ArticlePDF Available

Mapping attack hotspots to mitigate human–carnivore conflict: approaches and applications of spatial predation risk modeling

August 2015
Biodiversity and Conservation 24(12)

Authors:

Defenders of Wildlife

A major challenge in carnivore conservation worldwide is identifying priority human–carnivore conflict sites where mitigation efforts would be most effective. Spatial predation risk modeling recently emerged as a tool for predicting and mapping hotspots of livestock depredation using locations where carnivores attacked livestock in the past. This literature review evaluates the approaches and applications of spatial risk modeling for reducing human–carnivore conflict and presents a workflow to help conservation practitioners use this tool. Over the past decade 18 studies were published, most which examined canid and felid (10 and 8 studies on each group, respectively) depredation on cattle (14) and sheep (12). Studies employed correlation modeling, spatial association and/or spatial interpolation to identify high-risk landscape features, and many (but not all) validated models with independent data. The landscape features associated with carnivore attacks related to the species (carnivore and prey), environment, human infrastructure and management interventions. Risk maps from most studies (14) were used to help livestock owners and managers identify top-priority areas for implementing carnivore deterrents, with some efforts achieving >90 % reductions in attacks. Only one study affected policy, highlighting a gap where risk maps could be useful for more clearly communicating information to assist policymakers with large-scale decisions on conflict. Studies were used to develop a six-step workflow on integrating risk modeling into conservation. This review reveals a need for future predation risk modeling to focus more on validating models, accounting for feedbacks and impacting conflict-related policy in order to reliably improve the mitigation of human–carnivore conflict globally.

Content uploaded by Jennifer R. B. Miller

Content may be subject to copyright.

Content uploaded by Jennifer R. B. Miller

Content may be subject to copyright.

REVIEW PAPER

Mapping attack hotspots to mitigate human–carnivore

conﬂict: approaches and applications of spatial predation

risk modeling

Jennifer R. B. Miller

1,2

Received: 12 June 2015 / Revised: 7 August 2015 / Accepted: 12 August 2015

!Springer Science+Business Media Dordrecht 2015

Abstract A major challenge in carnivore conservation worldwide is identifying priority

human–carnivore conﬂict sites where mitigation efforts would be most effective. Spatial

predation risk modeling recently emerged as a tool for predicting and mapping hotspots of

livestock depredation using locations where carnivores attacked livestock in the past. This

literature review evaluates the approaches and applications of spatial risk modeling for

reducing human–carnivore conﬂict and presents a workﬂow to help conservation practi-

tioners use this tool. Over the past decade 18 studies were published, most which examined

canid and felid (10 and 8 studies on each group, respectively) depredation on cattle (14)

and sheep (12). Studies employed correlation modeling, spatial association and/or spatial

interpolation to identify high-risk landscape features, and many (but not all) validated

models with independent data. The landscape features associated with carnivore attacks

related to the species (carnivore and prey), environment, human infrastructure and man-

agement interventions. Risk maps from most studies (14) were used to help livestock

owners and managers identify top-priority areas for implementing carnivore deterrents,

with some efforts achieving [90 % reductions in attacks. Only one study affected policy,

highlighting a gap where risk maps could be useful for more clearly communicating

information to assist policymakers with large-scale decisions on conﬂict. Studies were used

to develop a six-step workﬂow on integrating risk modeling into conservation. This review

reveals a need for future predation risk modeling to focus more on validating models,

Communicated by Karen E. Hodges.

Electronic supplementary material The online version of this article (doi:10.1007/s10531-015-0993-6)

contains supplementary material, which is available to authorized users.

&Jennifer R. B. Miller

jennie.r.miller@gmail.com

Yale School of Forestry & Environmental Studies, 195 Prospect Street, New Haven, CT 06511,

USA

Wildlife Institute of India, Post Box 18, Chandrabani, Dehra Dun, Uttarakhand 248001, India

123

Biodivers Conserv

DOI 10.1007/s10531-015-0993-6

accounting for feedbacks and impacting conﬂict-related policy in order to reliably improve

the mitigation of human–carnivore conﬂict globally.

Keywords Attack hazard !Carnivore conservation !Grazing management !

Livestock depredation !Nonlethal carnivore control !Predator–prey interactions

Introduction

Carnivore attacks on livestock cause substantial conﬂicts between humans and carnivores,

making livestock depredation a major challenge for pastoral communities and conservation

practitioners worldwide (Treves and Karanth 2003; Woodroffe et al. 2005; Inskip and

Zimmermann 2009). Though cost-efﬁcient nonlethal techniques exist for reducing carni-

vore attacks (McManus et al. 2014; Lichtenfeld et al. 2014), these tools are often time-

intensive and difﬁcult to implement across the expansive landscapes where carnivores and

livestock interact (Shivik 2006). As a result, many livestock owners continue to use lethal

measures to reduce carnivore attacks (Ogada et al. 2003; Inskip et al. 2013), contributing to

rapid carnivore population declines and loss of attendant ecosystem service values for

humans (Ripple et al. 2014). However, a spatial statistical approach known as predation

risk modeling that identiﬁes high-priority ‘conﬂict hotspots’ where carnivores are likely to

attack livestock is rapidly emerging as a tool for informing livestock management and

carnivore conservation. A small yet notable number of modeling studies on carnivore–

livestock predation risk have been published over the past decade, and the accelerating use

of predation risk modeling suggests that the technique may become more widely used for

guiding mitigation of human–carnivore conﬂict. Thus, there is now a critical need to

evaluate the ﬁndings and impacts of predation risk modeling studies and prescribe best

practices for future studies.

Predation risk modeling was originally developed to quantify the ecological interactions

between predators and their natural prey. Models built in a spatially explicit context

(hereafter referred to as spatial risk models) use site-speciﬁc data on past species inter-

actions to quantify and map associations between predator–prey encounters and landscape

attributes (Hebblewhite et al. 2005). The concepts underlying spatial risk modeling,

including models involving carnivore–livestock interactions, stem from principles of

spatial ecology, predation risk and optimal foraging theory. How carnivores and livestock

interact spatially is largely shaped by the interplay between predator hunting strategies and

prey avoidance tactics (Sih 1984; Kluever et al. 2009; Laporte et al. 2010). To increase the

likelihood of a successful attack, carnivores tend to search for prey at locations where prey

are in high densities and in an exposed environment (Hopcraft et al. 2005; Balme et al.

2007; Atwood et al. 2009). To reduce vulnerability, prey exhibit antipredator behavior and

select locations that balance a trade-off between maximizing basic biological and social

needs (e.g., nutritional intake, physiological stress, mating) and minimizing exposure to

predators (Brown et al. 1999; Stankowich and Blumstein 2005; Laundre

´et al. 2010).

Though livestock tend to show attenuated antipredator responses and people often restrict

their spatial movements, risk avoidance and optimal foraging choices nonetheless can

affect their resource use decisions. In experiments with livestock, predator cues resulted in

increased vigilance, grouping and movement variation, decreased foraging rates and

association with more open habitats (Kluever et al. 2008,2009; Shrader et al. 2008;

Biodivers Conserv

123

Laporte et al. 2010). Hence, the sites where livestock are present on a landscape (prey

availability) may differ from locations where carnivores can successfully kill (prey ac-

cessibility) (Hebblewhite et al. 2005; Hopcraft et al. 2005; Trainor and Schmitz 2014). The

spatial nature of carnivore and livestock decisions also means that their interactions change

with environmental attributes across heterogeneous landscapes (Wydeven et al. 2004;

Gorini et al. 2012), forming gradients of predation risk where carnivores are more or less

likely to attack (Brown et al. 1999; Laundre

´et al. 2010).

Spatial risk models quantify the landscape attributes correlated with sites where car-

nivores kill prey and in doing so, reveal the spatial features associated with where car-

nivores are most likely to kill livestock. However, since spatial risk models are (most

often) based on livestock kill data, they describe the landscape attributes associated with

only a subset of overall predation risk: locations where carnivores successfully killed

livestock but not where attacks were unsuccessful or where attacks did not occur. In other

words, spatial risk models quantify the realized predation risk (where direct mortality

occurs) rather than the overall fundamental predation risk (where mortality and indirect

non-consumptive effects may occur) (Lima and Dill 1990; Schmitz et al. 1997; Hebble-

white et al. 2005). This distinction is particularly important to recognize if one wishes to

understand the non-consumptive effects of predators on livestock stress, which may affect

animal foraging, weight gain, reproduction and other factors tied to overall productivity

and ﬁnancial value (Howery and Deliberto 2004). In this case, telemetry data on carnivore

and livestock movement would be a better measure of fundamental predation risk than

livestock kill sites (Gorini et al. 2012). The literature on predation risk modeling of

carnivore–livestock interactions has not addressed the difference between realized and

fundamental predation risk as clearly as has the literature on wild predator–prey interac-

tions (Hebblewhite et al. 2005; Atwood et al. 2009; DeCesare 2012), though this topic

merits detailed discussion. However, since the non-consumptive effects of carnivores on

livestock have yet to be extensively demonstrated but carnivore depredation on livestock

(mortality) provokes human–carnivore conﬂict with serious implications for human

livelihoods and carnivore conservation, this review focuses only on spatial predation risk

modeling with livestock kills. The term ‘predation risk’ is used for consistency with current

literature and here refers to the realized predation risk, or the threat of mortality as mea-

sured by livestock kills.

