terrestrial habitat. Aquat Biol. 2013;18(1):81–9. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Animal Biotelemetry Springer Journals

Understanding and geo-referencing animal contacts: proximity sensor networks integrated with GPS-based telemetry

Loading next page...
 
/lp/springer-journals/understanding-and-geo-referencing-animal-contacts-proximity-sensor-0fgv3Zdb1V
Publisher
Springer Journals
Copyright
Copyright © 2016 by The Author(s)
Subject
Life Sciences; Animal Systematics/Taxonomy/ Biogeography; Conservation Biology/Ecology; Terrestial Ecology; Bioinformatics; Freshwater & Marine Ecology
eISSN
2050-3385
DOI
10.1186/s40317-016-0111-x
Publisher site
See Article on Publisher Site

Abstract

Background: In animal ecology, inter‑ individual encounters are often investigated using automated proximity log‑ gers. However, data acquired are typically spatially implicit, i.e. the question ‘Where did the contact occur?’ remains unanswered. To resolve this issue, recent advancements in Wireless Sensor Network technology have facilitated the geo‑ referencing of animal contacts. Among these, WildScope devices integrate GPS‑ based telemetry within fully distributed networks, allowing contact‑ triggered GPS location acquisition. In this way, the ecological context in which contacts occur can be assessed. We evaluated the performance of WildScope in close‑ to‑ real settings, whilst controlling for movement of loggers and obstacles, performing field trials that simulated: (1) different scenarios of encounters between individuals (mobile–mobile contacts) and (2) patterns of individual focal resource use (mobile– fixed contacts). Each scenario involved one to three mobile and two fixed loggers and was replicated at two differ ‑ ent radio transmission powers. For each scenario, we performed and repeated a script of actions that corresponded to expected contact events and contact‑ triggered GPS locations. By comparing expected and observed events, we obtained the success rate of: (1) contact detection and (2) contact‑ triggered GPS location acquisition. We modelled these in dependence on radio power and number of loggers by means of generalized linear mixed models. Results: Overall we found a high success rate of both contact detection (88–87%: power 3 and 7) and contact‑ triggered GPS location acquisition (85–97%: power 3 and 7). The majority of errors in contact detection were false negatives (66–69%: power 3 and 7). Number of loggers was positively correlated with contact success rate, whereas radio power had little effect on either variable. Conclusions: Our work provides an easily repeatable approach for exploring the potential and testing the perfor‑ mance of WildScope GPS‑ based geo‑ referencing proximity loggers, for studying both animal‑ to‑ animal encounters and animal use of focal resources. However, our finding that success rate did not equal 100%, and in particular that false negatives represent a non‑ negligible proportion, suggests that validation of proximity loggers should be under‑ taken in close‑ to‑ real settings prior to field deployment, as stochastic events affecting radio connectivity (e.g. obsta‑ cles, movement) can bias proximity patterns in real‑ life scenarios. Keywords: Contact‑ triggered GPS, Fully distributed Wireless Sensor Networks ( WSNs), Movement ecology, False positives and false negatives, Proximity loggers *Correspondence: francesca.cagnacci@fmach.it Biodiversity and Molecular Ecology Department, IASMA Research and Innovation Centre, Fondazione Edmund Mach, Via Mach 1, 38010 San Michele all’Adige, TN, Italy Full list of author information is available at the end of the article © The Author(s) 2016. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/ publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Ossi et al. Anim Biotelemetry (2016) 4:21 Page 2 of 14 date. One is a so-called master–slave configuration where Background only the ‘master’ logger has the capacity to communicate Proximity loggers are increasingly used animal-borne with other loggers (the ‘slaves’) and triggers GPS schedul- devices [1] (see also Table  1), with great potential for ing when contact occurs, but contacts and GPS locations behavioural ecology studies (e.g. predator–prey interac- between the ‘slaves’ are not recorded [26]. Another is a tions [22]; intra-specific relationships [23]; social systems fully distributed configuration where all loggers commu- [24]) and wildlife epidemiology (i.e. by evaluating contact nicate equally with each other, with both contacts and rate [10]). Proximity loggers contain a radio unit that is triggered GPS positions recorded by all loggers in the able to receive and transmit a low-power signal (ultra- network [19]. The fully distributed configuration char - high frequency, UHF, range 300  MHz–3  GHz) from/ acterizes the prototype WildScope [19] and has recently to other nearby loggers, thus forming a Wireless Sensor also been proposed for some commercial loggers [27], Network (WSN). In this way, loggers ‘detect’ the proxim- although so far empirical applications in animal ecology ity of one another. A micro-controller and a power unit remain limited. are essential complements to the radio sensor for deploy- Fully distributed configurations of WSNs somewhat ing proximity loggers on animals, so that radio transmis- parallel those of acoustic tags in the marine environ- sion and detection are self-powered, and recording and ment (e.g. detection of pelagic fish in harbours [28]), in storing of information automated. Proximity loggers that they combine a ‘mixed’ network of mobile and fixed can be used as both biologging units, i.e. deployed on loggers. Hence, they also provide new opportunities to animals for subsequent data retrieval via recapture [5], simultaneously investigate both geo-referenced encoun- and as biotelemetry devices, i.e. equipped with a remote ters between individuals (mobile–mobile contacts) data retrieval system, such as a GSM modem [9, 19], or and the use of specific habitat resources by individuals a WSN comprised of mobile proximity loggers and fixed (mobile–fixed contacts). The latter has long been recog - base stations (see Table 1 and [2, 9]). However, proximity nized as an important application of WSNs [5] but has loggers frequently provide only spatially implicit informa- thus far seldom been used in animal ecology studies (but tion, often processed with network-based analytical tools see [10, 11] and Table  1). In this work, we explore the (e.g. Social Network Analysis [6, 12]), where the sampling functionality and potential of WildScope WSN ([19] and units are the contacts between individuals (or ‘edges’) Additional file 1) for animal ecology investigation. rather than the individuals wearing the devices. The value Proximity loggers—and GPS-based geo-referencing of proximity information can be further enhanced by proximity loggers—should be carefully calibrated prior to making animal-to-animal contacts spatially explicit [25], use [13], as with all biologging devices. To this end, sev- providing ecologists with the added value of knowing eral authors have attempted to identify both the potential where animal encounters occur. Thus, the ecological con - and the limitations of WSNs, including contact reciproc- text of a given contact can be inferred (e.g. which habitat ity [15], effect of antenna orientation [5] and RSSI/dis - features characterize the contact site). Recent advance- tance relationship ([11, 13]; see