Department of the Environment

About us | Contact us | Publications

Header imagesHeader imagesHeader images

Publications archive - Biodiversity

Disclaimer

Key departmental publications, e.g. annual reports, budget papers and program guidelines are available in our online archive.

Much of the material listed on these archived web pages has been superseded, or served a particular purpose at a particular time. It may contain references to activities or policies that have no current application. Many archived documents may link to web pages that have moved or no longer exist, or may refer to other documents that are no longer available.

Review of existing Red Fox, Wild Dog, Feral Cat, Feral Rabbit, Feral Pig, and Feral Goat control in Australia. I. Audit

Ben Reddiex, David M. Forsyth, Eve McDonald-Madden, Luke D. Einoder, Peter A. Griffioen, Ryan R. Chick, and Alan J. Robley.
Department of the Environment and Heritage, 2004

8. Discussion

Reliability of knowledge about pest animal control in Australia

The field of wildlife management has been strongly criticised for its heavy reliance on information generated from descriptive studies (Romesburg 1981) and its slow adoption of experimental designs to examine the effects of management actions (Johnson 2002). Study areas - and the animals that are studied within that area - are rarely selected at random, and seldom are treatments applied independently (Hurlbert 1984). This review indicates that these criticisms apply equally well to our understanding of the consequences of pest animal control in Australia.

Of the 2095 control actions across the six targeted pest species that we reviewed which had indicated whether monitoring was being undertaken, only 3 incorporated treatment and non-treatment areas and pre- and post-control monitoring of both pest and resources (e.g., native species). We understand that few of these control actions had replicated sites, and that none had randomly assigned their treatment and non-treatment areas, although we note that randomisation is considered of lesser importance than replication (Walters and Green 1997; Johnson 2002). Hence, only 0.1% of the pest animal control actions have a design that can yield the most reliable inferences about pest animal control. The remainder of the control actions yield unreliable, or weak, inferences (Table 8.1). These conclusions mirror those of a recent review of the effectiveness of agri-environment schemes in conserving and promoting biodiversity in Europe (Kleijn and Sutherland 2004).

This review focused on public 'conservation' based organisations, and it is probable that control undertaken by the agricultural sector (i.e., agricultural based organisations or farmers) is over a larger area than that reviewed here. However, it is highly unlikely that any agricultural focused pest animal control would have an adequate experimental design to provide information on the benefits of control to native species or ecological communities.

These findings have implications for our understanding of the consequences of pest animal control in Australia. First, based on the survey results there is almost no reliable knowledge concerning the consequences of pest animal control, including its benefits or otherwise, to native species. There is also limited reliable knowledge concerning the consequences of pest animal control based on published studies (see below). This situation will only be remedied if agencies fund control programs that are designed to deliver reliable knowledge. (An associated problem is the collation, publication and/or retention of information, and we will return to this later). We therefore recommend that agencies fund work that can deliver reliable knowledge about the consequences of pest animal control (see Reddiex and Forsyth 2004 for recommended experiments to address knowledge gaps identified in this report). In other words, the proposed control must (where possible) have replicated treatment and non-treatment areas and have suitable monitoring designs for both pests and resources, and if possible the treatment and non-treatment should be randomly assigned. The importance of proper experimental and monitoring design cannot be emphasised enough.

We acknowledge that based on the available funding for pest animal control and research, and logistic constraints (e.g. no suitable non-treatment areas may exist) it is not possible to incorporate the above elements of experimental and monitoring design to all pest animal control programs. Therefore, we recommend focusing on a limited number of properly designed experiments (i.e., "evaluation" experiments). In the short term, these experiments will aid management decisions by enabling general extrapolation of the inferences gained from the evaluation sites to sites with less experimental rigour (i.e., sites that provide weaker inference). The utility of these experiments will increase with time as further evaluation sites are undertaken, and meta-analysis can then be undertaken on the key outcomes of the control operations (see below).

