Portal de Periódicos da CAPES

Will dingoes really conserve wildlife and can our methods tell?

2014; Wiley; Volume: 51; Issue: 4 Linguagem: Inglês

10.1111/1365-2664.12250

1365-2664

Matt W. Hayward, Nicky Marlow,

Rangeland and Wildlife Management

Debate rages in Australia about the merits of using apex predators (dingoes Canis lupus dingo) to limit the impact of introduced mesopredators, which have been identified as the primary cause of Australia's mass mammal extinction since European arrival. All studies in this debate have ignored detectability issues and have instead relied primarily on unvalidated indices of footprints or photographs from trails as a measure of abundance or activity or both. We illustrate how detectability is likely to vary using these methods in ways that predict the outcome before any data are collected. We argue that until robust methods (e.g. distance sampling, mark–recapture, occupancy modelling) coupled with experimentation are used to assess this issue (and those of competition and trophic cascades in general), practitioners will be unwilling to implement the conservation actions suggested (in this case the reintroduction of dingoes). These issues are relevant to carnivore studies more generally. Debate is raging in Australia surrounding the merits of using apex predators (dingoes C. lupus dingo) to conserve native fauna by reducing the abundance or area of impact of mesopredators (introduced red foxes Vulpes vulpes and feral cats Felis catus), which have largely been attributable for the numerous declines and extinctions of Australian wildlife (Burbidge & McKenzie 1989). There is an increasing accumulation of literature reporting on the mesopredator-suppressive effects of dingoes that is aimed at altering the current management practices of broad-scale dingo and/or wild dog control in conservation areas and pastoral land (Ritchie & Johnson 2009; Letnic et al. 2011). There is also an ever-increasing literature highlighting the weaknesses in the methodology used to illustrate mesopredator suppression, such as a lack of replication, methodological flaws, sampling bias and experimental design constraints (Fleming, Allen & Ballard 2012; Allen et al. 2013). In essence, this follows the debate about the impacts of wolves C. lupus in Yellowstone National Park, USA, and concerns about the actual cause of those impacts, and like this debate, ‘misrepresentation of published work has become commonplace’ (Massey, Cubaynes & Coulson 2013; p 1273), making it very difficult for conservation practitioners to detect the wheat from the chaff. Despite this debate, questions arise as to whether a reduction in fox/cat predation pressure coupled with additional dingo predation will improve conditions for Australian wildlife. Australian mammals weighing between 35 g and 5·5 kg are considered to occur within a critical weight range because these species have been decimated by introduced foxes and cats (Burbidge & McKenzie 1989). Even low densities of introduced predators drive critical-weight-range mammals extinct (Short et al. 1992). Despite hopping mice Notomys sp. indices being higher1 in cattle lands where dingoes are present compared to sheep lands where they are not (Letnic, Crowther & Koch 2009), the vast majority of fauna outside the critical weight range is secure and not threatened with extinction. These species outside the critical weight range are therefore unlikely to benefit overly from a reduction in fox and cat density, as opposed to complete eradication. Indeed, the only way Australian native fauna will evolve strategies to cope with predation by introduced predators is to coexist with them as it seems highly unlikely that introduced predators will be eradicated in the foreseeable future given their likely evolution of tolerance to the poisons used and the inability to eradicate them nationally to date, despite attempts over millions of hectares across Australia. Dingoes will never completely eradicate foxes and cats for three reasons. First, these three species (C. lupus, V. vulpes and F. catus) successfully coexist throughout Eurasia. Secondly, wild cats coexist with the ecological equivalents of wolves (African wild dogs Lycaon pictus) and foxes (jackals) throughout sub-Saharan Africa. Finally, following their introduction, foxes and cats were ultimately able to spread across Australia despite the presence of dingoes. At best, dingoes can structure ecosystems to create safer areas as predation refuges for native species. Questions remain as to whether this will be enough to allow Australia's threatened species to persist and whether lowering fox and/or cat density will yield sufficient benefits in the face of the likely human–wildlife conflicts between sheep graziers and dingoes (Fleming, Allen & Ballard 2012). Dingoes could become a valuable tool for conservation management; however, methodological problems need to be resolved before this is clear. The collection methods and data analysis of studies in this debate may well be the source of this controversy. Our first concern is that this voluminous literature invariably ignores detectability. Just because a cat footprint is not detected on a sand plot does not mean