When applied to carnivore–livestock interactions, spatial risk models reveal locations

and associated habitat features where carnivores kill livestock, providing both quantitative

and visual guides for targeting conﬂict mitigation interventions (Marucco and McIntire

2010; Treves et al. 2011). However, it is not yet clear from published studies whether

predation risk models actually serve as effective conservation tools and at which levels of

decision-making they are being implemented (household, management or policy). Fur-

thermore, case studies are needed to illustrate how risk maps can be practically integrated

into education and intervention efforts and whether the guidance provided by risk models

signiﬁcantly reduces livestock depredation.

Spatial risk models may be especially appealing to natural resource managers and

conservation practitioners because they offer a potentially low-effort and low-cost tech-

nique for understanding the spatial patterns of livestock depredation using already existing

data. Such models require data on where carnivores attack and kill livestock. Conveniently,

due to the ﬁnancial impacts of livestock losses on people’s livelihoods, livestock owners in

many regions report attacks to a local government agency or non-proﬁt organization as

evidence for ﬁnancial compensation or insurance payments (Inskip and Zimmermann

2009; Dickman et al. 2011). These records are often archived in long-term databases and

Biodivers Conserv

123

contain the same data needed to build spatial risk models. When such databases exist, risk

modeling may be integrated into management as a relatively simple step to quantify

spatiotemporal patterns of human–carnivore conﬂict. A clear workﬂow of this process is

necessary to make risk modeling more accessible to practitioners. In addition, because

multiple statistical approaches have been used to produce spatial risk models without clear

rationale for which methods to employ, guidance is needed to elucidate techniques for

modeling carnivore–livestock interaction data.

Here I review the approaches and applications of spatial risk modeling for mitigating

carnivore attacks on livestock. Using the emerging literature on spatial risk modeling of

carnivore–livestock interactions, I critically examine statistical methods and model valida-

tion and evaluate contributions to the management of livestock husbandry and carnivore

populations. Based on lessons learned, I develop guidelines for integrating spatial risk

modeling into the process of managing human–carnivore conﬂict and provide real-world

examples of how study authors worked with stakeholders, managers and policymakers to use

risk maps during decision-making. Finally, I compare the locations of current studies to the

spatial distribution of potential carnivore–livestock interactions to identify where mitigation

tools might be needed worldwide. Given that reducing livestock depredation is critical for

enabling coexistence between people and carnivores (Treves and Karanth 2003; Ripple et al.

2015), and that mistakes in conﬂict management can have serious consequences for people’s

livelihood and carnivore survival, this review increases the accessibility of spatial risk

modeling as a tool to enhance the effectiveness of human–carnivore conﬂict mitigation.

Methods

Literature search

I found peer-reviewed journal articles on spatial risk modeling and livestock depredation by

searching the Web of Science (www.webofknowledge.com) and Google Scholar (www.

scholar.google.com) using the keywords: (‘risk’) AND (‘spatial’ OR ‘map’ OR ‘model’ OR

‘hotspot’) AND (‘livestock’ OR ‘cattle’ OR ‘goat’ OR ‘sheep’ OR ‘buffalo’) AND (‘car-

nivore’ OR ‘predat*’ OR ‘depredation’ OR ‘predator-prey’). Articles were deemed relevant

if they involved spatial modeling and mapping of carnivore depredation on domestic live-

stock. Articles that focused solely on attacks on humans or pets (including dogs) were

excluded because the landscape variables associated with these attacks may differ from

livestock. Articles were also excluded if they did not include a spatially explicit analysis

technique that included mapping risk hotspots. For each paper I recorded the study year,

location and carnivore and livestock species. These data were used to assess temporal and

spatial trends. I also recorded landscape attributes associated with carnivore attacks in the

models and noted the relationship (positive or negative) between each variable and carnivore

risk. These data were examined to identify common landscape attributes associated with

livestock depredation. Finally, I also noted the type of model and the steps of the research

process, which I synthesized into a workﬂow for creating and applying risk models.

Impacts on conservation and management

None of the studies identiﬁed in the literature review contained speciﬁc information on

how risk models and maps were applied to mitigate human–carnivore conﬂict. I therefore

Biodivers Conserv

123

contacted the corresponding authors of all studies for information on how they commu-

nicated research ﬁndings to stakeholders, managers and policymakers and whether results

had been incorporated into conservation or management actions or the policies that control

them.

Spatial analysis of studies

To explore where mitigation tools might be needed worldwide, I compared the spatial

overlap of livestock and depredating carnivores with the locations of studies. I identiﬁed

areas of high potential livestock depredation by comparing maps of global livestock

density data (from Gridded Livestock of the World v.2.0; Robinson et al. 2007) to the

global presence of the world’s main carnivore species responsible for depredating livestock

(species list in Table S1 adapted from Breitenmoser et al. 2005; species range maps from

the IUCN Red List of Threatened Species version 2014.2 and Letnic et al. 2012).

Results

Overview of literature

The literature search revealed 18 studies involving spatial risk modeling of carnivore–

livestock interactions (Table S2). The oldest study was published in 2004 and thereafter

one or two articles were consistently published each year, with a noticeable increase to ﬁve

articles published in 2014 (Fig. 1). Papers mostly addressed Canids (n =10) and Felids

(n =8), with fewer studies addressing impacts from species in the families Ursids (n =2)

and Hyaenids (n =1; note that some studies included multiple species). Studies especially

focused on wolves (n =6; Canis lupus) and wild dogs (n =2; Cuon alpinus) and on tigers

(n =4; Panthera tigris), leopards (n =4; Panthera pardus), lions (n =2; Panthera leo)

and pumas (n =2; Puma concolor). Less-studied species were caracal (Caracal caracal),

jaguar (Panthera onca), spotted hyena (Crocuta crocuta), black-backed jackal (Canis

mesomelas), American black bear (Ursus americanus) and grizzly bear (Ursus arctos;

n=1 on each species). The majority of studies investigated depredation on cattle

(n =14), sheep (n =12) and goat (n =10) but the literature also reported attacks on

alpaca, buffalo, camel, horse, llama, pig, poultry and rabbit.

Modeling approaches

Most spatial risk modeling studies (13 out of 18; see Table S2 for studies) used cor-

relation modeling approaches, including ensemble modeling (common models included

Maxent and ecological niche factor analysis), generalized linear models, logistic

regression and single-sample discriminant-function analysis (see ‘‘Discussion’’ section

for information on models). Three studies employed spatial association approaches using

Getis–Ord G and G* statistics and Mahalanobis D

(k) (Baruch-Mordo et al. 2008; Davie

et al. 2014; Meena et al. 2014). One study used spatial interpolation to extrapolate risk

probabilities at known sites across a larger landscape (Shrader et al. 2008). One study

mapped hotspots by scaling the probability of wolf pack presence by the number of

available livestock pastures rather than explicitly modeling livestock depredation (Mar-

ucco and McIntire 2010).

Biodivers Conserv

123

Most (11 of the 18) studies conducted validation tests to measure the accuracy of model

predictions, and these studies used a mixture of external and internal validation methods

(see ‘‘Discussion’’ section for details on validation tests). To conduct external validation,

which requires independent data for comparison against model predictions, authors either

retained a portion of the original dataset explicitly for validation testing (Treves et al. 2004;

Marucco and McIntire 2010; Zarco-Gonza

´lez et al. 2013; Abade et al. 2014) or used data

from a later time period (Treves et al. 2011; Behdarvand et al. 2014; Miller et al. 2015).

Cross-validation (e.g., 10-folds or leave-one-out) was commonly used for internal vali-

dation (Treves et al. 2004; Kaartinen et al. 2009; Edge et al. 2011; Soh et al. 2014;

Behdarvand et al. 2014).

Landscape attributes associated with depredation hotspots

Studies commonly found that the risk of a large carnivore attack on livestock was asso-

ciated with several of four main factors: species (carnivores and wild and domestic prey),

environment (e.g., vegetation type and elevation), human infrastructure (e.g., presence of

roads and villages) and management (e.g., park boundaries and preventative husbandry

techniques; Table 1).

Species

The density of carnivores and/or livestock was one of the strongest predictors of livestock

depredation. In reviewed studies that included densities of predators or prey as variables in

risk models, all but one found that higher densities of carnivores and livestock led to

greater predation risk (Baruch-Mordo et al. 2008; Kaartinen et al. 2009; Karanth et al.

2013). Zarco-Gonza

´lez et al. (2013) found a negative relationship because dense livestock

production areas were located away from forested areas and protected with fencing.

Studies also found positive associations between wild prey densities and livestock

depredation, suggesting that greater wild prey availability may lead to increased frequency

of attacks on livestock (Treves et al. 2004; Karanth et al. 2013). However, the connection

between livestock depredation and wild prey densities is currently under debate in the

human–wildlife conﬂict literature due to its complexity (Gervasi et al. 2013) and is likely

Fig. 1 The accumulative number of spatial risk modeling studies on carnivore–livestock interactions

published through March 2015

Biodivers Conserv

123

Table 1 Top-ranked landscape attributes used in spatial predation risk models for predicting carnivore

attacks on livestock

Carnivores Livestocks Landscape

attributes

Relationship to

predation risk

Studies

Puma (Puma

concolor)

Cattle, goat,

sheep

Forest ?Zarco-Gonza

´lez et al. (2013)

Altitude or altitudinal

range

?Kissling et al. (2009) and

Zarco-Gonza

´lez et al.

(2013)

Distance from

paddock to puma

habitat

-Kissling et al. (2009)

Livestock density -Zarco-Gonza

´lez et al. (2013)

Jaguar (Panthera

onca)

Cattle, goat,

sheep

Percent tree cover ?Zarco-Gonza

´lez et al. (2013)

Arid vegetation -

Altitude ?