Table 8.1. The relationship between reliability of inferences and the design of pest animal control programs, and the percent of pest animal control actions surveyed for each level of inference.
Reliability of inferences
Design
Monitoring
Actions
  Treatment areas (number) Non-treatment areas (number) Resource Pest
Percent surveyed (any form of monitoring)
Percent surveyed (pre- and post-control monitoring only)
Least reliable 1 0 No No
66.0
90.4
1 0 No Yes
21.2
6.3
1 0 Yes No
4.7
1.2
1 0 Yes Yes
4.5
1.1
1 1 No Yes
0.7
0.4
1 1 Yes No
0.4
0.5
1 1 Yes Yes
2.3
0.1
1, random 1, random Yes Yes
0
0
≥2 ≥2 Yes Yes
0
0
Most Reliable ≥2, random ≥2, random Yes Yes
0
0

Second, our ability to address the second objective of this study is severely limited (i.e., review measures of the outcomes of differing levels or types of control activities in terms of reductions in pest species and recovery of native species/ecological communities). Very few of the operations attempted to monitor the abundance of pests and/or resources, and only a very small proportion of these made the data available to us.

We have not attempted to assess the quality of monitoring undertaken in the control actions reviewed. However, standard protocols for monitoring the efficacy of pest animal control operations and the resulting benefits to conservation resources are required. Of the 305 fox control actions where survey respondents said they monitored pest species, 75% of these used bait-take for their monitoring. The suitability of bait-take as a monitoring method is questionable, as bait-take is not independent of the control method, and changes in bait-take does not necessarily relate to the number of targeted animals killed (e.g., some baits are cached by foxes; Saunders et al. (1999), or not taken due to bait shyness or neophobia to baits; van Polanen Petel et al. (2001)). Therefore, this type of monitoring may provide little information on the actual success or otherwise of the baiting operation in terms of changes in pest animal abundance, and on assessing whether control actions benefit native species or ecological communities.

Our results (Figure 7.7) are primarily derived from control actions undertaken over the last 6 years. There are two possible explanations for the absence of information on control actions conducted =7 years ago. First, the number of control actions may have increased over time. Several survey participants indicated that this was true for their 'patch', especially as a result of increases in public land area that many of the organisations manage. However, it is unlikely that pest animal control has increased 8-25 fold over the period 1990 to 2003! Second, information that was available to be collected was often limited by a lack of 'institutional memory' of the participants we surveyed. Our interviews indicated that within most agencies, knowledge about pest animal control programs resided with the staff who managed the programs on a day-to-day basis. Hence, when these people moved on from that position and/or location, all knowledge about that program was lost. Until recently, none of the organisations surveyed had systems that could record pest animal control actions (i.e., where, when and how). Although Parks Victoria has recently developed a database for recording this information (data available post 1998), it does not contain a facility for collating monitoring information. Thus, despite the hundreds of millions of dollars that must have been spent on pest animal control in Australia, there is limited records of how and where most of that money was spent.

Maintaining institutional knowledge is difficult. The only solution that we can recommend is that all major pest animal control programs must report their actions and monitoring results in a way that can be accessed by others in the future. Independent of the way that information on control actions and monitoring is stored (e.g., computer spreadsheets or databases), it is important that the correct variables are collected and stored to enable evaluation of control action benefits.

Johnson (2002) has argued that entire studies should be replicated; so-called 'metareplication'. We agree with this, and argue that evaluation sites that specifically aim to address the benefits to native species and ecological communities from pest animal control should only be funded if they have an adequate experimental design and the data is made available through either publication of the study and/or by depositing the data with the funding agency with the understanding that it will be analysed by other researchers in the future.

The best example of metareplication in pest animal control is a recent analysis of factors affecting the kill rates of introduced brushtail possums (Trichosurus vulpecula) in New Zealand (Veltman and Pinder 2001). Possums in New Zealand are controlled using sodium monofluoroacetate ('1080') aerially broadcast in cereal baits. Veltman and Pinder (2001) collated information on 48 such poisoning operations conducted throughout New Zealand between 1994 and 1999. In each operation the kill rate was estimated from the difference in trap catch rates before and after the operation. Kill rates varied from 61-100% and were shown to be highest during months with the coldest average temperatures. Although Veltman and Pinder's study did not evaluate the conservation benefits of the control, it demonstrates the usefulness of the metareplication approach to understanding aspects of pest animal control. Veltman and Pinder's study was only possible because there was a national protocol for estimating changes in the abundance of possums (Warburton 2000) and operational staff recorded and stored their data adequately.

