- Email: [email protected]
Low-Cost Livestock Global Positioning System Collar from Commercial Off-the-Shelf Parts
Rangeland Ecology & Management 72 (2019) 954e958
Contents lists available at ScienceDirect
Rangeland Ecology & Management journal homepage: http://www.elsevier.com/locate/rama
Low-Cost Livestock Global Positioning System Collar from Commercial Off-the-Shelf Parts* Jason W. Karl a,*, James E. Sprinkle b a b
Department of Forest, Rangeland, and Fire Sciences, College of Natural Resources, University of Idaho, Moscow, Idaho, USA Department of Animal and Veterinary Science, College of Agricultural and Life Sciences, University of Idaho, Moscow, Idaho, USA
a r t i c l e i n f o
a b s t r a c t
Article history: Received 21 February 2019 Received in revised form 31 July 2019 Accepted 5 August 2019
Global Positioning System (GPS) tracking devices are a fundamental technology for quantifying the distribution and movement of livestock across landscapes. Although costs of GPS devices have decreased, it is still cost prohibitive to implement a large number of collars per study. Our objective was to develop and test a low-cost GPS collar using commercial off-the-shelf (COTS) electronic components to study livestock distribution and movement. Our COTS GPS tracker was built using the popular Arduino opensource microcontroller and a low-power timer board to cycle a GPS at defined intervals. Location data were saved to a data card in an open format for easy analysis. Total cost per COTS GPS device (including housing and collar) was $54.78. Average displacement from a known location and 95% circular error probability was 4.58 m, commensurate with other GPS collars. We tested durability and field performance of 25 COTS GPS collars against 24 existing GPS collars recording data at 5-min intervals in a southwest Idaho, United States study area. Our COTS GPS design and test showed that it is possible to manufacture low-cost location tracking devices, but the limitations of such devices must be considered relative to study objectives and duration. Low-cost location trackers will encourage collection of a higher density of location information to better understand patterns of livestock use in rangeland landscapes. © 2019 The Society for Range Management. Published by Elsevier Inc. All rights reserved.
Key Words: animal tracking Arduino Global Positioning System GPS location livestock distribution
Introduction GPS tracking devices are a fundamental technology for quantifying the distribution and movement of livestock at landscape scales (Turner et al., 2000; Ganskopp and Bohnert, 2009; Augustine and Derner, 2013; Bailey et al., 2018). Although costs of commercial GPS collars have declined (McGranahan et al., 2018), the number of location-tracking devices implemented on livestock in rangeland studies is still often determined by per-device cost. Prior research has demonstrated the feasibility of lower-cost GPS tracking devices for wildlife and livestock using custom electronics (e.g., Clark et al., 2006; ~$500/collar), modifying existing GPS loggers (e.g., Knight et al., 2018; ~$180/device) or commercially available off-the-shelf (McGranahan et al., 2018; ~$125/device) electronics components. Although general-use patterns of livestock herds or behavior of animals may be adequately tracked with a small number of devices if the animals are tightly grouped, factors affecting livestock
* This study was funded by a grant from the David Little Livestock Range Management Endowment. * Correspondence: Jason W. Karl, PhD, Dept of Forest, Rangeland, and Fire Sciences, University of Idaho, 875 Perimeter Dr, MS 1135, Moscow, ID 83844-1135, USA. Tel.: þ 1 208 885 0255. E-mail address: [email protected] (J.W. Karl).
distribution and herd dispersion are varied (Bailey et al., 1996). Thus, tracking movement and distribution of herds in variable conditions may require a larger number of devices. This argues for development of even lower-cost GPS collars than are presently available. The rise of the “Maker Movement” (Bajarin, 2014) has increased the availability and diversity of low-cost electronics components like programmable microcontrollers and GPS receivers and resources for how to use and combine them. This offers the opportunity to develop GPS tracking devices at much lower costs than currently available. Our objective was to develop a low-cost GPS collar using commercial off-the-shelf (COTS) electronic components to study livestock distribution and movement and evaluate its application in a pasture grazing study. Design goals for our GPS collars were that it should 1) provide GPS locations at user-defined intervals, 2) record data in a simple text file format on easy-to-obtain data cards, 3) be weather resistant and durable, 4) operate for at least 20 d on a battery, and 5) cost < $50 per collar to manufacture excluding labor.
Materials and Methods The GPS location tracker described as follows was built on the popular Arduino open-source microcontroller platform
https://doi.org/10.1016/j.rama.2019.08.003 1550-7424/© 2019 The Society for Range Management. Published by Elsevier Inc. All rights reserved.
J.W. Karl, J.E. Sprinkle / Rangeland Ecology & Management 72 (2019) 954e958
955
Figure 1. Schematic and prototype of a livestock Global Positioning System tracking collar built from low-cost commercial-off-the-shelf parts.
(https://www.arduino.cc). Detailed parts lists and assembly instructions for the COTS GPS devices, Arduino scripts, description of the GPS data format, and collar construction are available at https:// github.com/jkarl/COTS_GPS_Collars.
