Intraguild relationships between sympatric predators exposed to lethal control: predator manipulation experiments

Overall patterns in abundance

We found no indication that mean dingo, fox or cat passive tracking index (PTI) values were substantially greater or became greater in areas subjected to periodic poison-baiting for dingoes (Figure 2). The overall mean PTI values for dingoes were demonstrably less (43–77% lower) in baited areas than in paired unbaited areas for five of the six experimental sites and one of the three Blackall sites (Table 1). PTI values for dingoes were similar in both treatments at all other sites. Overall mean fox PTI values were also demonstrably less in the baited areas for four of the eight sites where foxes are found, and were similar in both treatments at the remaining four sites. At no site was a demonstrable difference found between the baited and unbaited areas in overall mean PTI values for cats (Table 1). Treatment differences in PTI values for goannas were found at three sites, with higher overall mean PTI values in the baited area at two of these sites (Table 1). Thus, the greater overall abundance of goannas in the baited areas of Tambo and Blackall were the only two instances (of 26 possible site × mesopredator combinations) where a sympatric mesopredator was detected more frequently in a paired dingo-baited area at any site. Stratifying the data by season indicated that each predator was in similar abundance in both baited and unbaited areas in 76 of 88 cases (Table 2). In every case where demonstrable differences between treatments were detected for any season (12 of 88 possible site × mesopredator × season combinations), dingoes, foxes, cats and goannas were each in greater abundance in unbaited areas (Table 2), indicating no overall benefit to mesopredators from top-predator control.

Short-term behavioural responses

A total of 25 baiting events from all sites included post-baiting surveys conducted within four months of baiting from both treatments (mean number of days since baiting = 51). Assessing the short-term responses of predators between surveys conducted just prior and subsequent to baiting showed no indication of short-term behavioural increases or decreases in mesopredator PTI following these dingo control events (Tables 3 and 4). There was a demonstrable effect of time on dingoes (F_1,18 = 14.37, p = 0.0013) and cats (F_1,18 = 3.91, p = 0.0636), and a demonstrable effect of treatment on dingoes only (F_1,18 = 46.01, p = 0.0001). However, no demonstrable time x treatment interactions were detected for any species (Table 3), indicating that the observed post-baiting decreases in dingo and cat PTIs were independent of treatment. No effects of time, treatment or their interaction were found for foxes or goannas (Tables 3 and 4). We likewise found no indication of short-term behavioural increases of mesopredators by assessing the mean net changes in predator PTI between pre- and post-baiting surveys at Mt Owen (N = 8), Quinyambie (N = 4), Strathmore (N = 5) or Todmorden (N = 5) (Figure 3). An insufficient number of pre- and post-baiting pairs to reliably run this analysis were obtained from the other sites. Using this approach, demonstrable changes were only found for dingoes, which were reduced by baiting, and only at Mt Owen, Quinyambie and Strathmore, but not Todmorden. Combining the data from all sites showed no overall short-term changes in PTI for any predator except cats, which slightly declined following dingo control (Figure 3).

		Dingo		Fox		Cat		Goanna
Source	df	F	p	F	p	F	p	F	p
E	18 (14)
T	1	14.37	0.0013	0.39	0.5424	3.91	0.0636	0.07	0.7893
T × E*	18 (14)
B	1	46.01	0.0001	2.22	0.1580	0.92	0.3504	0.52	0.4818
B × E*	18 (14)
T × B	1	0.12	0.7315	0.33	0.5755	0.31	0.5829	2.41	0.1381
T × B × E*	18 (14)

* = Error terms (with adjusted values for those of foxes in parentheses). See Table 4 for mean PTI values and sample sizes used in these analyses.

