Balancing act: evaluating non-target risks of feral cat management in conservation reserves

Study area

The main study area, Tutanning Nature Reserve (~2380 ha), is located 150 km south-east of Perth in the Wheatbelt region of WA and included both a control and treatment zone (Fig. 1). A second control zone was selected in Boyagin Nature Reserve (~4900 ha), situated 40 km to the east of Tutanning (Fig. 1). The area is characterised by Eucalyptus (powder-bark wandoo, brown mallet) and Allocasuarina and Acacia (rock sheoak and jam) woodlands interspersed with patches of kwongan heath and shrublands (Brown and Hopkins 1983). The reserves experience a Mediterranean climate with a mean annual rainfall of ~444 mm (Pingelly Weather Station; www.bom.gov.au; accessed 22 November 2023), and dominant winter rainfall between May and September. Mean long-term maximum and minimum temperatures are 32 and 5.5°C respectively (www.bom.gov.au). Commencing in the mid-1980s, both reserves have been regularly baited for foxes using a dried meat bait injected with 1080 (up to 2010) and then with Pro-bait (2011–2024), a 40 g dry sausage bait containing 3 mg of 1080 also developed and produced by DBCA, which is more inaccessible to small marsupials (Marlow et al. 2015a). Prior to our study, there was no active management of feral cats within either reserve. It is important to note here that our study was not designed to measure the efficacy of Eradicat® baiting on feral cats, or even necessarily to reduce their abundance. We were focused on the impact of Eradicat® baiting on the red-tailed phascogale.

Fig. 1.

Location of the study area in the agricultural zone of south-western Western Australia, showing Tutanning and Boyagin nature reserves.

Bait deployment

Eradicat® baits were deployed by hand to approximately half of Tutanning Nature Reserve at a rate of one bait every 100 m along existing vehicle tracks (i.e. treatment zone). In months when Eradicat® was not applied, Pro-bait continued to be delivered at a rate of one bait every 100 m. This resulted in two Eradicat® applications and 10 Pro-bait applications per annum in the treatment area. The remaining half of the reserve continued to receive its existing prescription of Pro-bait (one bait every 100 m) each month and acted as a control zone. A second control zone was established in the western half (~2819 ha) of Boyagin Nature Reserve. This reserve continued to receive Pro-bait at the same prescription as for the control zone in Tutanning.

Baits are typically delivered along ground transects by being thrown from a slow-moving vehicle; however, for these trials, we required that baits be placed in close proximity to the camera array. Prior to deployment, all Eradicat® baits were thawed and placed in direct sunlight to ‘sweat’ the oils and lipid soluble digest material to the surface of the bait. During this process, the baits were sprayed with an ant-deterrent compound (Coopex®) at a concentration of 12.5 g per litre to prevent bait degradation by ants.

Monitoring approach

Red-tailed phascogales were monitored before and after bait delivery by using Reconyx HP2X (Reconyx, Wisconsin, USA) camera traps mounted on 4 cm wide by 0.8 m long wooden platforms attached to tree trunks by right-angle metal brackets approximately 1.6 m above ground level. Camera traps were programmed to high motion-sensor sensitivity, with three rapid fire images taken per trigger event, no delay between trigger events and fast shutter for night-time exposures. Because red-tailed phascogales are primarily nocturnal, camera traps were programmed to trigger only between 1800 hours and 0600 hours. An olfactory lure (PVC container containing peanut butter and oats) was placed in front of each camera trap (except for the last trial, see below). Thirty camera traps at a minimum spacing of 500 m were established in each of the control zones, and within the Eradicat® treatment zone (Supplementary Figs S1, S2), and data collated for a minimum of 16 nights before, and again after, the baiting application. A bait was initially placed within 20 m of the tree with the mounted camera for the first trial, but was subsequently changed to within 5 m for the remaining trials to increase the chance that red-tailed phascogales active on monitored trees would encounter them. All images from the camera traps were uploaded into the program CPW Photo Warehouse (Ivan and Newkirk 2016) for processing. Date- and time-stamp information from each image was captured. For analysis, we used only the first camera trap detection within a 24-h period as an independent detection event, defined as from midday to midday, unless multiple individuals were present in a single image, in which case the number of individuals in that image was used. To investigate seasonal differences, trials were undertaken in autumn and spring 2021, and autumn and spring 2022.