GPS Location Tracker Construction The core of our COTS GPS collar is an inexpensive and readily available Arduino Pro Mini clone (Fig. 1, Table 1), which is
Table 1 List of parts, sources, and costs for construction of GPS tracking collars using commercial, off-the-shelf electronics components. Part GPS unit Arduino Pro Mini 16MHz 5v 6-pin female header Sparkfun Shifting MicroSD Card Reader SanDisk 8GB MicroSD Card uBlox M8N GPS Adafruit Low Power Timer Board Pololu 5v step-up voltage regulator 43k-Ohm resistor 100mF capacitor IMREN 18650 LiPo battery (3500 mAh) 18650 battery holder Battery connector clip 22g hookup wire 50-mm heat-shrink tubing Misc supplies (solder, hot glue, assembly backing) GPS DEVICE TOTAL Collar 1.500 Nylon webbing belt with aluminum buckle 2.500 hose clamps 1.500 PVC 1.500 PVC Cap 200 PVC removable cap (with hose clamp) COLLAR COST TOTAL COST
Quantity
Per unit cost1
Part number
Source2
1 1 1 1 1 1 1 1 2 1 1 1 30cm 8cm
$3.20 $0.75 $4.46 $5.58 $13.04 $4.62 $3.98 $0.04 $0.42 $6.68 $0.35 $0.74 $0.17 $0.22 $0.50 $45.13
XCSOURCE TE362 S5481-ND DEV-13743 SDSDQAB-008G RM8862 TPL5510 U1V11F5 43KQBK-ND 493-1548-ND N/A Sackorange 4330196205 CAB-14574 CAB-10647 Uxcell a12080700ux0466 N/A
Amazon DigiKey Electronics SparkFun Electronics Amazon RCMoment Adafruit Pololu Robotics & Electronics DigiKey Electronics DigiKey Electronics IMR Batteries Amazon SparkFun Electronics SparkFun Electronics Amazon Local hardware store
1 1 10cm 1 1
$5.99 $0.93 $0.20 $0.58 $1.95 $9.65 $54.78
Black Dog Distributing B073YYCHBR 2208032 N/A 447-015 QC-101
Amazon SupplyHouse Local hardware store SupplyHouse SupplyHouse
1 Per-unit cost includes the purchase price plus shipping. Purchase prices are variable and should be viewed as approximate values. In most cases, per-unit costs decrease as purchase quantity increases. Significant cost savings may be possible by sourcing parts from suppliers who offer free shipping if longer delivery times are acceptable. 2 URLs for part suppliers are: Adafruit (https://www.adafruit.com/), Amazon (https://amazon.com), DigiKey Electronics (https://www.digikey.com/), IMR Batteries (https:// www.imrbatteries.com/), Pololu Robotics & Electronics (http://www.pololu.com), RCMoment (https://www.rcmoment.com), SparkFun Electronics (https://www.sparkfun. com/), SupplyHouse (https://www.supplyhouse.com/)
956
J.W. Karl, J.E. Sprinkle / Rangeland Ecology & Management 72 (2019) 954e958
compatible with the Arduino Integrated Development Environment (IDE) for programming the microcontroller. The Arduino initiates and polls the GPS unit and writes the location information to a data card. To maximize battery life and control timing of GPS readings, we used an AdaFruit low-power timer board (after McGranahan et al., 2018). The timer board passes current from the voltage regulator to the Arduino until the Arduino signals the time board that it has completed recording GPS information. At that point, the timer board enters a sleep cycle and cuts off power to the Arduino. The duration of the sleep cycle is defined by the resistance across the delay pin and ground. For the COTS GPS collars, we chose a resistance of 43K-Ohms, which equated to a 5-min sleep cycle (Adafruit, 2017). The COTS GPS collars were powered by a single-cell 3.7v lithium polymer battery. A step-up voltage regulator converted the voltage from the battery to a consistent 5v for the Arduino, GPS, and data card reader. Initial tests on prototype units found that instantaneous power consumption during operation was 57 mA. Given an estimated efficiency of 70% to account for issues related to operating temperature and uncertainty in calculating power consumption, we estimated total daily power consumption for 5-min sampling intervals of z195 mA. Using a 3 500-mAh battery, we expected a lifespan of z18 d. Although less than our goal of a 20-d lifespan, shorter duration was accepted in order to minimize perunit costs. Longer durations could easily be achieved with highercapacity batteries. Connections between individual components were made with 22-gauge stranded wire soldered to provided terminals. Components were affixed to a 15 mm 75 mm 3 mm nonconductive backing (plastic or wood) to provide support for the soldered connections, and the entire unit was enclosed with 55-mm-diameter heat-shrink tubing to secure it. The Arduino microcontroller requires a custom script to activate and poll the GPS, write resulting data to the SD card, and signal the low-power timer to start the next cycle. The script was written and uploaded using the Arduino IDE, which is based on C/Cþþ (https:// www.arduino.cc/reference/en/), using the SD, SPI, SoftwareSerial, and TinyGPSþþ libraries (see Appendix A for library source information). Source files for the script loaded onto each COTS GPS collar are available at: https://github.com/jkarl/COTS_GPS_Collars/tree/ master/firmware/COTS_GPS_Logger.ino. Upon receiving power from the low-power timer board, the COTS GPS script initiates the GPS unit and polls it every second for its status (i.e., valid coordinate fix), coordinate values, and horizontal dilution of precision (HDOP, a metric of GPS signal quality). If no valid GPS-fix was obtained within 60 s of powering up, the device records a “no data” line to the data card and signals the lowpower board to initiate its next sleep cycle. If a valid GPS fix is obtained, the device polls the GPS unit until the HDOP drops below 2.5 or until 60 s from device startup have elapsed. GPS locations are recorded in a simple comma-delimited text format and include latitude and longitude (in decimal degrees, WGS84 datum), date and time of location, positional dilution of precision (PDOP), and elapsed time from device startup to GPS fix. A complete data format description is available on the GitHub repository. Collar Construction The collar assembly for the COTS GPS units was made from 38 mm (1.500 ) nylon webbing belts (1.37 m [54”] long) with aluminum “ladderloc”-style buckles and 102 mm (4”) of 38-mm (1.5”)diameter polyvinyl chloride (PVC) tubing (see Table 1, Fig. 1). A fixed PVC cap was glued to one end of the tubing, and a removable neoprene rubber cap was fit over the other end. The PVC housing was secured to the nylon belt using 63.5-mm (2.5”) hose clamps