			N=	Dingo	Fox	Cat	Goanna
T	Pre-baiting		38 (30)	0.408	0.025	0.064	0.070
	Post-baiting		38 (30)	0.238	0.022	0.046	0.055
B	Baited		38 (30)	0.189	0.014	0.052	0.027
	Unbaited		38 (30)	0.456	0.032	0.058	0.030
	Time	Treatment	N=	Dingo	Fox	Cat	Goanna
T × B	Pre-baiting	Baited	19 (15)	0.267	0.017	0.060	0.028
		Unbaited	19 (15)	0.549	0.032	0.069	0.026
	Post-baiting	Baited	19 (15)	0.112	0.011	0.045	0.026
		Unbaited	19 (15)	0.364	0.032	0.048	0.034

Longer-term abundance trends

Triangular relationships

Analysing our data using an alternative approach described by Johnson and VanDerWal [38] provided inconclusive evidence of negative relationships between dingoes and any of the mesopredators we assessed at our study sites (Figure 4). The shape of the relationships also changed depending on whether or not the data were transformed (compare Figures 4 and 5). However, both approaches were consistent in showing that ‘triangular relationships’ (sensu[38]) typically did not exist between dingoes and foxes, cats or goannas at the sites we monitored. In other words, we found little evidence that variability in fox, cat or goanna PTI decreased as dingo PTI increased using this approach. In contrast, mesopredator PTI typically varied little despite relatively large fluctuations in dingo PTI (Figures 4 and 5), suggesting that mesopredator abundances fluctuate independent of dingo abundances or dingo control (as shown in Figure 2). Regardless, we question the utility and reliability of the approach taken by Johnson and VanDerWal [38] for several reasons:

Given these issues, reducing our experimental data to approach the subject matter in the correlative way described by Johnson and VanDerWal [38] would be inferior to the primary analytical approaches we have taken, which do permit conclusive and demonstrable statements about cause (i.e. dingo control) and effect (i.e. mesopredator release; [35]).

Evidence for lethal control-induced mesopredator release

Our results provide demonstrable and conclusive evidence that in almost all cases and no matter which analytical approach was used to evaluate these experimental data, neither foxes, cats or goannas responded positively to contemporary dingo control practices. The three sympatric mesopredators we assessed were typically in similar or greater abundance in unbaited areas relative to baited areas (Tables 1 and 2). No short-term increases in mesopredator PTI values were observed in baited areas (Tables 3 and 4, Figure 3). Longer-term mesopredator PTI trends typically were either uncorrelated or positively correlated with dingo PTI trends over time (Table 5, Figure 2). If either fox, cat or goanna populations gained an overall benefit from lethal dingo control, then (1) overall mean mesopredator PTI should have been higher in baited areas, and/or (2) mesopredator PTI values should have increased in baited areas following baiting, and/or (3) mesopredator PTI trends should have diverged from dingo PTI trends over time. Rarely did any of these occur for any mesopredator at any site.

As found in wolves (e.g. [14, 45]), it is possible that dingo populations subjected to lethal control may also undergo demographic and social changes that might benefit mesopredator populations, such as the loss of experienced adults and an associated reduction in dingoes’ ability to repel mesopredators [16, 24]. However, dingo populations subjected to baiting are likely to contain a larger proportion of adults and adult-sized animals, exhibit temporarily elevated activity levels and territorial behaviour, and prey more heavily on species smaller than themselves [46]. Each of these behavioural changes would be highly unlikely to create a favourable environment for mesopredators attempting to establish in dingo-baited areas, where only individual dingoes from small and newly-formed packs are needed to kill invading foxes or feral cats [47]. Assessment of demographic perturbations to dingo populations was outside the scope of the present study, but if demographic and social changes to dingo populations occurred, they did not appear to benefit foxes, cats or goannas.

The study of Eldridge et al. [48] is the only other completed and true predator manipulation experiment investigating in situ dingo-mesopredator relationships [18, 30]. Using a different PTI methodology, Eldridge et al. [48] likewise showed that fox and cat population trends did not diverge from dingo trends over time (Figure 6), nor were dingo and fox PTI values negatively correlated (r = 0.049, p = 0.588). Viewed together or separately, both these and our large-scale, multi-year, predator-manipulation experiments at 12 sites across the beef-cattle rangelands of continental Australia provide consistent and demonstrable results that do not support, and indeed contradict, perceptions (reviewed in [28, 29]) that: (1) contemporary dingo control practices facilitate immediate or subsequent increases in mesopredator abundances, that (2) ceasing dingo control leads to reduced mesopredator abundances, or that (3) mesopredator populations are negatively associated with dingo populations over time. Long-term (10–28 years) correlative studies of dingoes support these experimental results (e.g. [49, 50]). A large and growing body of evidence from other countries similarly report neutral or positive relationships between terrestrial top-predators and mesopredators (e.g. [7, 51–55]). Thus, not only is there a clear absence of reliable evidence for dingo control-induced mesopredator release, but there is a strong and growing body of demonstrable evidence of absence for the same.

