Introduction

The use of Global Positioning System (GPS) telemetry has been widely applied in the field of wildlife research (Hebblewhite and Haydon 2010). Recently, technological advancements have resulted in the miniaturisation of GPS devices with the production of lightweight receivers increasing the range of taxa that can be monitored through GPS telemetry (Cagnacci et al. 2010).

Compared to very high frequency (VHF) telemetry, GPS possesses many inherent advantages such as increased frequencies and quantities of data acquisition, increased spatial accuracy, reduced survey effort and the ability to work independently all hours of the day, in all weathers and in remote or inaccessible habitats, whilst removing the risk of observer-induced bias or behavioural modification in the focal animal (Wegge et al. 2007; Urbano et al. 2010). The limitations of GPS devices include battery capacity, leading to a trade-off between the frequency of data acquisition and the size/weight of the device in determining functional longevity; cost; technological factors such as satellite geometry and atmospheric interference; environmental factors such as local topography and habitat type; and individual animal behaviour which can reduce accuracy and performance (Frair et al. 2010; Recio et al. 2011). Furthermore, the size of small terrestrial mammals often means that they are close to the ground increasing the possibility of satellite obstruction which could potentially reduce performance of any GPS device deployed. Whilst GPS devices collect vast quantities of spatial data at frequencies/intervals unachievable through VHF telemetry, concerns regarding independence and spatial and temporal autocorrelation are apparent (Boyce et al. 2010; Fieberg et al. 2010). However, with removed risk of observer-induced modification, GPS can provide accurate unbiased details of movement and resource use.

Performance testing prior to animal GPS deployment is essential. Static tests reveal reduced accuracy and performance in obstructive (e.g. forests) compared to open habitats (Guthrie et al. 2011) potentially biasing conclusions of habitat use in the focal animal though misclassification or underrepresentation of use (for review, see Frair et al. 2010).

In this study, we evaluate the use and performance of micro GPS devices through a series of static and field tests using hedgehogs (Erinaceus europaeus), a species of conservation concern in the UK (Wembridge 2011) and an invasive species which threatens native wildlife in New Zealand (Shanahan et al. 2007; Recio et al. 2013). We aim to evaluate performance and accuracy of the GPS devices in a variety of habitats frequented by hedgehogs in a rural landscape and to determine whether such devices are appropriate and cost effective for use on similar-sized small mammals.

Methods

Study site

The study was conducted at Nottingham Trent University's Brackenhurst campus, Southwell, UK (Grid Reference SK 6946 5243). The campus is 200 ha of mixed agricultural land dominated by arable and pasture fields.

GPS device

We used GPS Avian Bugs (hereafter termed ‘bug’) manufactured by Biotrack Ltd. (Wareham, Dorset, UK), fitted with a small ancillary VHF transmitter with a combined weight of 13 g. The bug incorporated an internal data logger, appropriate firmware with flexible schedule programming and rechargeable battery power. The bug had a universal serial bus (USB) port permitting communication between the device and a personal computer. The maximum fix attempt period is programmable and was left at the default setting of 70 s. If the bug failed to acquire a location fix before 70 s, it recorded a failed attempt and started again at the next scheduled fix attempt. Once deployed, the bug records the time and date of each scheduled fix attempt, and the latitude, longitude and altitude of each successful fix, and other information (see Quaglietta et al. 2012).

Static tests

Static tests were used to assess the accuracy and performance of four GPS bug devices in five different habitat types within a rural landscape that are utilised by hedgehogs (adjacent to buildings (<5 cm), hedgerow, open pasture, reedmace beds and woodland), each having different levels of cover obstructing satellite detection. Sites of mixed deciduous woodland with extensive canopy cover were chosen to offer the greatest obstruction. Leaf coverage was relatively high, as the study took place before the onset of winter. Similarly, we chose tall extensive hedgerows typical of hedgehog nesting habitat. Open pasture fields with short cut grass offered little obstruction. Ideally, we would have tested arable fields but due to harvesting, reedmace beds consisting of densely packed stands c. 1 m tall, where used in lieu. Two-storey academic buildings were selected to simulate urban habitats.

Static tests were carried out during October and November 2012. Bugs were fully charged and scheduled to record location fixes at 2-min intervals, placed side by side at ground level in each habitat type and left in place for a minimum of 4 h (>120 fix attempts) with two separate trial locations tested for each habitat. Data was downloaded after each individual trial. The high frequency of recording location fixes was chosen to limit the time bugs were in the field to reduce the chance of theft or disturbance, and bugs were deployed between the hours of 0900 and 1700 randomly in each location to ensure a wide range of satellite orientations were covered by the study. Since the average number of satellites detected by bugs at each fix attempt was 6.1 ± 0.2 (range 2–12), we are confident that our test was a representative of the daily variation of satellite configuration cycles. The actual known location (reference point: chosen as a prominent feature in the landscape that is easily identifiable within ArcGIS (e.g. the corner of a building or field)) where the bugs were placed and the bug derived location data was plotted onto a map in ArcMap (ArcGIS 10.0, ESRI, Oaklands, CA, USA). To assess performance, we calculated the percentage fix success rate (FSR) as the proportion of successful fix attempts against the total number of attempted fixes. Accuracy was determined as the mean location error (LE) calculated as the Euclidean distance (m) between each recorded fix and the reference point.