also Table 1). However, no ments in proximity logging technology have incorporated published studies have been presented so far to investi- indirect geo-referencing of animal contacts (see Table 1). gate performance of GPS-based geo-referencing proxim- In particular, Encounternet [12] and BATS [20] are WSN ity loggers in animal ecology and in particular to describe systems that infer spatial contextualization of proximity (1) how to design field calibration studies based on GPS- patterns by post-processing, and specifically by estimat - based geo-referencing proximity loggers, (2) what type of ing the distance between mobile loggers worn by individ- information GPS-based geo-referencing proximity log- uals, and fixed loggers deployed in the study area. In this gers collect and (3) how reliable they are at collecting it. scenario, the geo-referencing of proximity patterns essen- To address these issues, we developed and tested a pro- tially depends on the spatial overlap between a network tocol to derive data on the occurrence and location of of fixed loggers and animal movements, i.e. ultimately on animal encounters in semi-controlled conditions using the range of the target species. Where monitored animals WildScope GPS-based geo-referencing proximity log- move over wide areas (as is typical of medium to large gers. The work represents an extension of a preliminary mammals), coverage with fixed loggers becomes unfea - field calibration [19] that measured the average contact sible. In this case, geo-referencing of proximity patterns distance threshold of loggers at different radio transmis - between individuals is possible via integration of a GPS sion powers, in static and controlled open-air conditions sensor in the logger [13]. This capability is provided by [10]. Here, we introduced two well-recognized sources of those WSNs which integrate proximity detection with stochasticity in contact detection, i.e. the movement of GPS-based telemetry to simultaneously provide GPS loggers [11] and the presence of obstacles in the network location acquisition with animal-to-animal contacts [19]. [29]. Specifically, in our calibration exercise we mimicked Two different configurations have been developed to Ossi et al. Anim Biotelemetry (2016) 4:21 Page 3 of 14 ‑ ‑ ‑ ‑ ‑ ‑ ‑ Table 1 Review of proximity logger deployments Study Logger Logger Target species Mobile Static loggers Target Radio fre- Geo- Remote data down- Calibration scenario Power con- brand or weight (g) loggers detection quency referencing load sumption project name range (m) of encounters analysis [2] Zebranet Not Zebra (Equus Yes (only Yes (base sta Not reported 900 MHz No Yes Not reported Yes reported burchellii) for data tions ‘daily A GPS module Via radio between retriev drive’) is embedded mobile loggers and ing) in the logger base stations to track zebra move ments [3] Sirtrack Ltd. 30 Common Yes No 0.3–0.5 160 MHz No No (recapture neces Tests on captive Possums No brushtail sary) [4] possum (Trichosurus vulpecula) [5] Sirtrack Ltd. 125 Racoon (Pro- Yes No 1–1.5 916.5 MHz No No (recapture neces Laboratory test to assess Yes cyon lotor) sary) (1) threshold contact distance in dependence of antenna orientation; (2) contact duration and robustness Field tests on raccoons to evaluate (1) interlogger distance variation pre and post deployment; (2) contacts reciprocity between loggers [6] Sirtrack Ltd. 150 Eurasian badger Yes No 1.5–2.5 916.5 MHz No No (recapture neces Laboratory test to assess No (badger); (Meles meles)/ sary) threshold contact not domestic distance reported cattle (Bos the taurus) weight of the cattle logger [7] Sirtrack Ltd. 120 Tasmanian devil Yes No 0.3–0.5 916.5 MHz No No (recapture neces Laboratory test to assess No (Sarcophilus sary) threshold contact harrisii) distance Field test on captive Tas manian devils to assess contact distance, using video camera as ground truth [8] Sirtrack Ltd. Not White tailed Yes No 0–1.0 916.5 MHz No No (recapture neces Field test on captive white No reported deer sary) tailed deer to compare (Odocoileus observed and recorded virginianus) contacts Contacts reciprocity assessed Ossi et al. Anim Biotelemetry (2016) 4:21 Page 4 of 14 ‑ ‑ ‑ ‑ ‑ ‑ ‑ Table 1 continued Study Logger Logger Target species Mobile Static loggers Target Radio fre- Geo- Remote data down- Calibration scenario Power con- brand or weight (g) loggers detection quency referencing load sumption project name range (m) of encounters analysis [9] CraneTracker ~100 Whooping Yes (only Yes (base sta Not reported 2.4 GHz No Yes Field test on wild turkeys Yes crane (Grus for data tions) A GPS is Via radio between (Meleagris gallopavo) and americana) retriev embedded mobile loggers on crane species other ing) in the logger and base stations than Grus americana, to to track placed at wintering evaluate logger function longrange and breeding sites; ing prior to deployment migratory via GSM modem on target species movements during migration phase [10] Sirtrack Ltd. Not Eurasian badger Yes Yes 0.5–2 916.5 MHz No No (recapture neces Laboratory test to evaluate No reported (Meles meles)/ sary) (1) interlogger distances domestic at different heights cattle (Bos and combinations; (2) taurus) contacts reciprocity; (3) contacts duration Field test on cattle to compare observed and recorded contacts [11] Encounternet 1 Longtailed Yes Yes 5–30 433 MHz No Yes Field test to evaluate (1) No manakin Via radio between RSSI/distance relation (Chiroxiphia mobile loggers ship; (2) duration and linearis) and base stations, robustness of contacts; then via radio to a (3) effect of antenna ori PC by means of a entation on connectivity; masternode (4) effect of movement on connectivity (simula tion) [12] Encounternet 10 New Cal Yes Yes (base sta 20–40 for 433 MHz Yes Yes Field test on a fixed No edonian tions) mobile Via cross Via radio between network made up of crow (Corvus loggers triangulation mobile loggers dead quails, to model the monedu- 100 for static between and base stations, probabilistic relation loides) loggers base stations then via radio to a ship between RSSI and and mobile PC by means of a distance in function of loggers. RSSI masternode height from ground, values used antenna orientation, to estimate habitat type. Details of interlogger calibration model are distance provided in [13, 14] [15] Sirtrack Ltd. Not Cattle (Bos Yes Yes (base sta 2.0–3.5 916.5 MHz No No (recapture neces Laboratory test to assess No reported taurus) tions) sary) interlogger distance (mobile–base station) Field test on dairy cows to measure loggers reciprocity in measuring contact duration Ossi et al. Anim Biotelemetry (2016) 4:21 Page 5 of 14 ‑ ‑ ‑ ‑ ‑ Table 1 continued Study Logger Logger Target species Mobile Static loggers Target Radio fre- Geo- Remote data down- Calibration scenario Power con- brand or weight (g) loggers detection quency referencing load sumption project name range (m) of encounters analysis [16] Encounternet 65–70 Galapagos sea Yes Yes (base sta 10 433 MHz No Yes Laboratory test to assess No lion (Zalophus tions) Via radio between (1) RSSI/distance relation wollebaeki) mobile loggers ship; (2) effect of antenna and base stations, orientation on connectiv then via radio to a ity; (3) duration and PC by means of a robustness of contacts masternode Field test (1) to measure loggers reciprocity; (2) to compare observed and recorded contacts; (3) to assess RSSI/distance relationship in outdoor conditions [17] Bat Monitor 30–50 