Review of existing Red Fox, Feral Cat, Feral Rabbit, Feral Pig and Feral Goat control in Australia. II. Information Gaps

Final report
Ben Reddiex, David M. Forsyth.
Department of the Environment and Heritage, 2004

7. Results and discussion

The first stage of this review (audit of existing pest animal control programs in Australia; Reddiex et al. 2004), found that there was little reliable knowledge about the benefits of pest animal control in Australia. Few control programs monitored changes in the pest species targeted for control and the native species of interest. In addition, monitoring designs rarely included non-treatment areas or were randomly allocated, and few had assessed the species/habitat of interest prior to control.

Experimentation is required to gain reliable knowledge: that is, the proposed control should (where possible) have replicated treatment and non-treatment areas, have suitable monitoring designs for both pests and resources, randomly assigned treatment and non-treatment areas, and be undertaken over an appropriate scale and duration for both the pest animal and native species of interest. Failure by pest control programs to meet the basic experimental requirements makes it impossible to separate an observed response in native species/habitat following pest control from climatic factors and/or a suite of potentially threatening processes, namely; habitat change and degradation, impact of introduced animals and plants, disease, exploitation, and climate change. We acknowledge that few experiments/control programs have met all of the requirements for reliable knowledge due to the fact that large-scale manipulative experiments are difficult to implement in the field, and require long-term support and significant financial investment from management agencies. In nearly all cases the most expensive component of an appropriately designed experiment is monitoring.

The following section outlines proposed experiments that enable the benefits of pest animal control to be investigated for feral goats, feral pigs, feral rabbits, and foxes. We strongly recommend that the monitoring design is based on an underlying modelling framework, thereby ensuring the correct information is collected for system models (see Robley et al. 2004). The development of dynamic system models allow the effects of changing parts of a system (e.g., pest animal control) to be predicted. Such a model has been developed for the interactions between foxes, cats, feral rabbit and native prey species in Australia (see Robley et al. 2004). If experiments are designed with an underlying modelling framework in place, they can be used to continually update system models and decrease the amount of time it takes to improve the reliability of management decisions.

This report does not provide exact locations of proposed experiments, and therefore the costs can only be estimated. Experiments that include varied levels of control are only proposed for feral rabbits. The recommended experiments for feral goats, feral pigs and foxes could be replicated with different frequencies or intensities of control. However, there are significant costs associated with this (see below). We believe the first priority is to identify whether benefits of pest animal control exist when control is maximised. If benefits do exist, then future studies can investigate the effects of various frequencies or intensities of control on those benefits.

7.1 Feral goats

7.1.1 Information gaps

The first stage of this review (Reddiex et al. 2004) concluded that there was little reliable knowledge about the benefits of feral goat control for native species and ecological communities. In contrast, there are reliable methods for estimating the absolute abundance of feral goats in open or semi-open habitats (i.e., aerial survey; Pople et al. 1998), and the effectiveness and costs of control are well-known (e.g., Table 2 in Environment Australia 1999a).

7.1.2 Native species for which there is limited information

There is very little information on the impacts of feral goats on, or benefits of feral goat control for, native species or ecological communities. There is some information from an exclosure study in the Flinders Ranges (Henzell 1991). It was shown that feral rabbits (rather than feral goats or euros Macropus robustus) were a critical factor in determining mulga (Acacia aneura) regeneration because they killed nearly all of the seedlings.

The eleven plant, one invertebrate and eight fauna listed threatened species under the EPBC Act for which feral goats are a known or perceived threat appear to have that status because goats either have been observed feeding on those species, or because browse on those species has been attributed to feral goats, or because feral goats compete with native herbivores for resources or cause land degradation. Hence, there is extremely limited information for all of the listed threatened native species (EPBC Act) for which feral goats are either a known or perceived threat.

7.1.3 Priorities for filling information gaps

We consider that the greatest priority is understanding the benefits of feral goat control for native plant species. Aerial survey (e.g., Pople et al. 1998) appears to be an adequate method for estimating the abundance of feral goats in most habitats except gorges and forests with a closed canopy.

7.1.4 Recommended experimental design

We advocate an experiment that assesses the benefits of feral goat control for a combination of listed threatened species (EPBC Act) for which feral goats are known threat and common native species. Our preferred experimental design (outlined in Figure 2) first involves identifying as many (minimum of 5; see below) potential feral goat control programs around Australia. Each potential control program must contain native plant species that are predicted to respond to feral goat control (either in abundance and/or condition). Our preferred option is that both EPBC Act listed threatened species and common plant species are present in each potential feral goat control program. The more programs that can be included in the experiment the more reliable the inferences will be. We suggest five is the minimum number of feral goat control programs that should be included.

Each feral goat control program is then divided into pairs of potential control areas, with each pair of areas being as similar as possible in terms of vegetation composition and structure, and soil types; this is usually achieved by selecting adjacent areas. The most reliable inferences will come from the most pairs of areas. However, as long as there are at least five control programs there can be a minimum of one pair in each control program. The paired areas are then randomly assigned as treatment or non-treatment. Within all paired areas (i.e., both treatment and non-treatment) at least 20 similar pairs of sites will then be selected (Figure 2). These pairs of sites should be selected so that they each include the plant species predicted to increase in abundance following pest animal control, and at least half of the pairs (but as many as possible) should include the EPBC Act listed threatened species predicted to respond to control. These sites should be a minimum of 10´10 m and a maximum of 25´25 m. One of each pair of sites is randomly assigned an exclosure. The purpose of the exclosure is to exclude feral goats, and several fence designs achieve this (review in Long and Robley 2004). If it is believed that feral rabbits are also affecting the species of interest, then there are two options. Our favoured option is to select trios of similar sites rather than pairs of sites, and a goat exclosure is randomly assigned to one of the trio, a goat+rabbit exclosure to another, and the remaining site is open to both goat and rabbit herbivory.

