V-Track: software for analysing and visualising animal movement from acoustic telemetry detections

CSIRO PUBLISHING

Marine and Freshwater Research, 2012, 63, 815–820 http://dx.doi.org/10.1071/MF12194

V-Track: software for analysing and visualising animal movement from acoustic telemetry detections Hamish A. Campbell A,B, Matthew E. Watts A, Ross G. Dwyer A and Craig E. Franklin A A B

School of Biological Sciences, The University of Queensland, St Lucia, Qld 4072, Australia. Corresponding author. Email: [email protected]

Abstract. The tagging of aquatic and semi-aquatic animals with acoustic transmitters and their detection by passive underwater receivers has gained huge popularity over the past decade. This technology offers researchers the opportunity to monitor the finite- to broad-scale movements of multiple individuals over many years; however, the sheer scale and spatial complexity of these datasets are often beyond the capabilities of routine database and spread-sheet applications. In the present paper, we describe software (V-Track) that greatly facilitates the assimilation, analysis and synthesis of animallocation data collected by underwater passive acoustic receivers. The principal features within V-Track are the behavioural event qualifier (BEQ) and the receiver-distance matrix (RDM) calculator. The BEQ identifies and catalogues horizontal movements from receiver detection data, or vertical movements from transmitter sensor data (depth or temperature). The RDM is generated from the geographical location of the acoustic receivers and is utilised by V-Track to illustrate the behavioural event information in a spatial context. V-Track is a package written within the R-programming language, and a graphical user interface is also provided. Here, we feature two case studies to demonstrate software functionality for defining and quantifying behaviour in acoustically tagged marine and freshwater vertebrates. Additional keywords: behaviour, R, tracking VEMCO, VR2W. Received 19 July 2012, accepted 22 August 2012, published online 8 October 2012

Introduction The use of underwater acoustic telemetry to monitor the movements and behaviours of aquatic and semi-aquatic animals has proliferated over the past decade. The animal is fitted or implanted with a device that transmits an acoustic pulse encoded with a unique ID code and often sensor information (depth and temperature). The pulse can be detected through the water for hundreds of metres, and by actively tracking the tagged animal, or by deploying an array of static underwater receivers, animal movement, combined with physiological and environmental information, can be recorded (Clements et al. 2005; Humston et al. 2005; Heupel et al. 2006; Grothues 2009). The most widely used underwater acoustic receivers are those produced by VEMCO (Amirix Systems Ltd, Nova Scotia, Canada). The company estimates that over 15 000 of the VR2-W units are currently deployed worldwide (VEMCO employee, pers. comm.), with thousands of animals carrying the corresponding transmitters. The receivers operate independently on a single channel and multiple receivers can be deployed in a variety of formations, as curtain lines across estuaries or continental shelves (Melnychuk et al. 2010), as a series of gates throughout a river (Childs et al. 2008; Campbell et al. 2010) or in grid formation within areas of expected site fidelity (Mitamura et al. 2005; March et al. 2010). Acoustic transmitters are typically programmed to transmit more than one acoustic pulse Journal compilation Ó CSIRO 2012

every minute, and when multiple animals have been tagged, a single collection can comprise over a million acoustic detections. These datasets are beyond the capabilities of general spread-sheet applications (Heupel et al. 2006), and storing the data effectively while maintaining data integrity and accessibility is an involved process. We have addressed this issue through the creation of a database management system (called VEMCOtrack, abbreviated to V-Track), with a series of bespoke query functions and spatial capabilities. This system has been constructed within R (R Development Core Team 2011), a widely used, free and open-source software, to provide users with the flexibility to manage and plot their data and even add their own subroutines. We anticipate that many users of acoustic telemetry equipment may not be familiar with the R program, and we have therefore created a graphical user interface to call and run the V-Track R functions. The present paper details the functionality of V-Track for interpreting behavioural events from acoustic telemetry data for both marine and freshwater vertebrates. Materials and methods V-Track can support acoustic-detection data exported directly from the VEMCO software platform (VUE), as well as data collected by the VEMCO VR100 mobile receiver units, or from third party data repositories. These features provide users with the option of amalgamating detection data from several sources www.publish.csiro.au/journals/mfr

816

Marine and Freshwater Research

H. A. Campbell et al.

to facilitate batch processing. The main functions of V-Track are listed in Table 1. The database V-Track incorporates a database management system with query functions, to subset the data by time, receiver, station, transmitter ID and/or transmitter type.