Flying fox fam Yes Yes (base sta Not reported < 1 GHz No Yes Not reported Yes ing Project ily (Pteropodi tions) A GPS is Via radio between dae) embedded mobile loggers and in the logger base stations, then to track bats via GSM modem movements to the central database [18] Encounternet 1.3 Barn swallow Yes Yes (base sta 0–40 for 433 MHz No Yes Field tests to evaluate (1) No (Hirundo tions) mobile Via radio between body effect, antenna rustica) loggers mobile loggers orientation and environ 100 for static and base stations, ment effect on RSSI/ loggers then via radio to a distance relationship; (2) PC by means of a interlogger variability; masternode (3) contacts reciprocity [19] WildScope 440 Roe deer Yes Yes 5–30 2.4 GHz Yes Yes Field test ‘in vitro’ to evalu Yes (Capreolus A GPS is trig Via GSM modem ate effect of height from capreolus)/ gered in case from mobile log ground, casing and domestic of contact gers to the central radio power on contact horse (Equus between database distance caballus) proximity Field test on horses to loggers compare observed ver sus recorded contacts in relation to distance and radio power Field test on roe deer to evaluate functioning of contacttriggered GPS acquisition in case of contact Ossi et al. Anim Biotelemetry (2016) 4:21 Page 6 of 14 ‑ ‑ Table 1 continued Study Logger Logger Target species Mobile Static loggers Target Radio fre- Geo- Remote data down- Calibration scenario Power con- brand or weight (g) loggers detection quency referencing load sumption project name range (m) of encounters analysis [20] BATS 2 Mouse eared Yes Yes (base sta 50 868 MHz Yes Yes Not reported Yes bat (Myotis tions) Via cross Via radio between myotis) triangulation mobile loggers and between base stations base stations and mobile loggers. RSSI values used to estimate interlogger distance [21] BATS 2 Fringelipped Yes Yes (base sta 10 868 MHz No Yes Field test ‘in vitro’ to model No bat (Trachops tions) Via radio between relationship RSSI/dis cirrhosis) mobile loggers and tance, also in function of base stations antenna orientation Field test ‘in vivo’ on mouse eared bats to link RSSI variation to individual movement, using video camera as ground truth The table collects information on some of the most well-known proximity logger deployments. Details on target species, proximity logging features, calibration attempts and geo-referencing functionalities are provided We report the detection range of interest for the specific study, which does not necessarily pair the maximum distance potentially covered by the loggers of any given brand Ossi et al. Anim Biotelemetry (2016) 4:21 Page 7 of 14 stereotyped individual encounters and individual use of download. The fixed loggers are equipped solely with a focal resources [11], designing and performing field tri - radio unit, to allow contact detection with mobile log- als in semi-controlled conditions. During the trials we gers (see Additional file  1 and Picco et al. [19] for further described and predicted logger performance (in terms of details). contact detection and contact-triggered GPS locations WildScope adopts a fully distributed configuration of acquisition) based on the results provided in Picco et  al. the WSN: in contrast to the master–slave configuration, [19] (objective 1). Eventually, we compared the predicted all loggers transmit and receive a radio signal on the same and observed data to derive estimates of contact success channel, so that connectivity is allowed throughout the rate and contact-triggered GPS location success rate, network (see Additional file  1). A contact between two thus testing the reliability of WildScope GPS-based geo- mobile loggers (i.e. a dyad) denotes an event of animal referencing proximity loggers (objective 2). proximity, representing an animal encounter [13]. When a contact occurs between two mobile loggers, WildScope Methods triggers the acquisition of a GPS location (i.e. ‘contact- WildScope proximity sensor network triggered GPS location acquisition’) that is independent WildScope GPS-based geo-referencing proximity log- from the scheduling of periodic GPS acquisition, which gers are designed to be deployed in a network of mobile happens in parallel (see Additional file  1 and Picco et al. and fixed sensors (Fig.  1). Mobile loggers are commonly [19] for details). Finally, a contact occurring between a deployed on animals, whilst fixed loggers are deployed at mobile and a fixed logger represents the interaction of an static stations. The mobile devices include, among other individual with a static feature, e.g. use of a focal resource components, a radio unit for contact detection that can [19]. be set at different transmission powers, a Global Posi - Lifetime properties, details on power settings, data tioning System component for individual localization storage and data processing of WildScope are described and a quadriband GSM/GPRS modem for remote data extensively in Picco et al. [19] (see also Additional file 1). Fig. 1 Conceptual scheme of a mixed Wireless Sensor Network. The network is composed of mobile loggers deployed on animals (circles), equipped with GPS unit for localization, fixed loggers deployed on focal resources (squares) and a base station (triangle). Grey-filled forms denote the presence of a modem on the unit for remote data retrieval. Mobile loggers can be in contact via radio with each other (dash‑dot line). Data download can also happen via radio through contact detection (dotted lines), either between mobile and fixed loggers (and from here to the base station) or directly from the mobile loggers to the base station Ossi et al. Anim Biotelemetry (2016) 4:21 Page 8 of 14 Wireless Sensor Networks at work: network design and data recording (objective 1) The primary goal of our experiments was to design field trials in close-to-real settings that simulated stereotyped encounter patterns between individuals and between an individual and a focal resource [11], and to describe logger performance. In particular, we designed three dif- ferent behavioural scenarios (Fig.  2). The first simulated use of focal resources by one individual (mobile–fixed loggers) and the others additionally mimicked behav- ioural modes that imply proximity between individuals (mobile–mobile loggers; Fig.  3), specifically: encounter and avoidance (e.g. territory defence); paired movement in parallel directions (e.g. mother–calves); and random, Fig. 3 Photograph of deployment scenarios. Two people move the mobile loggers on an established path to simulate a pair of individu‑ als moving side by side independent movement of individuals in the same sur- roundings (e.g. for feeding purposes). The experiment took place in San Michele all’Adige, Italy, from 14 May 2014 to 20 May 2014, in dry and windless conditions, for a total of 7  days. The study site encompassed a regular and flat terrain with some trees and bushes (Fig.  3). We intentionally included these environmental obstacles in our trials to account for the generic noise of connectivity typical of the environments in which loggers are deployed on animals [10, 29, 30]. We set the loggers to two radio transmission powers: 3 and 7, since they corresponded to average contact distances appropriate to studying the type of stereotypical encoun- ters described above (see Additional file  1 for details on the power settings; Picco et  al. [19] for average contact distances). We set up the experimental network of fixed and mobile loggers