Gold-standard experiments to deter predators from attacking farm animals

We dismiss the long-held belief that randomized, controlled trials (RCTs) are impossible in wild ecosystems with working livestock. Crossover designs reduce confounding variables between subjects and strengthen inference beyond the RCT, yet obstacles exist, which we describe qualitatively. Non-lethal methods can be effective in preventing carnivore approaches and attacks on farm animals in fenced pastures or open rangelands. The relationship between approaches and attacks remains uncertain. Lethal methods of predator control have been subjected to less robust study designs that suggest mixed results including increases in livestock losses. Non-lethals promise the elusive triple-win for wildlife, domestic animals, and livelihoods. Climate change and extinctions now pose the major threats to life on our planet (Ripple et al., 2017; Chapron et al., 2018; Ceballos et al., 2020). Humans cause many extinctions by killing animals or by transforming habitats. Among mammals, large carnivores have faced higher than average rates of population extirpation because of direct and indirect competition with humans (Woodroffe and Ginsberg, 1998; Chapron et al., 2014; Ripple et al., 2014). Humans respond to real and perceived threats from carnivores with lethal action and with sociopolitical pressure against protecting the last remaining carnivores. Therefore, interest groups and individuals focused on preserving carnivore populations and minimizing harm to individual carnivores have prioritized non-lethal methods to prevent conflicts between humans and carnivores in recent years. In addition to reducing damage to human property by carnivores, non-lethal methods offer potential benefits to many actors involved, by saving animal lives and benefiting human health, safety, and income. Here we describe lessons learned from gold-standard, randomized, controlled trials (RCTs) with crossover designs, which we have conducted in four countries to protect farm animals from wild carnivores of many species. We synthesize lessons learned in four categories: experiences with randomized, controlled trials (RCTs), design recommendations, effectiveness of non-lethal methods to prevent wild carnivore predation on farm animals, and conclusions. We place these in a global context with similar trials. We discuss gaps in evidence that should motivate investments in research and precautions among decision-makers at all levels. Inspired by experiments in the United Kingdom on badgers (Meles meles) to evaluate the effect of two interventions on transmission of bovine tuberculosis to cattle (Donnelly et al., 2003) and by Australian experimenters killing red fox (Vulpes vulpes) with poison, to evaluate the effect on predation of sheep (Greentree et al., 2000), we conducted our first predator-control RCT in Wisconsin, United States (Shivik et al., 2003). This first effort did not involve domestic animals (defined as commercial or subsistence, not captive colonies), hereafter farm animals or livestock. Still, this initial study proved the feasibility of robust RCT designs under field conditions, including crossover. Crossover occurs when treatment and control are reversed midway through a study so each subject experiences each condition. In the interim period before our next RCT, other studies proved the utility of randomized experiments to examine the effectiveness of non-lethal methods to influence wild, medium- to large-bodied carnivores preying on livestock (Bromley and Gese, 2001; Davidson-Nelson and Gehring, 2010; Gehring et al., 2010). Also, numerous experiments on American black bears (Ursus americanus) damaging non-mobile property suggested a need for RCTs with predators of livestock. Despite an RCT by Greentree et al. (2000) on fox poisoning in Australia, to date, no further reliable, peer-reviewed RCTs have been conducted on lethal intervention against predators of farm animals. See web panel 1 in Treves et al. (2016) for discussions of other quasi- or apparently experimental work that was considered unreliable because of design flaws or statistical biases. Greentree et al. (2000) concluded that there was no significant effect of poisoning red foxes and concluded much effort was wasted. It remains unclear if insights gained from the experimental killing of non-native carnivores preying on sheep in Australia can be generalized to native carnivores elsewhere. We raise this question because these and other non-natives are descendants of animals held in captivity by humans, which might affect their attraction to or fear of, humans in ways that modify the effectiveness of predator control. Our crossover RCTs build upon global research on non-lethal methods that used less robust designs (Stone et al., 2017). The second RCT by members of our group focused on large carnivores interacting with working llamas, alpacas, and sheep in Chile’s remote Andean altiplano (Figure 1). This study built on Ohrens et al. (2015) previous work in the region interviewing domestic animal owners by using participatory intervention planning methods for an RCT, citations in TrevesWallace and White, (2009) and Ohrens et al., (2019a). The recruited landowners participated by choosing to evaluate the non-lethal deterrent Foxlights, a commercially available random light projector triggered by nightfall (Ohrens et al., 2019a). Ohrens published his PhD dissertation as our first RCT with crossover design on working farm animals, aimed at protecting 11 herds of alpacas and llamas from pumas (Puma concolor), also known as cougar or mountain lion, using a light deterrent (Figure 2). His study paved the way for the next 4, including efforts at replication, while it also revealed the surprising attraction of Andean foxes (Lycalopex culpaeus) that we