Our findings are in accord with what is known from other predator manipulation experiments worldwide: fauna at lower trophic levels are unlikely to respond positively to lethal control where multiple predators are removed (i.e. dingoes and foxes in our case), where the efficacy of predator removal is low (i.e. where predator populations quickly recover), and where the fauna are not the primary prey species of the predator [18]. That fox populations did not respond positively to poison-baiting is intuitive given that they are widely known to consume baits intended for dingoes and have similar rates of increase to dingoes (e.g. [19, 56, 57]). ‘Dingo baiting programs’ should therefore be better thought of as ‘dingo and/or fox baiting programs’ or ‘canid baiting programs’. It is theoretically possible that mesopredators may have increased in response to the removal of dingoes but that this response was masked by the fact that mesopredators may also have been simultaneously removed by baiting. Regardless, our data demonstrate that the overall population-level responses of mesopredators to the distribution of non-selective poison-baits were not positive. Though not susceptible to the toxin at the doses used in canid baits, goannas also consume baits. That foxes (e.g. [58]) and goannas (e.g. [59]) can reduce the number of baits available to dingoes may be one reason why dingoes were not demonstrably reduced at all sites. Our results for cats might also have been expected given that dingoes and cats are known to have mixed (usually neutral) relationships, which are typically weaker in places where foxes are present [29]. Where studied, neutral relationships between dingoes and goannas are also typical and independent of baiting [50].

Mesopredator release theory and reality

Our findings do not contradict mesopredator release theory, but merely indicate that contemporary dingo control practices do not produce the conditions required to generate a mesopredator release effect. In other words, and apart from targeting foxes as well, periodic poison-baiting across areas up to 4,000 km² does not appear to suppress dingo populations to levels low enough and long enough for mesopredators to exploit the situation (Tables 1, 2 and 3, Figures 2 and 3). That contemporary dingo control practices do not trigger a trophic cascade is likely due to rapid reinvasion of dingoes back into baited areas, which typically occurs within weeks or months after baiting ([19]; present study). In the beef-cattle rangeland systems we studied, dingoes from higher-density populations are known to migrate >550 km in 31 days or >1,300 km in four months into areas with lower dingo densities, with approximately 15% of dingoes dispersing over 100 km [60]. However, such long-distance migrations are not usually required for dingoes to recolonise baited areas because baiting programs seldom reduce extant dingo populations by even 50% and source populations of dingoes may only be a few kilometres away, and even within the baited area ([22, 61, 62]; present study). This is dissimilar to historical dingo control practices (which were reliant on extensive exclusion fencing to inhibit recolonisation) which did enable extermination of dingoes from some areas [20, 21]. The differences in overall efficacy (at reducing dingo abundance) between historical and contemporary control strategies could have contributed to the coarse continental-scale pattern of inverse relationships between dingoes and foxes often touted (e.g. [33, 63]).

Contemporary canid control practices at our study sites, which represent contemporary land use across much of the Australian continent, may be defined as the spatiotemporally sporadic application of relatively minor amounts of poisoned bait throughout a mosaic of baited and unbaited areas [22]. Though we cannot tell, we expect the results we describe to occur commonly in places where contemporary canid control is practiced in this way. Although canid control temporarily removes some dingoes (and might change their social structures), as intended, this does not imply their complete or sustained eradication, which might produce a positive response from mesopredators in places it could actually be achieved. Thus, there might be a baiting-induced low-point where top-predator populations become ecologically ineffective in their roles [17, 24]. However, our data suggest that this low-point may have to be extremely low given that dingoes were suppressed by up to 77% (Table 1) without corresponding evidence of positive responses from any mesopredator (Tables 3 and 5, Figures 2 and 3). Allowing dingo populations to recolonise following periodic suppression may provide livestock producers with a window of opportunity to reduce livestock depredation during high-risk times (such as peak cattle calving season) while retaining the ecological functions of dingoes over longer timeframes.