Percent free grazing

livestock

Asiatic lion

(Panthera leo

persica)

Buffalo, camel,

cattle,

donkey, goat,

horse, sheep

Distance to lion

dispersal corridors

-Meena et al. (2014)

Distance to protected

area

Tiger (Panthera

tigris)

Cattle, buffalo,

goat, pig

Tree cover ?Soh et al. (2014)

Deciduous forest ?

Distance to very

dense forest

-Miller et al. (2015)

Distance to

moderately dense

forest

Distance to scrub ?

Distance to scrub

Distance to river -Soh et al. (2014)

Aspect -

Protected area

(within or outside)

?when within

Distance to protected

area core

?Miller et al. (2015)

Distance to snare -Soh et al. (2014)

Distance to road ?Miller et al. (2015) and Soh

et al. (2014)

Distance to road

-Miller et al. (2015)

Distance to village ?Miller et al. (2015) and Soh

et al. (2014)

Distance to village

-Miller et al. (2015)

Biodivers Conserv

123

Table 1 continued

Carnivores Livestocks Landscape

attributes

Relationship to

predation risk

Studies

Tiger and leopard

(Panthera

pardus)

(combined)

Cattle, buffalo,

goat, pig

Distance to forest

cover

-Karanth et al. (2012)

Distance to water ?Karanth et al. (2012,

2013)

Distance to protected

area

Park wildlife density ?Karanth et al. (2013)

Number of livestock ?

Grazing cows inside

protected area

?Karanth et al. (2012,

2013)

Use of any mitigation

measure

Leopard, African

lion (Panthera

leo), spotted

hyaena

(Crocuta

crocuta)

(combined)

Cattle, goat, sheep Percent tree cover -Abade et al. (2014)

Distance to river -

Altitude -

Slope -

Annual precipitation -

Normalized

difference

vegetation index

(NDVI)

Black-backed

jackal (Canis

mesomelas),

caracal (Felis

caracal)

(combined)

Goat Visibility ?Shrader et al. (2008)

Wolf (Canis

lupus)

Bison, cattle, duck,

poultry, rabbit,

sheep, white-

tailed deer

Forest ?Kaartinen et al. (2009)

Coniferous forest ?/-Edge et al. (2011) and

Treves et al. (2004)

Distance to forest ?Treves et al. (2011)

Tall vegetation ?Davie et al. (2014)

Shrubland ?

Forbland -

Wetland -Treves et al. (2004) and

Kaartinen et al. (2009)

Open water ?/-Treves et al. (2004)

Distance to waterway ?Behdarvand et al. (2014)

Plantation ?Kaartinen et al. (2009)

Cropland -Treves et al. (2004) and

Edge et al. (2011)

Grass/pasture/

hayﬁeld

?/-Treves et al. (2004,2011)

and Edge et al. (2011)

Dry farms ?Behdarvand et al. (2014)

Number of sheep in

ﬂock

?Kaartinen et al. (2009)

Biodivers Conserv

123

Table 1 continued

Carnivores Livestocks Landscape attributes

Relationship to

predation risk

Studies

Pasture size -

Distance to wolf pack -Treves et al. (2011)

Distance to wolf

pack * distance to

forest

Wolf density ?Kaartinen et al. (2009)

Denning probability -Behdarvand et al. (2014)

Deer density ?Treves et al. (2004)

Distance to protected

area

?Behdarvand et al. (2014)

Distance from Russian

border

-Kaartinen et al. (2009)

Distance to human

settlement

?/-Behdarvand et al. (2014) and

Davie et al. (2014)

Road density -Treves et al. (2004)

Distance to road ?Behdarvand et al. (2014)

Grizzly bear

(Ursus arctos)

Cattle,

sheep

Percent riparian cover ?Wilson et al. (2005)

Distance to spring,

summer or fall

pastures

Unmanaged boneyard

present

Beehive present (either

fenced or unfenced)

Unfenced beehive *

unmanaged

boneyard

Distance to lambing

area

Calving area (inside or

outside)

?when inside

American black

bear (Ursus

americanus)

Sheep,

poultry

Sheep and farm density ?Baruch-Mordo et al. (2008)

Note that results from Marucco and McIntire (2010) were omitted from table because depredation risk was

based on assumptions rather than empirical data on livestock kill cases (refer to the study for details)

Variables are organized by theme, not ranking—see studies for coefﬁcients and importance to the model

?Indicates positive relationship, -indicates negative relationship, ?/-indicates mixed results from

studies; binary variables are explained with text; for non-linear or interaction variables, see speciﬁc studies

for details

Biodivers Conserv

123

more dynamic than a simple inverse or direct relationship and tied to many other factors,

such as the spatial distribution of domestic versus wild prey species and denning or resting

sites for predators (Stahl et al. 2001; Treves et al. 2004; Suryawanshi et al. 2013).

Environment

Studies typically quantiﬁed the environment at kill sites in terms of vegetation or habitat

type, percent vegetation cover, distance to vegetation or visibility. The relationship

between these factors largely depended on the hunting mode of the carnivore. Active-

roaming hunters such as wolves, which often kill in ﬂat, open areas conducive to long

pursuits, tended to exhibit higher predation risk in shrubland, ﬁeld or pasture and at farther

distances from denser forest (Treves et al. 2011; Davie et al. 2014), although not in all

cases (Kaartinen et al. 2009; Edge et al. 2011; Treves et al. 2011). Greater risk from

stalking hunters such as tigers, jaguars and pumas, which commonly use thick vegetation

with heavy cover to ambush prey, was associated with dense forest and low visibility

(Shrader et al. 2008; Kissling et al. 2009; Karanth et al. 2012; Zarco-Gonza

´lez et al. 2013;

Soh et al. 2014; Miller et al. 2015). Water also featured prominently in risk models and was

associated with lower risk of depredation by wolves (Treves et al. 2004,2011) but greater

risk of depredation by leopards (though effects were weak; Karanth et al. 2012,2013).

Finally, elevation was often ranked as an inﬂuential variable in systems where a large

topographic gradient exists, such as for jaguar in Mexico and puma in Argentina, with

higher elevations linked to greater predation risk (Kissling et al. 2009; Zarco-Gonza

´lez

et al. 2013).

Human infrastructure

Human infrastructure had mixed associations with predation risk in several carnivore

species. Tigers were more likely to kill livestock farther from roads and villages in China

(linear relationship; Soh et al. 2014) but in India, risk peaked around 1 km from roads and

villages and decreased very near and far from infrastructure (nonlinear relationship; Miller

et al. 2015). Wolves in Mongolia and the US tended to kill livestock away from towns and

roads (linear relationship; Treves et al. 2004; Davie et al. 2014) but in Iran wolf risk

showed a nonlinear relationship to human settlement, spiking very close to settlements but

decreasing at distances greater than 1.0 km. In these cases, the effect of human infras-

tructure may be linked to the density of roads and villages in the area. Neither human

presence (density and distance to camp) nor road (density and distance from) were ranked

as top predictors of attack on livestock for jaguar or puma (Kissling et al. 2009; Zarco-

Gonza

´lez et al. 2013) and human infrastructure was not explicitly explored for other

species in the reviewed studies.

Management

Many studies found strong connections between land-use zoning or preventative husbandry

interventions and risk of livestock depredation. Close proximity between livestock grazing

sites and high-quality carnivore habitat was one of the strongest indicators of attacks by

nearly all carnivores. Livestock depredations more often occurred in livestock paddocks

located near ideal habitat for puma (Kissling et al. 2009) and near dispersal corridors for

Asiatic lion (Meena et al. 2014). Greater predation risk by tigers, Asiatic lions and wolves

Biodivers Conserv

123

occurred close to protected areas (Karanth et al. 2012,2013; Soh et al. 2014; Meena et al.

2014; Behdarvand et al. 2014; Miller et al. 2015), and risk of attack increased if cattle

grazed within the park (Karanth et al. 2012,2013). Similarly, allowing livestock to freely

graze increased predation risk from jaguar (Zarco-Gonza

´lez et al. 2013). Livestock owners

lost fewer animals to tigers, leopards and wild dogs in India if they generally used miti-

gation measures, although no individual techniques were strongly associated with reduced

depredation (Karanth et al. 2012,2013). Grizzly bears were particularly drawn to certain

attractants and more likely to kill livestock near beehives (no matter whether they were

fenced or not), lambing and calving areas and unmanaged boneyards (Wilson et al. 2005).

Impacts on conservation and management

Authors from 16 of the 18 reviewed studies provided information about how risk models

had impacted conservation, management and policy. The authors of 14 papers shared study

results by actively and regularly engaging with local herders and villagers, government

agencies, conservation organizations and research institutes, and most knew of direct

outcomes on conservation, management and policy resulting from their efforts. The authors

of two studies did not engage extensively in outreach and were not aware of any impacts at

the study site. Overall, authors shared risk model results with three levels of decision-

makers: (1) stakeholders, (2) management and conservation authorities and (3) policy-

makers (Table 2).

At the stakeholder level, authors shared results directly with livestock owners and local

households through educational events and efforts to subsidize and improve the effec-

tiveness of carnivore deterrents (Table 2). In Mexico, Zarco-Gonza

´lez et al. (2013) pre-

sented risk maps to farmers during workshops on how to adapt grazing locations, design

pastures and handle animals to reduce attacks, and how to apply for ﬁnancial compensation

after an attack (Martha Zarco-Gonza

´lez personal communication). In Tanzania, Abade

et al. (2014) worked with the Ruaha Carnivore Project to share risk maps with local

villagers through Swahili PowerPoint presentations delivered during educational DVD

nights to help households recognize distributions of risk in their areas (Dickman personal

communication). The group also used risk maps to prioritize households for receiving

guard dogs and assistance constructing predator-proof enclosures, the latter which

reportedly reduced carnivore attacks by over 90 %. Maps also provided a baseline for

focusing mitigation efforts and examining stock loss trends over time. In the US, Wilson

et al.’s (2005) efforts to share grizzly bear conﬂict hotspot maps with livestock owners

through PowerPoint presentations, informal meetings and regular workgroups motivated

the community to help install electric fences around 18 calving areas and 85 beehives,

phase out 120 boneyards through a livestock carcass removal program and provide 100

bear-resistant trashcans to residents in high-risk areas (Wilson personal communication).