describe below, including an insight from his second experiment not yet published. Dr. Omar Ohrens at study site, Tarapaca, Chile with alpacas. Credit A. Treves. Puma in Chile. Credit O. Ohrens. The second peer-reviewed RCT aimed to evaluate herders using low-stress livestock handling methods, citations in Louchouarn and Treves (2023)—hereafter range riders (Figure 3). Range riders aimed to protect cattle from brown (grizzly) bears, gray wolves (C. lupus), pumas, black bears, and coyotes (C. latrans) in the Canadian Rockies of Alberta. This method aims to enhance and promote natural herding and antipredator behavior in cattle. This study incidentally shed light on the use of a pseudo-control rather than placebo control. Louchouarn’s work also sheds the most light on any previous work on the design of range rider interventions (Louchouarn and Treves, 2023). Namely, that a single experienced range rider could deter large carnivores as effectively as several range riders with less experience. Nevertheless, the frequency of range rider visits (dose effect) seemed important to understanding range rider effectiveness. Range riders in Alberta. Credit. N. X. Louchouarn. The third RCT as yet not peer-reviewed but published (Fergus, 2020), evaluated the same deterrent lights as Ohrens et al., (2019a) combined with fladry a visual deterrent composed of flagging hung at regular intervals (Figures 4 and 5) to deter black bears, coyotes, and gray wolves from five herds of diverse livestock types in Wisconsin, United States (Fergus, 2020). Another RCT replicated (Fergus, 2020) in hopes of combining datasets (Hermanstorfer, 2023). Hermanstorfer (2023) aimed to protect five herds of various livestock from coyotes, pumas, black bears, foxes, and free-ranging, domestic cats in Colorado, United States, using the same deterrents (Figures 4 and 5). The latter two studies have been combined to glean more insight into indeterminate effects and the role of inter-subject variation in treatment effects (Fergus et al. 2023; see also Treves and Khorozyan, pre-print). Fladry hung around a vehicle-killed deer carcass used in our earliest RCT (Shivik et al. 2003). Credit: A. Treves. Foxlight mounted on a stationary fence post, Cimitarra, Colombia. Credit: A. Pineda-Guerrero. Pineda-Guerrero’s dissertation RCT (Pineda Guerrero, 2023) aimed to protect 32 herds composed of a variety of livestock from approach by pumas and jaguars (Panthera onca) (Figures 5-8) at two Colombian forested sites, using stationary lights, as evaluated by Ohrens et al. (2019a), and a novel method never tested before: mobile deterrent lights (Figure 6). Her study offers the largest sample size subjected to our RCTs with crossover design (n = 25–32 depending on which effect was tested). Her study also offers insights into the use of true placebo controls versus inactive controls without placebo and an insight into taking farmers’ ideas about deterrents and implementing them within an RCT. Foxlight mounted on the back of a domestic animal, San Luis, Colombia. Credit: A. Pineda-Guerrero. Jaguar at Cimitarra, Colombia. Credit A.A. PinedaGuerrero. Puma at Cimitarra, Colombia. Credit A.A. Pineda-Guerrero. We present qualitative information and lessons we learned from field experiments in the following sections. Although some of the information may appear anecdotal because it arose from one or two studies, all of the qualitative information presented here and recommendations achieved consensus among the authors, unless otherwise noted and discussed below. We offer these lessons and insights as new hypothetical explanations or as methodological tips. We call repeatedly for replication of our findings and tests of our new hypotheses because we are aware that single studies or even the replicated ones we report, are not sufficient to guide regional policy or make confident recommendations to farmers. We hope, with these insights, we can help other researchers navigate randomized experimental trials with crossover designs in field conditions. Although we were not the first to complete RCTs under such conditions, our crossover designs and strict attention to avoiding sampling, treatment, measurement, and reporting biases (Treves et al., 2019) led to consistent methods that allow a systematic review of lessons learned. The noninvasive methods used in our RCTs allowed us to obtain an exemption from the Institutional Animal Care and Use Committee at the University of Wisconsin-Madison. Also, our protocols fs 2016-1071-CP005, 2019-0194, and 2021-0923-CP002 protected human subjects. Recruitment—Our general first step was to recruit owners or managers of livestock as participants. We use one or both individual recruitment interviews or group workshop-style recruitment methods (Treves et al. 2009). Some of the recruitment or attendance at workshops followed a snowball method or word-of-mouth between owners encouraging neighbors to attend and participate. We are aware this creates a potential bias due to non-independence in our attitudes dataset. Namely, word-of-mouth and snowball methods are likely to bring together participants of like mind and attitudes. Farms may also be managed similarly. Therefore, we emphasize the importance of replicating our findings in both RCTs and human dimensions. Once recruitment is complete, maintaining communication with participants becomes an ongoing task to establish trust. Participants may request to view photographs of wildlife on their farms and to stay informed about project updates. All but one RCT (Louchouarn and Treves 2023) involved our team conducting semi-structured interviews before, during, and after implementation so the ‘during’ interview could measure response to treatment and control separately. All interviews measure attitudes to carnivores, perceived effectiveness of non-lethal