cats are absent – rather, they were not detected, and five cat footprints on a sand plot does not mean five cats were present. Yet the vast majority of studies in Australia on placental predators have ignored this issue and have assumed a constant or predictable relationship between index value and abundance. This is because these studies invariably rely on indices of abundance or activity or both or potentially neither as they have not been validated on a large scale or related to the factor they aim to measure (cf. the validation of the tiger Panthera tigris index of abundance; Jhala, Qureshi & Gopal 2011). Indeed, the low detection rate for carnivores generally makes indices of abundance unreliable (Royle, Stanley & Lukas 2008). There are methods to incorporate detectability into studies of mesopredator interactions. Distance sampling, mark–recapture (either with animals in hand or via photographs) and occupancy modelling (with a revised study aim to investigate changes in area of occupancy, that is, predation refuge) account for differential detectability, and it is highly probable that detectability will vary in these mesopredator studies even in cross-fence, pairwise comparisons as have been regularly used in studies of dingo benefits and mesopredator suppression (Letnic, Crowther & Koch 2009; Letnic et al. 2011). For instance, mesopredators could rapidly alter their movement patterns when apex predators are removed, so detectability will be important even over short temporal and spatial scales. Perhaps most concerning, the major assumptions of distance sampling, mark–recapture and occupancy models are the same as for these indices (demographic closure; independent sites; no unexplained heterogeneity in occupancy or detectability), except indices also assume a constant proportion of the study population is counted in each survey and site, and the relationship between the index and actual abundance is linear or can be made so (MacKenzie et al. 2006). Yet some of the authors of these index-reliant studies have rejected the use of occupancy modelling because their data were not adequate to meet the model's assumptions (name withheld personal communication to M. W. Hayward). Several studies have shown large increases in footprint index value of foxes, kangaroos or rodents in the absence of dingoes, even though these trail-based footprint index values can vary by almost 500% between days or seasons, and there is no evidence that a large increase in index value relates to actual abundance, particularly as microscale movements (on/off trail) could drive such results. Furthermore, although surprisingly rarely attempted (but see Jhala, Qureshi & Gopal 2011), studies that have validated footprint indices have shown how variable they can be. For example, a 3% increase in Cook Strait giant weta Deinacrida rugosa footprint index was recorded despite a 66% decrease in actual population size (data from Fig. 7 in Watts et al. 2011); a 3% increase in black rat Rattus rattus footprint index occurred despite a 400% decrease in actual density (Fig. 2 in Blackwell, Potter & McLennan 2002), and there was no relationship between a footprint index of house mice Mus musculus and their actual abundance (Ruscoe, Goldsmith & Choquenot 2001). Even the validation of the tiger footprint index exhibited large variation with a pugmark index value of 0·37 equating to actual population estimates of 1·6 and 7·4 tigers per 100 km2 (360% variation in actual population size without any variation in index value; Fig. S1 in Jhala, Qureshi & Gopal 2011). A second concern relating to detectability is that these indices are generally derived from trail-based surveys. It is almost ubiquitous throughout the world that apex predators walk on trails, while mesopredators avoid them in the presence of apex predators. So when an apex predator is removed from an area, trails become the optimal movement route and mesopredators rapidly respond, making them more detectable. These biases are well known as unrepresentative sampling units (Mahon, Banks & Dickman 1998) and so cannot be used to make inferences beyond a local study area/species. The apparent ignorance of this issue (based on the continued use of unvalidated indices, the lack of attempts to validate them and the continued acceptance of papers relying on unvalidated indices in journals) suggests a simple analogy is worthwhile. For example, during peak hour in a city, a trail-based index would show vehicles as the most abundant or active member of the city community, whereas outside these times, humans would appear more abundant or active. Implementing vehicle-free areas (akin to removing predators) and the index shows that humans rapidly become the most abundant or active member of the city. Humans may be more aware of the risks in our landscape than other organisms, but natural selection for intraguild predator avoidance is likely to be strong. Reliance on unvalidated indices of abundance has led to major problems in conservation management throughout the world. The national index of footprint counts used in India to monitor tiger populations turned out to grossly overestimate actual population sizes when robust methods were used to