At least 12 months prior to feral goat control beginning, the plant species of interest are sampled in all of the pairs (or trios) of sites. Because monitoring protocols have not been developed (or at least, published) for most plant species of interest, we do not attempt to prescribe methods for monitoring changes in the abundance and condition of native species; rather, these must be developed as appropriate on a case-by-case basis. However, it is important to think about the life-history of each plant and how the feral goats (and possibly feral rabbits) might be affecting the population dynamics of the plant. Monitoring of plant abundance/condition should be conducted at least annually.

The feral goat control should aim to achieve residual densities that are as low as possible in the treatment area, but should not affect goat density in the non-treatment area. This may mean that treatment and non-treatment areas are some distance apart. Common techniques for controlling feral goats are aerial and ground shooting, trapping, and aerial and ground mustering (see Reddiex et al. 2004).

The abundance of feral goats (and feral rabbits if thought to be important) should be estimated in each area at least annually, and in the same month each year. We recommend aerial survey for feral goats (and kangaroos) and spotlight counts for feral rabbits, though the suitability of this technique will depend on the habitat at the treatment areas. Exclosures should be inspected at least every six months, depending on the potential for incursion (e.g., overhanging branches that might fall on the fences, or proximity to creeks that might erode the fence).

It will be important to measure other covariates at each pair or trio of sites. For example, rainfall is thought to be important for the germination of some seeds. Hence, a response to pest animal control might not occur until a threshold soil moisture has been exceeded. Other covariates might be the abundance of introduced mice, native herbivores (e.g., kangaroos), or domestic livestock.

How long should such an experiment run for? The answer will depend on the plant species monitored and the environmental conditions that occur during the work. And there is always the possibility of 'demonic intrusion' (e.g., the goats in the supposed non-treatment area are actually controlled) ruining even the best design. However, we believe that there should be at least 1 year of pre-control monitoring and at least 5 years of control before the experiment is reviewed.

Figure 2. The key elements of the experimental design for understanding the benefits of feral goat control for native species/communities.

Figure 2. The key elements of the experimental design for understanding the benefits of feral goat control for native species/communities.

The key parameter of interest is the mean difference in abundance (or condition) of plant species between the treatment area and the non-treatment area in the sites without exclosures (i.e., exposed to goat herbivory). This is the mean effect (or benefit) of feral goat control. The abundance (or condition) of plant species in the exclosures is the outcome if feral goats (and feral rabbits if that species was also excluded) had been eradicated from the area. (However, we note that eradication of feral goats is impossible for most of the feral goat range in mainland Australia; Parkes et al. 1996). One possibility is that there is no difference between the sites exposed to goats and the exclosures in both the treatment and non-treatment areas. If this was observed then we would conclude that there is no benefit to either controlling goats to low densities or eradicating goats.

Each pair of treatment and non-treatment sites contributes one data point to the final comparison. We estimate that at least five pairs of treatment and non-treatment sites are needed to provide a reasonable confidence interval around the benefits of feral goat control for native species.

One potential problem with assessing the benefits of feral goat control for native species/ecological communities is the ability to partition the benefits of feral goats from sympatric large herbivores (i.e., kangaroos and in some rangelands, domestic sheep). Forsyth and Parkes (2004) recommended that feral goats and kangaroos be incorporated into stocking rates in the rangelands. Feral goats should thus be considered one component of herbivory affecting native species/communities, and benefits to those species/communities may not accrue if feral goats - but not the other herbivores - are controlled to low densities. Thus, some programs may need to control kangaroos (and other herbivores) within the treatment area(s). In this situation the abundance of kangaroos should also be monitored (both kangaroos and feral goats can be monitored simultaneously with aerial survey; Clancy et al. 1997; Pople et al. 1998).

Until study sites are identified, the cost of this experiment can only be considered indicative. An indication of the cost of the experiment is shown in Table 2. Note that these costs are for one pair of areas, and that at least five pairs of areas (i.e., five control programs each with a minimum of one pair of areas) are recommended. Hence, the start-up cost of the experiment would be c. $325K, and the annual cost of running such a design $250K. The final-year costs are higher because of the need to analyse the data and publish the work.

Table 2. Indicative start-up and ongoing costs of an experiment assessing the benefits of feral goat control for biodiversity. The costs are for one pair (i.e., treatment and non-treatment) of areas (see Figure 2 for detail).
Item Start-up (year 1) costs ($000) Ongoing (year 2 and beyond) costs ($000) Final year costs ($000)
Labour1 $40 $20 $40
Materials $10 $0 $0
Transport $15 $10 $20
Feral goat control $0 $20 $0
TOTAL $65 $50 $60

1 Assumes 100% overheads, but not all organisations charge for these.