Briar Cook - January 2021
INTRODUCTION/BACKGROUND
Zero Invasive Predators (ZIP) is undertaking a research and development programme of work in the Perth River valley, South Westland. The aim is to develop the knowledge, tools and techniques to completely remove possums and potentially rats from the 12,000 ha remote mainland site. The site will then be protected from reinvasion and re-establishment of possum populations in perpetuity. The Perth River valley is home to an estimated population of 75-120 kea (Nestor notabilis), the world’s only true alpine parrot.
Earlier work by ZIP, and others, has found that live capture leghold traps are the most effective tool for catching possums. Kea (amongst other non target species) are susceptible to being caught and injured in leghold traps. The use of leghold traps is permitted on the West Coast, under approval by the Department of Conservation (DOC). Work contracted by DOC in 2012 to reduce the risk of leghold traps to kea in non-forested areas (through trap modifications) resulted in reduced efficacy at catching possums. Unfortunately, despite the modified traps, two kea were caught in leghold traps and had to be euthanised in DOC’s 2014/2015 National Biodiversity Monitoring season. Alternative monitoring methods for possums have been in use since the 2015/2016 season.
While protection of the Perth River valley site from possum re-establishment is crucial to the success of the project, it is important to mitigate the risks to non-target species – as well as their effect on device functionality. Kea are curious and intelligent birds (Huber and Gajdon, 2006; Heaney et al., 2017; Bastos and Taylor, 2020) with a reputation for thoroughly investigating novel objects in their environment.
The ZIP SafeLock trap consists of: the ZIP PosStop, a Pest Control Research #1™ leghold trap (similar to the Victor leghold trap) set in a custom plastic moulded platform; paired to a ZIP-designed automated reporting system with a daytime lockout function (henceforth known as ‘SafeLock’). The SafeLock system means that during daylight hours the traps is locked open and unable to be triggered, then full operation and sensitivity of the trap is restored when it is unlocked during the night. Each SafeLock trap has a short length of bungee cord tied into the chain at the trap end, to help absorb the initial ‘yank’ when a possum is caught. ZIP has found that having a bungee cord decreases the escape rate of possums from leghold traps from around 10% to below 3% (unpublished data).
Possum ranges tend to increase with reduced density (Sweetapple and Nugent 2009; Margetts et al 2020). ZIP has successfully used a low-density trap network to mop up invading individuals looking for conspecifics. At ZIP’s Bottle Rock field site, a lean network of leghold traps at a density of 1 per 50 ha was deployed to target possums that had leaked through a virtual barrier of traps. This network prevented possums from re-establishing on the peninsula for nearly two years (before our work there ended).
SafeLock traps were proposed to be permanently deployed in the Perth River valley site in an approximate 700 m by 300 m grid (i.e. 1 per 21 ha) within all forested areas (143 traps in total). Before undertaking the full deployment of this network, we sought to identify and mitigate any risks to (and trap interference by) kea from the SafeLock traps.
OBJECTIVES
The objectives of this trial were to:
Assess the impacts of kea interference on the technical functionality of ZIP SafeLock traps for catching possums in the Perth River valley
Build confidence in the robustness of the SafeLock set-up in kea habitat, and its ability to keep these birds safe
TRIAL METHODS - PERTH RIVER VALLEY
Figure 1: Location of the ZIP SafeLock traps at the headwaters of the Barlow River, within the Perth River valley project site
Ridge Traps and Camera Installations
On 20/12/18, five SafeLock traps (known as R1-R5) were installed on an exposed alpine ridge in the upper Barlow River valley, at elevations ranging from 13201350 m asl (Figure 1). Due to the lack of trees at the site, each set-up incorporated a 1.8 m waratah with a timber extension, upon which the plastic trap platform and a white powder coated aluminium visual lure was attached (Figure 2).
The traps were installed approximately 30 m apart, descending down the ridge, and each trap was monitored with two Browning Dark Ops™ trail cameras – one set close to the trap for detailed recording of the trap area, and a second camera to observe the whole set-up (Figure 2). All cameras on the ridge were mounted on waratahs, enclosed in kea-resistant cases, and programmed to take 30 second videos upon triggering, with a 5 second delay between each video.