(see Fig. 1). Total mass of the COTS GPS collar was 442 g (26 g for the GPS device, 55 g for the battery, and 361 g for the collar assembly). GPS Collar Testing Twenty-five COTS GPS collars were assembled and each run for 24 h to verify functioning. To evaluate accuracy and precision, we placed a COTS GPS collar at a location that was identified with realtime kinematic (RTK) GPS (± 3 cm) and measured the displacement between the COTS GPS collar locations and the known point. We calculated the mean displacement of the COTS GPS collar coordinates from the known point as a measure of accuracy and 95% circular error probability (CEP, Clark et al., 2006; McGranahan et al., 2018) as our measure of precision. Both Clark et al. (2006) and McGranahan et al. (2018) reported 95% CEP of GPS fixes taken at 1-s intervals. However, because precision of GPS measurements varies with fix interval length (Johnson and Ganskopp, 2008), we first recorded GPS readings at 1-s intervals and then performed a 24-h accuracy test at 5-min fix intervals. We calculated mean displacement and 95% CEP for both intervals. To test the durability and field performance of the GPS collars, we deployed the 25 COTS GPS collars along with 24 existing GPS collars using an iGotU GPS tracker (Knight et al., 2018) in a study area southwest of Malta, Idaho, United States (Fig. 2). Both collar types were set to record at 5-min intervals. From a herd of 100 heifers with calves, 49 heifers were selected at random and fitted with either a COTS or iGotU GPS collar. Cattle were released into a 65-ha (141-acre) holding pasture (see Fig. 2) for 4.12 d before being turned out to a larger pasture for 6 wk. Location data from the holding pasture and overall device duration results are presented here. Results Total cost including per COTS GPS collar was $54.78 (see Table 1). This included $3.45 per collar in parts shipping charges (6.3% of total cost). Cost savings could have been realized by sourcing components from different suppliers or prioritizing free shipping. However, due to project timing constraints, quick availability was prioritized over shipping cost. Actual costs will fluctuate over time with supplier costs and parts availability. At 1-s intervals, average displacement from the RTK test location for the COTS collars was 2.21 m. The 95% CEP at 1-s intervals for the COTS collars was 4.58 m, similar to that reported by McGranahan et al. (2018). At 5-min intervals, average displacement was 10.28 m and 95% CEP was 26.3 m. The iGotU collars, tested only at 5-min intervals, yielded average displacement of 9.27 m and 95% CEP of 24.5 m. Success of the COTS GPS collars was mixed. During the holding pasture test, 1 of the COTS collars experienced failure due to a faulty GPS device and recorded no data (i.e., device functioned, but recorded only zeros for latitude/longitude). All remaining collars lasted through the holding pasture trial, and the Arduino script and hardware components of each device functioned as intended. Average lifespan of the COTS GPS devices was 16 d (range 10 23 d). Several factors contributed to this shortened lifespan. In three devices, inconsistent grounding caused by faulty soldering or wiring caused inconsistent performance. In five devices, packaging of the GPS device in the collar led to an exposed reset button becoming depressed when the collar was jostled (e.g., when the cows walked). This had the effect of interrupting GPS fixes, causing loss of data and shortening the interval between GPS readings, causing the battery to deplete faster. In most cases, however, battery depletion occurred sooner than anticipated. The COTS and iGotU collars recorded a total of 17 315 and 23 275 location points in the holding pasture, respectively. Distribution
J.W. Karl, J.E. Sprinkle / Rangeland Ecology & Management 72 (2019) 954e958
957
Figure 2. Distribution of cattle Global Positioning System (GPS) locations shown as location density from iGotU GPS collars (left, n ¼ 22) and low-cost COTS GPS collars (right, n ¼ 25) in a 65-ha holding pasture in southern Idaho, United States. inset maps show example locations from one collar of each type.
patterns were similar for both collars (see Fig. 2). Because the COTS and iGotU collars recorded missing data values differently, the rate of failed GPS fixes was calculated from the theoretical maximum number of GPS readings at 5-min intervals while within the holding pasture. Excluding the collars with grounding or reset button issues, the COTS collars experienced a 19.9% average rate of missed GPS fixes compared with the iGotU collars, which experienced a 10.9% average rate of missed GPS fixes. Rates of spurious points (i.e., locations falling outside the pasture) were low for both collars, 0.8% and 0.1% for the COTS and iGotU collars, respectively. None of the COTS GPS collars were lost during the study, but two iGotU collars were lost and not recovered. After removal from the cows, the COTS collars were completely submerged in water for up to 5 min for cleaning before opening them to remove the GPS device. No COTS collar showed any sign of water penetration into the housing. In addition, no damage was observed to the COTS collars or housing. At the conclusion of the study, no evidence of rubbing or chafing from the COTS or iGotU collars on the cows was observed. No behavioral changes were noted on the cows with GPS collars. Discussion The results presented here demonstrate that GPS devices built from COTS parts are a potential option for tracking livestock locations in grazing studies, though the limitations of such devices must be weighed against study objectives. Three main limitations of the COTS GPS collars were identified: 1) reliability of the devices due to design and construction issues, 2) battery life and power consumption, and 3) GPS performance relative to other available GPS devices. Reliability of the COTS GPS devices was affected by design issues and the quality of soldered connections, which are easily addressed. One significant point of weakness for our COTS GPS collars was soldering together the various power leads from each device. A
small, custom circuit board could be used for this purpose and would reduce the potential for weak solders joining wires. Custom circuit boards can now readily be designed and ordered from online vendors (e.g., https://www.seeedstudio.com/). In addition, the configuration of the components should be changed to protect (or disable) the reset button from being accidentally pressed. Aside from design and construction issues, longevity of the COTS GPS devices was the largest limitation. Arduino microcontrollers were not originally designed for low-power applications, and while options exist for decreasing power consumption (see later), it is unlikely that an Ardunio-based device could achieve the longevity of a commercial device that was designed with low power consumption in mind. However, there are two easy opportunities for improving longevity of the COTS GPS devices. Although many studies have pointed to the advantages of shorter sampling intervals for describing animal distribution and movements (Johnson and Ganskopp, 2008; Rowcliffe et al., 2012; McGavin et al., 2018), increasing the GPS logging interval would extend battery life. Likewise, a higher-capacity battery could be used for longer implementations. Accuracy of the COTS GPS devices was commensurate with the iGotU devices and with results reported by McGranahan et al. (2018). However, the COTS GPS devices experienced a higher rate of missed GPS readings than the iGotU devices. This could be due to inherent differences in the GPS antenna used by each device or to the collar and housing for each unit. The metal hose clamps of the COTS collars may have impeded GPS signals (see Yi et al., 2012), and the weight of the housing caused it to typically hang below the cow’s neck. Conversely, the iGotU devices were housed in a nylon pouch stitched to the collar (after Knight et al., 2018) that would not impede GPS signal. The lighter weight of the nylon housing allowed the iGotU devices to slide on the cow’s neck, and they were often on the side or top of the neck. Options for nonmetallic housings or counterweighting the collar so that the device is always on top could be explored.