These actions reduced bear conﬂicts by 96 % from 2003 to 2010 (Wilson et al. 2014).

Through active communication with authorities, authors also integrated risk maps into

the decision-making process within government agencies and conservation organizations

(Table 2). In western India, Meena et al. (2014) developed conﬂict management manuals

on Asiatic lion attacks on livestock which were adapted and then distributed to high-risk

villages by the Gujarat Forest Department during lion conservation awareness programs

(Meena personal communication). In central India, tiger and leopard depredation hotspot

maps developed by Miller et al. (2015) were shared with the management community

through written reports, oral presentations and one-on-one meetings with agency ofﬁcials

and ultimately used by the Madhya Pradesh Forest Department to strategically build fences

Biodivers Conserv

123

Table 2 Outreach actions and impacts of risk models and maps on conservation, management and policy

related to human–carnivore conﬂict at the stakeholder, management/conservation authority and policymaker

level

Levels of

decision-

making

Outreach actions Impact on

conservation/

management

Regions Studies

Stakeholder Worked with households

in highest-priority/risk

areas to strengthen

mitigation efforts

Predator-prooﬁng

enclosures reduced

carnivore attacks by

[90 %; increased

use of guard dogs

Ruaha, Tanzania Abade et al.

(2014)

Increased awareness

about ways of re-

designating and

designing pastures

and handling

livestock to reduce

risk of predation

Mexico Zarco-Gonza

´lez

et al. (2013)

Shared published papers

with local herders

Education Northern Cape,

South Africa

Shrader et al.

(2008)

Continue to measure

livestock loss data at

high-risk households

after initial risk

modeling/mapping

Risk model/map serves

as baseline to

measure change

Ruaha, Tanzania Abade et al.

(2014)

Shared risk maps with

villages using local-

language PowerPoint

presentations and/or

DVD nights

Education Alps ecosystem,

Europe

Marucco and

McIntire

(2010)

Education Ruaha, Tanzania Abade et al.

(2014)

Conducted classes and

workshops for livestock

owners on ﬁnancial

compensation process

Education Mexico Zarco-Gonza

´lez

et al. (2013)

Distributed awareness

posters or leaﬂets with

results, especially in

communities with high

attack risk

Education Gujarat, India;

Hamadan, Iran

Meena et al.

(2014) and

Behdarvand

et al. (2014)

Posted interactive risk

map online

Education Wisconsin, USA Treves et al.

(2011)

Management

conservation

authority

Formatted results into

awareness materials for

each village (e.g.,

pamphlets) for

distribution by

management agency

Education Gujarat, India Meena et al.

(2014)

Used results to acquire

additional funding for

large-scale carnivore

conservation efforts

EU Life?Project

‘WOLFALPS’

Alps ecosystem,

Europe

Marucco and

McIntire

(2010)

Interactive online risk

map

Education Wisconsin, USA Treves et al.

(2011)

Biodivers Conserv

123

to exclude livestock grazing from high-risk areas (Jhala personal communication). In the

Italian Alps, maps showing potential livestock depredation (Marucco and McIntire 2010)

led to the EU Life?Project ‘WOLFALPS’, which supports natural wolf recolonization

over the Alps ecosystem through partnerships between agencies and conservation orga-

nizations (www.lifewolfalps.eu; Marucco personal communication). Hotspot maps also

helped identify regions where conﬂict management required more information: Baruch-

Mordo et al.’s (2008) risk model of black bear–human conﬂicts across Colorado in the

USA prompted interest from a local agency in the high-risk hotspot of Aspen to collect

ﬁner-scale habitat quality data that revealed greater insights into how to mitigate conﬂicts

(Baruch-Mordo et al. 2014). Several authors expressed the importance of regular inter-

action with authorities, especially government agencies, for developing trusted relation-

ships that eventually impacted management of human–carnivore conﬂict.

Though many authors sought to affect conservation at the policy-level, the authors of

only one reviewed study reported success (Table 2). In Hunchun, China, where the

Wildlife Conservation Society and local government authorities are jointly working

towards conserving the Amur tiger, hotspot results from Soh et al. (2014) were included in

a policy document that reportedly contributed to the government agency increasing efforts

to resolve human–tiger conﬂict (Wildlife Conservation Society personal communication).

Spatial distribution of studies and potential livestock depredation

Mapping these studies revealed a global distribution of research with clusters in several

countries (Fig. 2). Most studies were conducted in North America (n =6 studies; 5 in the

USA and 1 in Mexico) and Asia (n =7; 4 in India and 1 each in China, Iran and

Mongolia). Africa (n =2; 1 each in South Africa and Tanzania), Europe (n =2; 1 each in

Finland and Italy) and South America (n =1 in Argentina) were also represented.

Table 2 continued

Levels of

decision-

making

Outreach actions Impact on

conservation/

management

Regions Studies

Shared results as written

materials and/or

presentations to

government agency and

conservation

organizations

Results included in

management plan

Gujarat, India Meena et al. (2014)

Fencing built to keep

livestock out of high-

risk areas

Madhya Pradesh,

India

Miller et al. (2015)

Education Colorado, USA Baruch-Mordo

et al. (2008)

Education Dornogobi Aimag,

Mongolia

Davie et al. (2014)

Education Karnataka, India Karanth et al.

(2013)

Education Mexico Zarco-Gonza

´lez

et al. (2013)

Education Ruaha, Tanzania Abade et al. (2014)

Policymaker Shared results with

government agency via

policy document

Increased recognition

of the need to address

human–tiger conﬂict

at the study site

Hunchun, China Soh et al. (2014)

Biodivers Conserv

123

Comparing study locations against areas of potential livestock losses (livestock densi-

ties and presence of carnivore species) revealed a lack of studies in areas with the greatest

potential human–carnivore conﬂict (Fig. 2). Eighty-three percent of studies were con-

ducted in areas of low–moderate levels of potential livestock depredation (carnivores

present and 0–300 livestock/km

) and the remaining 17 % occurred in areas of high

potential conﬂict (carnivores present and 300–1000 livestock/km

). No studies were

located in regions where conﬂict-prone carnivores overlapped with the highest densities of

livestock (1000–13,000 livestock/km

). The highest levels of potential livestock depre-

dation (carnivores present and [300 livestock/km

) were distributed across a contiguous

belt stretching through eastern Europe, western Asia and southern and eastern Asia. High

potential conﬂict was also apparent in western (Nigeria), eastern (Ethiopia and Kenya) and

southern (South Africa) Africa, eastern South America (Ecuador and Peru), southern

Mexico and the midwestern US. Reviewed studies addressed hotspots in Italy, India,

Mexico and the US but spatial risk modeling has yet to be published from many areas with

high potential rates of livestock depredation. It is also worth noting that several regions

with signiﬁcantly dense livestock populations ([50 livestock/km

) no longer support

conﬂict-prone carnivores (e.g., parts of western Europe, southeastern Argentina, south-

eastern Australia and eastern southeast Asia; arrows in Fig. 2).

Fig. 2 Global potential livestock depredation hotspots compared to published spatial risk studies. Hotspots

of livestock depredation are likely to occur where high densities of livestock (shown for buffalo, cattle,

sheep, goats and pigs combined) overlap with the presence of the primary carnivores responsible for

livestock depredation worldwide (see Table S1 for species). Arrows are meant to clarify certain regions

where carnivores are absent (see Fig. S1 for separate livestock and carnivore maps). Data sources livestock

densities from FAO’s Gridded Livestock of the World and species ranges from IUCN Red List of

Threatened Species, Version 2014.2 (Letnic et al. 2012)

Biodivers Conserv

123

Discussion

This literature review demonstrates the utility of spatial predation risk modeling as a

technique for identifying high-risk priority areas in order to maximize the effectiveness and

minimize the costs of conﬂict mitigation. The number of published spatial risk studies is

currently small yet will continue to increase as conservation practitioners share successes

with colleagues and spatial resources become more freely accessible (Garcı

´a-Rangel and

Pettorelli 2013). It is therefore important to assess the gaps in current risk modeling and

outline the process for integrating hotspot mapping into management, conservation and

policy decision-making.

Guidelines for integrating spatial risk modeling and mapping

into conservation

The reviewed studies collectively indicated that six main steps were necessary for

developing conservation-relevant spatial risk models: (1) identify objectives, (2) collect

data, (3) build risk models, (4) map predation risk, (5) apply results to conservation and

management and (6) account for feedbacks in the system (Fig. 3). Due to the context-

speciﬁc nature of human–carnivore conﬂict, these steps should be considered ﬂexible and

adapted to speciﬁc systems, species and conservation goals.