Factors affecting predator responses to lethal control

Although our experiments were conducted over similar timeframes to most other predator manipulation experiments [18], it might be argued that 2–5 years is not long enough to detect positive mesopredator responses to dingo control. However, three lines of evidence suggest this is not the case for our data. First, the PTI methodology we applied was sufficient to detect the responses of predators to the bottom-up effects of rainfall within the timeframe covered ([22, 46, 64]). Some also claim that the top-down effects of dingo control can be greater than the bottom-up effects of rainfall in the systems we studied [65], so the predicted positive responses of mesopredators to baiting should have been observable. Second, from small-scale observational studies conducted in similar habitats to ours, Pettigrew [27], Christensen and Burrows [25] and Lundie-Jenkins et al. [26] each reported detectable positive responses of foxes and/or cats within a few weeks or months after single dingo-control events, implying that 2–5 years of repeated baiting and population monitoring across spatial scales several orders of magnitude larger should have readily detected both acute and chronic mesopredator releases. Third, having been exposed to the same treatments for at least 10 years at the three Blackall sites (where predator abundances might be expected to have stabilised), mesopredator abundances should have been higher in baited areas, but they were not (Tables 1 and 2). In contrast, the only two (out of a possible 26) instances where a sympatric mesopredator was detected more frequently in baited areas at any site was for goannas at Blackall and Tambo – sites where dingo abundances were not demonstrably less in baited areas (Table 1). These lines of evidence indicate that our methodology was sufficient to detect immediate and longer-term increases in baiting-induced mesopredator activity or abundance if they were occurring.

Although we undertook our study in an experimental framework inclusive of buffer zones to maintain treatment independence, it is also important to remember that our approach was an evaluation of the overall population-level responses of mesopredators to contemporary top-predator control practices under real-world environmental conditions where dingoes, foxes, cats and goannas are each capable of dispersal and migration between treatments over time. In other words, we sought not to compare nil-treatment areas to paired treated areas with ‘X% reduction of dingoes’ or ‘X density of baits’, but with ‘contemporary dingo control practices’. This applied-science focus therefore produces results that reflect the in situ outcomes of contemporary dingo control practices in the beef-cattle rangelands present across much of the Australian continent. Alternative dingo control strategies which actually achieve complete and sustained dingo removal from the landscape (such as exclusion fencing) may yield different results, as may studies interested in smaller spatial scales where physical interactions between predators might be observed.

Viewed collectively, possible explanations for our observations might include that dingoes do not interact strongly with foxes, cats or goannas and/or that abundances of these mesopredators are associated primarily with bottom-up factors (such as rainfall, primary productivity, habitat complexity or prey availability), as has been found in other fox and cat studies (e.g. [66, 67]). These bottom-up factors likely affect dingoes similarly [20, 68]. Mesopredators (especially scavenging foxes and goannas) may also derive substantial benefit from dingoes through kleptoparisitism [19, 69], which may also have contributed to our observations. The relative strength of top-down and bottom-up processes affecting predator populations in Australia has not been well studied, though it seems clear from our results that mesopredator populations in the rangelands do not appear to be enhanced by contemporary dingo/fox control practices. Indeed, we found no empirical evidence to support the supposition that cessation of periodic top-predator control will somehow suppress sympatric mesopredators, nor did we find evidence to suggest that commencement of top-predator control increases mesopredator activity or abundance. It is clear that a far greater understanding of context-dependant top-down and bottom-up processes must be acquired before biodiversity restoration is to be achieved simply by bolstering top-predator populations in Australia [19]. Thus, proposals to cease dingo control are presently unjustifiable on biodiversity protection grounds [70].

Study sites and design

We conducted a series of large-scale, multi-year, predator-manipulation experiments (from [22, 46, 64]) on extensive beef-cattle producing properties in five different land systems representing the breadth of the beef-cattle rangelands of Australia, where mean rainfall varied from 160–772 mm annually, or from arid to tropical areas (Figure 7, Table 6).