958
J.W. Karl, J.E. Sprinkle / Rangeland Ecology & Management 72 (2019) 954e958
The primary advantages of the COTS GPS devices, however, are their low cost, opportunities for customization, and simplicity. Our final per-unit cost exceeded our initial goal of $50 per collar. This was largely due to shipping costs associated with the parts. Because of the tight timeline of our project, however, we needed to obtain parts quickly and thus paid for shipping when we could have opted for free shipping with a longer delivery time. There are other opportunities to reduce costs further, including building custom circuit boards that minimize unnecessary parts (which would also reduce power consumption) and configuring the device to run on 3.3V, thereby eliminating the voltage converter (see https://github.com/jkarl/COTS_ GPS_Collars for examples). Low-power sleep-cycling software libraries are available for Arduino and could be implemented in place of the timer board to save cost. With default Arduino settings, software sleep cycling gains only modest power savings. Customizing the fuse settings on Arduino’s microcontroller allows for more significant lowpower sleep, but this step involves low-level reprogramming of the Arduino. Thus, the trade-offs for these steps are increased complexity in development and building of the devices. Unlike commercial collars, or collars based on commercial GPS trackers that are limited to what functionality the device has at time of purchase, the requirement to write custom scripts for the COTS GPS collars presents unique opportunities for improving data quality and adding functionality. For example, more sophisticated GPS signal processing could be implemented to improve data accuracy and reliability. In a test of simple GPS point averaging, average displacement and 95% CEP for 5-min interval logging that used the average of 20 consecutive GPS readings decreased to 2.68 m and 7.76 m, respectivelydmuch closer to the 1-s-interval CEP values than the standard 5-min values. In addition, with GPS point averaging, the incidence of failed GPS readings also decreased. Finally, the simplicity of the COTS GPS collars can be both an advantage and a limitation. The COTS GPS devices can be built by anyone with modest technical knowledge and soldering skills, and the basic, open data format is easy to process with GIS and statistical software. However, the fact that the COTS GPS devices perform only one functiondrecord locations at a predefined intervaldcan be a drawback. The loss of two of the iGotU collars (which also are solely GPS recorders) during the study, presumably due to failure of the collar’s strap or buckle, highlights that data are contingent on recovering the device. An independent VHF transmitter could be added to aid in relocating lost devices (but would add to device cost). Management Implications The design and test presented here demonstrate that it is possible to manufacture low-cost location tracking devices that perform similarly to more expensive devices. In addition, the COTS collar approach has several advantages over other commercial devices including the ability to 1) build in only necessary features to save power and reduce complexity, 2) easily add other sensors, and 3) develop custom scripts to allow for advanced data processing like point averaging to overcome limitations of low-cost sensors. Given the limitations discussed earlier, though, the COTS GPS collars would be best suited for studies where a large number of collars were needed for a short duration (weeks with existing batteries or several months with larger-capacity batteries). Despite their simplistic design and relatively limited functionality compared with many commercial collars (e.g., no wireless data download), the COTS GPS collars could encourage location tracking in many more livestock rangeland studies. Although battery life for a COTS GPS collar may be shorter than for a similar commercial device, a primary advantage of the COTS GPS collar described here is the ability to easily and inexpensively develop custom GPS tracking applications. The lower cost of the COTS GPS collars will also facilitate collection of a much higher density of location
information (i.e., tracking more animals in a herd) to better understand patterns of livestock use in rangeland landscapes and how livestock interact with wildlife populations and rangeland uses. Acknowledgments We are grateful to R. Ward and C. Ward for making their cattle herd available for this study and for assistance in installing and removing the collars. This study was approved by the University of Idaho Institutional Animal Care and Use Committee under study IACUC-2018-25. Appendix A. Source information for Arduino libraries used in Global Positioning System collar script
Library name and version
Description
Source
SPI, 1.0.0
Enables communication with devices that use the serial peripheral interface (SPI) bus. Enables reading and writing to secure digital cards through the SPI interface.
Arduino: https://www. arduino.cc/en/Reference/SPI, Accessed 5 November, 2018 Arduino, Sparkfun: https:// www.arduino.cc/en/ Reference/SD, Accessed 5 November, 2018 Arduino: https://www. arduino.cc/en/Reference/ SoftwareSerial, Accessed 5 November, 2018
SD, 1.1.1
SoftwareSerial, Enables serial communication 1.0.0 on any digital Arduino pin. SoftwareSerial is used for communication with the GPS device. TinyGPS þþ, Arduino library for parsing 0.95a NMEA data streams from GPS devices.
Mikal Hart: http://arduiniana. org/libraries/tinygpsplus/, Accessed 5 November, 2018
References Adafruit, 2017. Adafruit TPL5110 Power Timer breakout. Available at: https://learn. adafruit.com/adafruit-tpl5110-power-timer-breakout/. Accessed 5 November 2018. Augustine, D., Derner, J., 2013. Assessing herbivore foraging behavior with GPS collars in a semiarid grassland. Sensors 13, 3711e3723. Bailey, D.W., Gross, J.E., Laca, E.A., Rittenhouse, L.R., Coughenour, M.B., Swift, D.M., Sims, P.L., 1996. Mechanisms that result in large herbivore grazing distribution patterns. Journal of Range Management 49, 386e400. Bailey, D.W., Trotter, M.G., Knight, C.W., Thomas, M.G., 2018. Use of GPS tracking collars and accelerometers for rangeland livestock production research. In: Translational Animal Science. Available at: http://academic.oup.com/tas/ advance-article/doi/10.1093/tas/txx006/4824982. Accessed 21 February 2018. Bajarin, T. 2014. Why the maker movement is important to America’s future. Time. Available at: http://time.com/104210/maker-faire-maker-movement/. Accessed 23 July 2018. Clark, P.E., Johnson, D.E., Kniep, M.A., Jermann, P., Huttash, B., Wood, A., Johnson, M., McGillivan, C., Titus, K., 2006. An advanced, low-cost, gps-based animal tracking system. Rangeland Ecology & Management 59, 334e340. Ganskopp, D.C., Bohnert, D.W., 2009. Landscape nutritional patterns and cattle distribution in rangeland pastures. Applied Animal Behaviour Science 116, 110e119. Johnson, D.D., Ganskopp, D., 2008. GPS collar sampling frequency: effects on measures of resource use. Rangeland Ecology & Management 61, 226e231. Knight, C.W., Bailey, D.W., Faulkner, D., 2018. Low-cost global positioning system tracking collars for use on cattle. Rangeland Ecology & Management 71, 506e508. McGavin, S.L., Bishop-Hurley, G.J., Charmley, E., Greenwood, P.L., Callaghan, M.J., 2018. Effect of GPS sample interval and paddock size on estimates of distance travelled by grazing cattle in rangeland, Australia. The Rangeland Journal 40, 55. McGranahan, D.A., Geaumont, B., Spiess, J.W., 2018. Assessment of a livestock GPS collar based on an open-source datalogger informs best practices for logging intensity. Ecology and Evolution 8, 5649e5660. Rowcliffe, M.J., Carbone, C., Kays, R., Kranstauber, B., Jansen, P.A., 2012. Bias in estimating animal travel distance: the effect of sampling frequency: estimating animal travel distance. Methods in Ecology and Evolution 3, 653e662. Turner, L.W., Udal, M.C., Larson, B.T., Shearer, S.A., 2000. Monitoring cattle behavior and pasture use with GPS and GIS. Canadian Journal of Animal Science 80, 405e413. Yi, T.-H., Li, H.-N., Gu, M., 2012. Effect of different construction materials on propagation of GPS monitoring signals. Measurement 45, 1126e1139.