Fig. 3 The six-step process of creating and applying predation risk maps to mitigating human–carnivore

conﬂict: (1) identify objectives, (2) collect spatial data on predation risk and landscape attributes that may

inﬂuence carnivore–livestock interactions, (3) model data using a spatially explicit modeling approach and

validate results, ideally with independent data, (4) use model to map predictions of risk while considering

the purpose, type of media and risk level scheme, (5) apply results using various media formats to

stakeholders, managers and conservation practitioners and policymakers and (6) update models and maps

regularly to account for feedbacks within the system

Biodivers Conserv

123

Identifying objectives

The modeling process should be closely tied to the local community, carnivore species and

the ultimate objectives of management and conservation (Fig. 3-1). In situations where

existing data collected by agencies or conservation organizations will be used for risk

modeling, the spatial (extent and resolution) and temporal limitations of the data must be

considered when determining how risk models can inform the desired objectives. For

instance, models may focus on the spatial patterns of current or future livestock depre-

dation across an entire country (Marucco and McIntire 2010) or a single park (Abade et al.

2014), or for a carnivore community (Karanth et al. 2013) or a single carnivore species

(Miller et al. 2015).

Collecting data

Modeling carnivore predation risk on livestock requires two types of information (Fig. 3-

2). First, a model requires data on the locations of carnivore–livestock interactions. In

many landscapes globally, conservation schemes exist which enable owners to report

livestock death to authorities for ﬁnancial compensation or insurance (e.g., Schiess-Meier

et al. 2007; Agarwala et al. 2010). These programs generate databases of kill data that—if

accurate and unbiased—may be appropriate for modeling carnivore attacks. Most of the

reviewed studies used data from livestock kill sites (15 studies) or villager surveys about

livestock mortalities (2 studies). Databases on livestock mortality may contain detailed

information on the livestock (species, sex, age, health), carnivore (species, sex, age), owner

(name, village, income) as well as spatial (habitat type, distance from village) and temporal

(time of attack, date) characteristics. The long-term nature of these databases can also

provide insight into the historical patterns of attacks and changes with management

interventions or natural events (e.g., drought, native prey declines). However, livestock

mortality databases may include strong reporting biases if social tension or spatial isolation

prevent certain livestock owners from reporting losses (Karanth et al. 2012) or if managers

do not consistently record conﬂict incidences (Baruch-Mordo et al. 2008). It is essential

that researchers understand these biases and limitations when analyzing and interpreting

livestock kill data. Furthermore, risk models generated from kill data represent the odds of

a successful kill given an encounter between carnivore and livestock but do not account for

failed hunting attempts, encounters where hunting did not occur or livestock fear (Heb-

blewhite et al. 2005; Gorini et al. 2012).

Besides kill data, several other types of information may be used to analyze carnivore–

livestock spatial interactions, although these methods are currently underdeveloped. First,

livestock fear may serve as a proxy for the likeliness of a carnivore attack (Brown et al.

1999; Laundre

´et al. 2010). Since prey must choose between allocating precious energy

resources between feeding and anti-predator efforts, the energetic cost involved in staying

aware of predators can make vigilance an indicator of predation risks (Lima and Bednekoff

1999). Vigilance can be quantiﬁed through measures like alertness, ﬂight initiation dis-

tance or the willingness to give up food in order to remain on guard, the latter which is also

known as the giving-up-density (Brown 1999; Stankowich and Blumstein 2005; see

Shrader et al. 2008 for an example). However, because prey can misinterpret environ-

mental cues or misjudge carnivore behavior, and livestock in particular are known to show

attenuated responses to carnivore threats (Laporte et al. 2010), fear responses may not

serve as an accurate prediction of the likeliness of livestock depredation (Creel and

Biodivers Conserv

123

Christianson 2008). Second, data on general carnivore–livestock encounters (deﬁned as a

spatial and/or temporal overlap) may provide insight into a broader spectrum of predation

risk beyond realized risk based on mortality alone (Hebblewhite et al. 2005). Encounters

can include various elements of a hunt (e.g., searching, approaching, watching and

attacking prey; Kluever et al. 2008) and may present the opportunity for a carnivore to

attack livestock, although carnivores may not always attack or with success (only 25 % of

observed leopard attacks resulted in a wild prey kill; Balme et al. 2007). Furthermore,

measuring encounters is time-consuming and expensive, often requiring scientiﬁc equip-

ment and monitoring teams, because animals typically interact quickly and within small

areas, making it essential to record encounters and very ﬁne spatial and temporal resolu-

tions (Gorini et al. 2012).

The second type of data required for modeling is information on the landscape where

livestock are killed. Gathering spatial data on the landscape attributes, organismal biology

and land cover at and surrounding livestock kill sites strengthens the ability of risk models

to identify the speciﬁcally variables associated with carnivore attacks (Treves et al. 2011;

Garcı

´a-Rangel and Pettorelli 2013). The four categories of factors identiﬁed in the

reviewed studies (species, environment, human infrastructure and management) represent a

starting point of biologically meaningful variables for modeling risk, representing both

prey availability (species presence and abundance) and accessibility (vulnerability based

on the surrounding environment). On-the-ground observations can also be highly infor-

mative for understanding the site-speciﬁc context, including the heterogeneity that may

exist throughout the landscape and inﬂuence species interactions (Gorini et al. 2012). Local

peoples’ experiences can offer long-term insights into livestock husbandry and grazing

patterns, species population trends, socio-economic factors, policy or governance changes,

and other interactions that might affect risk (Treves et al. 2006; Dickman 2010). In

addition, the unit of a variable must be carefully chosen to precisely express biological

impact (Treves et al. 2011). Once the appropriate landscape attributes have been identiﬁed,

many remote sensing and geographic information data and software are available (often

free of cost) for accessing spatial data (for a list of sources see Garcı

´a-Rangel and Pettorelli

2013).

Building models

Reviewed studies revealed that three primary approaches are commonly used for mapping

carnivore risk to livestock: correlation modeling, spatial association and spatial interpo-

lation (Fig. 3-3).

Correlation modeling was the most popular method and is often used in the wildlife

biology ﬁeld for examining animal resource use. Resource selection functions (RSFs) are a

subset of correlation models that estimate the probability of an animal selecting a resource

(e.g., carnivore targeting livestock) and are often applied to generate spatially explicit

predictions of predation risk (Manly et al. 2002; Boyce et al. 2002). RSFs use regression

models to reveal associations between landscape characteristics and selection frequencies

(Lele et al. 2013), such as the tendency of carnivores to kill livestock in certain envi-

ronmental conditions. Many spatial software use correlation modeling to predict gradients

of animal interactions relative to spatially explicit environmental or anthropogenic vari-

ables (e.g., Maxent, open modeller and biomapper; Behdarvand et al. 2014; Zarco-Gon-

´lez et al. 2013). Correlation models may be built with presence–absence data (locations

where events did and did not occur) or presence-‘availability’ data following a use-

availability design (locations where events occurred and random locations representing

Biodivers Conserv

123

where events could have occurred; Johnson et al. 2006). For analyzing livestock depre-

dation hotspots, correlation modeling is recommended if data are available on livestock

kills as well as the surrounding landscape where kills did not or could have occurred but

are not known.

Several studies used spatial association to examine the clustering of kill sites for

identifying signiﬁcant conﬂict hotspots. These techniques work well with relatively smaller

datasets and apply well-established statistical approaches for assessing autocorrelation to

obtain insights on biology (Dale and Fortin 2014). However, spatial association strictly

requires independent data samples (i.e., kills that are unrelated to each other), which may

be impossible for ecological studies. The process may also over-emphasize risk hotspots

due to autocorrelation depending on how sampling was distributed (Dale and Fortin 2014).

Spatial association techniques include statistics such as Moran’s I, which reveals spatial

clustering between points to identify associations between response and covariate variables

(Willems and Hill 2009). The Getis–Ord statistics examines concentrations of points with

respect to a central area of interest (Getis and Ord 1992), such as livestock kills located

around a protected area (Meena et al. 2014), and provides information regarding high and

low clusters (hotspots and coldspots; Baruch-Mordo et al. 2008; Dale and Fortin 2014).

Mahalanobis D

(k) distance presents an alterative method for cluster analysis that requires

only presence data (e.g., kill site locations), thus alleviating problems associated with

gathering absence data (where kills did not occur) or random data (where kills could have

occurred) as well as assumptions of data normality (Clark et al. 1993). Since these statistics

speciﬁcally examine spatial autocorrelation and point clustering, they are useful for

addressing whether areas of high or low livestock depredation densities cluster on the

landscape relative to certain landscape attributes and spatial extents (Bump et al. 2009;

Dale and Fortin 2014). Models assign signiﬁcance values to neighborhood units (e.g.,

deﬁned village or district area) to provide information about clustering across the study

extent; scores over a certain signiﬁcance threshold indicate high clustering (e.g., areas

where carnivores tend to depredate) and scores under a certain threshold indicate low

clustering (areas where carnivores do not tend to depredate; Baruch-Mordo et al. 2008;

Meena et al. 2014). Once these statistics reveal clustering patterns, additional analysis is

required to identify spatial associations with landscape attributes (Baruch-Mordo et al.

2008).

Spatial interpolation was used by several studies for quantifying continuous spatial

gradients of environmental factors based on sampled points. Interpolation techniques, such

as kriging or inverse distance weighing, ﬁt a statistical model through spatially separated

data points to characterize the remaining landscape between the known sites (Dale and

Fortin 2014). Spatial interpolation is useful for adapting spatially limited datasets (e.g.,

villages with household socio-economic surveys or speciﬁc sites where livestock numbers

were estimated) for use in combination with other spatially explicit data (e.g., GIS data;

Alessa et al. 2008; Bryssinckx et al. 2012) but does not examine correlations between

multiple variables or their associations with species interactions. Models output a spatial

layer of continuous data for a given variable across a ﬁxed area. Therefore, the technique is

most appropriate for initial, descriptive examination of carnivore risk to livestock or for

converting spatially implicit variables but not for complex multivariate analyses of live-

stock depredation hotspots. Additionally, interpolation models are very sensitive to the

sample density, data variation and sampling design such that sampling must adequately

capture variation in the underlying environmental conditions for the model to accurately

represent the study extent (for details see Li and Heap 2011).

