Towards a unified generic framework to define and observe contacts between livestock and wildlife: a systematic review

Literature search and data extraction

We defined livestock as ‘farmed domesticated mammals’ (FAO, 2020), wild animals as ‘free-ranging non-domesticated mammals’, and contact as ‘activity implying an interaction or association between species including the shared use of resources such as farmland’. The terms interaction and contact were used synonymously within the literature, but contact is used here for consistency. The systematic review question was “Which methods have been used to assess the frequency of, and types of, contacts between wild animals and livestock or livestock farms worldwide?”.

Search terms for wildlife, livestock and type of contact were combined by the Boolean operators ‘OR’ and ‘AND’ to identify publications that investigated contact between any wild and domestic mammal (Table S1). Search terms were based on common species names, and generic terms such as ‘feral’, ‘wildlife’, ‘livestock’ and ‘farm’. Searches were conducted in CAB Abstracts, Scopus and Pubmed. CAB Abstracts is a comprehensive database of life science research with broad coverage of veterinary literature in particular, and Scopus has a broad coverage of interdisciplinary journals (Grindlay, Brennan & Dean, 2012; Aghaei Chadegani et al., 2013).

Search results were consolidated into Microsoft Excel and duplicates were identified and removed using queries followed by manual inspection. Titles, abstracts and full texts of the retrieved publications were evaluated by SAB against pre-specified exclusion and inclusion criteria (Table 1). Any papers for which the criteria were not clear were also evaluated by JAD. In all such cases both authors agreed on the final decision. We wished to capture publications that collected, used or analysed data to investigate direct or indirect contacts between farmed livestock and terrestrial wild mammals whose adult bodyweight is typically >5 kg. Specifically, publications were included if they attempted to quantify, characterise, or identify risk factors for livestock-wildlife contacts. Only articles in English and those accessible to researchers were included. All reasonable efforts were made to access papers that passed abstract screening. We excluded studies in which predation events were the sole indicator of livestock-wildlife contacts, and studies of wild animals that were not free-living, were tamed or were relocated for the purpose of the study. Publications until 11 November 2019 were included, and no time restrictions were applied to the start of the search. Working definitions of direct and indirect contact were developed before performing the literature search and used to avoid ambiguity when evaluating publications for inclusion.

Direct contact was provisionally defined prior to reviewing the papers as physical contact between at least one wild animal and one farm animal. Indirect contact was provisionally defined as contact between at least one wild animal and a resource used by at least one farm animal including, but not limited to, food, water and space. Therefore, studies that investigated wildlife and livestock shared resource use, but did not explicitly investigate contacts, were included. These definitions were used throughout the process of identifying and analysing the papers in this review. Study data was extracted and livestock and wildlife species, observation methods and definitions were categorised. Where available, the power of each study, defined as the likelihood of detecting contacts, was recorded. Themes that emerged during data extraction were grouped into seven broad study themes, namely behavioural, competition, conservation, disease, human-wildlife conflict, methods papers and wildlife management (Fig. S1). Where studies had more than one theme, themes were subjectively allocated as dominant (primary) or secondary based on the aims of the study. Results were visualised and plotted using R (version 3.6.3 (R Core Team, 2020)) and R packages listed in Table S2.

Development of a generic unified framework

Following categorisation of definitions, a generic unified framework was developed by grouping and identifying commonalities in definitions of ‘direct’ and ‘indirect’ contact, namely relating to space and time. The spatial and temporal limits separating relevant contacts from inconsequential contacts and non-contact events were identified for each study, and a framework was developed based on defining contacts in relation to both space and time. Using this framework, relevant contacts were defined using the parameters of critical space (S_C) and critical time (T_C). We defined S_C as the critical space (distance or area) between animals below which a contact relevant to the study is considered to have occurred, and T_C as the critical time window within which a relevant contact is considered to have occurred.

Search results, quality appraisal and themes

A total of 43,032 papers were identified by the search terms across all three databases, of which 30,080 were unique results. After screening using the exclusion and inclusion criteria in Table 1, 122 publications remained in the final analysis (Fig. 1). Publication date ranged from 1980 to 2019, with 117 (96%) published in the last 20 years (Fig. 2). Studies conducted in Europe, North America and Africa made up 89% of the results (Table S3) with the USA and UK producing the most publications (21% and 18%, respectively).

