Set free: an evaluation of two break-away mechanisms for tracking collars

Keywords: animal welfare, bio-logging, drop-off, environmentally-degradable link, radio collar, tracking device, TRD, weak-link.

Tracking devices are an important tool to study animal movements, use of habitat, and survival (Cochran and Lord 1963; Schladweiler and Tester 1972; Storm 1972; Kenward 1987; White and Garrott 1990). Since the 1990s, there has been increasing use of Global Position System (GPS) devices to quantify animal movement, which provide greater accuracy than older Very High Frequency (VHF) technology (Bunnefeld et al. 2011; Recio et al. 2011; Lanzone et al. 2012). However, even within advances in technology, some researchers have experienced difficulties with GPS collars (Matthews et al. 2013; Dore et al. 2020).

Over the years, tracking collars have become more reasonably priced, lighter, and have better performance (Kays et al. 2015; Portugal et al. 2018; Batsleer et al. 2020; Dore et al. 2020; Katzner and Arlettaz 2020). Collection of higher-resolution data and advances in processing of big data (Hooten 2017; McMahon et al. 2017; Browning et al. 2018) have improved our ability to retrieve and analyse large and complex data (Kays et al. 2015; Foley et al. 2020; Katzner and Arlettaz 2020). These advancements have enabled finer scale analyses of animal movement than possible previously or increasingly smaller animals. This is also likely to increase the number of animals being fitted with what can be a permanent collar or harness attachment (Hebblewhite and Haydon 2010; Kays et al. 2015; Hughey et al. 2018; Buil et al. 2019).

The most appropriate method to retrieve a tracking device is re-capturing the animal, where the animal can also be checked for signs of injury that could arise from being collared. Re-trapping may work for resident or philopatric populations, but for species or populations that are trap-shy or more mobile over space and time, the ability to re-capture or even re-locate animals with tracking devices is not always certain. At the conclusion of the study period, if the animal cannot be re-captured, then the collar will remain on the animal indefinitely. Some researchers have recorded problems with tracking devices, including potential impacts of attachment such as skin trauma (Merino et al. 2007; Berg et al. 2010; Fitzgibbon et al. 2011; Coetsee et al. 2016), vegetation and limb entanglement (Barron et al. 2010; Juarez et al. 2011; Coetsee et al. 2016), changes in behaviour (Wilson et al. 2004; Brooks et al. 2008; Dennis and Shah 2012; Gibson et al. 2013), and decreased survival (Xiong et al. 2009; Severson et al. 2019). Such issues make indeterminate collar attachment a significant animal welfare issue.

For animals that are not readily re-trapped, an option is to include a break-away mechanism in the collar. One method involves the inclusion of materials that expand according to neck growth and eventually break; e.g. expanding break-away with low density polyethylene (Strathearn et al. 1984), rubber (Soderquist 1993), elastic and Velcro (Robertson and Harris 1996). Another method is to include an environmentally-degradable link, such as corrodible bolts (Thalmann 2013; Povh et al. 2019), or cotton thread (Karl and Clout 1987; Hellgren et al. 1988; Merrill et al. 1998; Casper 2009; Cawthen and Munks 2011; Collins et al. 2014). Programmable collar release systems (timed-release devices; TRD) are increasingly available through commercial providers (Evans 1996; Merrill et al. 1998; Kochanny et al. 2009; Purcell 2010; Ruykys et al. 2011; Matthews et al. 2012; Cowan et al. 2020), or through open-source designs (Buil et al. 2019; Rafiq et al. 2019), and the recent miniaturisation of TRD (micro-TRD ~10 g) allows this approach to be used on smaller animals.

Few studies have provided adequate details about the mechanism function, retention and release time of break-away mechanisms (Evans 1996; Kochanny et al. 2009; Collins et al. 2014; Dore et al. 2020). The paucity of such knowledge makes it difficult to make an informed decision about the selection of break-away type for a given species. This is a particularly important consideration when working with conservation significant species and those that are exposed to public scrutiny (e.g. animals that are viewed as part of ecotourism ventures).