Biodivers Conserv

123

Many reviewed studies did not carry out model validation, the process of comparing

model output against independent data (Fig. 3-3). Measuring a model’s predictive accuracy

is an essential part of conﬁrming its strength as a tool for conservation decision-making

(Sinclair 1991; Schmolke et al. 2010). High accuracy is particularly important when

dealing with human–carnivore conﬂict (Treves et al. 2011), where inaccurate guidance

could provoke additional livelihood losses and species declines. The most rigorous method

of validating data is to conduct external validation tests using independent data. These data

may have been separated from the data used to train models and reserved for testing

purposes, or may have been collected in the ‘‘future’’ from a time period after the data used

to build the model was collected, or may have been collected outside the study area extent

and time period if the model is applicable elsewhere. When independent data are not

available or the dataset is too small to withhold testing data, internal validation offers an

alternative process of validation. Internal methods, such as k-folds cross-validation, use the

complete dataset for both calibration and testing by partitioning and cross-checking data

multiple times (Boyce et al. 2002). However, these methods commonly produce overly

optimistic assessments because they use the same data for model training and testing

(Dormann et al. 2012).

Mapping predicted risk

Authors used maps as a primary tool when communicating the results of risk modeling.

Maps are considered highly effective tools for community engagement in management due

to their ability to bridge communication challenges through a visual language composed of

colors and symbols (Brown and Raymond 2007; Rambaldi et al. 2006). Especially in the

case of human–carnivore conﬂict, maps depicting color gradients of risk are more likely to

be more interpretable than model statistics by stakeholders from diverse educational and

cultural backgrounds (Rambaldi et al. 2006). The reviewed studies portrayed risk using a

variety of visualization approaches, ranging from black-and-white to color, from discrete

to continuous gradients, and at different spatial scales and management units (Fig. 4).

Producing a meaningful map of risk is an important step that requires consideration as to

how the map will be ultimately used for conservation (Fig. 3-4). For instance, when color

gradients are used, the color scheme should be carefully chosen to emphasize the most

biologically meaningful levels of risk. Also, maps should account for varying visual and

cultural interpretations of color, such as the effects of red–green color blindness or the

association of color with certain social cues (e.g., trafﬁc lights).

With increasing global access to mobile phones and internet, maps are no longer

restricted to printed media, offering researchers the opportunity to develop engaging

interactive maps that provide people with insights into carnivore risks in their immediate

proximity. The Carnivore Coexistence Lab at the University of Madison-Wisconsin used

the open access platform Leaﬂet (www.leaﬂetjs.com) to create an interactive online map of

wolf risk to livestock in Wisconsin, where users can type in an address and view to risk

from several years around their property (www.faculty.nelson.wisc.edu/treves/wolves/

interactiveRiskMap.php; Treves et al. 2011). Such creativity in showcasing risk models

and maps should play a greater role to increase access by a broader audience.

Applying results to conservation and management

Studies revealed that risk model output can successfully be applied to conservation and

management at multiple stages of decision-making, from farm-level management to

Biodivers Conserv

123

region-level policies, through the use of diverse forms of communication (Fig. 3-5).

Organizing meetings or classes created an opportunity to share results directly with

stakeholders who could personally incorporate model outcomes into their decision-making.

Risk models helped identify livestock owners in high-priority areas where implementing

preventative husbandry techniques (predator-proof enclosures, guard dogs and livestock

carcass removal programs) and deterrents (electric fences and predator-proof enclosures

and trash cans) resulted in dramatic reductions in carnivore attacks. Likewise, directly

communicating results to managers helped integrate risk model results into management

plans, village awareness programs and on-the-ground actions for preventing livestock

grazing in high-risk areas. Risk models and maps signiﬁcantly strengthened outreach

efforts, supporting previous evidence that outreach increases the likeliness of stakeholders

implementing conﬂict mitigation interventions (Pienaar et al. 2015).

The fact that only one study impacted policy-level decision-making highlights a sig-

niﬁcant gap in current efforts and a critical target for future outreach. One of the greatest

advantages of spatial risk modeling is the visual output, and maps could be showcased in

policy documents and presentations to help policymakers understand patterns of human–

carnivore conﬂict across constituents’ properties. Such success is currently underway in

Sweden, where risk maps helped the Environmental Protection Agency, Department of

Agriculture and Regional County Administration Boards allocate subsidies for farmers

building fences to deter carnivores from attacking livestock (Karlsson personal commu-

nication). Swedish agencies similarly used risk maps to identify regions with high pre-

dation risk and high sheep farm density where culling carnivores might be more effective

in the short term than fencing subsidies. Though not yet published, the utility of risk maps

for advising policy-level budgetary decisions in Sweden demonstrates the potential impact

that future risk modeling efforts could have on policy decision-making.

Management, conservation and policy decisions occur over a range of scales, spanning

from individual farms to entire states and nations. Carefully thinking through how model

results will be used can help avoid mismatches between the scale of analysis and imple-

mentation (Chapin et al. 2010). Will regional managers use the analysis implement pre-

ventative interventions according to management units (Baruch-Mordo et al. 2008) or

townships (Treves et al. 2011)? Would a policy leader prefer to understand the risks from

recolonizing carnivores to livestock across the entire country or a particular region

(Marucco and McIntire 2010)? Consulting with both stakeholders and decision-makers to

understand how risk models will be used for management and conservation will facilitate

modeling at a practical spatial scale (Treves et al. 2006).

Success in applying risk models and maps to on-the-ground conservation ultimately

depends on the interest of decision-makers. Authorities sometimes feel they intuitively

understand carnivore threats so well that they do not need risk models, yet later see

unexpected patterns and insights emerge from quantitative analysis (Author personal

observation; Treves personal communication). Many authors stressed the importance of

building long-term relationships with management authorities to develop trust, which

created opportunities for presenting risk maps. Especially within government agencies,

shifting constituent concerns and political priorities can also affect managers’ interest in

risk modeling. In 2011 after wolves were delisted from the US Endangered Species Act,

the impetus to manage high-risk hotspots was lost once wolf hunting was legalized as a

management tool (Treves personal communication).

Biodivers Conserv

123

Keeping risk models relevant: accounting for behavioral feedbacks

The literature review revealed a lack of attention on the behavioral feedbacks that may

arise from the use of spatial risk models in management and conservation. Spatial risk

models represent dynamic systems in which carnivores and livestock (and humans that

manage livestock) continuously respond to one another. Management efforts that consider

carnivore risks when applying interventions, such as in shifting livestock to low-risk

grazing sites, may prompt feedbacks: carnivores could adjust attack patterns to account for

different livestock distributions, which could change risk distributions from the original

modeled predictions. This ‘arms race’ of human–carnivore conﬂict mitigation—whereby

carnivores adapt behavior to override mitigation strategies that in turn must adapt to deter

carnivore attacks—is the reason why the effective period of most conﬂict mitigation tools

is short (Shivik 2006).

To account for feedbacks, models must be regularly updated to reﬂect changes in the

system (Marucco and McIntire 2010; Fig. 3-6). Yet none of the reviewed studies addressed

Fig. 4 Risk maps showing different approaches for illustrating livestock depredation hotspots. aContour

probability plot for grizzly bears in Montana, USA, where more widely spaced lines represent lower levels

of conﬂict (Wilson et al. 2005). bGetis–Ord hotspots illustrating spatial clustering of livestock kills by

African lion by village area in western India (Meena et al. 2014). cLogistic regression model representing

the relative probability of tiger attacks on livestock in central India (Miller et al. 2015). dPuma risk to

livestock by paddock (livestock pasture) calculated using generalized linear regression (Kissling et al. 2009).

Darker shades (a,d) or warmer colors (b,c) represent greater probability of carnivore attack.

Figures published with permission. (Color ﬁgure online)

Biodivers Conserv

123

changes in predation risk over time or ties to behavioral feedbacks. Regularly revising

spatial risk models with recent ﬁeld data on carnivore–livestock interactions (e.g., sea-

sonally or annually) will be necessary to ensure that hotspot maps offer up-to-date guid-

ance on carnivore attack patterns. This is an especially critical step if risk maps are

incorporated into management, conservation and policy.

Conclusions

Based on the review, ﬁve key issues require consideration to strengthen the utility of spatial

predation risk modeling as a tool for mitigating carnivore attacks on livestock. First, models

should be quantitatively and externally validated with independent data to ensure the integrity

of model output for accurately guiding mitigation interventions. Second, risk modeling

should be integrated into long-term monitoring to assess the impacts of associated mitigation

efforts on livestock depredation. As part of this process, risk models must be updated with

recent data to account for behavioral feedbacks and changes in carnivore–livestock inter-

actions. Third, innovative methods for displaying, sharing and applying results from risk

models should be used to reach new audiences among stakeholders, managers and policy-

makers. Fourth, outreach with policymakers should be prioritized so that risk maps may

inform large-scale decisions on conﬂict. Finally, there is great opportunity for more risk

modeling across the world, particularly in the areas of high potential livestock depredation

such as eastern Europe, western, southern and eastern Asia western, eastern and southern

Africa and eastern South America. Efforts to establish regular risk modeling efforts in these

regions should be prioritized for ﬁnancial and logistical support, especially since human–

carnivore conﬂict is currently contributing to the rapid of greatly endangered carnivores such

as the tiger and the Andean bear (Tremarctos ornatus).