Contents lists available at ScienceDirect
Rangeland Ecology & Management journal homepage: http://www.elsevier.com/locate/rama
Low-Cost Livestock Global Positioning System Collar from Commercial Off-the-Shelf Parts* Jason W. Karl a,*, James E. Sprinkle b a b
Department of Forest, Rangeland, and Fire Sciences, College of Natural Resources, University of Idaho, Moscow, Idaho, USA Department of Animal and Veterinary Science, College of Agricultural and Life Sciences, University of Idaho, Moscow, Idaho, USA
a r t i c l e i n f o
a b s t r a c t
Article history: Received 21 February 2019 Received in revised form 31 July 2019 Accepted 5 August 2019
Global Positioning System (GPS) tracking devices are a fundamental technology for quantifying the distribution and movement of livestock across landscapes. Although costs of GPS devices have decreased, it is still cost prohibitive to implement a large number of collars per study. Our objective was to develop and test a low-cost GPS collar using commercial off-the-shelf (COTS) electronic components to study livestock distribution and movement. Our COTS GPS tracker was built using the popular Arduino opensource microcontroller and a low-power timer board to cycle a GPS at defined intervals. Location data were saved to a data card in an open format for easy analysis. Total cost per COTS GPS device (including housing and collar) was $54.78. Average displacement from a known location and 95% circular error probability was 4.58 m, commensurate with other GPS collars. We tested durability and field performance of 25 COTS GPS collars against 24 existing GPS collars recording data at 5-min intervals in a southwest Idaho, United States study area. Our COTS GPS design and test showed that it is possible to manufacture low-cost location tracking devices, but the limitations of such devices must be considered relative to study objectives and duration. Low-cost location trackers will encourage collection of a higher density of location information to better understand patterns of livestock use in rangeland landscapes. © 2019 The Society for Range Management. Published by Elsevier Inc. All rights reserved.
Key Words: animal tracking Arduino Global Positioning System GPS location livestock distribution
Introduction GPS tracking devices are a fundamental technology for quantifying the distribution and movement of livestock at landscape scales (Turner et al., 2000; Ganskopp and Bohnert, 2009; Augustine and Derner, 2013; Bailey et al., 2018). Although costs of commercial GPS collars have declined (McGranahan et al., 2018), the number of location-tracking devices implemented on livestock in rangeland studies is still often determined by per-device cost. Prior research has demonstrated the feasibility of lower-cost GPS tracking devices for wildlife and livestock using custom electronics (e.g., Clark et al., 2006; ~$500/collar), modifying existing GPS loggers (e.g., Knight et al., 2018; ~$180/device) or commercially available off-the-shelf (McGranahan et al., 2018; ~$125/device) electronics components. Although general-use patterns of livestock herds or behavior of animals may be adequately tracked with a small number of devices if the animals are tightly grouped, factors affecting livestock
* This study was funded by a grant from the David Little Livestock Range Management Endowment. * Correspondence: Jason W. Karl, PhD, Dept of Forest, Rangeland, and Fire Sciences, University of Idaho, 875 Perimeter Dr, MS 1135, Moscow, ID 83844-1135, USA. Tel.: þ 1 208 885 0255. E-mail address: [email protected] (J.W. Karl).
distribution and herd dispersion are varied (Bailey et al., 1996). Thus, tracking movement and distribution of herds in variable conditions may require a larger number of devices. This argues for development of even lower-cost GPS collars than are presently available. The rise of the “Maker Movement” (Bajarin, 2014) has increased the availability and diversity of low-cost electronics components like programmable microcontrollers and GPS receivers and resources for how to use and combine them. This offers the opportunity to develop GPS tracking devices at much lower costs than currently available. Our objective was to develop a low-cost GPS collar using commercial off-the-shelf (COTS) electronic components to study livestock distribution and movement and evaluate its application in a pasture grazing study. Design goals for our GPS collars were that it should 1) provide GPS locations at user-defined intervals, 2) record data in a simple text file format on easy-to-obtain data cards, 3) be weather resistant and durable, 4) operate for at least 20 d on a battery, and 5) cost < $50 per collar to manufacture excluding labor.
Materials and Methods The GPS location tracker described as follows was built on the popular Arduino open-source microcontroller platform
https://doi.org/10.1016/j.rama.2019.08.003 1550-7424/© 2019 The Society for Range Management. Published by Elsevier Inc. All rights reserved.
J.W. Karl, J.E. Sprinkle / Rangeland Ecology & Management 72 (2019) 954e958
955
Figure 1. Schematic and prototype of a livestock Global Positioning System tracking collar built from low-cost commercial-off-the-shelf parts.
(https://www.arduino.cc). Detailed parts lists and assembly instructions for the COTS GPS devices, Arduino scripts, description of the GPS data format, and collar construction are available at https:// github.com/jkarl/COTS_GPS_Collars.