Field tests

A Natural England Licence (number: 20121788) was obtained to permit the capture and handling of hedgehogs during April to October 2012. Small VHF radio transmitters (Biotrack Ltd.) weighing c. 7 g were attached to individuals hedgehogs (n = 11; 3 males, 8 females) weighing over 600 g (Hof et al. 2012). Hedgehogs were tracked up to three nights per week and where possible located at least twice per night, and their locations were recorded using a handheld GPS (Garmin GPS60). GPS bugs were attached to the radio tagged individuals from June onwards. Two schedules were programmed into the bugs to reflect changes in daylight through the summer. Schedule 1 operated between 2200 hours British Summer Time (BST) and stopped at 0400 hours, whilst in schedule 2 (mid-August onwards), the bugs operated between 2000 hours and 0600 hours, with location fixes recorded at 20-min intervals. Twenty-minute time intervals between fixes were chosen as a compromise between gathering accurate data on hedgehog nightly movement and battery life of the bugs, to ensure several nights' data were recorded in one deployment, thereby minimising disturbance to the animal. We attached the bugs for a total of five full nights after which they were removed, data downloaded, recharged and reattached to a different individual the following night. Where possible, we aimed to attach a GPS bug to each individual twice during the season (mid-June–October).

Analysis

Statistical testing was performed using Minitab 16 (Minitab Inc. 2010) with all data tested for normality using a Kolmogorov–Smirnov test. A Kruskal–Wallis test was used to test FSR between the different habitat types, and LE differences were tested using a one-way ANOVA.

To determine whether FSR was influenced by the possibility of hedgehogs being in the nest, and therefore unavailable to satellite coverage, we calculated the average number of failed fixes per hour within the first hour after sunset (Sunset + 1) (the time period where hedgehogs are likely to be still in their day nests) with the hour before sunrise (Sunrise − 1) (the time period when hedgehogs are likely to have returned to their day nests), and between these two periods (Mid) (the time period when hedgehogs are most likely to be active) using a one-way ANOVA. Home range estimations were calculated for hedgehogs with >50 VHF radio tracking locations across the season (April–October) that also had two GPS bug attachments (GPS fix data was combined prior to analysis). The 100 % minimum convex polygon (MCP) and 95 % kernel density estimation (KDE) for each hedgehog were calculated, and we compared the estimates of each survey method using Paired T Tests.

We performed a cost analysis comparing the project costs of generating home range estimations for 1, 5 and 10 hedgehogs through GPS bug and VHF radio tracking methods where we assumed >50 location fixes were required for the estimation. A single GPS bug (£1,700) is ten times more expensive than a VHF bug (£170). GPS software was £350, whilst tracking equipment essential for both methods was £1,800. Labour costs were fixed at £11 per hour for attaching/removing the GPS bug or manually locating the hedgehog through VHF tracking once an hour.

Results

Static tests

During static tests of GPS bug performance, a total of 3,538 location fixes were collected from 4,484 attempts with an FSR of 78.9 %. LE and FSR were significantly different between habitat types (LE: one-way ANOVA; F 4, 34 = 9.85, P < 0.001; FSR: Kruskal–Wallis test; H = 29.79, DF = 4, P < 0.001) with accuracy and performance greatest in open pasture and worst in woodland habitat (Table 1).

Table 1 A comparison of mean location error in metres (LE) and median fix success rate expressed as a percentage (FSR) recorded from Global Positioning System (GPS) bug static tests in five different habitat types
Full size table

Field tests

Mean (± SE) FSR across all attachments was good at 84.6 % ± 2.4, with an average of 96 ± 4.5 successful location fixes recorded per attachment. The mean (± SE) number of failed fixes per hedgehog (n = 10) was significantly higher during the hour after sunset (n = 10, 8.25 ± 1.2) and the hour before sunrise (n = 4, 7.75 ± 1.3) compared to the time period in between (n = 10, 1.99 ± 0.3) (one-way ANOVA; F 2, 21 = 14.53, P < 0.001). There was no difference in the mean home range size between the VHF and combined GPS estimates for the MCP (VHF: 18.3 ha ± 3.7; GPS: 21.2 ha ± 4.7; n = 9, Paired T Test; T = −0.65; P = 0.536) and 95 % KDE (VHF: 23.9 ha ± 4.8; GPS 21.4 ha ± 4.7; n = 9, Paired T Test; T = 1.33, P = 0.219) estimations (Table 2).