A factor that will determine the uptake and application of collar break-away methods is whether the mechanisms will release in the desired timeframe, balancing adequate data collection and prevention of potential long-term welfare impact (Matthews et al. 2013; Buil et al. 2019; Dore et al. 2020). For example, Cawthen and Munks (2011) used cotton thread weak-link on collars on brushtail possums (Trichosurus vulpecula), but showed short mean retention time and increasing project costs to replace the radio-collars that fell off prematurely. A similar outcome was reported for a study with juvenile red foxes (Vulpes vulpes) that necessitated regular recapture of animals and resulted in high costs (Robertson and Harris 1996). Although release failures might be reported, only one recent study mentioned a minor injury caused by tracking devices equipped with cotton thread weak-link and recommended adjustments (Sims et al. 2021).

Timed-release devices may also have issues with failures either through releasing earlier than scheduled, or not at all (Matthews et al. 2013; Dore et al. 2020). For example, Kochanny et al. (2009) deployed 21 store-on-board GPS telemetry collars with a releasable mechanism (model G2000; Advanced Telemetry System Inc. Isanti, Minnesota, USA) on white-tailed deer (Odocoileus virginianus); of these, eight collars failed to release. Furthermore, a recent study deployed four GPS collars with Sirtrack LiteTrack 140 TRD on red foxes but only one collar was retrieved (Main 2020).

The quokka is a marsupial species endemic to south-western Australia (Kitchener 1995). These medium-sized wallabies are listed under State legislation as Rare or Likely to Become Extinct, and as Vulnerable both nationally (EPBC 1999) and internationally on the IUCN red list (Burbidge and Woinarski 2020). In the northern part of their range (Spencer et al. 2019), quokka populations in the jarrah forest are fragmented and subject to a range of threats, including competition and predation by invasive species (feral pig Sus scrofa, red fox and feral cat Felis catus), decreased rainfall, and habitat loss due to Phytophthora dieback (Department of Environment and Conservation 2013). Altered fire regime is also a significant threat to quokka populations. Wildfire can devastate quokka habitat, and therefore carefully managed prescribed burning can be an important tool to manage habitat and reduce fuel load that could lead to potential wildfire (Bain et al. 2016).

As part of a broader study, we sought to identify the response of quokka to habitat change caused by prescribed burning. We fitted tracking collars to quokkas to study their movements before and after prescribed burning. Mainland quokkas are present in low numbers in the northern jarrah forest, which reduces overall trap success, therefore increasing the costs and efforts required to track these animals. Quokkas can also be elusive, trap shy, and can travel long distances outside their home range subject to habitat connectivity (Bain et al. 2020), reducing the likelihood of re-capturing the same individual multiple times. Therefore, to ensure animal welfare was not compromised, we equipped VHF and GPS collars with two types of break-away mechanism: cotton thread (CT) and micro-TRD (mTRD). This provided the opportunity to compare the fates of collars with these two break-away mechanisms.

Between July 2018 and September 2019, quokkas were captured with Thomas soft-wall traps (360 × 480 × 800 mm (LWH) Sheffield Wire Works, Welshpool, Western Australia) at five sites (Churchman, Wungong, Gordon, Chandler, and Marrinup forest blocks) in the northern jarrah forest of south-western Australia (Fig. 1). Individuals trapped overnight were removed from traps the following morning. Quokkas were weighed, microchipped, and basic details were recorded. Adult quokkas weighing >2 kg were fitted with a tracking collar (Fig. 2). After fitting collars, quokkas were released at point of capture. All handling was undertaken by experienced ecologists and this project was approved by Murdoch University Animal Ethics Committee (R3058/18) and Department of Biodiversity, Conservation and Attractions (FO25000082-2).

Fig. 1.  Mainland quokka (Setonix brachyurus) study areas, in the northern jarrah forest of south-western Australia.

Fig. 2.  The VHF and GPS collars that had been deployed on mainland quokkas (Setonix brachyurus). (a) Retrieved GPS Litetrack 30 collar without micro timed-release device, (b) the same type of GPS collar with micro-TRD (mTRD), (c) VHF collar belt cut to prepare for cotton thread (CT) weak-link insert and (d) cut sections sewn together with CT.

Seventeen VHF and three GPS collars were deployed with a cotton thread (CT) weak-link. VHF collars (M1820 ATS, Australia) weighed 27 g, with housing dimensions of 37 × 12 × 13 mm (L × W × H) and a battery life estimation of 390 days. GPS collars (model LiteTrack 30 Sirtrack, New Zealand) weighed 35 g, with housing size of 20 × 37 × 24 mm, a battery life estimation of 1 year, and with remote UHF download combined with VHF signal. CT weak-links were added to the collars by cutting the collar belt and stitching the two parts together with cotton sewing thread (Fig. 2) using a different number of stitch loops on each collar. The variation in number of loops aimed to test the best stitch configuration to avoid early release.