GPS Location Tracker Construction The core of our COTS GPS collar is an inexpensive and readily available Arduino Pro Mini clone (Fig. 1, Table 1), which is
Table 1 List of parts, sources, and costs for construction of GPS tracking collars using commercial, off-the-shelf electronics components. Part GPS unit Arduino Pro Mini 16MHz 5v 6-pin female header Sparkfun Shifting MicroSD Card Reader SanDisk 8GB MicroSD Card uBlox M8N GPS Adafruit Low Power Timer Board Pololu 5v step-up voltage regulator 43k-Ohm resistor 100mF capacitor IMREN 18650 LiPo battery (3500 mAh) 18650 battery holder Battery connector clip 22g hookup wire 50-mm heat-shrink tubing Misc supplies (solder, hot glue, assembly backing) GPS DEVICE TOTAL Collar 1.500 Nylon webbing belt with aluminum buckle 2.500 hose clamps 1.500 PVC 1.500 PVC Cap 200 PVC removable cap (with hose clamp) COLLAR COST TOTAL COST
Quantity
Per unit cost1
Part number
Source2
1 1 1 1 1 1 1 1 2 1 1 1 30cm 8cm
$3.20 $0.75 $4.46 $5.58 $13.04 $4.62 $3.98 $0.04 $0.42 $6.68 $0.35 $0.74 $0.17 $0.22 $0.50 $45.13
XCSOURCE TE362 S5481-ND DEV-13743 SDSDQAB-008G RM8862 TPL5510 U1V11F5 43KQBK-ND 493-1548-ND N/A Sackorange 4330196205 CAB-14574 CAB-10647 Uxcell a12080700ux0466 N/A
Amazon DigiKey Electronics SparkFun Electronics Amazon RCMoment Adafruit Pololu Robotics & Electronics DigiKey Electronics DigiKey Electronics IMR Batteries Amazon SparkFun Electronics SparkFun Electronics Amazon Local hardware store
1 1 10cm 1 1
$5.99 $0.93 $0.20 $0.58 $1.95 $9.65 $54.78
Black Dog Distributing B073YYCHBR 2208032 N/A 447-015 QC-101
Amazon SupplyHouse Local hardware store SupplyHouse SupplyHouse
1 Per-unit cost includes the purchase price plus shipping. Purchase prices are variable and should be viewed as approximate values. In most cases, per-unit costs decrease as purchase quantity increases. Significant cost savings may be possible by sourcing parts from suppliers who offer free shipping if longer delivery times are acceptable. 2 URLs for part suppliers are: Adafruit (https://www.adafruit.com/), Amazon (https://amazon.com), DigiKey Electronics (https://www.digikey.com/), IMR Batteries (https:// www.imrbatteries.com/), Pololu Robotics & Electronics (http://www.pololu.com), RCMoment (https://www.rcmoment.com), SparkFun Electronics (https://www.sparkfun. com/), SupplyHouse (https://www.supplyhouse.com/)
956
J.W. Karl, J.E. Sprinkle / Rangeland Ecology & Management 72 (2019) 954e958
compatible with the Arduino Integrated Development Environment (IDE) for programming the microcontroller. The Arduino initiates and polls the GPS unit and writes the location information to a data card. To maximize battery life and control timing of GPS readings, we used an AdaFruit low-power timer board (after McGranahan et al., 2018). The timer board passes current from the voltage regulator to the Arduino until the Arduino signals the time board that it has completed recording GPS information. At that point, the timer board enters a sleep cycle and cuts off power to the Arduino. The duration of the sleep cycle is defined by the resistance across the delay pin and ground. For the COTS GPS collars, we chose a resistance of 43K-Ohms, which equated to a 5-min sleep cycle (Adafruit, 2017). The COTS GPS collars were powered by a single-cell 3.7v lithium polymer battery. A step-up voltage regulator converted the voltage from the battery to a consistent 5v for the Arduino, GPS, and data card reader. Initial tests on prototype units found that instantaneous power consumption during operation was 57 mA. Given an estimated efficiency of 70% to account for issues related to operating temperature and uncertainty in calculating power consumption, we estimated total daily power consumption for 5-min sampling intervals of z195 mA. Using a 3 500-mAh battery, we expected a lifespan of z18 d. Although less than our goal of a 20-d lifespan, shorter duration was accepted in order to minimize perunit costs. Longer durations could easily be achieved with highercapacity batteries. Connections between individual components were made with 22-gauge stranded wire soldered to provided terminals. Components were affixed to a 15 mm 75 mm 3 mm nonconductive backing (plastic or wood) to provide support for the soldered connections, and the entire unit was enclosed with 55-mm-diameter heat-shrink tubing to secure it. The Arduino microcontroller requires a custom script to activate and poll the GPS, write resulting data to the SD card, and signal the low-power timer to start the next cycle. The script was written and uploaded using the Arduino IDE, which is based on C/Cþþ (https:// www.arduino.cc/reference/en/), using the SD, SPI, SoftwareSerial, and TinyGPSþþ libraries (see Appendix A for library source information). Source files for the script loaded onto each COTS GPS collar are available at: https://github.com/jkarl/COTS_GPS_Collars/tree/ master/firmware/COTS_GPS_Logger.ino. Upon receiving power from the low-power timer board, the COTS GPS script initiates the GPS unit and polls it every second for its status (i.e., valid coordinate fix), coordinate values, and horizontal dilution of precision (HDOP, a metric of GPS signal quality). If no valid GPS-fix was obtained within 60 s of powering up, the device records a “no data” line to the data card and signals the lowpower board to initiate its next sleep cycle. If a valid GPS fix is obtained, the device polls the GPS unit until the HDOP drops below 2.5 or until 60 s from device startup have elapsed. GPS locations are recorded in a simple comma-delimited text format and include latitude and longitude (in decimal degrees, WGS84 datum), date and time of location, positional dilution of precision (PDOP), and elapsed time from device startup to GPS fix. A complete data format description is available on the GitHub repository. Collar Construction The collar assembly for the COTS GPS units was made from 38 mm (1.500 ) nylon webbing belts (1.37 m [54”] long) with aluminum “ladderloc”-style buckles and 102 mm (4”) of 38-mm (1.5”)diameter polyvinyl chloride (PVC) tubing (see Table 1, Fig. 1). A fixed PVC cap was glued to one end of the tubing, and a removable neoprene rubber cap was fit over the other end. The PVC housing was secured to the nylon belt using 63.5-mm (2.5”) hose clamps