Table 1. An overview of the most important functions of V-Track V-Track function

The function’s scope

ReadInputData

Imports the VEMCO VUE detection data into the V-Track structured data frame. Extract a subset of data (i.e. transmitters, receivers, stations, time period, sensor type). Calculates distance between the detection fields of adjacent receivers through a series of waypoints. Calculates the straight-line distance between all receivers in the array. Calculates time periods when tagged animals are within or between receiver detection fields. Calculates depth and temperature events based on increasing or decreasing values within a set period. Classifies residence, movement or sensor events by temporal conditions. Creates a keyhole markup language (KML) animation of horizontal movement that can be displayed in Google EarthTM.

ExtractData

GenerateCircuitousDistance

GenerateDirectDistance RunResidenceExtraction

RunSensorEventExtraction

RunTimeProfile GenerateAnimationKMLFile

The receiver-distance matrix The receiver-distance matrix (RDM) is used when multiple static underwater receivers have been deployed in an array formation. The RDM is essentially the distance of the outer limits of each detection field from all the receivers in the array. It is created from the geographic coordinates for each receiver within the array, together with their detection field radii and is used by V-Track to calculate and define animal trajectories and rate of movement. In some instances, the minimum-route distance undertaken by a tagged animal between two receivers may be the shortest route, but in deployments around coastlines and reefs, or throughout rivers and estuaries, the minimum distance may follow a more circuitous route between two receivers. V-Track provides users with the option of creating a distance matrix using either the shortest distance between each two receivers (Fig. 1a) or with a user-defined circuitous route passing through other receivers and additional waypoints (Fig. 1b). Moreover, the receiver detection-field radius will vary in its extent due to the environmental conditions, and thus the inter-receiver detection gap will not be a constant. The detection-field radius for each receiver is usually measured directly in the field and V-track allows the user to alter the detection field radius for each receiver before calculation of the RDM. The behavioural event qualifier The BEQ function is the most noteworthy feature of V-Track. This function qualifies presence, absence or sensor data into spatially and temporally labelled behavioural events. This condenses the detection database by orders of magnitude and greatly speeds up the data synthesis process. Each event is qualified through a series of user-defined threshold and timeout parameters, allowing the user to qualify and quantify behavioural events across a wide range (b) ⫺12.25

⫺12.30

⫺12.30

⫺12.35

⫺12.35

⫺12.40

⫺12.40

Latitude

(a) ⫺12.25

141.95

142.00 142.05

142.10 142.15

142.20 142.25

141.95

142.00 142.05

142.10 142.15

142.20 142.25

Longitude Fig. 1. (a) V-Track graph, displaying the receiver-distance matrix as calculated using the V-Track GenerateDirectDistance function. Crossed circles indicate the geographical locations of each receiver. The lines show that the V-Track calculated the minimum distance between receivers. (b) V-Track graph, displaying the receiver-distance matrix as calculated using the V-Track GenerateCircuitousDistance function. Crossed circles indicate the geographical locations of each receiver. The empty grey circles indicate user-defined waypoints. The lines indicate the minimum direct or circuitous distance between receivers.

V-Track

Marine and Freshwater Research

of temporal scales. Once the user defines the event, V-Track will search the detection-data archive and catalogue the qualified events. The spatial information for each behavioural event is derived from the RDM, and the temporal information from the time when the tagged animals arrived or departed from within the receiver’s detection field. It is important that users are able to examine the acoustic detections that comprise the events that V-Track has recorded, and each event is provided with a unique index code for cross-referencing. In this manner, the BEQ function can be optimised to improve event recognition. Erroneous detections are often generated during acoustic tracking and V-Track mitigates for these by only qualifying events that contain more than two acoustic detections in a 24-h period. This is the default setting and can be altered by the user. Creation of animations Visualising the horizontal movements of a tagged animal between the receivers within the array greatly aids in understanding event information, and also in identifying unexpected movement patterns. For this purpose, V-Track creates animations in the keyhole mark-up language (KML), which can be run in real time in Google EarthTM (S1, available as Supplementary data on the web). Case study 1 Determining patterns in horizontal movement from presence and absence data Thirty five VR2W receivers were spaced 1–2 km apart along a 65-km stretch of a tidal-river system and estuary. The geographical coordinates from each receiver were recorded by handheld GPS, and the detection radius of each receiver was determined by towing an activated transmitter behind a boat, and then matching up detections with boat location (for methods, see Campbell et al. 2012). Four bull sharks (Carcharhinus leucas, 1.4
0.2 m) were implanted with acoustic transmitters (V9 VEMCO, Halifax Canada), with a nominal delay time of 20 s. The tagged sharks were released within the confines of the receiver array and tracked for 6 months. The geographical coordinates of each VR2W receiver were uploaded into V-Track and the RDM was calculated using the circuitous method (Fig. 1b). The acoustic-detection data were downloaded from each receiver into a database within the VEMCO VUE software platform. The detection data were then exported as a .csv file (version 1.0 format) and imported into V-Track. The event-analyser function in V-Track was used to define the total duration spent by each shark within the detection