Trends in monitoring changes in pest and resource abundance/condition

More than 66% of all control actions did not monitor the abundance of the pest or resource (Table 8.1). Hence, the vast majority of pest animal control undertaken in Australia has no capacity for determining whether or not the resource has been protected/enhanced; we believe that, if a choice has to be made, it is more important to monitor the resource ('performance monitoring'; Choquenot et al. 1996) rather than the pest ('operational monitoring'; Choquenot et al. 1996). Interestingly, over four times as many actions monitored only the pest (21.2%) rather than only the resource (4.7%). Only 6.8% of control actions monitored both the resource and pest. The higher level of pest monitoring probably reflects the availability of well-publicised techniques for monitoring changes in the abundance of some of the pest animal species (c.f. resources). We also suspect that many planners are intrinsically more interested in estimating how many pests they have killed rather than the status of conservation resources.

Current knowledge of the benefits of pest animal control in Australia

A series of publications funded by the Bureau of Rural Sciences (BRS) reviewed the benefits of controlling feral pigs (Choquenot et al. 1996), wild dogs (Fleming et al. 2001), feral goats (Parkes et al. 1996), feral rabbits (Williams et al. 1995), and foxes (Saunders et al. 1995). Since we collated relatively little information on pest animal control prior to 1998 (see Figure 7.7), we briefly review the findings of those BRS publications, and other published studies, to determine our current knowledge of the benefits of pest animal control in Australia.

Foxes

Saunders et al. (1995:28) stated that "except for some detailed studies of fox predation on a limited range of Western Australian native mammals and… malleefowl, there is little quantitative information on the damage that foxes cause to native fauna". We consider those cited examples plus several more recent studies in detail.

Five populations of rock-wallabies (Petrogale lateralis) in the central wheatbelt region of Western Australia were studied by Kinnear et al. (1998). At two of the colonies ground-based 1080 baiting was conducted with the intention of killing foxes; dogs were not apparently present at any of the sites. The trends in rock-wallaby abundance at the sites that did and did not receive 1080 were markedly different; rock-wallabies apparently increased greatly at the two sites that received 1080 but either did not increase (1 site) or decreased (2 sites) at the sites that did not receive fox control. No data were presented on the financial costs of control. Caughley and Gunn (1996) criticised the design because the treatment sites (i.e., those that received 1080) and non-treatment sites were 'not interspersed'. Unless there was an a priori reason that spatial effects were likely, then a more appropriate criticism is that the treatment and non-treatment sites were not randomly assigned. We agree with Caughley and Gunn (1996: 259) that "only weak inferences can be drawn from this design". Hone (1999) criticised Kinnear et al.'s conclusions for different reasons. Hone argued that, since no data were presented on trends in fox abundance as a result of the fox baiting, it cannot be assumed that the mechanism for the increase in rock-wallabies at the baited sites was reduced fox predation. We agree, and emphasise the importance of proper experimental design and monitoring of both pests and resources.

Friend and Scanlon (1996) report on the effect of fox control on populations of red-tailed phascogale in the Western Australian wheatbelt. Trapping grids were established to monitor the abundance of red-tailed phascogale (Phascogale calura) on 9 reserves in 1993, two of which have received no fox control, three have received fox control since 1985, and four have received fox control since 1994. Trap success data from 1994-1996 suggests that fox control benefits populations of red-tailed phascogale. However, it was also noted that rainfall and population abundance from the previous year were strongly related, which obscures the effect of other factors.