(see Fig. 1). Total mass of the COTS GPS collar was 442 g (26 g for the GPS device, 55 g for the battery, and 361 g for the collar assembly). GPS Collar Testing Twenty-five COTS GPS collars were assembled and each run for 24 h to verify functioning. To evaluate accuracy and precision, we placed a COTS GPS collar at a location that was identified with realtime kinematic (RTK) GPS (± 3 cm) and measured the displacement between the COTS GPS collar locations and the known point. We calculated the mean displacement of the COTS GPS collar coordinates from the known point as a measure of accuracy and 95% circular error probability (CEP, Clark et al., 2006; McGranahan et al., 2018) as our measure of precision. Both Clark et al. (2006) and McGranahan et al. (2018) reported 95% CEP of GPS fixes taken at 1-s intervals. However, because precision of GPS measurements varies with fix interval length (Johnson and Ganskopp, 2008), we first recorded GPS readings at 1-s intervals and then performed a 24-h accuracy test at 5-min fix intervals. We calculated mean displacement and 95% CEP for both intervals. To test the durability and field performance of the GPS collars, we deployed the 25 COTS GPS collars along with 24 existing GPS collars using an iGotU GPS tracker (Knight et al., 2018) in a study area southwest of Malta, Idaho, United States (Fig. 2). Both collar types were set to record at 5-min intervals. From a herd of 100 heifers with calves, 49 heifers were selected at random and fitted with either a COTS or iGotU GPS collar. Cattle were released into a 65-ha (141-acre) holding pasture (see Fig. 2) for 4.12 d before being turned out to a larger pasture for 6 wk. Location data from the holding pasture and overall device duration results are presented here. Results Total cost including per COTS GPS collar was $54.78 (see Table 1). This included $3.45 per collar in parts shipping charges (6.3% of total cost). Cost savings could have been realized by sourcing components from different suppliers or prioritizing free shipping. However, due to project timing constraints, quick availability was prioritized over shipping cost. Actual costs will fluctuate over time with supplier costs and parts availability. At 1-s intervals, average displacement from the RTK test location for the COTS collars was 2.21 m. The 95% CEP at 1-s intervals for the COTS collars was 4.58 m, similar to that reported by McGranahan et al. (2018). At 5-min intervals, average displacement was 10.28 m and 95% CEP was 26.3 m. The iGotU collars, tested only at 5-min intervals, yielded average displacement of 9.27 m and 95% CEP of 24.5 m. Success of the COTS GPS collars was mixed. During the holding pasture test, 1 of the COTS collars experienced failure due to a faulty GPS device and recorded no data (i.e., device functioned, but recorded only zeros for latitude/longitude). All remaining collars lasted through the holding pasture trial, and the Arduino script and hardware components of each device functioned as intended. Average lifespan of the COTS GPS devices was 16 d (range 10 23 d). Several factors contributed to this shortened lifespan. In three devices, inconsistent grounding caused by faulty soldering or wiring caused inconsistent performance. In five devices, packaging of the GPS device in the collar led to an exposed reset button becoming depressed when the collar was jostled (e.g., when the cows walked). This had the effect of interrupting GPS fixes, causing loss of data and shortening the interval between GPS readings, causing the battery to deplete faster. In most cases, however, battery depletion occurred sooner than anticipated. The COTS and iGotU collars recorded a total of 17 315 and 23 275 location points in the holding pasture, respectively. Distribution
J.W. Karl, J.E. Sprinkle / Rangeland Ecology & Management 72 (2019) 954e958
957
Figure 2. Distribution of cattle Global Positioning System (GPS) locations shown as location density from iGotU GPS collars (left, n ¼ 22) and low-cost COTS GPS collars (right, n ¼ 25) in a 65-ha holding pasture in southern Idaho, United States. inset maps show example locations from one collar of each type.
patterns were similar for both collars (see Fig. 2). Because the COTS and iGotU collars recorded missing data values differently, the rate of failed GPS fixes was calculated from the theoretical maximum number of GPS readings at 5-min intervals while within the holding pasture. Excluding the collars with grounding or reset button issues, the COTS collars experienced a 19.9% average rate of missed GPS fixes compared with the iGotU collars, which experienced a 10.9% average rate of missed GPS fixes. Rates of spurious points (i.e., locations falling outside the pasture) were low for both collars, 0.8% and 0.1% for the COTS and iGotU collars, respectively. None of the COTS GPS collars were lost during the study, but two iGotU collars were lost and not recovered. After removal from the cows, the COTS collars were completely submerged in water for up to 5 min for cleaning before opening them to remove the GPS device. No COTS collar showed any sign of water penetration into the housing. In addition, no damage was observed to the COTS collars or housing. At the conclusion of the study, no evidence of rubbing or chafing from the COTS or iGotU collars on the cows was observed. No behavioral changes were noted on the cows with GPS collars. Discussion The results presented here demonstrate that GPS devices built from COTS parts are a potential option for tracking livestock locations in grazing studies, though the limitations of such devices must be weighed against study objectives. Three main limitations of the COTS GPS collars were identified: 1) reliability of the devices due to design and construction issues, 2) battery life and power consumption, and 3) GPS performance relative to other available GPS devices. Reliability of the COTS GPS devices was affected by design issues and the quality of soldered connections, which are easily addressed. One significant point of weakness for our COTS GPS collars was soldering together the various power leads from each device. A
small, custom circuit board could be used for this purpose and would reduce the potential for weak solders joining wires. Custom circuit boards can now readily be designed and ordered from online vendors (e.g., https://www.seeedstudio.com/). In addition, the configuration of the components should be changed to protect (or disable) the reset button from being accidentally pressed. Aside from design and construction issues, longevity of the COTS GPS devices was the largest limitation. Arduino microcontrollers were not originally designed for low-power applications, and while options exist for decreasing power consumption (see later), it is unlikely that an Ardunio-based device could achieve the longevity of a commercial device that was designed with low power consumption in mind. However, there are two easy opportunities for improving longevity of the COTS GPS devices. Although many studies have pointed to the advantages of shorter sampling intervals for describing animal distribution and movements (Johnson and Ganskopp, 2008; Rowcliffe et al., 2012; McGavin et al., 2018), increasing the GPS logging interval would extend battery life. Likewise, a higher-capacity battery could be used for longer implementations. Accuracy of the COTS GPS devices was commensurate with the iGotU devices and with results reported by McGranahan et al. (2018). However, the COTS GPS devices experienced a higher rate of missed GPS readings than the iGotU devices. This could be due to inherent differences in the GPS antenna used by each device or to the collar and housing for each unit. The metal hose clamps of the COTS collars may have impeded GPS signals (see Yi et al., 2012), and the weight of the housing caused it to typically hang below the cow’s neck. Conversely, the iGotU devices were housed in a nylon pouch stitched to the collar (after Knight et al., 2018) that would not impede GPS signal. The lighter weight of the nylon housing allowed the iGotU devices to slide on the cow’s neck, and they were often on the side or top of the neck. Options for nonmetallic housings or counterweighting the collar so that the device is always on top could be explored.