Because lures may influence behaviour and this can vary over time, and given that earlier trials had shown that red-tailed phascogales remained active on platforms even without lures, we used Trial 4 to investigate this aspect further. We compared mean detection rate (per 100 trap-nights) for this trial with that of Trial 2 (i.e. same sampling season) at both Tutanning (mean ± s.e., 9.2 ± 1.3 lured, 8.8 ± 1.5 un-lured) and Boyagin (mean ± s.e., 10.6 ± 2.0 lured, 8.3 ± 1.4 un-lured) by using a two-tailed paired Student’s t-test, with no significant difference observed at either site (Tutanning: t-value = 0.33, P = 0.74; Boyagin: t-value = 0.94, P = 0.35).

Eradicat® uptake

To better understand animal interactions with Eradicat® baits, especially those of red-tailed phascogales, and the longevity of baits in the environment, a ground-based camera trap (Reconyx HP2X) mounted on a plastic stake at a height of 40 cm was placed 1.5 m from an Eradicat® bait at each of the 30 monitoring sites in the treatment zone during autumn and spring 2021 (Trials 1 and 2) and autumn 2022 (Trial 3) baiting events. These cameras used the same settings as did the tree mounted cameras, except that they were active both day and night. This resulted in the monitoring of 90 individual Eradicat® baits. At the monitoring sites where the bait was removed, we examined the number of detections on the tree-mounted camera traps for 3 weeks before and after bait removal, to identify the possibility of a negative impact at those sites by using a two-tailed paired Student’s t-test.

Statistical analysis

Changes in red-tailed phascogale activity were modelled using a generalised linear mixed model (GLMM) with a Poisson distribution for each reserve (Tutanning or Boyagin) and each trial separately. The response variable was the total number of independent detections of red-tailed phascogales recorded on each camera trap. For the analysis, only camera detection data from 21 nights pre- and post-baiting were used for all trials and in all zones, but only 16 nights pre- and post-baiting for Trial 4. Camera trap ID was modelled as a random intercept and fixed factors included zone (Tutanning only: control or treatment), session (pre- or post-baiting) and the interaction term (Tutanning only: zone × session). All analyses were performed using R Statistical Software (v4.2.2; R Core Team 2022). GLMMs were fitted using the ‘lme4’ package (v1.1-30; https://cran.r-project.org/web/packages/lme4/index.html; Bates et al. 2015). Post-hoc tests were used to examine pairwise combinations of all variables in the R package ‘emmeans’ (V1.8.0; https://cran.r-project.org/web/packages/emmeans/index.html; Lenth et al. 2020).

Additionally, occupancy from camera trap data was examined for each reserve and each trial, separately, to determine if there was a decrease in occupancy following baiting. Single-season occupancy analysis, without covariates, was employed to assess changes in peak occupancy and probability of detection by only using the seven consecutive day period with the highest occupancy before and after Eradicat® baiting. While we initially examined data for both autumn and spring, a large decrease in detections in spring, and consequently, low overall detection probability, resulted in 95% confidence intervals for occupancy which were too large to make any meaningful interpretations from these data (ranging from close to 0–1; Table S4). Subsequently, we restricted discussion of the occupancy analysis to autumn trials only (Trials 1 and 3).

Ethics statement

The study was approved by the DBCA Animal Ethics Committee (AEC 2020/24A).

Changes in detection rate

At Tutanning, the only significant treatment by session interaction observed was in Trial 1 (Fig. 2, Supplementary Table S1). Post hoc tests showed a significant increase in red-tailed phascogale detections in the control zone during the post-baiting period (where Eradicat® baits were not deployed). However, there was no significant difference pre- and post-baiting in the treatment zone (Table S2). The only significant decline in detections in the treatment zone (following baiting) was in Trial 2, but a decline was also observed in the control zone, the treatment by session interaction was not significant (Table S2), and very low detection rates were observed during this trial (Fig. 2). In the Boyagin control zone (again where Eradicat® baits were not deployed), a significant decline in red-tailed phascogale detections was also observed in Trial 2 during the post-baiting period (Table S1, Fig. 2). Post hoc tests similarly showed a significant decline in detections in Trial 3 in the Boyagin control zone during this period (Table S2).