as follows. Two fixed loggers were fitted on trees at 1 m height, within plastic water-proof boxes; mobile loggers (one to three, according to the scenario) Fig. 2 Graphical description of the three deployed scenarios to test were fitted onto two-litre bottles filled with a saline solu - WildScope. The path followed by each mobile logger is described by tion to mimic the body effect of an animal [10, 19] and a line of a different colour. The sequence of contemporary actions is then fixed to rigid wooden supports, with the antenna represented by letters, whereas numbers indicate different loggers. oriented upwards (Fig.  3). For consistency, we chose the Scenario 1: one mobile logger moving towards two fixed loggers same site and logger support used for the field calibration (mimic of focal resource use) (box 1, upper left); scenario 2: two mobile loggers (mimic of encounter and avoidance movements) and two described in Picco et al. [19] that we expanded to ‘noisier’ fixed loggers (box 2, upper right); scenario 3: three mobile loggers settings. (mimic of parallel and random movements) and two fixed loggers In each of the three scenarios described above, one or (box 3, bottom left). The shaded circles represent the expected range more experimenters moved the mobile logger(s) on a of contact detection for the fixed loggers (stars) at the two power predetermined path for a period of 17 min, which repre- settings (dark-shaded small circle: power 3; light-shaded large circle: power 7). The radii of these circles, which are not reported here on sented the trial (Fig. 3). Each trial comprised a ‘script’ of the proper scale, were derived from the outcome of previous tests [8]. actions controlling the position of every logger in space The dark green circles represent trees and bushes of various sizes in and time (Table  2), at 1-min intervals (i.e. the temporal the arena where we performed the experiments resolution of contact detection chosen or ‘epoch’; see Ossi et al. Anim Biotelemetry (2016) 4:21 Page 9 of 14 Table 2 Example of  scenario for  testing contact detection Table 3 Example of  match between  expectations and contact-triggered GPS location and observations Minute Action Expected event Reason for expectation Minute Expected event Observed event Typology event 1 Stop No contact Distance M1–L1 and M1– 1 No contact No contact True positive ( TP) L2 > 2.38 m 2 No contact No contact True positive ( TP) 2 Stop No contact Distance M1–L1 and M1– 3 No contact No contact True positive ( TP) L2 > 2.38 m 4 No contact No contact True positive ( TP) 3 Stop No contact Distance M1–L1 and M1– 5 Start contact with L1 Contact with L1 True positive ( TP) L2 > 2.38 m GPS triggered GPS location True positive ( TP) 4 Stop No contact Distance M1–L1 and M1– L2 > 2.38 m 6 In contact with L1 Contact with L1 True positive ( TP) 5 Movement Start contact with L1 Distance M1–L1 < 2.38 m 7 In contact with L1 No contact with L1 False negative (FN) GPS triggered Detection of a contact 8 End of contact with No contact with L1 True positive ( TP) L1 6 Movement In contact with L1 Distance M1–L1 < 2.38 m 9 Start contact with L1 Contact with L1 True positive ( TP) 7 Movement In contact with L1 Distance M1–L1 < 2.38 m GPS triggered GPS location True positive ( TP) 8 Stop End of contact with L1 Distance M1–L1 > 2.38 m 10 End of contact with No contact with L1 True positive ( TP) 9 Movement Start contact with L1 Distance M1–L1 < 2.38 m L1 GPS triggered Detection of a contact 11 Start contact with L2 Contact with L2 True positive ( TP) 10 Movement End of contact with L1 Distance M1–L1 > 2.38 m GPS triggered GPS location True positive ( TP) 11 Movement Start contact with L2 Distance M1–L2 < 2.38 m 12 End of contact with Contact with L2 False positive (FP) GPS triggered Detection of a contact L2 12 Movement End of contact with L2 Distance M1–L2 > 2.38 m 13 No contact Contact with L2 False positive (FP) 13 Movement No contact Distance M1–L2 > 2.38 m 14 Start contact with L2 Contact with L2 True positive ( TP) 14 Movement Start contact with L2 Distance M1–L2 < 2.38 m GPS triggered No GPS location True positive ( TP) GPS triggered Detection of a contact 15 In contact with L2 Contact with L2 True positive ( TP) 15 Stop In contact with L2 Distance M1–L2 < 2.38 m 16 In contact with L2 Contact with L2 True positive ( TP) 16 Stop In contact with L2 Distance M1–L2 < 2.38 m Example of matching the expected and observed data in terms of contact Example of a 17-min trial ‘script’ for testing contact detection and contact- detection and contact-triggered GPS location acquisition. The results refer to the first scenario, as for Table 2 (one mobile logger M1 and two fixed loggers L1 triggered GPS location acquisition with WildScope loggers, deployed in a mixed network (mobile and fixed loggers). The script refers to the first scenario (Fig. 2, and L2, power 3, see also Fig. 2), and specifically to the results obtained for the mobile logger. The recorded events depend on the comparison of expected and one mobile logger M1 and two fixed ones, L1 and L2), with loggers set at power 3 where the expected contact detection threshold is 2.38 m [19]. The table refers observed events only to the actions taken by the mobile logger. Both expected contact detection and contact-triggered GPS location expected events are reported, with the associated reason for expectation We performed three replicates of the trials for each experimental setting and power, for a total of 18 tri- als. We used three to five loggers to perform the trials, Additional file  1). u Th s, for each minute, we knew the depending on the scenario mimicked, and reused the respective position (and distance) of loggers from each same loggers across replicates to avoid biases arising other. We were able to express the ‘expected’ connectivity from a single-logger malfunctioning. between loggers based on the findings of Picco et al. [19], who determined the average contact distance in base- Estimation of contact detection and contact-triggered GPS line conditions (static loggers, open-air, no obstacles; see location success rate (objective 2) also [10]). Notably, Picco et  al. [19] determined an aver- We used the proximity and contact-triggered GPS loca- age contact distance of 2.38  ±  1.65  m for power 3 and tion data collected during the trials to measure the 7.31 ± 2.97 m for power 7. Thus, we expected loggers to reliability of contact detection and acquisition of con- be in contact if they were at or below the linear distance tact-triggered GPS locations by WildScope in close-to- of 2.38  m for power 3 and 7.31  m for power 7 (see also real settings. We did this by comparing expected and Fig.  2: such threshold distances are represented as buff - observed contact events and acquisitions of contact- ers around the loggers). Similarly, every observed contact triggered GPS locations during the trials, to eventually leads to the acquisition of an expected contact-triggered determine the success rate both for contact detection GPS location. For each trial, we were thus able to express and for acquisition of contact-triggered GPS locations. all expected events (Table  2), that we then cross-refer- Specifically, we transposed contacts and contact-trig - enced with the observed ones, based on the timestamp gered GPS locations into a list of true positives (TP; (Table 3). expected and occurred contact detection events or Ossi et al. Anim Biotelemetry (2016) 4:21 Page 10 of 14 expected and