817

field of each receiver. The timeout window for the termination of each residence event was set to 10 min. This resulted in the event creator triggering an event when a transmitter was first detected by a receiver, and terminating the event at the last recorded detection if no further transmissions were detected within the 10-min window. A residence event would also terminate if the transmitter was detected at least twice at another receiver. Each of the qualified residence events and their associated variables are catalogued in an output file (Table 2). The grouping of the residence information for each shark at each receiver revealed the distribution of time that each shark spent within each receiver’s detection field (Fig. 2). The receiver residence event data are utilised by V-Track to provide information about periods when the tagged animals were between receiver detection fields. These events are termed non-residence events and are catalogued in an output file shown in Table 3. Because the RDM contains the minimum-route distance between all the receiver detection fields within the array, V-Track can calculate minimum rates of movement from the arrival and departure times from each receiver. For the tagged sharks, V-Track was used to calculate the average minimum rate of movement for each hour over the diel cycle (Fig. 3). A tagged animal may show numerous residence and nonresidence events, and if arranged chronologically, these illustrate the horizontal movements of the tagged animal within the confines of the receiver array. Case study 2 Determining patterns in vertical movement from sensor data The purpose of this study was to determine the frequency distribution for dive duration in a freshwater snake (Acrochordus arafurae). A snake was captured by baited trap from a small lake (150  400 m), implanted with a VEMCO continuous depthsensitive acoustic transmitters (V13TP 4 L, VEMCO), and released back into the lake (300  150 m). The tag emitted an acoustic signal containing water depth information every 4 s. The transmissions were detected on a VEMCO omni-directional hydrophone and recorded on a VR100 mobile receiver. The data from the VR100 were exported from the VEMCO VR100 software platform as a .csv file and imported into V-Track (Pratt et al. 2010). The BEQ function was utilised to qualify and catalogue changes in depth sensor data that were indicative of diving behaviour. The BEQ settings for triggering the recording of an event were as follows: Trend ¼ increasing, Trigger Threshold ¼ 0.4 m and Trigger Window ¼ 20 s. Thus, a change in

Table 2. The first few entries of the V-Track behavioural-event output table, showing periods when a tagged Carcharhinus leucas is within the limits of each receiver’s detection area (residence events) START TIME

END TIME

2/09/2007 00 : 30 : 00 2/09/2007 01 : 14 : 06 2/09/2007 01 : 39 : 10

2/09/2007 00 : 48 : 23 2/09/2007 01 : 27 : 18 2/09/2007 01 : 45 : 10

EVENT INDEX

TRANSMITTER ID

RECEIVER ID

DURATION (S)

END REASON

NO. DETECTIONS

1

34

101158

1103

Timeout

65

2

34

101157

792

Receiver

48

3

34

101144

360

Receiver

19

818

Marine and Freshwater Research

H. A. Campbell et al.

e (h) Residency tim

300

200

100

0 29.00

60.00

Ri tan ver ce (km

dis

1

Sh

2

91.00

ar k

4 3

)

Fig. 2. The total duration of time that each tagged Carcharhinus leucas (n ¼ 4) occurred within the detection field of each VR2W receiver over a 6-month monitoring period. Table 3. The first few entries of the V-Track behavioural-event output table, showing periods when a tagged Carcharhinus leucas is between the detection areas of adjacent receivers (non-residence event) START TIME

END TIME

2/09/2007 00 : 48 : 23 2/09/2007 01 : 27 : 18

2/09/2007 01 : 14 : 06 2/09/2007 01 : 39 : 10

EVENT INDEX

TRANSMITTER ID

RECEIVER ID1

RECEIVER ID2

DURATION (S)

DISTANCE (M)