Using paired nil-treatment areas without dingo control for comparison (Figure 8A), we examined the relative abundances of predators in paired areas subjected to periodic broad-scale poison-baiting for dingoes at six of nine study sites (Strathmore, Mt Owen, Cordillo Downs, Quinyambie, Todmorden, and Lambina; Table 6), referred to as the six experimental sites. Aerial and/or ground-laid sodium fluoroacetate (or ‘1080’) poison-baits were distributed individually (spaced at least 300 m apart) along landscape features (e.g. drainage lines, ridges, fragment edges etc.) and/or unformed roads according to local practices and regulations up to five times each year (typically once in spring and again in autumn at the six experimental sites, and every 2–4 months continuously at the other three sites). Baits were distributed over a 1–2 day period to a midway point in the buffer zone between treatments (described below; Figure 8A). Each bait weighed 100–250 g and contained at least 6 mg of 1080, sufficient to kill adult dingoes, foxes or cats (but not goannas) if consumed soon after bait distribution [57]. Such baiting practices are common, occur widely across Australia, and are considered the only effective dingo and fox control tool used in rangeland areas [56]. Cat and goanna populations are not typically susceptible to such baiting practices because goannas are tolerant of the toxin (at the low-level doses used in canid baits) and cats rarely consume carrion-like baits, preferring live prey instead (e.g. [59, 74–76]). Opportunistic shooting of dingoes occurred at some sites during the study and historically (Table 6), but with negligible effects on dingo populations because very few dingoes were ever shot [22]. Experimental treatment (i.e. baited) and nil-treatment (i.e. unbaited) areas were randomly allocated. Treatment and nil-treatment areas were also replicated in some land systems (Table 6). Hone [35] defines this study design as an ‘unreplicated experiment’ or a ‘classical experiment’ for our site with replication (i.e. Todmorden and Lambina might be considered a single site with two treatments and two controls). Both treatment and nil-treatment areas at some of the six experimental sites were historically exposed to baiting up until the commencement of the experiment, whereas, both treatment and nil-treatment areas were not historically exposed to baiting at other sites (Table 6). Such histories were necessary to investigate the responses of predators to either the commencement or cessation of baiting, or to the ‘removal’ or ‘addition’ of predators (i.e. dingoes and foxes were killed at some sites or allowed to increase at others).

All procedures described were sanctioned by the relevant animal care and welfare authorities for each site (Queensland Department of Natural Resources’ Pest Animal Ethics Committee, PAEC 930401 and PAEC 030604; South Australian Department of Environment and Heritage’s Wildlife Ethics Committee, WEC 16/2008).

The other three sites (Barcaldine, Blackall and Tambo; referred to as the Blackall sites) were monitored for a similar length of time (Table 6), but differed from the six experimental sites in that the treatments and nil-treatments had already been established for over 10 years and they did not have buffer zones between them (Figure 8B). This allowed an assessment of the longer-term outcomes of lethal dingo control. Treatment size, independence and baiting practices therefore varied between the nine sites in order to deliver in situ tests which reflected contemporary dingo control practices within each bioregion. Experiments were conducted at large spatial scales, where the size of the total treatment and nil-treatment area at each of the nine sites ranged between 800 km² and 9,000 km², or 45,600 km² in total (Table 6). The largest contiguous baited area was ~4,000 km². By comparison, the combined size of the areas we assessed is approximately four times the size of Yellowstone National Park or half the size of the Greater Yellowstone Ecosystem, where a substantial amount of similar research on wolves has been conducted. Only one other predator manipulation experiment (on wolves and ungulates in Canada) has a larger spatial scale than ours (i.e. [12], a single site with one treatment area and three controls totaling ~60,000 km²), and very few such experiments ‘add’ predators; most ‘remove’ them [18]. Each site we studied was separated by 100–1,500 km, except in the case of Todmorden and Lambina, which were neighbouring properties (Figure 7).

Predator population monitoring

Dingo and sympatric predator populations were simultaneously monitored in treatment and nil-treatment areas using passive tracking indices (PTI; [77]). This technique is the recommended and standard monitoring technique used for assessing terrestrial predator populations in Australia [78] and has also been used to monitor a variety of other predators in other countries (e.g. [79–81]). Variants of this technique are commonly used to sample many different terrestrial fauna around the world [82].