occurred contact-triggered GPS loca- Results tions), false negatives (FN; expected events which were Contact detection and contact-triggered GPS location not recorded by the loggers) and false positives (FP; acquisition in simulated scenarios of encounters unexpected but observed events). In this way we derived and movement (objective 1) estimates of contact success rate (RTP ) by comput- For each combination of number of loggers per power, we contact ing the ratio of true positives to the total expected and matched the observed and expected events for both con- unexpected events (RTP   =  TP/[TP  +  FP  +  FN]). tact detection and contact-triggered GPS location acqui- contact We also computed the false-negative rate of contact sition. Hereafter we provide a description of the mobile detection (RFN ) as the ratio of false negatives to logger performance, as a guideline for users who wish contact total unexpected events (RFN   =  FN/[FN  +  FP]). to repeat this validation exercise (see Tables  2, 3 for the contact For the analysis on success rate of contact-triggered described example and Fig.  2 for the corresponding sce- GPS locations, we did not include false-positive events, nario). From the beginning until minute 6 we classified i.e. records acquired out of the expected contact-trig- only true-positive events, i.e. M1 performed as expected gered GPS locations, because these represented spuri- (no contact from minute 1 to minute 4; contact detec- ous information that would not bias the analyses. Thus, tion with L1 at minutes 5 and 6, with consequent acqui- we computed the success rate of contact-triggered sition of contact-triggered GPS location). At minute 7, GPS locations (RTP ) as the ratio of true positives M1 was moved away, but still within the expected range GPS to the sum of true-positives and false-negative events of contact detection with L1. Contrary to expectations, (RTP  = TP/[TP + FN]). M1 broke the contact with L1. We classified this unex - GPS We used RTP , RFN and RTP to evalu- pected end of contact as a false-negative event. From contact contact GPS ate the effect of radio transmission power on the perfor - minute 8 until minute 11, the expected and observed mance of WildScope whilst controlling for the number events matched. At minute 12 and 13, contrary to our of loggers in the network. Specifically, we fitted a gen - expectations, M1 kept the contact with L2. We classified eralized linear mixed model (GLMM) to evaluate the this unexpected duration of the contact as a false-positive dependence of RTP , RTP and RFN on event. At minute 14, the scenario predicted a new con- contact GPS contact power level (two levels, 3 and 7) and number of loggers tact between M1 and L2, with the consequent acquisition (three levels: 1, 2 and 3). We used the 3 replicates within of a contact-triggered GPS location. Since the contact each combination of power per number of loggers in between M1 and L2 had not been interrupted during the network to estimate the variance. Since the depend- minutes 12–13, M1 kept the contact with L2 at minute 14 ent variable is binary (i.e. contact recorded or not, 1 and (true-positive event), but did not acquire a contact-trig- 0, respectively), we fitted a binomial distribution with a gered GPS location. This was correct, since the contact logit link function. For all the three analyses, we prepared had been never interrupted. We thus classified the lack of a list of candidate models including all the potential acquisition of a contact-triggered GPS location as a true- combinations of the two covariates, fitted either as fixed positive event. In the last 2 min of the trial M1 performed or random effects. We adopted this procedure to evalu - as expected. ate whether the fitted covariates contributed to explain We adopted a similar approach for each combination of the variance of the intercept only, or whether they were power and number of loggers involved in the scenarios. significant predictors of the model (see also Additional file 2). Estimation of contact detection and contact-triggered GPS We then applied a model selection procedure based on location success rate (objective 2) AIC scores to determine the model of best fit [31]. We We derived estimates of contact success rate based on further evaluated the importance of each of the terms 3095 total expected contact events, of which 2757 were retained in the best-fit model to contribute to the good - true-positive events. We computed the false-negative rate ness-of-fit of the model by means of an ANOVA based from 231 false-negative and 108 false-positive events. We on deviance procedure (see Additional file  2). We applied based our assessment of contact-triggered GPS location an F test to account for overall differences within the success rate on 104 realized contact-triggered GPS loca- variances of the covariates levels retained in the best-fit tions out of 113 expected locations (Table 4). model. Lastly, we tested the significance of β-coefficients We found a high success rate of contact detection with respect to the reference level by means of a Student’s RTP with respect to the expected events that we contact t test. described for the simulated scenarios, for both powers All the statistical analyses were programmed in SAS tested (89% for power 3; 89% for power 7) and across software 9.3 [32]. the number of loggers involved in the trials (86% for one logger; 87% for two loggers; 91% for three loggers). Ossi et al. Anim Biotelemetry (2016) 4:21 Page 11 of 14 Table 4 Contingencies to  estimate contact success rate, false-negative rate and  contact-triggered GPS location success rate Trial Contact success rate False-negative rate Contact-triggered GPS location success rate Tot. events True pos. False neg. False pos. Tot. events True pos. 1L‑P3 204 184 20 0 14 14 1L‑P7 204 167 29 8 8 8 2L‑P3 480 414 46 20 17 11 2L‑P7 480 423 40 17 15 13 3L‑P3 864 774 52 38 29 28 3L‑P7 864 795 44 25 30 30 The table summarizes the contingency data used to estimate contact success rate, false-negative rate and contact-triggered GPS location success rate. ‘Tot. events’, ‘True pos.’, ‘False neg.’ and ‘False pos.’ denote, respectively, the total events, true positives, false negatives and false positives used to estimate the rates. 1L-P3 = trials with one logger set at power 3; 1L-P7 = trials with one logger set at power 7; 2L-P3 = trials with two loggers set at power 3; 2L-P7 = trials with two loggers set at power 7; 3L-P3 = trials with three loggers set at power 3; 3L-P7 = trials with three logger set at power 7 The model selection and the ANOVA based on devi - of loggers in the trials (86% of total failures for one log- ance procedure indicated that the number of loggers in ger; 70% of total failures for two loggers; 60% of total the network positively affected the contact success rate failures for three loggers). The false-negative rate was (Table  5a), whilst the effect of power was negligible (see negatively correlated with the number of loggers in the Additional file  2). The F test confirmed the significance of network, and this relationship approached significance the number of loggers in predicting the pattern of contact (F   =  3.57; P  =  0.0538; Table  5b, see also Additional 2,15 success rate (F  = 6.11; P < 0.05). file 2). 