ROM M/S

1

34

101 158

101 157

1543

2654

0.581

2

34

101 157

101 144

712

1345

0.529

0.6

0

1.0 Depth (m)

0.5 0.4

Proportion

ROM (m s⫺1)

0.8

0.3 0.2 0.1

2

4

0.6 6 04:18

0.4

04:21

04:24

Time

0.2

0 0100

0500

0900

1300

1700

2100

Time

0 0–5

5–10

10 –15

⬎15

Time (mins) Fig. 3. The rate of movement (mean
s.e.) exhibited by a Carcharhinus leucas (n ¼ 4) within each hour over the diel cycle.

subsequent depth measurements qualified as an event if the change occurred in an increasing direction and the difference was $0.4 m and occurred within a 20-s period. The BEQ settings for terminating the event were as follows: Termination Threshold ¼ 0.1 m and Termination Window ¼ 240 s. Thus, the recording of each event was terminated when

Fig. 4. The frequency histogram showing the distribution of event duration for 220 vertical movement events. Inset graph shows the depth recordings and acoustic detections comprising a single vertical movement event derived in a freshwater snake, qualified from depth-sensor information.

the depth value returned to within 0.1 m of the trigger value or if a period of 240 s occurred between consecutive acoustic detections. Each event was thus composed of several depth values, with the time of the first and last value determined by the depth values and

V-Track

Marine and Freshwater Research

819

Table 4. The first few entries of the V-Track behavioural-event output table, showing a vertical movement event in Acrochordus arafurae, qualified by changes in depth recordings START TIME

END TIME

18/09/2008 12 : 57 18/09/2008 14 : 37

18/09/2008 13 : 55 18/09/2008 14 : 47

SENSOR EVENT

TRANSMITTER ID

RECEIVER ID

DURATION (S)

START SENSOR (M)

19

139

103 551

3499

0.2

0.2

4.62

Return

6

20

139

103 555

600

0.2

0.2

5.11

Time out

8

the interval between consecutive depth detections (Fig. 4). V-Track catalogued all behavioural events within the detection archive that were qualified using these parameter values. Each was assigned temporal and spatial information as well as mean, maximum and minimum sensor values (Table 4). V-Track also designated an event with a true or false status. A true event was one that terminated because of a return in the sensor value, whereas a false event was one that terminated because a duration greater than that designated in the termination window was exceeded. False events may occur as a result of the tagged animal moving outside a receiver detection range during an event. False-event indexing is used so that events can be checked before further analyses and to optimise the eventtimeout duration. The events can then be categorised by temporal conditions or by sensor-value information (Fig. 4). The V-Track BEQ uses the same methodology for defining events in both depth and temperature data. Temperature is often used as a proxy for vertical movement in aquatic animals, but it could also be used to examine horizontal movement or periods when the animal is out of the water (Campbell et al. 2012). Discussion Essentially, V-Track enables the user to rapidly visualise spatial and temporal patterns within their acoustic-detection data. It can be used to identify and catalogue a broad range of behaviours in any animal that can be tagged with an acoustic transmitter (Campbell et al. 2010; Pratt et al. 2010; Campbell et al. 2012). In the present paper, we have demonstrated only a few of the features available within V-Track, and further examples of the information that can be synthesised from the acoustic-detection data are supplied in S2, available as Supplementary Material on the web. We also provide worked example files, additional help files and a user manual (S3 & S4). The sharing of acoustic-detection data has the potential to greatly increase the comprehensiveness and effectiveness of individual research studies. This has been recognised by the acoustic-tracking community and collaborations have been created at the national (Australian Animal Tracking and Monitoring System, Pacific Ocean Shelf Tracking) and international scale (Ocean Tracking Network; O’Dor and Stokesbury 2009). These collaborative ventures primarily take the form of acoustic-detection repositories which may contain millions of acoustic detections from multiple users. Defining ecologically relevant behaviours from within these vast collections is an involved, multi-faceted and time-consuming process, and one

END MAX SENSOR SENSOR (M) (M)

END NUM REASON RECS

that can be expected to intensify as receiver deployments grow and transmitter capability and longevity escalates. V-Track is a first-step in the creation of a universal suite of R-based analysis tools for the acoustic-telemetry community. It was coded in the R programming language and is composed in a modular format so that additional modules and routines can be added as new requirements and procedures are identified and created. Such a centralised source of freely available analysis tools will not only assist individual research groups but also facilitate collaboration through data-sharing. Acknowledgements This work was supported by the Australian Research Council linkage scheme with Australia Zoo as industry partners. V-Track and its supporting graphical user interface are available at http://www.uq.edu.au/eco-lab/v-track.