compare indices (Karanth et al. 2011). A change from indices of abundance to methods that accounted for detectability (occupancy modelling of photo captures) led to a switch in conservation concern in New Zealand when grand skinks Oligosoma grande turned out to have a much higher detectability than Otago skinks Oligosoma otagoense, but were actually less numerous and more restricted in distribution (Reardon et al. 2012). Methods that accounted for detectability (genetic mark–recapture) doubled the population estimate based on indirect sign of giant pandas Ailuropoda melanoleuca (Zhan et al. 2006). Similarly, robust population estimation methods were 326% greater than an index of observations for eastern imperial eagles Aquila heliaca (Katzner et al. 2011). There is no consistent over or under estimation with indices of abundance, but given the importance of accurate measures in all aspects of conservation, robust methods are fundamental (Katzner et al. 2011). Weight of evidence does not arise from the wealth of publications on a topic (particularly if all are flawed), but rather the robust nature of their methods, data, analysis and conclusions. In this debate, the methods and analysis problems identified above mean that the weight of evidence regarding dingo impacts on other predators and their prey is limited. This lack of robust evidence means that conservation managers are unlikely to use dingoes to limit the impacts of foxes and cats, despite the plethora of publications promoting such actions. Despite the above discussion, returning dingoes to all national parks in Australia that could support a viable population (or parks enlarged to ensure that) would yield multiple ecological, economic and societal benefits. Watching a dingo hunt a kangaroo or regurgitate meat for the pups would be as much of an attraction for ecotourists as watching wolves whelp in North America. Society could benefit by refreshing the waning interest of the general public in natural places, while ecosystems would benefit through the increased robustness apex predators provide. The actual motivation behind returning dingoes to natural areas is less important than having them there because it achieves a conservation benchmark more appropriate to the continent. To restore dingoes to Australian protected areas may necessitate complete predator-proof fencing of the largest national parks, but this may be required to avert human–wildlife conflict anyway and has been proposed as the most effective way to secure the conservation of other large predators (e.g. lions Panthera leo). The debate about the use of dingoes to control the impacts of mesopredators is doing nothing to close the implementation gap as there are too many problems with the published research for practitioners to be confident that dingoes will alleviate the threats caused by foxes and cats. Indices have very limited utility in good monitoring programmes because of their inability to meet necessary assumptions (MacKenzie et al. 2006; p10; Boitani, Ciucci & Mortelliti 2012; p 13) and their unreliability for carnivores (Royle, Stanley & Lukas 2008), and the problems with indices appear to have clouded the debate about mesopredator suppression in Australia. Science needs to be driven by experimentation, robust methods, appropriate analysis and adequate data. Only substantial field research using replicated, randomized experimental manipulations and methods that account for differential detectability will do that, and we suspect that these recommendations apply to the use of unvalidated indices of abundance throughout the world. Until this is done, it would be a brave conservation practitioner who leaps at the recommendations proffered. Practitioners will be fundamental participants in the fieldwork necessary to solve this problem through their role as site managers. Thus, we urge practitioners to be involved in planning and implementing innovative, ideally randomized, management manipulations despite the challenges these present. At issue is one of the major potential conservation challenges of our time. There are enormous difficulties in designing robust studies at appropriate scales to address this issue, but Australia has innovative conservation agencies that manage landscapes at a large enough scale to conduct this research and the initial experimental manipulations at South Australia's Arid Recovery illustrate it is possible. While we recognize that concerns over the adequacy of the analysis of fox control programs delayed the implementation of broad-scale fox control, particularly in eastern Australia, we urge all participants in this debate to get back into the field to collect the robust data necessary to provide greater certainty for management action. We thank Julia Jones, Phillip Hulme, Jacqueline Frair and two anonymous referees for comments that vastly improved this manuscript. Matt Hayward and Nicky Marlow are applied ecologists who have tried to utilise new methods to avoid the biases of indices in monitoring the responses of feral cats and foxes to control programmes. Matt did so while working for the Australian Wildlife Conservancy and Nicky for the Western Australian Department of Parks and Wildlife.