2,15 The analysis of false-negative rate RFN indicated The success rate of contact-triggered GPS location contact that the false negatives constituted the majority of total RTP was high at both powers (power 3: 88%; power GPS failures for both powers (67% of total failures for power 3; 7: 96%) and across the number of loggers in the trials 69% of total failures for power 7) and across the number (100% for one logger; 75% for two loggers; 98% for three loggers). The F test indicated a marginal significance of Table 5 Summary of  the best model accounting for  the the number of loggers in predicting the pattern of con- observed patterns of  contact success rate (a), false-neg- tact-triggered GPS location success rate (F   =  3.71; 2,15 ative rate (b) and  contact-triggered GPS location success P = 0.0492). In particular, RTP was significantly lower rate (c) GPS when two loggers were deployed, whilst there were no β-Coefficient SE df t P differences between trials with one and three loggers (Table 5c and Additional file 2). (a) Model contact success rate Intercept 2.2893 0.08323 15 27.51 <.0001 Discussion 1 logger −0.4716 0.1653 15 −2.85 0.0121 In this paper we describe contact detection by Wild- 2 loggers −0.3635 0.1277 15 −2.85 0.0123 Scope, a fully distributed proximity logging Wireless 3 loggers – – – – – Sensor Network that allows GPS-based geo-referencing (b) Model false-negative rate proximity detection, which we tested within a semi-con- Intercept 0.6931 0.1768 15 3.92 0.0014 trolled set-up. Previous studies have performed valida- 1 logger 1.1192 0.4203 15 2.66 0.0177 tion exercises of proximity loggers in order to evaluate 2 loggers 0.1386 0.2647 15 0.52 0.6082 the effect of body obstruction on radio connectivity (e.g. 3 loggers – – – – – [9]) or to measure the relationship between logger dis- (c) Model contact‑triggered GPS location success rate tance and radio connectivity (e.g. [13]). The final goal Intercept 4.0604 1.0086 15 4.03 0.0011 of the majority of these calibrations was to measure the 1 logger 11.2386 4.4765 15 0.03 0.9803 average threshold distances of contact detection (e.g. [6]) 2 loggers −2.9618 1.0881 15 −2.72 0.0157 in order to choose the correct transmission setting that 3 loggers – – – – – corresponds to the distance of biological interest (e.g. β-Coefficients, standard errors and significance for each level of the number of 1.5–2.0  m for perpetuation of bovine tuberculosis (bTB) loggers in the network, for the model of contact success rate (a), false-negative rate (b) and contact-triggered GPS location success rate. Since the number of in cattle herds [10]). Here, we also used average contact loggers was fitted as a categorical variable in the models, the β-coefficients refer threshold distance, as derived from Picco et  al. [19] and to the difference with respect to the reference level (three loggers) Ossi et al. Anim Biotelemetry (2016) 4:21 Page 12 of 14 made the assumption that contacts should occur at such Although mixed networks of fixed and mobile loggers a distance, within defined error bounds, i.e. we calculated represent the typical case for testing contact detection, expected events of contact detection. However, Picco they have rarely been included in animal-borne stud- et  al. [19] did not take into account the influence of fac - ies (but see [10, 11]). In this paper we propose a simple tors such as the presence of obstacles [29] and the move- mixed network and provide a description of its function- ment of loggers [11] on contact detection based on radio ing in the presence of moving subjects that mimic com- connectivity. Here, we built our experiments to control mon animal behaviours. We propose this exercise as a for movement and obstacles and demonstrated that, convenient way for users to understand how proximity despite the loggers being generally reliable, there was a loggers might work in typical deployment scenarios and noticeable rate of failures in contact detection. to easily test the reliability of a given proximity detection Furthermore, in our study the majority of failures were system whilst accounting for stochastic bias. Moreover, caused by false-negative events, i.e. expected contacts mixed networks have a much greater potential than being that did not occur. This is consistent with the increased ‘just’ a technical tool to remotely download data acquired noise arising from movement of loggers (as proposed by by mobile devices (e.g. [2, 9]). The integration of mobile Rutz et  al. [13]) and presence of obstacles [29]. Had the and fixed loggers in a fully distributed WSN can help effect of movement and obstacles been negligible, we address a variety of important themes in animal ecology, would have detected similar rates of false negatives and including focal resource use in a given habitat by an indi- false positives arising as a consequence of equal distri- vidual [11]; identification of crucial corridors for animal bution of errors within the variance of contact distance movement or cross-road sites [33]; and consequences of threshold determined by Picco et al. [19]. In contrast, the resource distribution on animal social networks [14]. majority of failures due to false negatives might be due In this work, we also demonstrate the reliability of a rel- to a reduced connectivity compared with controlled set- atively new function of GPS sensors—that is, contact-trig- tings, caused by logger movement and presence of obsta- gered GPS location acquisition, in application to animal cles in the network, i.e. an actual ‘lower’ threshold value ecology. Aside from our anomalous finding that contact- in the ‘noisier’ conditions compared with the baseline triggered GPS location success rate was reduced using ones of Picco et al. [19]. We also found that contact suc- two loggers, the overall performance of this GPS-based cess rate increased with the number of loggers, and false- geo-referencing proximity logger system was highly satis- negative rate decreased. This can be explained by the factory. Furthermore, the fully distributed (peer-to-peer) hypothesized stochasticity of the ‘noise’ of the system, capability of WildScope and other logging systems with which is characteristic of radio transmission and thus a fully distributed configuration considerably extends of WSN deployments [29, 30]. In particular, the noise the type of ecological question that can be addressed in caused by the environmental conditions [30] in addition comparison with master–slave configurations. All log - to the presence of obstacles and fine movements [29] is gers in the network can exchange proximity information likely to exert stochastic variation in connectivity that with all others, whilst contacts and contact-triggered GPS leads to the occurrence of errors (either false positives or locations are recorded in parallel by all loggers involved false negatives) in contact detection. Given the stochas- in the contact, thus permitting ‘encounter direct mapping’ tic nature of this ‘noise’, its overall effect on connectivity (sensu [23]). This has the potential to elucidate previously should be less than proportional to the number of loggers little-known ecological and behavioural processes. For included in the network. In other words, it should not example, social interactions in a group of individuals can strictly depend on the number of loggers deployed in the be detected and geo-referenced, as well as movement pat- network. Conversely, the number of contacts between terns of a group of individuals approaching or abandoning sensors (i.e. the true positives) increases proportionally a point resource (e.g. feeding stations). with the number of loggers