Native fauna response to large scale fox control in the northern jarrah forest of south-west Western Australia : Operation Foxglove

In Australia, and particularly Western Australia, increases in the abundance of several species of native fauna have been attributed to reductions in population density of the introduced predator, the red fox (Vulpes vulpes). Reductions in fox density have been achieved, or inferred, through the use of 1080 baiting (dried meat baits containing the toxin sodium monofluoroacetate). In most cases, the areas baited have been relatively small and the baiting regimes have incorporated a high intensity and/or a high frequency of baiting. In 1994, a large scale 1080 baiting program, Operation Foxglove, commenced within the northern jarrah forest of south-west Western Australia. The objectives of Operation Foxglove were to: implement fox control at a landscape scale to determine efficient and cost effective baiting regimes; determine the fox abundance/ density in this forest environment which would enable a sustained increase in native fauna abundance; and determine whether fox predation alone was limiting native fauna occurrence within the northern jarrah forest. The latter was deemed necessary as fox baiting programs, although not explicitly stated as such, are often seen to be the panacea for all fauna conservation issues. The research undertaken for this PhD program was the basis of Operation Foxglove. The study area was approximately 550,000ha and was aerially baited at different frequencies: 221,400ha baited twice yearly; 130,400ha baited four times yearly; and 88,600ha baited six times yearly. Vehicle based delivery of 1080 baits was also required to ensure the margins of the forest, i.e. the forest interface with agricultural land, were also baited. There was an unbaited treatment of 103,500ha. All baits were delivered at an intensity of five baits/km2. Some managerial and logistical constraints were imposed on the experimental design. These constraints were largely a function of conducting ecological research at a landscape scale within a production forest. Replication was limited to within-treatment replication (pseudoreplication) and the allocation of treatments was not random. To assess the faunal response to the different baiting frequencies, the predation sensitive potoroid marsupial, the brush-tailed bettong, or woylie (Bettongia penicillata), was used as an indicator species for predation sensitive, terrestrial, mammalian fauna. Four hundred and ninety two woylies were translocated from Dryandra Woodland to 19 sites within the three baited and the unbaited treatment groups. A large subset of the translocated population and a subset of recruits to the population (n=384, translocated individuals and recruits, combined) was fitted with movement sensitive ‘mortality’ radio-collars and survivorship was monitored almost daily for 37 consecutive months. The woylie survivorship data were analysed in an information-theoretic framework using the Known Fate model from Program MARK (White, 2001). The ΔAICc ‘best’, or preferred model, to describe the data included an effect from the baiting treatment groups (the frequency of baiting). The preferred model also included a temporal component, with survivorship reduced in winter. Model averaging of the best six models estimated woylie survivorship over the duration of the study to be 16.4% higher in the most frequently baited treatment group (six baitings per year) than in the treatment baited with the ‘standard operating procedure’ of four baitings per year. Baiting twice yearly conferred no benefit over not baiting at all. The fox and feral cat (Felis catus) were implicated as major predators of the monitored woylie population. The findings have significant implications for the department’s Western Shield Fauna Conservation and Recovery Program. A Bayesian modelling approach was adopted when analysing count data collected from monitoring 17 sandplot networks at which fox and cat spoor were detected. The methodology incorporated development of a site specific technique to derive abundance estimates for foxes. This technique was validated through a fox removal study and was subsequently further validated as part of research with the Invasive Animals Cooperative Research Centre’s Western Australian Demonstration Site. The latter validation incorporated use of molecular techniques, specifically, species and individual identification from DNA recovered from hair and scats. Fox abundance was lowest within the six baitings per year treatment group and increased in the four and two baitings per year treatment groups and the unbaited group. Although there was considerable overlap in the 95% equal-tailed Bayesian credible intervals (95%BCI), the reduction in fox abundance achieved within the six baitings per year treatment was interpreted as biologically significant for survivorship of the woylie. Estimating cat abundance was more problematic with little published data on cat abundance within forest areas and no reliable technique for estimating cat abundance generally. A Bayesian modelling approach was again adopted and the data from three consecutive years were analysed. In all years, the only model with any support modelled cat abundance without a treatment effect and without any explanatory variables, i.e. the estimate of cat abundance was constant across all treatment groups. The results were interpreted as reflecting the combination of the extremely sparse data and the lack of sensitivity of the sandplotting technique to detect cat presence. A revised technique is described and recommended. Occupancy modelling was used to identify the variables which best described the presence (occurrence/occupancy), within the northern jarrah forest, for each of two in situ mammalian species, the brushtail possum (Trichosurus vulpecula hypoleucus) and the southern brown bandicoot (Isoodon obesulus). The estimated probability of occurrence for the brushtail possum revealed the importance of wandoo (Eucalyptus wandoo and E. accedens) for its tree hollow bearing ability at the ‘pole’ size/age class. The presence of wandoo as a pole stand was interpreted as providing a very different resource from pole stands of jarrah (E. marginata) and marri (Corymbia calophylla). Wandoo is hollow bearing in this size class whereas jarrah and marri are not. The results did not support the hypothesis that the probability of brushtail possum occupancy increased with increased frequency of baiting. The estimated probability of occurrence for the southern brown bandicoot was highest in the four baitings per year treatment group and an increase in rainfall was positively associated with occupancy and detection. The effect from baiting on the southern brown bandicoot was equivocal. The conclusion, among others, was that baiting frequencies need to be in place for a longer period of time to enable the low densities and patchy distribution of native fauna to respond through naturally occurring recruitment. The results highlight the complexity of fauna interactions and the difficulty for conservation managers to ‘get it right’. There is no single prescription which will provide the requirements of all species. A 1080 baiting frequency of six baitings per year will benefit woylies, however, presence of bandicoots was maximised in the four baitings per year treatment group, however there was a trend of decreasing occurrence over time in this treatment group. Fox control appears beneficial for the woylie. However, it is hypothesised in some circumstances that it may lead to mesopredator release of cats. This may result in increased predation pressure on the woylie and other native fauna. This highlights the necessity of monitoring the native fauna response to baiting and, in particular, the need to monitor not only fox abundance, but the abundance of other predators (native and introduced) which may show indirect effects from any fox density reduction. Specific research and management recommendations are made in light of these findings with particular reference to assessing whether the feral cat and native predators show a mesopredator release response to fox density reduction