Eleven GPS collars with mTRD were deployed. The Sirtrack mTRD went out in the field for beta trials in 2016, but have had nothing published on them to date. The mTRD added 10 g to the weight of the collar. The mTRD had an independent battery and clock and were programmed to release at 300 or 350 days after deployment (the dates were scheduled differently for independent study sites).

Following collar deployment, quokkas were tracked twice per week. The period of attachment (retention time) was calculated from the date of initial deployment until the last known occasion of attachment. Re-trapping was carried out to replace VHF with GPS collars (year 2 of the study) and to retrieve collars that had not released at completion of the study, placing traps at the same initial trapping point the individual had been located and around nearby known refuge locations determined by radio-tracking. Additionally, 80 camera traps (Reconyx© HC600 Hyperfire, Reconyx Inc., Holmen, USA; 10–20 camera traps at each site, depending on the size of quokka habitat) were deployed to monitor quokka wellbeing and to confirm collar release. Camera trap locations were selected based on areas with presence of animal trails and quokka scats. Camera traps were attached to trees with cable locks (Master Lock Company Phython™, Wisconsin, USA) at a height of approximately 0.5 m and remained in the same position for 2 years. Camera traps were baited daily with cut-up pieces of apples during live trapping, and once per month when not trapping. Camera traps were set on high sensitivity passive infra-red trigger, rapid fire at five photos per trigger with no quiet period, to ensure as many photos as possible were taken of each quokka to facilitate individual identification. Individual quokkas were distinguished through unique markings and general home range locations. Camera trap photos for each individual were examined to determine the last date the collar was present and the first date that the collar was absent.

To determine the likely factors influencing retention time of collar attachments, we carried out multiple regression analysis to describe the relationship between CT weak-link retention time (dependent variable) against models including the number of cotton loops applied and the body mass (kg) and pes length (mm) of quokkas. We used Akaike’s information Criterion (AIC) with a small sample correction (AICc) model selection and considered models with delta AICc values <2 to have strong support to distinguish among a set of possible linear models. To compare the retention time for the collars fitted with the two weak-link devices, retention time was compared by Kaplan–Meier curve with package ‘survival analysis’ in R (Therneau 2021). To test whether there were detrimental impacts of wearing a collar, changes in body weight before and after deployments were analysed using Shapiro–Wilk to test normality in a small sample size, followed by Paired Samples t-test.

Statistical analyses were performed using RStudio ver. 4.0.3. Significance values for all tests were set at α = 0.05, and values of response variables are reported as means ± standard deviation (s.d.).

We live-trapped 75 mainland quokkas across the five sites and 33 adult males weighing >2 kg were fitted with tracking-collars. Of those, 28 collars were modified with a break-away (Table 1).

Table 1.  The fate of 20 collars with a cotton thread weak-link (CT) and 11 collars with a lightweight Sirtrack (~10 g Sirtrack LiteTrack 30) automatic micro timed-release device (mTRD) deployed on quokkas (Setonix brachyurus).

Of the 17 radio collars with a CT weak-link, the fates of 15 collars contributed to collar retention data. Seven collars fell off at an average of 148 ± 64 (range 13–199) days after deployment. There was no correlation in retention time, animal morphometrics, or number of cotton loops, with the null model being the best model describing collar retention time (Table 2). Six collars were retrieved, while one collar was in mortality mode but could not be located (ID#2569; the animal was re-trapped a year later confirming that the collar had released). Two collars did not release for a minimum of 1 year after deployment (data from tracking): one animal (ID#2579) was last seen on camera trap 186 days after deployment (with collar) and was tracked for 398 days, and the second animal (ID#2628) was last seen on camera trap 389 days after deployment (with collar) but could not be re-trapped. The other eight quokkas were recaptured (161 ± 109 days after deployment) and collars removed; six of these collars had been on animals for more than 5 months.