Fig. 2.

Mean detection rate (per 100 camera trap nights per camera trap site) at (a) the Tutanning treatment and (b) control zones, and (c) Boyagin control zone, prior to (before_bait) and after (post_bait) Eradicat® baiting across four trials (1: autumn 2021; 2: spring 2021; 3: autumn 2022; 4: spring 2022). Error bars represent standard error.

Changes in occupancy

Peak occupancy remained high and stable in all zones and across the two autumn trials (Trials 1 and 3), with slight increases in the Tutanning treatment zone following baiting, and slight decreases or no differences in the two control zones, and considerable overlap of the 95% confidence intervals (Fig. 3, Table S3).

Fig. 3.

Occupancy and detection probabilities in (a) the Tutanning treatment and (b) control zones, and (c) Boyagin control zone, prior to (before_bait) and after (post_bait) Eradicat® baiting across two trials (1: Autumn 2021; 3: Autumn 2022). Error bars represent 95% confidence intervals.

Eradicat® bait uptake

Brush-tailed possums (koomal; Trichosurus vulpecula) were responsible for the majority of Eradicat® bait uptake (37%; 33 baits) with red-tailed phascogales removing 8% (7 baits) (Table 2). After 28 days post-baiting, the maximum length of time camera traps were deployed, the proportion of baits not consumed was 10% (nine baits). These baits had become hard and dry and were likely to be unpalatable and inaccessible to red-tailed phascogales. Baits that were removed or consumed lasted on the ground for an average of 2.8 ± 0.3 days. There were interactions at seven sites where red-tailed phascogales removed the bait from the view of the camera, but none was observed to eat the bait. There were a further eight sites at which red-tailed phascogales either investigated or passed baits but had no other interaction. Where baits were removed, a comparison of the pre-bait (mean ± s.e. = 3.9 ± 1.5) and post-bait (mean ± s.e. = 3.7 ± 1.6) red-tailed phascogale detections on the tree-mounted camera traps showed no significant difference (t-value = 0.28, P = 0.79).

Arboreal camera-monitoring

In our assessment, the first challenge was to devise a monitoring approach that would provide enough detections of red-tailed phascogales to undertake a robust analysis based on a before–after–control–impact (BACI) design. Red-tailed phascogales can be difficult to trap, and often require a period of trap conditioning whereby the trap mechanism is disabled and traps are pre-baited before attempts to capture them (e.g. Pierson et al. 2023). While red-tailed phascogales are reported to forage on the ground, they are largely arboreal (Short et al. 2011). Our alternative approach of using tree-mounted camera traps proved to be highly effective and efficient at detecting the species, in some cases, up to three individuals were concurrently detected on the same camera, as well as it being a low-impact method from an animal-welfare perspective. Arboreal camera trapping is a recent approach that has shown to be effective elsewhere for monitoring other tree-dwelling species (Moore et al. 2021). A further advantage of using camera traps is that they allow assessments over much larger scales both temporally and spatially than does live-trapping.

The use of a lure is likely to have increased our ability to detect red-tailed phascogales; however, lures can confound results if the effectiveness of the lure diminishes over time (Barcelos et al. 2023). This potential bias can be managed by restricting the duration of the exposure to the lure during each monitoring period and being consistent across treatments. Our examination of lured versus un-lured cameras showed that the mean detection rate did not differ between the spring 2021 (lured) and spring 2022 (un-lured) trials. As such, over extended periods, the detection rate of red-tailed phascogales without the use of a lure may be sufficient to monitor trends, but this requires further validation, given the low number of detections during the spring trials. In terms of maximising detections, results also showed that the optimal time for monitoring red-tailed phascogales using camera traps is during autumn, prior to male die-off.

The disadvantage of relying on camera traps for monitoring trends of red-tailed phascogales is that changes in abundance cannot be readily determined. However, cameras can identify declines in red-tailed phascogale activity over time and trigger the need to undertake more intensive investigations should a decline be observed. In further studies, another approach that could be trialled is the use of remote-logging microchip readers (e.g. Ross et al. 2022). Because red-tailed phascogales readily use the camera trap platform attached to trees (in this case, using lures would be advantageous), a microchip reading antenna could be mounted to these platforms to capture data from microchipped individuals. The only live-trapping effort required would be to microchip as many individuals as possible prior to baiting, and all subsequent monitoring would occur remotely, allowing continuous data capture over extended timeframes.