Quantitative, scientiﬁc tools are a critical part of wildlife management (Sinclair 1991).

Yet despite the serious ramiﬁcations of carnivore attacks for local livelihoods and carni-

vore conservation, surprisingly few quantitative techniques exist to assist stakeholders,

managers and policymakers in understanding and mitigating human–carnivore conﬂicts

such as livestock depredation (Treves and Karanth 2003; Woodroffe et al. 2005). The

studies in this review demonstrate that spatial predation risk modeling can serve as a

practical tool for guiding on-the-ground decision-making about where to implement pre-

ventative husbandry interventions and carnivore deterrents. Especially in situations where

authorities routinely collect information on livestock mortalities as part of ﬁnancial

incentive or compensation programs, risk modeling can use existing data to offer additional

insight into the spatiotemporal patterns and socio-ecological drivers of human–carnivore

conﬂict. As conservation practitioners increasingly recognize and use spatial predation risk

modeling as a valuable tool in the conservation toolkit, the technique will continue to

improve the effectiveness of mitigation efforts for reducing livelihood losses and

strengthening carnivore conservation.

Acknowledgments Oswald Schmitz, Adrian Treves, Y. V. Jhala, Walter Jetz, David Skelly, J.

S. Chauhan, Rakesh Shukla, Mark Hebblewhite, Meghna Agarwala, Anne Trainor, Colin Donihue, Wesley

Hochachka and several anonymous reviewers provided input that greatly improved this manuscript. Thank

you to the authors of the reviewed studies for sharing their experiences in applying risk models to con-

servation. Funding was provided by the American Institute for Indian Studies, American Philosophical

Society, Association of Zoos and Aquariums, John Ball Zoo Society, Yale Tropical Resources Institute

Endowment Fellowship and the United States National Science Foundation.

Biodivers Conserv

123

References

Abade L, Macdonald DW, Dickman AJ (2014) Assessing the relative importance of landscape and hus-

bandry factors in determining large carnivore depredation risk in Tanzania’s Ruaha landscape. Biol

Conserv 180:241–248

Agarwala M, Kumar S, Treves A, Naughton-Treves L (2010) Paying for wolves in Solapur, India and

Wisconsin, USA: comparing compensation rules and practice to understand the goals and politics of

wolf conservation. Biol Conserv 143:2945–2955

Alessa LN, Kliskey AA, Brown G (2008) Social–ecological hotspots mapping: a spatial approach for

identifying coupled social–ecological space. Landsc Urban Plan 85:27–39

Atwood TC, Gese EM, Kunkel KE (2009) Spatial partitioning of predation risk in a multiple predator–

multiple prey system. J Wildl Manag 73:876–884

Balme GA, Hunter LTB, Slotow R (2007) Feeding habitat selection by hunting leopards Panthera pardus in

a woodland savanna: prey catchability versus abundance. Anim Behav 74:589–598

Baruch-Mordo S, Breck SW, Wilson KR, Theobald DM (2008) Spatiotemporal distribution of black bear–

human conﬂicts in Colorado, USA. J Wildl Manag 72:1853–1862

Baruch-Mordo S, Wilson KR, Lewis DL et al (2014) Stochasticity in natural forage production affects use of

urban areas by black bears: implications to management of human–bear conﬂicts. PLoS One 9:1–10

Behdarvand N, Kaboli M, Ahmadi M et al (2014) Spatial risk model and mitigation implications for wolf–

human conﬂict in a highly modiﬁed agroecosystem in western Iran. Biol Conserv 177:156–164

Boyce MS, Vernier P, Nielsen S, Schmiegelow F (2002) Evaluating resource selection functions. Ecol

Model 157:281–300

Breitenmoser U, Angst C, Landry J-M et al (2005) Non-lethal techniques for reducing depredation. In:

Woodroffe R, Thirgood S, Rabinowitz A (eds) People and wildlife: conﬂict or coexistence? Cambridge

University Press, Cambridge, pp 49–71

Brown JS (1999) Vigilance, patch use and habitat selection: foraging under predation risk. Evol Ecol Res

1:49–71

Brown G, Raymond C (2007) The relationship between place attachment and landscape values: toward

mapping place attachment. Appl Geogr 27:89–111

Brown JS, Laundre

´JW, Gurung M (1999) The ecology of fear: optimal foraging, game theory, and trophic

interactions. J Mammal 80:385–399

Bryssinckx W, Ducheyne E, Muhwezi B et al (2012) Improving the accuracy of livestock distribution

estimates through spatial interpolation. Geospat Health 7:101–109

Bump JK, Tischler KB, Schrank AJ et al (2009) Large herbivores and aquatic-terrestrial links in southern

boreal forests. J Anim Ecol 78:338–345

Chapin FS III, Carpenter SR, Koﬁnas GP et al (2010) Ecosystem stewardship: sustainability strategies for a

rapidly changing planet. Trends Ecol Evol 25:241–249

Clark JD, Dunn JE, Smith KG (1993) A multivariate model of female black bear habitat use for a geographic

information system. J Wildl Manag 57:519–526

Creel S, Christianson D (2008) Relationships between direct predation and risk effects. Trends Ecol Evol

23:194–201

Dale MR, Fortin M-J (2014) Spatial analysis: a guide for ecologists, 2nd edn. Cambridge University Press,

Cambridge

Davie HS, Murdoch JD, Lhagvasuren A, Reading RP (2014) Measuring and mapping the inﬂuence of

landscape factors on livestock predation by wolves in Mongolia. J Arid Environ 103:85–91

DeCesare NJ (2012) Separating spatial search and efﬁciency rates as components of predation risk. Proc R

Soc B 279:4626–4633

Dickman AJ (2010) Complexities of conﬂict: the importance of considering social factors for effectively

resolving human–wildlife conﬂict. Anim Conserv 13:458–466

Dickman AJ, Macdonald EA, Macdonald DW (2011) A review of ﬁnancial instruments to pay for predator

conservation and encourage human–carnivore coexistence. Proc Natl Acad Sci USA 108:13937–13944

Dormann CF, Schymanski SJ, Cabral J et al (2012) Correlation and process in species distribution models:

bridging a dichotomy. J Biogeogr 39:2119–2131

Edge JL, Beyer DE, Belant JL et al (2011) Adapting a predictive spatial model for wolf Canis spp. predation

on livestock in the Upper Peninsula, Michigan, USA. Wildl Biol 17:1–10

Garcı

´a-Rangel S, Pettorelli N (2013) Thinking spatially: the importance of geospatial techniques for car-

nivore conservation. Ecol Inform 14:84–89

Gervasi V, Nilsen EB, Odden J et al (2013) The spatio-temporal distribution of wild and domestic ungulates

modulates lynx kill rates in a multi-use landscape. J Zool 292:175–183

Biodivers Conserv

123

Getis A, Ord JK (1992) The analysis of spatial association. Geogr Anal 24:189–206

Gorini L, Linnell JDC, May R et al (2012) Habitat heterogeneity and mammalian predator–prey interactions.

Mamm Rev 42:55–77

Hebblewhite M, Merrill EH, McDonald T (2005) Spatial decomposition of predation risk using resource

selection functions: an example in a wolf–elk predator–prey system. Oikos 111:101–111

Hopcraft JGC, Sinclair ARE, Packer C (2005) Planning for success: Serengeti lions seek prey accessibility

rather than abundance. J Anim Ecol 74:559–566

Howery LD, Deliberto TJ (2004) Indirect effects of carnivores on livestock foraging behavior and pro-

duction. Sheep Goat Res J 19:53–57

Inskip C, Zimmermann A (2009) Human–felid conﬂict: a review of patterns and priorities worldwide. Oryx

43:18–35

Inskip C, Ridout M, Fahad Z et al (2013) Human–tiger conﬂict in context: risks to lives and livelihoods in

the Bangladesh Sundarbans. Hum Ecol 41:169–186

Johnson CJ, Nielsen SE, Merrill EH et al (2006) Resource selection functions based on use-availability data:

theoretical motivation and evaluation methods. J Wildl Manag 70:347–357

Kaartinen S, Luoto M, Kojola I (2009) Carnivore–livestock conﬂicts: determinants of wolf (Canis lupus)

depredation on sheep farms in Finland. Biodivers Conserv 18:3503–3517

Karanth KK, Gopalaswamy AM, DeFries RS, Ballal N (2012) Assessing patterns of human–wildlife con-

ﬂicts and compensation around a central Indian protected area. PLoS One 7:e50433

Karanth KK, Gopalaswamy AM, Prasad PK, Dasgupta S (2013) Patterns of human–wildlife conﬂicts and

compensation: insights from Western Ghats protected areas. Biol Conserv 166:175–185

Kissling WD, Ferna

´ndez N, Paruelo JM (2009) Spatial risk assessment of livestock exposure to pumas in

Patagonia, Argentina. Ecography 32:807–817

Kluever BM, Breck SW, Howery LD et al (2008) Vigilance in cattle: the inﬂuence of predation, social

interactions, and environmental factors. Rangel Ecol Manag 61:321–328

Kluever BM, Howery LD, Breck SW, Bergman DL (2009) Predator and heterospeciﬁc stimuli alter

behaviour in cattle. Behav Process 81:85–91

Laporte I, Muhly TB, Pitt JA et al (2010) Effects of wolves on elk and cattle behaviors: implications for

livestock production and wolf conservation. PLoS One 5:e11954

Laundre

´JW, Hernandez L, Ripple WJ (2010) The landscape of fear: ecological implications of being afraid.