Low study power was mentioned briefly in only 11 (9%) publications and statistical power calculations were not performed. The level of uncertainty was acknowledged in 64 (53%) publications.

Disease was the dominant theme and featured in 80 of 122 studies (66%), followed by human-wildlife conflict (22/122; 18%), competition between wildlife and livestock (17/122; 14%), conservation (16/122; 13%), wildlife management (11/122; 9%), behavioural studies (3/122; 2%) and methods validation (2/122; 2%) (Fig. S1). Within the disease-themed papers, Mycobacterium bovis was the most studied pathogen (49/80; 61%) followed by foot-and-mouth disease virus (8/80; 10%) (Tables S4 and S5). Wildlife-cattle contacts were the focus of 98 of the 122 studies (80%) and a further 22 studies (18%) focussed on sheep, pigs, farmed deer and camelids. The most studied wildlife species were deer (30/122; 25%), wild pigs [including wild boar] (26/122; 21%) and badgers (25/122; 20%: Figs. S2 and S3). The wildlife species were not specified in 11 papers, some of which studied wild ungulates competing for livestock grazing (Mizutani, Kadohira & Phiri, 2012; Sitters et al., 2009; Crawford et al., 2019), others that concerned wildlife as hosts of cattle diseases such as bovine tuberculosis (Munyeme et al., 2010; Witmer et al., 2010; Katale et al., 2013) and foot-and-mouth disease (Brahmbhatt et al., 2012; Molla et al., 2013), and the remainder that were completely unspecified.

Methods used to observe livestock-wildlife contacts

Methods that monitored both livestock and wildlife species were used in 88 publications (72%) whereas 34 studies (28%) monitored wildlife only. Camera trapping was the most frequent method of monitoring wildlife (37 studies, 31%), and was most prominently used in badgers, deer and wild pigs (Fig. 3). GPS collars were the second most used method to monitor wildlife (29 studies, 24%), and while they were also used predominantly on badgers, deer and wild pigs, they were used proportionally more than cameras to monitor predators such as big cats and wolves and large herbivores such as buffalo, wild horses and elephants. Other methods used to monitor wildlife were direct visualisation (21; 17%), farmer questioning (20; 16%), radio-transmitters (17; 14%), activity signs (15; 12%) and proximity loggers (7; 6%). Some studies utilised more than one observation method, and therefore the numbers of studies exceed 122 (100%) Studies that monitored livestock tended to use the same methods as for wildlife, although 10 studies dedicated fewer resources to monitor livestock; for example (Pruvot et al., 2014) used GPS collars to monitor wild deer and farmer questioning to monitor cattle behaviour. Studies that did not monitor livestock tended to infer wildlife-livestock contact from monitoring only the activities of wildlife on or around livestock holdings, such as on pasture, in buildings and the shared use of resources such as livestock feed.

A variety of methods were used to observe different types of contact data (Fig. S4). Methods such as GPS collars and radio-tracking (telemetry) were used to collect the locations of wildlife (e.g., Barasona et al., 2014; Raizman et al., 2013; Cooper et al., 2008), whereas proximity loggers were used to detect close proximity contacts between livestock and wildlife or with postulated high-risk disease transmission areas such as badger latrines (e.g., Drewe et al., 2013). Camera traps and direct visualisation were used to observe behavioural activity, such as nose-to-nose contacts between cattle and badgers (Tolhurst et al., 2009), foxes taking piglets from farrowing huts (Fleming et al., 2016) and wild boar eating from cattle troughs (Kukielka et al., 2013). Some methods were used to detect the presence of wild animals on farms or on pasture only, such as surveys of activity signs to detect wild boar rooting on sheep pasture (Guillermo Bueno et al., 2010) and GPS collars to demonstrate the avoidance of livestock pasture by lions (Oriol-Cotterill et al., 2015). Thirty studies combined more than one method to monitor wildlife, such as (Wyckoff et al., 2012) which combined activity signs, GPS collar data and camera traps to monitor feral swine activity at and around domestic pig pens. The majority of studies, however, used only one method and were able to collect information about the type of contact defined by the study.