References Campbell, H. A., Watts, M. E., Sullivan, S., Read, M. A., Choukroun, S., Irwin, S. R., and Franklin, C. E. (2010). Estuarine crocodiles ride surface currents to facilitate long-distance travel. Journal of Animal Ecology 79, 955–964. doi:10.1111/J.1365-2656.2010.01709.X Campbell, H. A., Hewitt, M., Watts, M. E., Peverell, S., and Franklin, C. E. (2012). Short- and long-term movement patterns in the freshwater whipray (Himantura dalyensis) determined by the signal processing of passive acoustic telemetry data. Marine and Freshwater Research 63, 341–350. doi:10.1071/MF11229 Childs, A. R., Cowley, P. D., Naesje, T. F., Booth, A. J., Potts, W. M., Thorstad, E. B., and Okland, F. (2008). Do environmental factors influence the movement of estuarine fish? A case study using acoustic telemetry. Estuarine, Coastal and Shelf Science 78, 227–236. doi:10.1016/J.ECSS.2007.12.003 Clements, S., Jepsen, D., Karnowski, M., and Schreck, C. B. (2005). Optimization of an acoustic telemetry array for detecting transmitterimplanted fish. North American Journal of Fisheries Management 25, 429–436. doi:10.1577/M03-224.1 Grothues, T. M. (2009) A review of acoustic telemetry technology and a perspective on its diversification relative to coastal tracking arrays. In ‘Tagging and Tracking of Marine Animals with Electronic Devices. Vol. 9’. (Eds J. L. Nielsen, H. Arrizabalaga, N. Fragoso, A. Hobday, M. Lutcavage and J. Sibert.) pp. 77–90. (Springer Media B.V.: Houten, The Netherlands.) Heupel, M. R., Semmens, J. M., and Hobday, A. J. (2006). Automated acoustic tracking of aquatic animals: scales, design and deployment of listening station arrays. Marine and Freshwater Research 57, 1–13. doi:10.1071/MF05091 Humston, R., Ault, J. S., Larkin, M. F., and Luo, J. G. (2005). Movements and site fidelity of the bonefish Albula vulpes in the northern Florida

820

Marine and Freshwater Research

H. A. Campbell et al.

Keys determined by acoustic telemetry. Marine Ecology Progress Series 291, 237–248. doi:10.3354/MEPS291237 March, D., Palmer, M., Alo´s, J., Grau, A., and Cardona, F. (2010). Shortterm residence, home range size and diel patterns of the painted comber Serranus scriba in a temperate marine reserve. Marine Ecology Progress Series 400, 195–206. doi:10.3354/MEPS08410 Melnychuk, M. C., Welch, D. W., and Walters, C. J. (2010). Spatio-temporal migration patterns of pacific salmon smolts in rivers and coastal marine waters. PLoS ONE 5(9), e12916. doi:10.1371/JOURNAL.PONE. 0012916 Mitamura, H., Arai, N., Mitsunaga, Y., Yokota, T., Takeuchi, H., Tsuzaki, T., and Itani, M. (2005). Directed movements and diel burrow fidelity patterns of red tilefish Branchiostegus japonicus; determined using ultrasonic telemetry. Fisheries Science 71, 491–498. doi:10.1111/ J.1444-2906.2005.00992.X

O’Dor, R. K., and Stokesbury, M. J. W. (2009) The ocean tracking network – adding marine animal movements to the global ocean observing system. In ‘Tagging and Tracking of Marine Animals with Electronic Devices. Vol. 9 (Eds J. L. Nielsen, H. Arrizabalaga, N. Fragoso, A. Hobday, M. Lutcavage and J. Sibert.) pp. 91–100. (Springer Media B.V.: Houten, The Netherlands.) Pratt, K. L., Campbell, H. A., Watts, M. E., and Franklin, C. E. (2010). Environmental and ecological factors influencing dive behaviour in the freshwater snake Acrochordus arafurae: a field-based telemetric study. Marine and Freshwater Research 61, 560–567. doi:10.1071/ MF09194 R Development Core Team (2011). ‘R: a Language and Environment for Statistical Computing.’ (R Foundation for Statistical Computing: Vienna.)

www.publish.csiro.au/journals/mfr