Bush Traps and Camera Installations
A further five sets of cameras and SafeLock traps (known as B1-B5) were installed on 08/01/19 on trees in the valley below the ridge traps. These were located approximately 400 m downhill from the ridge traps, at elevations ranging from 960-1000 m asl (Figure 1) and set up as described above, minus kea-resistant camera cases. We considered the limited exposure (comparative to the ridge sites) and our records to date of very low kea activity below the bush line would result in minimal camera damage.
Figure 2: SafeLock trap and camera set-up on the ridge above the Barlow River
Ensuring Kea Safety
Each SafeLock trap was set, then the lockout motor was turned on and activated to ‘lock out’ the trap. The mechanism for this involved a motor driven locking cam that moved forward to pull down on a yoke, forcing the treadle plate up and preventing the trap from being able to spring off (Figure 3). Once a trap was locked, the motor was turned off, ensuring that the motor could not cause the trap to become unlocked. As an additional measure, D-shackles glued with lock-tight were initially installed to cover the unlikely possibility that kea could somehow break the lock mechanism and expose themselves to a set trap.
Figure 3: Cross section showing components and mechanism of the ZIP SafeLock system, in ’locked’ position
Camera Servicing
Cameras were serviced at intervals dependent on weather and site access, but often enough to minimise footage loss through flat batteries or full SD cards. Services consisted of an SD card change, battery change (as required) and a visual inspection of the trap, platform and SafeLock mechanism for any signs of damage that could compromise the trap functionality or the safety of kea.
Data Collection and Curation
Kea interactions were collated with a focus on interaction timing, duration and an assessment of whether or not the trap would have been triggered under the proposed ZIP SafeLock regime (daytime lockout). Camera settings were not altered to account for daylight savings, so footage collected after 07/04/2019 was adjusted during data curation. Trigger events were defined as an interaction where the trap may have been sprung off if the SafeLock was not activated i.e. if the trap was a standard leghold trap without modification. Trigger events were classified into ‘day triggers’ and ‘night triggers’. Trigger events were recorded at a maximum of one per bird per interaction. Background kea numbers were counted where visible, to indicate numbers that were present, but not interacting with the traps. Damage caused, demographics and any other unusual behaviour were also recorded where possible.
The trial was concluded on 19/07/19, after running for almost seven months.
RESULTS
Activity Summary
Over 1838 trap nights, 480 records of kea were made (Table 1). With the exception of a single ‘day trigger’ interaction on one of the bush traps (Table 2), all of these occurred at the five ridge traps, with 277 records classed as direct interactions with the ridge trap set-ups (Table 3).
Table 1: Summary of kea records on all traps between 21/12/18 and 19/07/19
Table 2: Bush trap interactions broken into 4 categories (total and per trap night)
NB: Due to the lack of activity on the bush traps (while a significant finding in itself) further results and activity discussion refer only to kea activity that occurred on the ridge traps.
Table 3: Ridge trap interactions broken into 4 categories (total and per trap night)
*In the real-life scenario where SafeLock traps would be livened across the block, these night triggers would be unsafe for kea. During the trial, virtual triggers at night were safe for kea due to all traps being locked down at all times.
Kea Activity Timing
From December 2018 through to the end of April 2019, only 34 incidences of kea activity were recorded at the SafeLock traps and all fell within the hours of sunrise and sunset (Figure 7). During May 2019, there was a noticeable increase in activity (90 records), and kea began to interact within the hour before sunrise and after sunset. This trend further increased in June 2019 (141 records), including interactions that occurred throughout the night.
Kea interactions between 21 December 2018 and 22 April 2019 were limited to less than five minutes in duration, with the exception of one 10-minute interaction on 2 January (Figure 8). From 24 April 2019, interaction duration increased significantly, peaking at 58 minutes on 3 June 2019. There was no minimum interaction duration before a trigger event occurred – these frequently happened within seconds of a kea appearing at a trap (Figure 8).
Trap Interference and Night Activity
While kea visited the traps multiple times between 21 December 2018 and 22 April 2019, their interactions were largely restricted to landing on the trap or platform, or chewing the ramp or visual lure. The first record of kea accessing and playing with the trap chain occurred on 23 April 2019, after which a further 40 interactions described kea accessing, playing with or attempting to eat the chain and bungee cord. Through this behaviour, kea were able to compromise the traps, rendering them ineffective at capturing possums (Figure 4).