Definitions of direct and indirect contacts

Definitions for both direct contact and indirect contact were provided by 27 studies, with a further four defining direct contact only and 54 defining indirect contact only (Tables 2 and 3). Definitions of direct contact tended to focus on the spatial distance between wildlife and livestock at one point in time (Table 2). Definitions of indirect contact tended to focus on the use of space or resources by wildlife in a location previously or subsequently occupied by livestock, within a certain time frame (Table 3). There were some variations to these trends: two studies specified a time frame longer than one time point to define direct contact (Lavelle et al., 2016; Cooper et al., 2010). The amount of time was usually determined by the context of the study, such as the survival time of a specified pathogen in the environment, known as the critical time window of a contact (Cowie et al., 2016). Contacts were also defined in 15 studies as the shared use of resources between livestock and wildlife, such as feed and water. There were large variations between studies in the defined distances and time windows, with direct contact distances ranging from physical contact (seven studies) to within 120 metres of each other (one study), and indirect definitions ranging from within the same camera image (two studies) to within 50 kilometres of a location (one study). There was less variation in definitions between studies with similar contexts and aims. For example, among M. bovis transmission studies in cattle and badgers, the definition of direct contact ranged from physical contact to within two metres (six studies), and indirect contacts were defined as presence on farmland, sharing of resources and visits to badger latrines by cattle (20 studies). Importantly, no definition of contact was provided in 25 studies (44%) that reported direct contacts, and 34 studies (29%) that reported indirect contacts.

Regardless of the contact definitions or methods used to observe contacts, direct contacts were detected much less frequently than indirect contacts. For example, one study (Lavelle et al., 2016) found no instances of cattle within two metres of deer, compared to over 40,000 indirect contacts of deer with cattle via shared feed. Overall, the median number of direct contacts between wildlife and livestock was in single figures, whereas the median number of indirect contacts occurred in the order of hundreds or even thousands. The types of contacts reported between livestock and wildlife, and the methods used to observe contacts, are summarised in Table 4. Low study power was acknowledged, but not calculated, by 11 studies (9%), and is likely to be a feature of many more which did not report it. No studies reported adequate power. The low power of studies to observe rare contacts, coupled with the variation in, or lack of, contact definitions, makes it very difficult to compare the effectiveness of the methods used to observe wildlife-livestock contacts.

Proposed unified framework to define direct and indirect contacts

Space (area or distance between animals) and time were crucial components of the varied definitions of direct and indirect contact in this review. In an effort to unify these parameters, a novel generic framework to categorise wildlife-livestock contacts is proposed in Fig. 4, based on the locations of individuals in space and time. Using this framework, we propose that the contact type (direct or indirect) is defined using the two parameters S_C and T_C. Multiple critical thresholds can be used within the framework to differentiate between definitions of direct contact (S_C1 and T_C1) and indirect contact (S_C2 and T_C2). For a direct contact to occur, two individuals are within the same pre-specified critical space (distance or area: S_C1) within a pre-specified critical time window (T_C1). Similarly, for an indirect contact to occur, animals are within another pre-specified critical space (S_C2) within another pre-specified critical time window (T_C2). The reader is directed to Fig. 4 for examples from the literature of possible combinations of S_C and T_C. T_C2 may be the same as T_C1 (if S_C2 is larger than S_C1: compare example A with example B in Fig. 4) or T_C2 may be different from T_C1(in which case T_C2 will usually, but not always, be larger than T_C1: compare example A with examples C, D, E and F in Fig. 4). Similarly, S_C2 may be the same as S_C1 (if T_C2 is larger than T_C1: compare example A with examples C and E in Fig. 4) or S_C2 may be different from S_C1 (in which case S_C2will usually, but not always, be larger than S_C1: compare example A with examples B, D and F in Fig. 4). Same, near and different are used here to illustrate spatial and temporal differences between examples. These terms are relative and will vary along with S_C and T_C depending on the system being studied, the objectives of the study and other factors such as host behaviour and the biology of the pathogen, in the case of disease studies; therefore, values for T_C1, T_C2, S_C1 and S_C2 should be decided in advance of a study being conducted, and they should be clearly reported when data are presented.

Although the exact values of the critical distance between animals and the critical time window over which this happens will depend on the system being studied as well as the specific objectives of each study, the adoption of this generic framework to define direct and indirect contacts will help identify studies with similar definitions where results are more easily comparable.

