Stoats have been labelled “public enemy number one for New Zealand birds” by the Department of Conservation.
Detecting stoats when they are at low density is challenging. This finding describes the interim results of a project with the primary objective to detect stoat density in the Perth River valley, using trail cameras paired with automatic lure dispensers. A secondary objective was to determine the level of interaction of stoats with the lure dispenser, to help us to understand the potential of this device to dispense toxin.
The standard method to determine the relative abundance of stoats is to place tracking tunnels along survey lines throughout the area of interest, bait the tunnels with a piece of rabbit meat bait for a period of three fine days, and then check the tracking card/paper inside the tunnels to see whether any stoat footprints have been recorded (National Pest Control Agencies 2015).
However, the standard method generally does not detect stoats to the desired degree of sensitivity when they are at low density (Smith and Weston 2017). That’s because the three-day period of monitoring is not long enough to detect stoats; in part because animals are usually reluctant to interact with a device. Longer periods are generally not feasible because it is too expensive to refresh the bait, and because a longer period increases the risk of damage to the tracking paper as a result of wet weather.
The lack of sensitivity when stoats are at low density is a significant limitation where the managers of a programme to protect a native species that is very susceptible to stoat predation (e.g. kiwi chicks) need a robust estimate of the stoats present at a site. It is also a limitation where the goal of the operation is the complete removal of stoats, as in the case of the Remove and Protect model that Zero Invasive Predators Ltd (ZIP) is seeking to develop. In these situations, a more sensitive method is needed to detect stoats and assess the effectiveness of an operation.
The utility of tracking tunnels versus trail cameras has been studied by others (Smith and Weston 2017). Using cameras is generally considered to be a more sensitive method because cameras can be deployed for longer periods without the operational requirement to frequently change the bait, the memory cards are less vulnerable to the adverse effects of wet weather, and stoats do not have to interact with the camera.
ZIP has developed an automated lure dispenser (ALD). 95 ALDs were paired with Browning Dark Ops (6HD-940) trail cameras on a 700 x 500m array throughout the research area in the Perth River valley, at altitudes ranging from 250 to 1,240m above sea level. This density of network was to ensure there was at least 1 ALD/camera within a stoat home range (which can vary from 30-300ha). In total, the array of dispensers and cameras covered approximately 3,700ha of rata-kamahi forest and sub-alpine vegetation.
The ALD can deliver a small amount of liquid lure at a desired volume and frequency (for up to one year). Egg mayonnaise has long been known to be palatable to stoats (e.g. King et al. 2007). For this project we used “Best Foods” brand mayonnaise; we confirmed its palatability through pen trials at our Lincoln predator behaviour facility. Each ALD delivered 0.15ml of egg mayonnaise, three times a day, for a minimum period of three weeks during June and July 2018.
The cameras in this project were programmed to take a burst of three images whenever an animal was detected, with each image 0.3 seconds apart, and a 5 second delay between each burst. These settings were designed to maximise the length of time that the camera could run for before the SD memory card was full.
Of the 95 cameras, 41% (i.e. 39 cameras) detected stoats on at least one occasion, and 19% (i.e. 18 cameras) detected stoats on multiple days (i.e. 46% of the cameras that detected stoats).
In comparison, when we used the standard method over the same area two months earlier, only 1.7% of 60 tracking tunnels (i.e. 1 tunnel) detected a single stoat.
38% of the cameras that detected stoats (i.e. 15 cameras) also recorded stoats interacting with an ALD on at least one occasion. Stoats were recorded as interacting with an ALD on multiple days at 10% of the cameras (i.e. 4 cameras). A stoat was assessed to have “interacted” with the dispenser if the images showed it clearly eating the lure, or investigating the ALD, or changing its behaviour by moving towards the ALD rather than past it. Owing to the constraints of the camera settings, we expect that the level of interaction recorded is lower than the actual total.
It cost approximately $17,000 of field team time to install the network of 95 paired lure dispensers and cameras, and to subsequently remove them. This equates to approximately $5/ha over the course of the project.
Each image recorded by the cameras was reviewed by a team member, at a rate of c.2,500 images per hour. Less than 1% of the images were of the target – i.e. stoats! The majority of the other images were of possums (59%) and rats (20%), or were unable to be identified (17%). Clearly the number of non-targets can significantly influence the time required for data processing.
Other related expenses included establishing the survey lines the devices are located on, helicopter transport (approximately 5 hours), and the capital cost of the ALDs and cameras (i.e. approximately $120 and $220 excl. GST each respectively).
This project confirmed, unsurprisingly, that a comprehensive network of paired ALDs and cameras, deployed for three weeks, will detect more stoats than using a standard tracking tunnel method over three days. That said, it is possible that some stoats still evaded detection during this project!
We plan to do the following related work:
estimate the optimal period using the ALD/camera method to detect all resident stoats
model the results to estimate the minimum number of stoats present in the research area
·measure the effectiveness of aerially deployed 1080 at removing stoats via secondary kills
catch, tag and release marked stoats, to understand how long before they are detected and whether any stoats are likely to not interact with the camera network.
King CM, McDonald RM, Martin RD, Tempero GW and Holmes SJ 2007. A field experiment on selective baiting and bait preferences of pest mustelids (Mustela spp.). International Journal of Pest Management 53(3): 227-235.
National Pest Control Agencies November 2015. Pest Mustelids: Monitoring and Control. Best practice guideline A8. Wellington. 38p.
Smith DH and Weston KA 2017. Capturing the cryptic: a comparison of detection methods for stoats (Mustela erminea) in alpine habitats. Wildlife Research 44(5): 418-426.