Table 2 Home range estimations for each individual hedgehog that was tracked using Very High Frequency (VHF) radio tracking and through Global Positioning System (GPS) bugs during 2012
Full size table

Cost analysis

Total home range project costs for one hedgehog tagged with GPS bug were £3,872 compared to £2,520 for VHF. Just 2 h of labour were required for the GPS bug (£22) compared to 50 h for VHF (£550). If five GPS and VHF bugs were purchased, the total costs would be £10,760 and £5,400, respectively, requiring just 10 h of labour (£110) for GPS compared to 250 h (£2,750) for VHF. For ten, total costs are £19,370 and £9,000, with 20 h (£220) and 500 h (£5,500) of labour, respectively.

Discussion

Static testing revealed that habitat type had a significant influence on both the performance and accuracy of the GPS bugs, with the physical characteristics of the habitat influencing line of sight visibility between the GPS bug and orbiting satellites (Frair et al. 2010). Accuracy and performance were reduced adjacent to buildings compared to more open habitats, whereas hedgerows were intermediate. Performance was worst in woodland, typically failing to obtain >50 % of attempted location fixes. We recommend researchers must assess the potential for habitat biased data loss for species utilising obstructive habitats, with static testing illustrating how the device may perform once deployed on the animal.

Individual animal movement and behaviour influence the performance and accuracy of GPS devices (Recio et al. 2011). We observed high rates of FSR from bugs deployed on hedgehogs in a rural landscape which is similar to other studies (Guthrie et al. 2011; Quaglietta et al. 2012; Recio et al. 2013). Nesting, denning and bedding activities (Kochanny et al. 2009; Mattisson et al. 2010) and the use of burrows (Price-Rees and Shine 2011) will increase the risk of fix failure. We observed more failed fixes in the hour after sunset and before sunrise when the hedgehogs were likely to be within their day nests compared to the active period. We suggest tailoring fix acquisition schedule to improve performance, and a pro-active approach modifying the activation and cessation of the bug in relation to the changing times of sunset and sunrise throughout the season to maximise performance potential for use in hedgehog home range studies.

Despite the inherent costs associated with the GPS bugs, survey effort and labour were an order of magnitude lower compared to VHF tracking. The ability to collect vast quantities of highly accurate spatial data is clearly advantageous and worth the expense, especially considering the effort required to manually locate each individual through VHF. If the animal enters an area inaccessible to the researcher, they may be unable to obtain a fix, further adding to costs. Studies using triangulation are often accurate to just 200–600 m (Frair et al. 2010), and the method is particularly time consuming (Guthrie et al. 2011). Since we used homing, our VHF locations were determined using a handheld GPS utilising the same technology as the bugs and accurate to 6.6 m ± 0.08 (625 VHF locations during 2012) which is comparable to the bugs.

Home range estimates from the GPS bugs were similar to those derived from an entire season of radio tracking as noted in other studies (Kochanny et al. 2009; Mustonen et al. 2012) which is clearly beneficial in reducing survey effort and labour costs. Combining location data from two relatively short periods of intensive data collection our estimates are similar to those of other researchers studying hedgehog home ranges through VHF tracking over a similarly short time period (Hof 2009). The flexibility to programme the fix acquisition schedule will make the bugs widely applicable to other species. The GPS bugs are also rechargeable and potentially have unlimited longevity and can be used to investigate home range on multiple animals further lowering costs, whereas the lifespan of the VHF bugs (c. 8 months) necessitates their replacement after each season of intensive survey effort.

Studies of hedgehog movement typically have involved the surveyor continuously tracking a radio tagged individual from a distance throughout the night, obtaining location fixes at regular intervals (Hof and Bright 2010). We were able to continually track some individuals for over 9 h per night with GPS bugs, with minimal effort and no risk of surveyor induced bias or behavioural modification. We suggest that GPS telemetry presents an opportunity to gain a truer, and unbiased, depiction of small mammal movements.

We have demonstrated the effectiveness of the use of GPS telemetry in the spatial study of a small terrestrial mammal and recommend its use in any spatial ecological study focused on mammals in rural landscapes. However, the reduced accuracy and performance within woodland are certainly concerning and may problematic for researchers studying species that exclusively reside within forested habitats. The device may also miss occasional or temporary use such as through nesting or predator avoidance. We also recommend GPS use on urban hedgehog populations and for other species that are urban adapters (see Bateman and Fleming 2012). Hedgehogs are increasingly utilising urban gardens in the UK (Dowding et al. 2010), and the GPS bugs will allow tracking through private areas inaccessible to the researcher.

The ability of GPS bugs to collect large volumes of highly accurate spatial data with greatly reduced survey effort and labour costs certainly outweigh the initial purchase costs. Without risk of surveyor induced bias, researchers are able to independently track small animal movement and behaviour, quantifying resource selection and home range in natural unbiased conditions. We expect the use of GPS bugs will likely increase in the coming years, with continued technological advancements likely to further expand the range of small mammals that can be tracked through GPS telemetry.