Figure 4: Lockout traps that were rendered ineffective by kea during the trial by either pulling the chain or bungee out between the jaws, or flipping one jaw up in order to fiddle with the bungee.
We recorded a total of 21 interactions that fell between an hour after sunset and an hour before sunrise, when the traps were proposed to be open. Of these, six were within an hour of our proposed open/close times. The first activity in the deep night was recorded on 6 June 2019 at 03:51, a 24-minute sequence of a single bird playing with the bungee cord. Further night activity ranged in duration from less than 30 seconds to 41 minutes, with the majority of activity focused on the bungee (Figure 5). This activity resulted in 12 night trigger events (where the trap may have been triggered in a real-life scenario).
Figure 5: A kea at trap R1 chewing on the bungee after digging it out of the snow on the trap platform. Note – camera settings were not adjusted to account for daylight savings, so the actual time was one hour earlier than shown here.
Kea Activity by Life Stage
Identifying kea sex and life stage was difficult, as black and white footage due to low light conditions diminished our ability to see the typical yellow beak and eye of the non-adults. Despite this, it appears that non-adults accounted for almost twice as many triggers as adults, and approximately twice as many non-trigger interactions (Figure 6). Aside from one record in late December 2018, non-adults were not recorded visiting traps until early February 2019 (the advent of fledging season), after which the majority of classified visits were made by non-adults.
Figure 6: Kea interactions with SafeLock traps by life stage (ridge only, n=277)
Dec Jan Feb Mar Apr May Jun Jul
Figure 7: Daily kea activity at SafeLock traps relative to sunrise, sunset and the originally proposed open and close times (one hour after sunset and one hour before sunrise, respectively). Red dots indicate interactions that fell within this time period, when traps would be open in a real-life scenario (n=21). The sharp kink in the linear time variables is due to the end of daylight savings, and blue shaded areas denote successive months (n=8).
Figure 8: Duration and level of kea activity over time. Red triangles (n=12) indicate virtual trigger events during night hours (total virtual triggers at the ridge traps = 215)
DISCUSSION - PERTH RIVER VALLEY FIELD TRIALS
The sudden increase in kea activity from April 2019, after several months of low interaction levels with the SafeLock traps, is likely to be linked with the end of the kea fledgling season. Young birds are known to be extremely curious as they learn about the environment around them for the first time. This increase in non-adult kea activity coincided with the kea discovering the chain and bungee components of the SafeLock traps. We believe that this discovery was the catalyst for a further, rapid increase in activity. The kea learned that there was a reward available through playing with the rubber bungee, and the removal of the chain was likely done to help facilitate this. We believe that these factors resulted in an atypical expansion of the kea activity window at the SafeLock traps from daylight hours to predawn, and ultimately at random times during the night, when they could be exposed to potential harm.
Pleasingly, there was very little interaction with the SafeLock traps within the bush. This is despite them being in place over the same time period, including the kea fledgling season. We suspect that the added cover that the bush provides for the traps meant that the kea did not find them to then interact with the traps; as opposed to the ridge traps that were very exposed and visible. It is possible that traps within the forest would not suffer the same interaction and disabling rate, and therefore could be left in place on a permanent basis (as originally proposed with the lean leghold trap network to target invading possums).
In saying that, the context where the goal is to detect and capture the individual possum invader, requires having the ability to target all locations where those invaders may be present. In the case of possums at the Perth River valley site, this is the headwaters of the river boundaries – areas that are typically less vegetated and exposed. The level of kea interference seen on the ridge leading to the disarming of the SafeLock traps was not acceptable. Traps that had their chains and bungees pulled out, or front jaws flipped up, were rendered ineffective. In the ‘live’ scenario, these traps would not have been able to catch invading possums. In fact, they would likely have served to create a trap shy possum, where the trap springs off ineffectually, and the possum is frightened or partially caught and escapes.
The above results led us to design an additional part to this trial, outlined below.
PART II – CAN WE MITIGATE THE EFFECTS OF KEA INTERFERENCE?