Friend and Thomas (2003) report changes in the sighting rates of numbats (Myrmecobius fasciatus) at Dryandra, Western Australia (see also Saunders et al. 1995). Dried meat baits containing 1080 were laid over the area from 1989 with the aim of killing foxes. The sighting rate increased from c. 5 numbats 100 km-1 in 1989 to c. 11 numbats 100 km-1 in 1992, but thereafter declined to pre-poison baiting levels. The cause of this decline was suspected to be a decline in the food supply induced by the increasing population (Friend and Thomas 2003).

Chuditch (or western quoll, Dasyurus geoffroii) apparently increased in abundance following the application of dried meat baits containing 1080 to kill foxes at Batalling forest, Western Australia (Morris et al. 2003). Chuditch trap success increased from c. 0.5% in December 1990, when fox control began, to a peak of 13% in July 1995. Thereafter, trap success has been lower but still an order of magnitude greater than when control began in 1990. Morris et al. (2003) describe examples of chuditch translocated to former range establishing populations when foxes were controlled.

Saunders et al. (1995) report of a 30-fold increase in the abundance of Rothschild's rock-wallaby (Petrogale rothschildi) following the removal of foxes from Dolphin Island by 1080 baiting. There was also an apparent increase in the abundance of bettongs (Bettongia penicillata) in the Tutanning Nature Reserve, Western Australia, following a 5 year baiting program.

The numbat, chuditch, Rothschild's rock-wallaby, and bettong case studies above all suffer from a lack of replication (Caughley and Gunn 1996). However, as Friend and Thomas (2003) argue, there are often no other areas available for replication. Also, some populations may be very close to extinction such that non-treatment sites may be ethically difficult to justify. Notwithstanding these limitations, the above observational studies are the major sources of information about the conservation benefits of fox control, and appear to be the motivation for large-scale fox control programs in Western Australia ('Operation Foxglove' and 'Western Shield') and Victoria ('Deliverance' and 'Southern Ark'). A recent review of interactions between feral cats, foxes, native carnivores, and feral rabbits in Australia noted several exceptions to the predicted responses of small mammals to fox control (Robley et al. 2004). We suspect that 'negative' responses are generally less well publicised than 'positive' responses, leading to a biased view of the benefits of pest animal control. This is our major criticism of observational studies (c.f. McArdle 1996); that unexpected responses are explained away (e.g., there was a drought which reduced food supply) and the results are not publicised (the 'file drawer problem'; Rosenthal 1979). Only a well-designed experiment allows the effects of the treatment to be estimated, but these are lacking in the field of pest animal control in Australia (see also Robley et al. 2004).

New South Wales National Parks and Wildlife Service has recently implemented the state's threat abatement plan for predation by the red fox (NSW National Parks and Wildlife Service 2001). Part of this plan includes experiments aimed at measuring the response of threatened species to fox control. Although some of these were included in this review, it is too early for results on the response of threatened species.

Wild dogs

Fleming et al.'s (2001) review did not mention any examples where the responses of native species to wild dog control had been estimated. We are not aware of any current studies that seek to answer this question.

Feral cats

Dickman (1996) noted that "no… experiments have been completed to determine the effects of cats on native fauna". Our review has indicated that this situation has not changed. In the absence of reliable information, Dickman (1996) tabulated the responses of birds to the eradication or control of feral cats at seven sites in Australia. Six of the 7 cases exhibited a "population increase" and one exhibited a "mixed" (i.e., increased and then declined) response. However, six of these seven responses were from feral cats being eradicated (primarily islands), and it is unclear if these responses would be observed in sustained control operations on mainland Australia. Dickman also noted that these inferences should "be treated with some caution", as many factors other than feral cats can affect population size. None of the species listed by Dickman are listed under the EPBC Act.

Feral rabbits

Williams et al. (1995) list examples of the impacts of feral rabbits on the distribution and abundance of some species of grass, herb and tree seedlings. For example, feral rabbits are believed to prevent regeneration of mulga (Acacia aneura) by feeding on seedlings (Henzell 1991). Mass germinations of white cypress pine (Callitris columellaris) were only reported in the 1950s, which coincided with a large reduction in the abundance of feral rabbits by myxomatosis. However, the 1950s also coincided with the end of a 60 year period of below-average rainfall, so the cypress regeneration may be a consequence of increased rainfall, feral rabbits, or both (Williams et al. 1995). However, there are few case studies investigating the benefits of control.