deployed in the network. Finally, contact-triggered GPS locations are addi- Therefore, if the number of contact events increases tional to periodic GPS acquisition for all individuals beyond the number of associated errors, so too does the equipped with loggers. In this way, studies on spatially success rate. Further research in this direction should explicit contact detection can be combined with ‘tradi- disentangle the overall stochasticity of the system with tional’ movement research [25]. The flexible scheduling the variability linked to specific deployment features (but of both contact-triggered and independent (periodic) see [29, 30]). In particular, future studies should address GPS location acquisition allows the user to adjust the the physics of transmission of WSN (but see [13]). Radio tool in accordance with study-specific requirements. For transmission power had a negligible effect on contact instance, mother–calf interactions can last a long time, success rate and false-negative rate, demonstrating that so the investigator of parental care might decrease the observed patterns are robust to the power adopted. scheduling frequency of the GPS device after the first Ossi et al. Anim Biotelemetry (2016) 4:21 Page 13 of 14 30  min of contact, to save battery lifetime. Conversely, Additional files the spatial contextualization of aggressive encoun- ters between individuals, which can be instantaneous, Additional file 1. WildScope proximity loggers. This additional file might require an intense but brief GPS sampling regime. describes in detail the technical components (hardware and software) of Indeed, we should note that when animal GPS detection WildScope GPS‑based geo ‑referencing proximity loggers. is possible at a very high rate, i.e. when battery consump- Additional file 2. Model selection procedure. This additional file tion constraints are irrelevant (e.g. large marine birds describes in detail the procedure to select the best model describing contact detection success rate, false negative rate and contact‑triggered provided with solar-powered GPS units) high-frequency GPS location success rate). animal trajectories can provide identical information on animal-to-animal contacts as proximity detection sen- sors [34]. However, GPS frequency trades-off with bat - Abbreviations tery lifetime (e.g. [19, 35]), with constraints including the TP: true positive; FP: false positive; FN: false negative; RTP : ratio of true‑ contact length of the monitoring period for each individual, and positive events of contact detection; RTP : ratio of true‑positive events of GPS contact‑triggered GPS location acquisition. recapture feasibility. If recapture is a viable option and/ or the monitoring period required is short, then GPS Authors’ contributions frequency can be kept at a maximum, thus guaranteeing FO, SF, GP, AM, DM, FC designed the experiments. FO, SF, DM, FC designed and carried out the analysis. FO, DM, NG carried out the experiments. FO, FC wrote semi-continuous monitoring from which it is possible to the paper, with all other authors’ contribution. All authors read and approved extrapolate proximity patterns between individuals. the final manuscript. Notwithstanding study-specific trade-offs, the spa - Authors’ information tial information associated with contacts can remarkably Our team is mainly constituted by academic researchers belonging to differ ‑ increase the robustness of inference on animal encounters ent fields. FO, SF, FC, NG, JG and BT are animal ecologists, whilst GP, AM and [36], enhancing a process-based interpretation of these DM are information engineers with long experience in Wireless Sensor Net‑ work applications. These two components of the team worked side by side observations [37]. This may prompt a completely new set to assess the potential of WSN for animal ecology investigation and wildlife of questions in behavioural ecology (e.g. do aggregation management. patterns in winter depend on resource distribution? Is pro- Author details vision of maternal care altered by habitat composition?), Biodiversity and Molecular Ecology Department, IASMA Research and Inno‑ which could not be addressed with previous technology. vation Centre, Fondazione Edmund Mach, Via Mach 1, 38010 San Michele all’Adige, TN, Italy. UMR CNRS 5558 “Biometrie et Biologie Evolutive”, Université Claude Bernard Lyon 1, Bat G. Mendel 43 Bd du 11 Novembre Conclusions 1918, 69622 Villeurbanne Cedex, France. Istituto dei Sistemi Complessi, CNR, We tested a novel biotelemetry tool that integrates prox- 4 Via Madonna del Piano 10, 50019 Sesto Fiorentino, FI, Italy. Department imity detection and contact-triggered GPS location of Information Engineering and Computer Science (DISI), University of Trento, Via Sommarive 9, 38123 Povo, TN, Italy. Center for Scientific and Technologi‑ acquisition, in a fully distributed mixed wireless net- cal Research, Bruno Kessler Foundation, Via Sommarive 18, 38123 Povo, TN, work. This technological advancement has the potential 6 Italy. Ecosystems and Environmental Management Group, The University to bring together two branches of ecology thus far kept of Brighton, Huxley Building, Lewes Road, Brighton, East Sussex BN2 4GJ, UK. Organismic and Evolutionary Biology Department, Harvard University, 26 relatively distinct, namely movement ecology [37] and Oxford St, Cambridge, MA 02138, USA. animal encounters [23]. Moreover, we suggest further enhancements of the application of mixed proximity Acknowledgements We acknowledge the work of Sandro Nicoloso and Michele Corrà in manufac‑ logger networks to more comprehensively understand turing WildScope. The research was conducted with the support of Forestry animal use of specific focal resources within a habitat. and Wildlife Service of the Autonomous Province of Trento. We are also grate‑ However, these wide-reaching and innovative applica- ful to three anonymous reviewers for insightful comments on previous drafts. tions of a novel GPS-based geo-referencing proximity Competing interests logger must take into account the limitations of their use. The authors declare that they have no competing interests. We present a simple and repeatable series of scenarios to Availability of data and materials test the functioning and reliability of GPS-based geo-ref- The data set supporting the conclusions of this article is available in the Dryad erencing proximity loggers, prior to deployment in field repository. conditions. We foresee further steps for calibrating these Consent for publication tools, including probabilistic modelling of error rates The subjects represented in Fig. 3 give their consent for publication of the associated with contact detection as a function of the dis- image. tance between loggers, in conditions as close as possible to final deployment [13], i.e. on wild animals. Ossi et al. Anim Biotelemetry (2016) 4:21 Page 14 of 14 Funding 17. Sommer P, Kusy B, McKeown A, Jurdak R. The big night out: experiences This project was partly funded by Autonomous Province of Trento (PAT ), under from tracking flying foxes with delay‑tolerant wireless networking. In: grant BEARNET, Protocol No. S044‑5/2012/226570. F.O. was granted three ‑ Real‑ world wireless sensor networks; 2014. p. 15–27. yearly scholarships financed by the European Union (European Social Funds), 18. Levin II, Zonana DM, Burt JM, Safran RJ. Performance of encounternet Aosta Valley Autonomous Region and the Italian Ministry for Work and Social tags: field tests of miniaturized proximity loggers for use on small birds. Politics. PLoS ONE. 2015;10(9):e0137242. 