INTRODUCTION/BACKGROUND
The Perth River valley field trial revealed that kea could (and would), once trained, disable a SafeLock trap so that it was rendered ineffective at catching possums during night hours. This was a key finding, and an important problem to solve if SafeLock traps were to be permanently deployed in the Perth River valley. To ensure the functionality of the trap, we required a solution to protect the chain and bungee from kea access, removing the reward provided by playing with moving parts.
In designing the solution, we required a setting where we could test, measure and respond rapidly as results occurred. ZIP has worked with Willowbank Wildlife Reserve (WWR) in the past, and they kindly granted us access again to their 12 captive kea to trial our solution. These captive birds, in a setting of reduced day to day stimulation (relative to wild kea) provided a particularly hard test.
METHODS - WILLOWBANK WILDLIFE RESERVE
We exposed the captive kea to three trial phases, to isolate any unexpected behavioural differences between the captive kea and the wild kea, and ensure that the test was robust.
Phase I: 16 – 18 July 2019
It was important to confirm that the captive kea would interfere with a bungee and chain in the first instance.
Two plastic ZIP leghold platforms were installed on posts in the kea enclosure at WWR. Both ends of a single piece of chain (without a trap attached) were screwed into the chain bucket of each platform. Each chain had two bungee cords threaded through and knotted on. This set-up was left for two days. A Browning trail camera (set to record 30 second videos) focused on each platform to confirm the kea were interacting as expected.
Phase II: 18 – 19 July 2019
The next step was to confirm that when a locked out SafeLock trap was presented to the captive kea, behaviour would continue to mirror that of the wild kea (i.e. trap functionality interference).
The chains were removed and replaced with a set, locked out, SafeLock trap with a single bungee, as per the field set-up in the Perth River valley. D-shackles glued with lock-tight were used as an additional security measure to ensure no kea was harmed (Figure 9). Set-ups were checked regularly over a two-day period, and if the bungees were missing or destroyed, they were replaced. Both platforms were monitored by camera to ensure the expected behaviour was occurring.
Figure 9: (left) ZIP engineer ensures a trap is locked open under the watchful eye of a resident kea; (right) an adult kea interacting with a chain and bungee during Phase I
Phase IIIA: 30 July – 02 Aug 2019
The final step was to test whether our solution could inhibit the interference behaviour.
The same SafeLock trap set up as Phase II was in place, with the addition of a custom fitted stainless steel plate into the platform between the chain bucket and the trap itself (Figure 10). This plate was designed to limit access to the chain and bungee, while permitting free run of the chain as would be required in live capture settings.
Phase IIIB: 05 – 09 August 2019
The SafeLock traps were set as for Phase IIIA. An alternative protection plate was inserted into the space between the trap and the chain bucket. This plate incorporated slightly more protection around the chain (Figure 11). This was trialled for four days.
RESULTS
Phase I
All four bungees were removed by the kea within hours on 16 July. Bungees were replaced on 17 July, some with double knots, and these were also removed by the kea within hours.
Phase II
Kea were able to pull the chains and bungees out from beneath the set, locked SafeLock traps within hours of install. This confirmed that their behaviour would provide a hard test for our protection plate designs.
Phase IIIA
The first steel plate had little effect on restricting kea access to the chain and bungee. Kea were able to hook their beaks into the corner gap and chain hole in the steel plate to access the bungee, and did this within hours of install (Figure 10). The traps were reset with the bungee buried beneath the chain, but again the kea were successful.
Phase IIIB
The second steel plate also had little effect on restricting kea access to the chain and bungee (Figure 11).
Figure 10: (left) stainless steel plate V1; (right) results of kea interaction
Figure 11: (left) stainless steel plate V2; (right) results of kea interaction
DISCUSSION - CAPTIVE KEA TRIALS
We deliberately chose to trial the protection plates in an environment that would provide a hard test, as any results rapidly discovered here were likely to occur, eventually, in the wild. Given the results of Phase IIIA and IIIB, we decided that further prototypes of this nature would be ineffective. It is likely that the trap would require significant modification so that the front jaw could not be flipped up, or the whole platform to be redesigned with a deeper chain bucket and in-built protection plate. Retooling, testing and production would likely take months to complete, with no full guarantees of effectiveness against the highly intelligent kea. We opted not to attempt this redesign. Instead, we adjusted our operational approach for managing possum invasion (to be initiated on a localised scale only once a possum was detected).