958
J.W. Karl, J.E. Sprinkle / Rangeland Ecology & Management 72 (2019) 954e958
The primary advantages of the COTS GPS devices, however, are their low cost, opportunities for customization, and simplicity. Our final per-unit cost exceeded our initial goal of $50 per collar. This was largely due to shipping costs associated with the parts. Because of the tight timeline of our project, however, we needed to obtain parts quickly and thus paid for shipping when we could have opted for free shipping with a longer delivery time. There are other opportunities to reduce costs further, including building custom circuit boards that minimize unnecessary parts (which would also reduce power consumption) and configuring the device to run on 3.3V, thereby eliminating the voltage converter (see https://github.com/jkarl/COTS_ GPS_Collars for examples). Low-power sleep-cycling software libraries are available for Arduino and could be implemented in place of the timer board to save cost. With default Arduino settings, software sleep cycling gains only modest power savings. Customizing the fuse settings on Arduino’s microcontroller allows for more significant lowpower sleep, but this step involves low-level reprogramming of the Arduino. Thus, the trade-offs for these steps are increased complexity in development and building of the devices. Unlike commercial collars, or collars based on commercial GPS trackers that are limited to what functionality the device has at time of purchase, the requirement to write custom scripts for the COTS GPS collars presents unique opportunities for improving data quality and adding functionality. For example, more sophisticated GPS signal processing could be implemented to improve data accuracy and reliability. In a test of simple GPS point averaging, average displacement and 95% CEP for 5-min interval logging that used the average of 20 consecutive GPS readings decreased to 2.68 m and 7.76 m, respectivelydmuch closer to the 1-s-interval CEP values than the standard 5-min values. In addition, with GPS point averaging, the incidence of failed GPS readings also decreased. Finally, the simplicity of the COTS GPS collars can be both an advantage and a limitation. The COTS GPS devices can be built by anyone with modest technical knowledge and soldering skills, and the basic, open data format is easy to process with GIS and statistical software. However, the fact that the COTS GPS devices perform only one functiondrecord locations at a predefined intervaldcan be a drawback. The loss of two of the iGotU collars (which also are solely GPS recorders) during the study, presumably due to failure of the collar’s strap or buckle, highlights that data are contingent on recovering the device. An independent VHF transmitter could be added to aid in relocating lost devices (but would add to device cost). Management Implications The design and test presented here demonstrate that it is possible to manufacture low-cost location tracking devices that perform similarly to more expensive devices. In addition, the COTS collar approach has several advantages over other commercial devices including the ability to 1) build in only necessary features to save power and reduce complexity, 2) easily add other sensors, and 3) develop custom scripts to allow for advanced data processing like point averaging to overcome limitations of low-cost sensors. Given the limitations discussed earlier, though, the COTS GPS collars would be best suited for studies where a large number of collars were needed for a short duration (weeks with existing batteries or several months with larger-capacity batteries). Despite their simplistic design and relatively limited functionality compared with many commercial collars (e.g., no wireless data download), the COTS GPS collars could encourage location tracking in many more livestock rangeland studies. Although battery life for a COTS GPS collar may be shorter than for a similar commercial device, a primary advantage of the COTS GPS collar described here is the ability to easily and inexpensively develop custom GPS tracking applications. The lower cost of the COTS GPS collars will also facilitate collection of a much higher density of location
information (i.e., tracking more animals in a herd) to better understand patterns of livestock use in rangeland landscapes and how livestock interact with wildlife populations and rangeland uses. Acknowledgments We are grateful to R. Ward and C. Ward for making their cattle herd available for this study and for assistance in installing and removing the collars. This study was approved by the University of Idaho Institutional Animal Care and Use Committee under study IACUC-2018-25. Appendix A. Source information for Arduino libraries used in Global Positioning System collar script
Library name and version
Description
Source
SPI, 1.0.0
Enables communication with devices that use the serial peripheral interface (SPI) bus. Enables reading and writing to secure digital cards through the SPI interface.
Arduino: https://www. arduino.cc/en/Reference/SPI, Accessed 5 November, 2018 Arduino, Sparkfun: https:// www.arduino.cc/en/ Reference/SD, Accessed 5 November, 2018 Arduino: https://www. arduino.cc/en/Reference/ SoftwareSerial, Accessed 5 November, 2018
SD, 1.1.1
SoftwareSerial, Enables serial communication 1.0.0 on any digital Arduino pin. SoftwareSerial is used for communication with the GPS device. TinyGPS þþ, Arduino library for parsing 0.95a NMEA data streams from GPS devices.
Mikal Hart: http://arduiniana. org/libraries/tinygpsplus/, Accessed 5 November, 2018
References Adafruit, 2017. Adafruit TPL5110 Power Timer breakout. Available at: https://learn. adafruit.com/adafruit-tpl5110-power-timer-breakout/. Accessed 5 November 2018. Augustine, D., Derner, J., 2013. Assessing herbivore foraging behavior with GPS collars in a semiarid grassland. Sensors 13, 3711e3723. Bailey, D.W., Gross, J.E., Laca, E.A., Rittenhouse, L.R., Coughenour, M.B., Swift, D.M., Sims, P.L., 1996. Mechanisms that result in large herbivore grazing distribution patterns. Journal of Range Management 49, 386e400. Bailey, D.W., Trotter, M.G., Knight, C.W., Thomas, M.G., 2018. Use of GPS tracking collars and accelerometers for rangeland livestock production research. In: Translational Animal Science. Available at: http://academic.oup.com/tas/ advance-article/doi/10.1093/tas/txx006/4824982. Accessed 21 February 2018. Bajarin, T. 2014. Why the maker movement is important to America’s future. Time. Available at: http://time.com/104210/maker-faire-maker-movement/. Accessed 23 July 2018. Clark, P.E., Johnson, D.E., Kniep, M.A., Jermann, P., Huttash, B., Wood, A., Johnson, M., McGillivan, C., Titus, K., 2006. An advanced, low-cost, gps-based animal tracking system. Rangeland Ecology & Management 59, 334e340. Ganskopp, D.C., Bohnert, D.W., 2009. Landscape nutritional patterns and cattle distribution in rangeland pastures. Applied Animal Behaviour Science 116, 110e119. Johnson, D.D., Ganskopp, D., 2008. GPS collar sampling frequency: effects on measures of resource use. Rangeland Ecology & Management 61, 226e231. Knight, C.W., Bailey, D.W., Faulkner, D., 2018. Low-cost global positioning system tracking collars for use on cattle. Rangeland Ecology & Management 71, 506e508. McGavin, S.L., Bishop-Hurley, G.J., Charmley, E., Greenwood, P.L., Callaghan, M.J., 2018. Effect of GPS sample interval and paddock size on estimates of distance travelled by grazing cattle in rangeland, Australia. The Rangeland Journal 40, 55. McGranahan, D.A., Geaumont, B., Spiess, J.W., 2018. Assessment of a livestock GPS collar based on an open-source datalogger informs best practices for logging intensity. Ecology and Evolution 8, 5649e5660. Rowcliffe, M.J., Carbone, C., Kays, R., Kranstauber, B., Jansen, P.A., 2012. Bias in estimating animal travel distance: the effect of sampling frequency: estimating animal travel distance. Methods in Ecology and Evolution 3, 653e662. Turner, L.W., Udal, M.C., Larson, B.T., Shearer, S.A., 2000. Monitoring cattle behavior and pasture use with GPS and GIS. Canadian Journal of Animal Science 80, 405e413. Yi, T.-H., Li, H.-N., Gu, M., 2012. Effect of different construction materials on propagation of GPS monitoring signals. Measurement 45, 1126e1139.