19. Picco GP, Molteni D, Murphy AL, Ossi F, Cagnacci F, Corrà M, Nicoloso S. Received: 8 August 2016 Accepted: 17 October 2016 Geo‑referenced proximity detection of wildlife with WildScope: design and characterization. In: Proceedings of the 14th international conference on information processing in sensor networks. 2015; p. 238–49. 20. Dressler F, Ripperger S, Hierold M, Nowak T, Eibel C, Cassens B, Mayer F, Meyer‑ Wegener K, Kolpin A. From radio telemetry to ultra‑low‑ power sensor networks: tracking bats in the wild. IEEE Commun Mag. References 2016;54(1):129–35. 1. Wilmers CC, Nickel B, Bryce CM, Smith JA, Wheat RE, Yovovich V. The 21. Ripperger S, Josic D, Hierold M, Koelpin A, Weigel R, Hartmann M, Page golden age of bio‑logging: how animal‑borne sensors are advancing the R, Mayer F. Automated proximity sensing in small vertebrates: design frontiers of ecology. Ecology. 2015;96(7):1741–53. of miniaturized sensor nodes and first field tests in bats. Ecol Evol. 2. Zhang P, Sadler CM, Liu T, Fischhoff I, Martonosi M, Lyons SA, Rubenstein 2016;6(7):2179–89. DI. Habitat monitoring with ZebraNet: design and experiences. In: Bulusu 22. Tambling CJ, Belton LE. Feasibility of using proximity tags to locate female N, Jha S, editors. Wireless sensor networks: a systems perspective. Nor‑ lion Panthera leo kills. Wildl Biol. 2009;15(4):435–41. wood: Artech House; 2005. p. 235–57. 23. Krause J, Krause S, Arlinghaus R, Psorakis I, Roberts S, Rutz C. Reality min‑ 3. Ji W, White PC, Clout MN. Contact rates between possums revealed by ing of animal social systems. Trends Ecol Evol. 2013;28(9):541–51. proximity data loggers. J Appl Ecol. 2005;42(3):595–604. 24. Prange S, Gehrt SD, Hauver S. Frequency and duration of contacts 4. Ji W, Clout MN, Douglas M, Day T, Hendra R. Mate ID: first trial of a novel between free‑ranging raccoons: uncovering a hidden social system. J device for measuring possum contacts. In: Sutherland G, editor. Advances Mammal. 2011;92(6):1331–42. in the biological control of possums: Report of a workshop sponsored 25. Bettaney EM, James R, St Clair JJH, Rutz C. Processing and visualizing by the National Science Strategy Committee for Possum & Bovine Tb association data from animal‑borne proximity loggers. Anim Bioteleme. Control. Wellington: Royal Society of New Zealand; 1999. p. 92–5. 2015;3(1):1–11. 5. Prange S, Jordan T, Hunter C, Gehrt SD. New radiocollars for the detection 26. Vectronic Aerospace GmbH. http://www.vectronic‑aerospace.com/ of proximity among individuals. Wildl Soc B. 2006;34(5):1333–44. wildlife‑monitoring/sensors/uhf‑id‑tags/. Accessed 29 Jul 2016. 6. Böhm M, Hutchings MR, White PC. Contact networks in a wildlife‑live ‑ 27. Zoological Society of London – Institute of Zoology. https://www.zsl.org/ stock host community: identifying high‑risk individuals in the transmis‑ science/research/badger‑ cattle‑ contact‑project. Accessed 29 Jul 2016. sion of bovine TB among badgers and cattle. PLoS ONE. 2009;4(4):e5016. 28. Klimley AP, Le Boeuf BJ, Cantara KM, Richert JE, Davis SF, Van Som‑ 7. Hamede RK, Bashford J, McCallum H, Jones M. Contact networks in meran S. Radio‑acoustic positioning as a tool for studying site ‑specific a wild Tasmanian devil (Sarcophilus harrisii) population: using social behaviour of the white shark and other large marine species. Mar Biol. network analysis to reveal seasonal variability in social behaviour and 2001;138(2):429–46. its implications for transmission of devil facial tumour disease. Ecol Lett. 29. Ceriotti M, Chini M, Murphy AL, Picco GP, Cagnacci F, Tolhurst B. Motes in 2009;12(11):1147–57. the jungle: lessons learned from a short‑term WSN deployment in the 8. Walrath R, Van Deelen TR, VerCauteren KC. Efficacy of proximity loggers Ecuador cloud forest. In: Real‑ world wireless sensor networks. Springer, for detection of contacts between maternal pairs of white‑tailed deer. Berlin; 2010. p. 25–36. Wildl Soc B. 2011;35(4):452–60. 30. Marfievici R, Murphy AL, Picco GP, Ossi F, Cagnacci F. How environmental 9. Anthony D, Bennett WP, Vuran MC, Dwyer MB, Elbaum S, Lacy A, Engels factors impact outdoor wireless sensor networks: a case study. In: 2013 M, Wehtje W. Sensing through the continent: towards monitoring migra‑ IEEE 10th international conference on mobile ad‑hoc and sensor systems; tory birds using cellular sensor networks. In: Proceedings of the 11th 2013. p. 565–73. international conference on information processing in sensor networks. 31. Burnham KP, Anderson DR. Information and likelihood theory: a basis for ACM; 2012. p. 329–40. model selection and inference. In: Burnham KP, Anderson DR, editors. 10. Drewe JA, Weber N, Carter SP, Bearhop S, Harrison XA, Dall SR, McDonald Model selection and multimodel inference: a practical information‑theo ‑ RA, Delahay RJ. Performance of proximity loggers in recording intra‑and retic approach. New York: Springer‑ Verlag; 2002. p. 49–97. inter‑species interactions: a laboratory and field‑based validation study. 32. SAS/STAT Software, Version 9.3 (2011) by SAS Institute Inc. PLoS ONE. 2012;7(6):e39068. 33. Garcia‑Sanchez AJ, Garcia‑Sanchez F, Losilla F, Kulakowski P, Garcia‑Haro J, 11. Mennill DJ, Doucet SM, Ward KAA, Maynard DF, Otis B, Burt JM. A Rodríguez A, Lopez‑Bao JV, Palomares F. Wireless sensor network deploy‑ novel digital telemetry system for tracking wild animals: a field test ment for monitoring wildlife passages. Sensor. 2010;10(8):7236–62. for studying mate choice in a lekking tropical bird. Method Ecol Evol. 34. Strandburg‑Peshkin A, Farine DR, Couzin ID, Crofoot MC. Shared 2012;3(4):663–72. decision‑making drives collective movement in wild baboons. Science. 12. Rutz C, Burns ZT, James R, Ismar SM, Burt J, Otis B, Bowen J, St Clair 2015;348(6241):1358–61. JJ. Automated mapping of social networks in wild birds. Curr Biol. 35. Tomkiewicz SM, Fuller MR, Kie JG, Bates KK. Global positioning system 2012;22(17):669–71. and associated technologies in animal behaviour and ecological 13. Rutz C, Morrissey MB, Burns ZT, Burt J, Otis B, St Clair JJ, James R. Calibrat‑ research. Philos Trans R Soc B. 2010;365(1550):2163–76. ing animal‑borne proximity loggers. Method Ecol Evol. 2015;6(6):656–67. 36. Tosa MI, Schauber EM, Nielsen CK. Familiarity breeds contempt: 14. St Clair JJ, Burns ZT, Bettaney EM, Morrissey MB, Otis B, Ryder TB, Fleischer combining proximity loggers and GPS reveals female white‑tailed deer RC, James R, Rutz C. Experimental resource pulses influence social‑ (Odocoileus virgilianus) avoiding close contact with neighbours. J Wildl network dynamics and the potential for information flow in tool‑using Dis. 2015;51(1):79–88. crows. Nat Commun 2015; 6: article number 7197. 37. Nathan R, Getz WM, Revilla E, Holyoak M, Kadmon R, Saltz D, Smouse 15. Boyland NK, James R, Mlynski DT, Madden JR, Croft DP. Spatial proximity PE. A movement ecology paradigm for unifying organismal movement loggers for recording animal social networks: consequences of inter‑log‑ research. Proc Natl Acad Sci USA. 2008;105(49):19052–9. ger variation in performance. Behav Ecol Sociobiol. 2013;67(11):1877–90. 16. Meise K, Krüger O, Piedrahita P, Mueller A, Trillmich F. Proximity loggers on amphibious mammals: a new method to study social relations in their terrestrial habitat. Aquat Biol. 2013;18(1):81–9.

Journal

Animal BiotelemetrySpringer Journals

Published: Nov 15, 2016

References