Susceptibility of red-tailed phascogales to Eradicat®

Our results supported the conclusion of Friend et al. (2020) that the risk posed to red-tailed phascogale populations from the use of Eradicat® baits in areas where the species occurs is likely to be low. Although there is a chance that some individuals may consume enough bait to receive a lethal dose, our results suggested that this may not translate to population-level impacts (e.g. Fancourt et al. 2021). Moreover, the benefit of reducing the impact of predation by feral cats on highly threatened and susceptible fauna populations is likely to outweigh the possible impact on a few individual red-tailed phascogales (Glen et al. 2007). Our study attempted to maximise exposure of red-tailed phascogales to Eradicat® by delivering baits near trees where animals were known to be present; yet, we could not detect changes attributable to the bait application. Operational baiting such as the aerial application of Eradicat® used by the Department of Biodiversity, Conservation and Attraction’s Western Shield Program would be likely to provide less overall exposure to baits than was the case in our study. Nevertheless, a precautionary approach should be taken as population-level impacts may be difficult to determine (Glen et al. 2007). This is particularly the case if the baiting prescription is more intensive, i.e. there is a higher frequency and density of baits in the landscape (than was used in our study) to improve the likelihood of feral cats encountering baits. Long-term monitoring that captures a range of seasonal conditions, and corresponding fluctuations in population size of red-tailed phascogales, including drought periods when natural food resources are likely to be scarce, is likely to be more informative. Although poison-baiting programs may be more effective during dry conditions when feral cats are hungry, non-target species are equally likely to be more vulnerable to toxic ingestion.

Bait uptake by non-target species

Examination of bait uptake by red-tailed phascogales supported the results of the arboreal camera-monitoring. Although we recorded a small number of bait removals by red-tailed phascogales, we were unable to detect a difference in tree-mounted camera detection rates at those sites where bait was removed before and after baiting (noting the small sample size). Friend et al. (2020) also documented baits being removed outside the field of view of cameras by red-tailed phascogales, but rhodamine B results indicated little evidence of baits being consumed. Also consistent with Friend et al. (2020), other non-target species were responsible for removing a large proportion of the baits, predominantly brush-tailed possums. Non-target 1080-bait uptake studies in WA have generally failed to demonstrate impacts to populations of native mammals (e.g. Dundas et al. 2014; Comer et al. 2020; Cowan et al. 2020; Bain et al. 2021; Doherty et al. 2022), which is probably not that surprising, given their naturally high tolerance to 1080, particularly in south-western WA (Twigg and King 1991). For example, the northern quoll, in theory, is at risk of lethal poisoning from the consumption of just a single Eradicat® bait. Using radio-collars, Cowan et al. (2020) examined the impact of Eradicat® on the northern quoll in the Pilbara region of WA. Their study demonstrated that there was no detectable risk to northern quolls from aerially delivered Eradicat® baits (applied at 50 per km²) and there were no sublethal effects on reproductive output. Palmer et al. (2021) further demonstrated that ongoing Eradicat® use had a positive effect on the northern quoll population, with evidence of range expansion in the baited area. Similarly, Moseby et al. (2021) reported that none of 37 collared chuditch (western quoll, Dasyurus geoffroii), a species that also occurs in the Wheatbelt region of WA, died as a result of consuming Eradicat® baits (also aerially applied at 50 per km²), despite their bait-uptake study showing that chuditch removed 26% of the baits.

It is also possible that only a small proportion of the bait (if any) is consumed, which may elicit a taste-aversion response, with animals exposed avoiding baits thereafter (Palmer et al. 2021). Learned avoidance of 1080-baits through a sublethal dose has been reported in introduced brush-tailed possums in New Zealand (Moss et al. 1998). However, removal of baits by non-target species is problematic because this reduces the number of baits available to the target species, and thereby may reduce the efficacy of the baiting program. Indeed, we recorded no baits being removed by feral cats, which is consistent with the results of Friend et al. (2020). This is an ongoing problem for introduced-predator control using baits in south-western WA (Dundas et al. 2014; Bain et al. 2021).