CONCLUSIONS
From the data gathered at the Perth River valley and at WWR, we concluded:
The ZIP SafeLock system works – despite 216 virtual trigger events in the field (and multiple at WWR), no traps were able to be triggered by kea while locked out.
Kea interference with the bungee and chain components of the traps regularly rendered them ineffective for catching possums (if used in a real world setting).
Kea can learn new behaviour, and sometimes this will occur over a long period of time – it took four months for kea in the Perth River valley to begin fiddling with the bungee and chain components of the ZIP SafeLock trap. This learning has also been seen with DOC200 traps in the Murchison Mountains. Between 2002 and 2007, no traps appeared to be set off by kea using sticks – they would instead roll the traps to set them off. However, between 2008 and 2017, kea eventually learned to set traps off using sticks they had whittled for purpose (Goodman et al 2018). Therefore, it is necessary for any further behavioural trials to run over a period of several months in order to truly test whether kea will develop undesirable behaviours.
The learned behaviour appeared to coincide with juvenile fledging season, which also overlaps the end of sub-adult possum dispersal season in the Perth River valley (December to April)
The lockout leghold traps installed below the bush line appear to be less obvious to kea than those on the ridge. Of the five bush traps in this trial, only a single kea interaction was recorded (an adult) over 895 trap nights. Of the 205 unset SafeLock traps awaiting full installation in the 12,000 ha Perth River valley site (almost all below the bush line) only two have been found with the chain pulled out and bungee chewed by kea
While there were night interactions on the exposed SafeLock traps on the ridge, these occurred at a rate of 0.01 interactions per trap night, an extremely low rate which we think was largely influenced by the learned attraction to the bungee and chain components.
RECOMMENDATIONS
Based on the conclusions described above, we initiated the following actions in the Perth River valley field site:
Minimise opportunities for kea to learn to play with the bungee and chain components of the SafeLock trap by moving away from a permanently installed network, minimising their exposure to traps in the site
Use trail cameras to detect invading possums, and respond to detections with targeted, temporary infrastructure deployment. There are two options to achieve this:
Use a small number of temporary SafeLock traps to respond to any possum detections below the bush line, on a case-by-case basis. These will be removed after the possum is caught.
Use automated-reporting cage traps, and/or a possum dog to respond to any possum detections in alpine or exposed areas
Between May and June 2020, we successfully employed the recommendations described above to remove three possums from the block.
ACKNOWLEDGEMENTS
We would like to thank Wayne Costello (Department of Conservation) for permission to carry out this work; Ken Hutchins and the team from Fox/Franz Heli Services for enabling access to the Perth River valley trial site; and Nick Ackroyd and staff at Willowbank Wildlife Reserve for their support and use of the captive kea for behavioural trials. Thanks to the Perth River valley field team for their field work and dedicated footage processing throughout this trial, Maggie Nichols for her assistance during the captive trials, and John Wilks for his adaptability in coming up with rapid prototypes of protection plates. Thanks to James Ross for reviewing this report.
REFERENCES
Bastos AP, Taylor AH. 2020. Kea show three signatures of domain-general statistical inference. Nature Communications, 11(1), 1-8
GOODBYE, POSSUMS! A lean network of traps prevents possums re-populating
Goodman M, Hayward T, Hunt GR. 2018. Habitual tool use innovated by free-living New Zealand kea. Scientific Reports, 8(1) 1-12.
Heaney M, Gray RD, Taylor AH. 2017. Kea perform similarly to chimpanzees and elephants when solving collaborative tasks. PloS one 12, e0169799.
Huber L, Gajdon GK. 2006. Technical intelligence in animals: the kea model. Animal cognition 9, 295-305.
Margetts BI, Ross JG, Buckley HL. 2020. Measuring Home‐Range Changes Following Density Reduction of Australian Brushtail Possum. The Journal of Wildlife Management, 84(1), 185-192.
Sweetapple P, Nugent G. 2009. Possum demographics and distribution after reduction to near–zero density. New Zealand Journal of Zoology, 36(4), 461-471.