methods, and satisfaction with the participatory experiments. Our studies ended with small sample sizes of attitudes, perforce, because recruitment to the RCT with crossover was our priority. We more and more to the sample sizes of both of such Some of our (Pineda Guerrero, 2023) are the first of their to measure in attitudes an RCT with crossover to evaluate predator control. our findings replication before the findings policy or sample sizes of owners or managers and their herds our initial designs in four of the five studies sizes at various in the experiment because one one in and or our (Pineda two herds could not be protected by fladry (Fergus, the around herds (Pineda and one allowed two sheep to between and wild them in the subject cattle protected by our deterrent lights (Pineda Guerrero, 2023). The of such from to and may be for some but not We are aware of two owners may have the placebo control by on deterrent lights (Pineda Guerrero, 2023). of two owners also on our in and or our (Pineda Guerrero, 2023). In the latter we conducted a herds to treatment bias (Pineda Guerrero, 2023). We or participants on the lights when in can bias the treatment we researchers to these experiences and approach in ways that biases. attitudes to the non-lethal methods, carnivores, and with carnivores have been The largest study (Pineda Guerrero, 2023) for carnivores after the experiment and more attitudes about non-lethal methods, or no evidence the methods protected their We to this as a between perceived effectiveness and effectiveness (Ohrens et al. Also, in were more for owners or managers in control in 1 and on the treatment in to owners the of (Pineda Guerrero, 2023). Hermanstorfer (2023) mixed effects on attitudes with two of five owners more to carnivores and of five remaining that farm owners may be of carnivore but are more with the which in them to be more or in the The findings of (2023) and Hermanstorfer (2023) both may be and might be the need for and replication with against non-independence in as Although with owners a of and in and we individual owners to work with in general and more to in experimental studies including than and in other consistent with research of interest groups more than their et al., the of with been project et al. for the of Therefore, we for farm animals and conducted RCTs on alpacas, llamas, and including and animals in small pastures or and in herds or herds of mixed animals that from to herds can about mixed effects of non-lethal of a significant effect of a treatment may from attraction to one subject and from subject which might the or of individuals in each subject (Fergus et al. 2023; 2023). Nevertheless, herds are not (Louchouarn and Treves, 2023). in subjects is an impossible and we impossible because individual of subject animals efforts at Crossover a more and against between subjects. not by the research that the or a treatment effect (Ohrens et Treves et al. 2023). We when is We also of confounding effects to a treatment or effect in RCTs with crossover design (Pineda Guerrero, 2023). our first RCT with domestic animals, which experienced of livestock (Ohrens et al. have been (Louchouarn Treves 2023) 1 in the (Pineda Guerrero, 2023) on and Hermanstorfer (2023) Therefore, of carnivore predation on livestock in our We followed for protecting their while treatment or control. the deterrent methods or even a novel in owners suggested deterrent lights to the of livestock animals. lights to the of domestic animals before the RCT with The of proved impossible for some animals also when cattle but more by some the of mobile lights Although we no significant in effects of mobile lights versus stationary lights herds (Pineda Guerrero, we suggest further research on mobile deterrents can by large forested or animals around subject but not wild not the RCT may involve owners and livestock managers in the of more to prevent studies also conducted indirect to approaches by carnivores (Louchouarn and Treves, 2023; Ohrens et al. and in of us individual between carnivores or between of carnivores may have our The non-lethal methods we tested were to deter medium- and large-bodied and in that also bears or and Ohrens et al. (2019a) a deterrent effect of on pumas but a to Andean His second study pumas at that were not by the same the in our of predator we see a major in understanding of the relationship between carnivore approaches to humans or domestic animals and the of wild carnivores are elusive are of are to by and Gese, et al., et al., methods such as and and that wild carnivores approach to humans or domestic animals without any Our experiments with of farm animals could with about the of approaches to subject approaches are in some and less so in in which we have suggest two general and studies be approaches are et al., 2009). variables should be with and two types that may not for domestic and such as or of domestic animals. the methods for approaches should approaches rather than approaches at because remote carnivores not pose a to domestic animals the potential predator the individual variation in to approach is likely to be a We that an individual of to such as human lights, that to a light to deterrent been and and effectiveness is a of both non-lethal and lethal interventions and or random deterrents may as might deterrents (Shivik et al., variables with deterrent may in ways with the of wild animals and their with the The of initial or initial followed by in individual carnivore behavior more Louchouarn Treves (2023) of gray wolves the first of the RCT in treatment but not the second which in the when wolves might range more We that by wolves could have them to the new range but this as wolves to the new humans (Louchouarn Treves 2023). emphasize and of deterrent effectiveness using methods and in our