The need for a generic unified framework

This review has found that definitions of contact are wide-ranging and highly dependent on the context of the study. Definitions can vary depending on the species and demographics of the wildlife and livestock involved, the methods used to detect contacts and the system being studied such as the environmental conditions and pathogen characteristics in studies where contacts are representative of disease transmission. Definitions of direct contact were extremely diverse, ranging from direct physical contact to animals being merely within a hundred metres of each other. Indirect contact ranged from animals sharing resources, being within five kilometres of each other or overlapping in home ranges, and the time window that these events occurred in varied from hours to weeks.

The aim of this generic unified framework is to promote consistent reporting of definitions of contacts enabling comparisons to be made between the approaches of wildlife-livestock contact studies, regardless of the species or pathogen studied or the context of the study. This is needed because our systematic review found that while wildlife-livestock contact data was collected in terms of space and time, some studies omitted space or time in their definitions, or there was a complete lack of a definition. Conflicting and overlapping definitions of direct and indirect contact were also identified. Making any sort of meaningful comparison between such studies is challenging. For example it is difficult to assess what, if any, implications there are for deer-cattle disease transmission from a behavioural study showing deer avoid cattle despite similar habitat preferences (Mattiello et al., 2002), without knowing what types of contact (e.g., direct or indirect; what specific types) were likely to be meaningful. It is even difficult to compare studies within the same system, for example establishing the relevance of cattle-badger contacts for bovine tuberculosis transmission when some studies define a contact as ‘presence on farm’ (Mullen et al., 2015; Sleeman, Davenport & Fitzgerald, 2008) and others define it as ‘presence in buildings’, and neither study defines the time window. Use of the generic unified framework would enable consistent reporting of definitions between studies and is an important step if the results of wildlife-livestock contact studies are to be comparable.

Applications of a generic unified framework

Models that incorporate empirical rather than theoretical information on the frequency and duration of contacts important for disease transmission are more likely to be useful for disease mitigation (Craft, 2015). The use of a standardised definition framework in future studies of livestock-wildlife contacts would enable consistency in datasets and enable the retrospective selection of contact data relevant to a particular model, which could then be incorporated in a similar way to the data used in recent bovine tuberculosis transmission models (Wilber et al., 2019; Silk et al., 2018). The generic unified framework proposed in this current paper could also be useful in designing livestock-wildlife contact studies, since defining the type of contact to be detected—in addition to practical considerations, such as an area’s signal strength affecting the viability of GPS device use—helps with the choice of detection method. The framework is also flexible and applicable to different contexts, species and diseases since it allows for the variation in definitions seen in this review, and it is hoped it will broaden the range of future livestock-wildlife contact studies.

To resolve human-wildlife conflicts usually requires robust livestock-wildlife contact studies. At least 120 studies that only used predation events to infer livestock-wildlife contacts were excluded from the review, yet predators –particularly wolves –were the second most commonly studied group of wild mammals. Given that predation studies appear to form a large proportion of wildlife-livestock contact studies, this is an area where adoption of the generic framework could help design meaningful contact studies to evaluate preventive measures without relying solely on predation events.

Further development of the generic unified framework

The generic unified framework does not provide an overall consensus on definitions of direct and indirect contact, but does provides a structure with which to start this process. While using the generic unified framework provides a standardised definition of contact in time and space, identifying the types of contact that are relevant to the study, and thus the values of S_C and T_C, will vary depending on the objectives and context of each study. While a universally accepted set of definitions for contacts is difficult to devise, we hope that by defining Sc and Tc here we will encourage the start of the debate around (and between) studies of similar contexts, and perhaps then acceptable ranges for these values will emerge. Developing a framework for deriving S_C and T_C, based upon the species studied, environment, pathogen and methodology is beyond the scope of this review, and would be a necessary next step so that wildlife-livestock contact rates could be comparable between studies of similar contexts. For example, for disease studies, it would be advisable that S_C and T_C were based on values below which transmission is likely to occur, such as aerosol dispersion distance and environmental survivability. For any system, there may be a range of appropriate values for S_C and T_C.