Table 2.  Comparison of fitted linear models (Gaussian distribution) for the deployment time (retention time in days) of cotton thread weak-link collars (CT), comparing the effect of average body weight (Model 1), cotton thread arrangement (Model 2) and pes length (Model 3). Dots indicate that the predictor variable was absent from the model.

Of the 11 radio collars with mTRD, nine collars released. Two fell off before their scheduled release and were recovered at 16 and 29 days after deployment (Fig. 3). Seven collars released around their scheduled release date. Two collars were removed from animals (one collar went into recovery mode 250 days after deployment – the animal was re-captured, and the collar removed; the second was a mortality – the carcass was retrieved, and the collar removed) and the mTRD boxes monitored until they both released on their scheduled release date. For the other five collars that released around their programmed schedule, the collars could not be located but camera trapping data (Fig. 4) indicated that they had released somewhere between 35 ± 48 before to 22 ± 26 days after schedule. We could not locate any of these collars, despite extensive searching of the study sites.

Fig. 3.  The retention time for two types of tracking collar break-away (CT cotton thread and mTRD timed release device), showing collar retention in number of days since deployment. The GPS collars equipped with mTRD were programmed to release at either 300 or 350 days; dates shown were the maximum length of time these collars were retained as release could only be confirmed by camera trap images of the animals without their collar.

Fig. 4.  Five individual quokkas fitted with a GPS collar equipped with micro timed-release device (mTRD). The mTRD released around their scheduled dates, but the VHF signal stopped working and the collar could not be retrieved. Collar release was therefore confirmed by camera trap photos, showing last sighting with collar (left hand column) and first sighting without collar (right hand column) for the same five animals uniquely identified by location and natural body marks (e.g. notches in the ear margin).

The fate of the last two collars could not be confirmed. One collar was either in mortality or release function (251 days after deployment), but the faint signal from the collar was insufficient to locate it; the animal was never re-trapped or seen on a camera trap subsequently. We do not know the fate of one other quokka, which was observed on camera trap just a few times after deployment, tracked for 142 days, but then not re-captured or seen on camera traps subsequently (Table 1).

CT collars had less reliability in terms of collar retention time, with three collars retained for more than 1 year. By contrast the seven mTRD boxes that did not fail early all released within a close range of their scheduled release date (Fig. 3). The difference in retention time between these two mechanisms was not statistically significant (P = 0.250).

Collared quokkas that were re-trapped for collar removal demonstrated no visible neck abrasion or other injuries that could be caused by the collar or break-away mechanism. Of the 33 collared quokkas, 26 were recaptured at the end of the study. Recaptured quokkas showed no significant change in body mass over the period that they had been collared (t₂₅ = 0.79, P = 0.442).

We compared two types of tracking collar break-away mechanism and found no visible neck abrasion or other injury that could be caused by either the collars or break-away mechanisms. This supports other studies that reported weak-links did not cause negative impacts on the mammals studied (Cawthen and Munks 2011; Thalmann 2013; but see Sims et al. 2021). For both break-away types, the release timing was not 100% predictable. CT collars were less reliable, with three retained on the animals for over a year. By contrast all the mTRD devices that did not fail early released close to their scheduled release date.

Although authors rarely report failures, many attachment systems result in premature equipment release (Evans 1996; Robertson and Harris 1996; Cawthen and Munks 2011; Kesler 2011), last longer than scheduled (Garshelis and McLaughlin 1998; Collins et al. 2014), or do not release at all (Kenyon et al. 2015; Sims et al. 2021). We aimed to determine the best cotton thread stitch arrangement to avoid early collar release, but found no significant relationship between collar retention time and the number of cotton thread loops stitched around the collar belt. The exact cause of breakage of the CT weak-link could not be determined, and the collars released independently of the cotton arrangement.

Our outcome is similar to the other studies that indicated the effectiveness of weak-links designed with cotton thread is highly variable (Cawthen and Munks 2011; Kenyon et al. 2015; Sims et al. 2021) and may not increase collar retrieval rates (Rayner et al. 2022). The relative retention time and breakage may be affected by alterations to the basic design (e.g. cotton thread thickness), habitat conditions (e.g. Thalmann 2013), animal strength, robustness and dexterity or behaviour (Garshelis and McLaughlin 1998). A review (Rayner et al. 2022) detailed that weak-links broke consistently during dasyurid studies, suggesting the breakages might be related to this group having strong forelimbs and exhibiting interactions between individuals through mating. Conversely, no weak-links broke to release the collars deployed on hare-wallabies, potentially due their weak forelimbs. Even though the mainland quokka is robust and exhibits interactions between individuals, our study found no correlation in collar retention time and animal morphometrics, with collars released or retained on the animals independently of body mass and pes length.