Sandell (2002) investigated the responses of semi-arid woodland vegetation in north-west Victoria following the arrival of RHD, which substantially reduced feral rabbit densities. It was predicted that following the decline in feral rabbit abundance due to RHD, there would be no significant difference in vegetation condition between plots where feral rabbits were excluded by fencing compared to those to which feral rabbits could access. There was a general increase in the abundance of species in the pasture layer preferred by feral rabbits, and exotic weeds also increased. There was no widespread regeneration of woody seedlings, but there was an increase in the sucker growth of cattle bush (Alectryon oleifolius subsp. canescens). Following the arrival of RHD in Australia, 10 monitoring sites were established throughout Australia. There was 'a measurable vegetation recovery' at five of the sites, particularly in an increased abundance of woody seedlings and shrub and tree regrowth (Sandell and Start 1999). At one site (Lake Burrendong, New South Wales) feral rabbits were still preventing seedling regeneration. Monitoring ceased after two years at the majority of these sites.

Feral pigs

Choquenot et al. (1996) reported several studies showing that feral pig control increased lamb survival in the semi-arid rangelands (see also Choquenot et al. 1997). However, no studies have evaluated the benefits of control for conservation resources. Since that review, Hone (2002) reported that the relationship between the percentage of plots with ground rooting and the percentage of plots with feral pig dung at Namadgi National Park (south-eastern Australia) was concave-down, suggesting that a large reduction in pig abundance is needed to achieve a substantial reduction in pig rooting. However, in the parlance of this report, no reliable study has been conducted (or to our knowledge underway) to evaluate the benefits of feral pig control for conservation resources.

Feral goats

Parkes et al. (1996) considered that none of the potential impacts of feral goats had been quantified, and similarly the benefits of control have still not been quantified (Forsyth and Parkes 2004). The most useful information comes 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. We are not aware of any study underway that aims to yield reliable information about the benefits of feral goat control for biodiversity.

National costs of pest animal control in Australia

Our estimates of the annual national expenditure on pest animal control by government agencies in Australia (Figure 7.16) should be considered conservative for three main reasons. First, our review focused primarily on 'conservation' based organisations, however, significant amounts of pest animal control is undertaken on private land throughout Australia. We are not aware of any nationwide or state summaries on pest control expenditure on private land. Second, the figures do not include expenditure by the Western Australian Department of Agriculture, and this organisation has spent large amounts on both feral rabbit and feral goat control in some years. Third, our estimates of costs were based on an assumed labour rate and did not include operational costs such as bait, vehicles, helicopter charter, firearms, and traps.

Notwithstanding the above caveats, the costs indicate some interesting patterns (Table 8.2). The greatest expenditure was on fox control ($5.3 million), followed by wild dogs ($3.2 million). Similar money was spent controlling the four other pest species (range: $0.4-1.4 million). This pattern contrasts with that reported by Bomford and Hart (2002) (Table 8.2). Unfortunately, Bomford and Hart (2002) did not indicate how their figures were estimated. According to Bomford and Hart, the greatest expenditure was on feral rabbits ($10 million), followed by wild dogs ($4 million). One potential explanation for the discrepancies in the estimates of expenditure for all species between the two studies, is that Bomford and Hart (2002) have included both environmental and 'agricultural' focused organisations in their calculations.

Table 8.2. Costs (millions of dollars) of pest animal control in Australia by government agencies estimated by this study (labour costs only) and Bomford and Hart (2002).
Species
This study1
Bomford and Hart (2002)
Fox
5.3
2.0
Wild dog
3.2
4.0
Feral cat
1.1
1.0
Feral rabbit
1.4
10.0
Feral pig
0.9
2.5
Feral goat
0.9
0.2
TOTAL
12.82
21.5

1 Highest reported for the years 1998-2003 (see Figure 7.16).
2 Labour costs are not mutually exclusive between pest species in this study