The generic unified framework presented in this paper is a step towards being able to compare observation methods and contact data in order to standardise and evaluate different monitoring methods. This is important as our systematic review revealed that the methods used to observe livestock-wildlife contacts to date have often had low detection rates and therefore been of low power due to the difficulty of monitoring cryptic wildlife species, and the rarity of some types of wildlife-livestock contacts, particularly direct contacts. Further considerations for the comparison of observation methods are the representativeness of individuals monitored, especially with methodologies that require the marking of individuals such as GPS and proximity loggers, and a standardised system for relativizing the number of contacts with regards to the total observation effort. For example, two studies will not be comparable if study A only uses 3 camera traps and study B uses 100 camera traps, or if study C collects GPS locations every hour when study D collects only one GPS fix per day. Reporting representativeness of individuals and relativizing contact rates in terms of total population will go some way to establishing the power of wildlife-contact studies. Furthermore, it may be useful for studies to indicate the detection limits of the methodology used, in terms of space and time.

Scope of existing wildlife-livestock contact studies

This review has identified the narrow scope and limited geographic range of livestock-wildlife contact studies, which means the data summarised in this review should not be considered representative of all wildlife-livestock contacts worldwide. The majority of studies focussed on cattle-wildlife contacts and diseases of cattle. Bovine tuberculosis (infection with M. bovis) featured prominently, indicative of the economic and potentially zoonotic importance of this disease to the USA and UK, where the most livestock-wildlife contact studies were conducted (De la Rua-Domenech, 2006; O’Brien et al., 2011). Foot-and-mouth-disease was the most studied viral pathogen and this is most likely explained by its broad geographical spread and high economic impact (Knight-Jones & Rushton, 2013). This demonstrates the human-centric view of the wildlife-livestock interface, with most focus on the impacts on humans and domestic animals, and very little (if any) focus on the value of wildlife (Chardonnet et al., 2002). There were, however, some livestock-wildlife contact studies of high impact conservation importance such as infection with Mannheimia spp. in bighorn sheep (Ovis canadensis) and Pasteurella spp. in saiga antelope (Clifford et al., 2009; O’brien et al., 2014; Beauvais et al., 2019). If we are to collect more (and better) wildlife-livestock contact data that include a broader range of species and contexts, careful consideration must be used when selecting the most effective and practical observational method for monitoring cryptic wildlife species.

This review highlights that observing contacts between multiple species is possible and can yield high quality information. Increasing the efficiency of monitoring methods would justify their use for more applications. Health surveillance systems at livestock-wildlife interfaces have been suggested as a method to detect and control emerging diseases along with preventing contact between wildlife and livestock (Gortazar et al., 2015). Preventing high-risk contacts may be more cost-effective than surveillance, but the effectiveness of prevention strategies will need to be evaluated by monitoring contacts, or lack thereof. More efficient monitoring will also allow for quantitative risk assessments of wildlife-livestock contacts which are presently difficult to conduct due to a limited understanding of potential contacts leading to pathogen transmission (Miller, Farnsworth & Malmberg, 2013). Some observation methods such as camera traps are likely to have the ability to identify new potential transmission routes between livestock and wildlife (e.g., the use of cattle salt licks by raccoons (Witmer et al., 2010)), and may identify livestock-wildlife contacts previously not considered (e.g., observing farm visits by foxes during a study focussing on badgers (O’Mahony, 2015)). Identifying wildlife species that may be the origin of rapidly emerging human diseases is a priority to prevent future pandemics (Morse et al., 2012). In situations where human infections are mediated by livestock, rapid implementation of observational methods to detect contacts between wildlife and livestock could more quickly identify wildlife hosts and risky behaviours. In order to determine the efficiency and efficacy of different observational methods, the methods used and data collected by them must be comparable, hence the need for a unified framework.

Limitations of this review

Our study has some limitations which we summarise here. At present, our generic unified framework does not explicitly account for disease transmission via vectors or fomites, although the latter will to some extent be captured within our definition of indirect contact. In order that observation methods were likely to be comparable between species, we focussed on terrestrial mammals so did not address diseases primarily hosted by birds or bats such as avian influenza, Nipah virus and Hendra virus. Small terrestrial mammals (<5 kg) were also not included for this reason, and because a disproportionate number of rodent studies focus on their roles as laboratory animals or farm pests, and not on contacts with livestock. While the generic unified framework may be applicable to these types of wildlife, it is unclear which observational methods seen in this review would be most effective or efficient, and the conclusions drawn from this review may not be reflective of systems that involve other taxa.