The devices with the CT weak-link had greater retrieval success than the collars equipped with mTRD, which failed to transmit VHF signals shortly after the scheduled release date and could not be located. After the scheduled release date, we carried out extensive ground searches for collars for several weeks, including adapting a Yagi antenna making it 3 m taller to improve receiver gain. It is not clear what caused the collars to fail to transmit a drop off signal. A recent review (Dore et al. 2020) has reported similar failure, where collars manufactured by Lotek and Tellus Micro had lower battery life, affecting early loss of VHF or UHF signals, impacting the ability to retrieve the collars. These collars have two independent batteries, one for the GPS beacon and another for the mTRD and, like another studies (e.g. Kochanny et al. 2009), we can only speculate that it could be either exhaustion of the beacon battery or weak recovery signal being absorbed by the ground and dense vegetation cover preferred by quokkas.

Of the 11 collars equipped with mTRD in our study, 18% released before their due date, a rate that is significantly greater than the average of 5% collars released before or after the due date in six studies reviewed by Matthews et al. (2013). Our 18% unknown collar failure rate is comparable to the average of 19% of collars with TRD failure reviewed by Matthews et al. (2013) but less than the 25% and 32% failure rates reported by Kochanny et al. (2009) and Cowan et al. (2020), respectively. A subsequent review of the functionality of 75 GPS collars equipped with electronic drop-off devices (Dore et al. 2020) reported that success of the drop-off mechanisms varied according to manufacturer, with ~50 to ~90% success. Most of our GPS collars released (64%), but because retrieval of mTRD collars was made difficult by their failure to transmit VHF signals shortly after the scheduled release date, we have had to rely on supplementary camera trapping to confirm the releases. This back-up added extra manual labour to create the quokka profiles (ID quokkas by unique markings) and to review 110 000 pictures. Despite the increased labour, we recommend the use of camera traps to monitor collared animals to similarly facilitate confirmation of collar release.

From the limited number of studies available, the outcomes of TRDs are highly variable. The main concern with these mechanisms is their confirmed release around the programmed date (Matthews et al. 2013), as the delays in activation of a TRD have implications for animal welfare and collar recovery, particularly in remote areas (Cowan et al. 2020). The specific time of collar release is also likely to be important for ensuring retrieval, as it would influence whether the collars fall off when the animals are active (midnight for nocturnal species such as quokkas) or located around rest areas (midday for nocturnal species). This could also influence the distance required to be searched to recover the devices.

Our study provides valuable information about <40 g collars equipped with CT weak links and mTRD. While most collars equipped with either mechanism released, CT weak links had less reliability and breakage of the cotton thread could not be predicted. Increasing the range of materials (e.g. different types of cotton threads) with various arrangement (e.g. the number of CT stitched around the collar belt) in different environments (wet/dry areas) may increase predictability of collars with this mechanism. Our poor mTRD collar recovery rates show the importance in adding camera traps to monitor the fate of collared animals. With the low numbers of individuals in our study sites, we were able to distinguish individuals from natural body marks and identify them from photos. For other species, marking animals (e.g. permanent ear tags) may be crucial to assist in individual identification from pictures to later confirm collar release; otherwise, it may be required to re-trap animals to retrieve collars.

We argue that tracking devices equipped with a break-away mechanism are essential for safeguarding animal welfare outcomes. Although weak-links do not always work as planned, because of material variations, habitat or animal behaviour, they do not cause additional problems compared to collars without links; however, they should work only as a back-up and not as a primary strategy for collar removal (Casper 2009; Matthews et al. 2013; Rayner et al. 2022). Here we contribute to knowledge about the fate of animals in tracking studies, and we call for continued systematic documentation of collar retrieval and potential impacts and mitigations of devices deployed on animals.

The data that support this study will be shared upon reasonable request to the corresponding author.

The authors declare no conflicts of interest.

Funding was received via a compensation arrangement between Alcoa Australia and Department of Biodiversity, Conservation and Attractions through the Forest Enhancement Committee.