is that non-lethal methods are on farms with previous such bias from and (Treves and Khorozyan, Treves and Treves, 2001; we randomized designs and for with treatment and We with other that individual carnivores may to deterrents and a of deterrent is than a single when two of us tested a light deterrent and a fladry the on carnivore approaches into and to evaluate between and (Fergus, et al., 2023; Hermanstorfer 2023). work is to our understanding of deterrents and even single interventions against a of (Louchouarn Treves 2023; et al., 2017). The effect of non-lethal deterrents may on by carnivores. first (Ohrens et al. was remote with lights and large between human so pumas might have been with lights in that altiplano In Andean foxes were more to lights because more human or foxes pumas, which lights more as in the of human the of second was more with both humans and pumas This was our first to and effects on one carnivore while the for for and an RCT using the same deterrent lights as Ohrens et al. (2019a). (2023) that jaguars the same deterrent lights at two of study but a second field revealed jaguars the deterrent lights at the first study when the experiment the next The attraction and of jaguars was not so that can some individual carnivores to approach deterrents and deter them for our four RCTs using light deterrents of the suggest these lights that at in are to deterrent effects on carnivores unless individuals are to Also, the of two of foxes and their (Ohrens et and should is one potential but the role of human should be considered at the same the of effect and results to non-lethal methods a bias if no by lethal methods also of effect and effects et al., 2018; and Treves et al., The field from more that all interventions have effects and from understanding the of such have to of our RCT with depending on conditions. Louchouarn Treves (2023) used a pseudo-control because owners the true placebo of no range rider were to an RCT and a pseudo-control in which range for a and were with an experienced range We this by maintaining the experienced range rider as the pseudo-control in all subject herds and his work with one to two range riders for the treatment condition. This us new to hypotheses about the of range frequency of their around and of as one was the period pseudo-control and treatment on the same and to our RCT there been and attacks on cattle in these same our results suggest range riders were We that a higher frequency of range rider visits be important for in the Colombian (2023) evaluated effects by the of light two treatment by their as mobile or and two types of control placebo control or inactive large sample size (n = (2023) was to evaluate while some statistical Nevertheless, treatment effects were (Pineda Guerrero, 2023). we the that any intervention is an experiment and that be to the effectiveness We are confident about two of non-lethal methods to prevent predation on livestock. The first is the of herders using low-stress (Louchouarn Treves 2023). The second is the of light deterrents when wild carnivores are not to human lights with the discussed or when lights are with a both can be farms or herds (Ohrens et al. et al. 2023). Our in fladry is by from other RCTs and studies and Gehring, 2010; et al., et al. 2020) and work in other countries is new insights into effective non-lethal methods with sample sizes than have been achieved et al., et al., 2020). The field to the of Gehring et al. to design non-lethal methods so that domestic animal owners can and them and offer managers a way to the triple-win for wild and domestic animals in addition to We that any review and of evidence by the authors, with which other may that to evidence as we have not dismiss or the evidence we have We are about a for individuals and conducting research and to evidence and dismiss research that at about Therefore, we open on methods, and Treves and and domestic animals carnivore and of the and of at the University of his PhD at his as a and in studies of before to animal The members of the peer-reviewed and an of other including of the and of and to work of the on experimental evidence for the of of large carnivores, and of wild carnivores to a in and through the at the University of the effectiveness of the deterrents fladry and at reducing carnivore visits to livestock farms within the to and deterrents carnivore and the deer seemed to be livestock on is now at the in the these to the on the as the wildlife also as a for the to the wildlife in and to study the Hermanstorfer was his of at the for at the University of His research carnivore in through trials of non-lethal deterrent and interviews with livestock a as a research with The (Stone also as a at for the of in X. Louchouarn PhD and as a research at the at the University of of from University with in wildlife and a of and at focused on Her PhD research focused on the effectiveness of and to in also and published peer-reviewed about carnivore Omar Ohrens is a from the University of Chile. his as an at the working on several wildlife the a and at the University of at the His research focused on the of carnivores, including pumas, and in Chile. His research been by the and and the is a for the Puma on in his work on of pumas and and and in is a Colombian from University with the to in and conducted research in with and in Colombia. been a of the at University of Wisconsin-Madison. In Her research focused on the and perceived effectiveness of a non-lethal method to prevent and attacks in two in Colombia. also conducted a study in attitudes of participants in Her research carnivore the human of and research to the of a from the and in We are also for and from for and at the The to and of Animal of and available upon request of

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.