PTI surveys were conducted several times each year at each site and were repeated at similar times each subsequent year over a 2–5 year period (Table 6). At the Blackall sites, between 92 and 166 passive tracking plots (or ‘sand plots’) were spaced at 1 km intervals along unformed vehicle tracks. At the six experimental sites, 50 plots each were similarly established in both the treatment and nil-treatment areas (i.e. 100 plots per site). The number of plots we monitored is roughly double that used in most other similar studies of dingoes and we also monitored these plots for longer than most similar studies at most of our sites [30]. For any given survey, plots in both treatments were read and refreshed at the same time daily by the same experienced observer and were monitored for up to 10 successive days (usually 2–5). The location of the first tracking plot in each treatment area was randomly allocated and plots were distributed throughout a similar suite of microhabitat types in both treatment areas to minimise potential microhabitat differences in predator detectability between treatments. Plots rendered unreadable by wind, rain or other factors were excluded from analyses. All predator track intrusions were counted (i.e. a continuous measure). PTI values for a given survey therefore represented the mean number of predator track intrusions per sand plot tracking station per 24 hr period (i.e. the mean of daily means; [77]). Analysed appropriately, PTIs collected in this way can be interpreted as robust estimates of relative abundance (e.g. [40, 41, 83]).

At least one PTI survey was conducted before the imposition of treatments (i.e. before commencement or cessation of baiting in a given treatment) at the six experimental sites to identify any spatial variation in predator population abundances between treatments prior to manipulations. Tracking plot transects at these six sites were separated by a buffer zone 10–50 km wide (Figure 8) to achieve treatment independence during individual surveys. The appropriate width of the buffer zone at each site was based on the width of 1–2 dingo home ranges in the study areas (e.g. [61, 62]). Tracking plots were located no closer than 5–25 km from the edge of the treatment area (i.e. half the width of the buffer zone) to minimise potential edge effects. Overall, we obtained 35,399 plot-nights of tracking data from 128 surveys conducted over 31 site-years (Table 6).

Analytical approaches

We used three primary approaches to examine the effects of lethal dingo control on sympatric mesopredators. First, we compared the mean PTI of predators (both overall, and also stratified by season) between baited and unbaited areas at each site using repeated measures ANOVAs. Second, we determined the short-term changes in predator PTI values between pre- and post-baiting surveys (conducted within four months since dingo control) through a (1) 2-factor (time and treatment) repeated measures ANOVA, and for completeness, by also (2) assessing mean net changes in PTI (i.e. changes in the baited area after accounting for changes in the unbaited area) with two-tailed t-tests for each site where at least four pre- and post-baiting surveys were conducted. Third, we assessed temporal correlations between abundance trends of dingoes and each sympatric mesopredator, separately for baited and unbaited areas. Data were not available for goannas in winter or foxes at one site in northern Australia, because ectothermic reptiles are typically inactive in winter and the national distribution of foxes did not extend to the northernmost site (Strathmore) in the Gulf of Carpentaria (Figure 7). Additional details on study site characteristics, dingo control history, experimental designs, bait types and densities, baiting regimes and efficacy, and application of passive tracking indices can be found in Table 6 or in Allen [46, 64] and Allen [22].

Extant mesopredators have been predicted to respond positively to dingo control through either (1) a numerical reduction in dingo abundance and/or, where sustained numerical reductions may not have occurred, through (2) behavioural or demographic changes to dingo populations which facilitate increases in mesopredator activity or abundance (e.g. [16, 24, 65]). These responses are predicted to occur rapidly, manifest first by immediate and then sustained increases in mesopredator PTI (e.g. [25–27]). Thus, our primary analytical approaches would detect dingo control-induced mesopredator release where: (1) mean mesopredator PTI is greater in paired baited areas (indicative of greater mesopredator densities in baited areas); where (2) mesopredator PTI increases in baited areas (relative to unbaited areas) shortly after dingo control (indicative of an immediate behavioural release of mesopredators); or where (3) mesopredator PTI trends diverge over time (indicative of a longer-term numerical release of mesopredators). An alternative approach to assessing dingo-mesopredator relationships has been suggested by Johnson and VanDerWal [38], which report the occurrence of non-linear ‘triangular relationships’ between dingoes and foxes; dingoes apparently setting an upper limit on fox abundance in temperate forests. This approach pools binary data across sites and surveys, transforms the values to generate a visual affect, selects only the extreme values, and then correlates these extreme values between predators. For completeness, we therefore explored the utility of this approach with our data as a supplementary exercise only.