Open Ecol J 3:1–7

Lele SR, Merrill EH, Keim J, Boyce MS (2013) Selection, use, choice and occupancy: clarifying concepts in

resource selection studies. J Anim Ecol 82:1183–1191

Letnic M, Ritchie EG, Dickman CR (2012) Top predators as biodiversity regulators: the dingo Canis lupus

dingo as a case study. Biol Rev Camb Philos Soc 87:390–413

Li J, Heap AD (2011) A review of comparative studies of spatial interpolation methods in environmental

sciences: performance and impact factors. Ecol Inform 6:228–241

Lichtenfeld LL, Trout C, Kisimir EL (2014) Evidence-based conservation: predator-proof bomas protect

livestock and lions. Biodivers Conserv 24:483–491

Lima SL, Bednekoff PA (1999) Temporal variation in danger drives antipredator behavior: the predation

risk allocation hypothesis. Am Nat 153:649–659

Lima SL, Dill LM (1990) Behavioral decisions made under the risk of predation: a review and prospectus.

Can J Zool 68:619–640

Manly BF, McDonald LL, Thomas DL (2002) Resource selection by animals: statistical analysis and design

for ﬁeld studies, 2nd edn. Springer, Boston

Marucco F, McIntire EJB (2010) Predicting spatio-temporal recolonization of large carnivore populations

and livestock depredation risk: wolves in the Italian Alps. J Appl Ecol 47:789–798

McManus JS, Dickman AJ, Gaynor D et al (2014) Dead or alive? Comparing costs and beneﬁts of lethal and

non-lethal human–wildlife conﬂict mitigation on livestock farms. Oryx 1–9. doi:10.1017/

S0030605313001610

Meena V, Macdonald DW, Montgomery RA (2014) Managing success: Asiatic lion conservation, interface

problems and peoples’ perceptions in the Gir Protected Area. Biol Conserv 174:120–126

Miller JRB, Jhala YV, Jena J, Schmitz OJ (2015) Landscape-scale accessibility of livestock to tigers:

implications of spatial grain for modeling predation risk to mitigate human–carnivore conﬂict. Ecol

Evol 5:1354–1367

Ogada MO, Woodroffe R, Oguge NO, Frank LG (2003) Limiting depredation by African carnivores: the

role of livestock husbandry. Conserv Biol 17:1521–1530

Pienaar EF, Telesco D, Barrett S (2015) Understanding people’s willingness to implement measures to

manage human–bear conﬂict in Florida. J Wildl Manag 79:798–806

Biodivers Conserv

123

Rambaldi G, Kyem P, Mccall M, Weiner D (2006) Participatory spatial information management and

communication in developing countries. Electron J Inf Syst Dev Ctries 25:1–9

Ripple WJ, Estes JA, Beschta RL et al (2014) Status and ecological effects of the world’s largest carnivores.

Science 343:1–11

Ripple WJ, Newsome TM, Wolf C et al (2015) Collapse of the world’s largest herbivores. Sci Adv

1:e1400103

Robinson TP, Franceschini G, Wint W (2007) The food and agriculture organization’s gridded livestock of

the world. Vet Ital 43:745–751

Schiess-Meier M, Ramsauer S, Gabanapelo T, Ko

¨nig B (2007) Livestock predation—insights from problem

animal control registers in Botswana. J Wildl Manag 71:1267–1274

Schmitz OJ, Beckerman AP, O’Brien KM (1997) Behaviorally mediated trophic cascades: effects of pre-

dation on food web interactions. Ecology 78:1388–1399

Schmolke A, Thorbek P, DeAngelis DL, Grimm V (2010) Ecological models supporting environmental

decision making: a strategy for the future. Trends Ecol Evol 25:479–486

Shivik JA (2006) Tools for the edge: what’s new for conserving carnivores. Bioscience 56:253–259

Shrader AM, Brown JS, Kerley GIH, Kotler BP (2008) Do free-ranging domestic goats show ‘‘landscapes of

fear’’? Patch use in response to habitat features and predator cues. J Arid Environ 72:1811–1819

Sih A (1984) The behavioral response race between predator and prey. Am Nat 123:143–150

Sinclair ARE (1991) Science and the practice of wildlife management. J Wildl Manag 55:767–773

Soh YH, Carrasco LR, Miquelle DG et al (2014) Spatial correlates of livestock depredation by Amur tigers

in Hunchun, China: relevance of prey density and implications for protected area management. Biol

Conserv 169:117–127

Stahl P, Vandel JM, Herrenschmidt V, Migot P (2001) The effect of removing lynx in reducing attacks on

sheep in the French Jura Mountains. Biol Conserv 101:15–22

Stankowich T, Blumstein DT (2005) Fear in animals: a meta-analysis and review of risk assessment. Proc

Lond R Soc B 272:2627–2634

Suryawanshi KR, Bhatnagar YV, Redpath S, Mishra C (2013) People, predators and perceptions: patterns of

livestock depredation by snow leopards and wolves. J Appl Ecol 50:550–560

Trainor AM, Schmitz OJ (2014) Infusing considerations of trophic dependencies into species distribution

modelling. Ecol Lett 17:1507–1517

Treves A, Karanth KU (2003) Human–carnivore conﬂict and perspectives on carnivore management

worldwide. Conserv Biol 17:1491–1499

Treves A, Naughton-Treves L, Harper EK et al (2004) Predicting human–carnivore conﬂict: a spatial model

derived from 25 years of data on wolf predation on livestock. Conserv Biol 18:114–125

Treves A, Wallace RB, Morales A (2006) Co-managing human–wildlife conﬂicts: a review. Hum Dimens

Wildl 11:383–396

Treves A, Martin KA, Wydeven AP, Wiedenhoeft JE (2011) Forecasting environmental hazards and the

application of risk maps to predator attacks on livestock. Bioscience 61:451–458

Willems EP, Hill RA (2009) Predator-speciﬁc landscapes of fear and resource distribution: effects on spatial

range use. Ecology 90:546–555

Wilson SM, Madel MJ, Mattson DJ et al (2005) Natural landscape features, human-related attractants, and

conﬂict hotspots: a spatial analysis of human-grizzly bear conﬂicts. Biol Conserv 16:117–129

Wilson SM, Neudecker G, Jonkel J (2014) Human-grizzly bear coexistence in the Blackfoot River

Watershed, Montana getting ahead of the conﬂict curve. In: Clark SG, Rutherford MC (eds) Large

carnivore conservation: integrating science and policy in the North American West. The University of

Chicago Press, Chicago, pp 177–214

Woodroffe R, Thirgood S, Rabinowitz A (2005) People and wildlife: conﬂict or coexistence?. Cambridge

University Press, Cambridge

Wydeven AP, Treves A, Brost B, Wiedenhoeft JE (2004) Characteristics of wolf packs in Wisconsin:

identiﬁcation of traits inﬂuencing depredation. In: Fascione N, Delach A, Smith M (eds) People and

predators: from conﬂict to coexistence. Island Press, Washington, DC, pp 28–50

Zarco-Gonza

´lez MM, Monroy-Vilchis O, Alanı

´z J (2013) Spatial model of livestock predation by jaguar and

puma in Mexico: conservation planning. Biol Conserv 159:80–87

Biodivers Conserv

123

Online Resource 1

Miller, JRB (2015) Mapping attack hotspots to mitigate human-carnivore conflict: Approaches and applications

of spatial predation risk modeling. Biodivers Conserv 24(12):2887-2911. DOI 10.1007/s10531-015-0993-6.

http://link.springer.com/article/10.1007/s10531-015-0993-6

Table S1 Main mammalian carnivore species in the world responsible for depredating livestock. Adapted from

Breitenmoser et al. 2005

Family

Common Name

Scientific name

Canidae

Golden jackal

Canis aurenus

Coyote

C. latrans

Grey wolf

C. lupus

Dingo

C. l. dingo

Black-backed

jackal

C. mesomelas

Red wolf

C. rufus

African wild dog

Lycaon pictus

Ursidae

Andean bear

Tremarctos

ornatus

American black

bear

Ursus americanus

Brown bear

U. arctos

Asiatic black bear

U. thibetanus

Mustelidae

Wolverine

Gulo gulo

Hyaenidae

Spotted hyaena

Crocuta crocuta

Brown hyaena

Hyaena brunnea

Striped hyaena

H. hyaena

Felidae

Cheetah

Acinonyx jubatus

Caracal

Caracal caracal

Puma

Puma concolor

Asiatic golden cat

Felidae temmincki

Eurasian lynx

Lynx lynx

Bobcat

L. rufus

African/Asiatic lion

Panthera leo

Jaguar

P. onca

Leopard

P. pardus

Tiger

P. tigris

Snow leopard

Uncia uncia

Table S2 Published spatial risk modeling studies on carnivore-livestock interactions through March 2015

Citation

Country

Continent

Specific

region

Carnivore

Carnivor

e family

alphabeti

cal)

Prey (alphabetical)

Modeling method

Abade et al. 2014

Tanzania

Africa

Rungwa-

Ruaha

National Park

Leopard (Panthera

pardus), lion

(Panthera leo),

spotted hyaena

(Crocuta crocuta)

Felidae,

Hyaenidae

Cattle, goat, sheep

Correlation modeling (ensemble modeling

using Maxent, Ecological Niche Factor

Analysis and Support Vector Machines)

Baruch-Mordo et

al. 2008

USA

North

America

Western

Colorado

American black bear

(Ursus americanus)

Ursus

Alpaca, cattle, goat, llama,

pig, poultry, sheep

Spatial association (Getis-Ord G* clustering)

Behdarvand et al.

2014

Iran

Asia

Hamedan

providence

Wolf (Canis lupus)

Canidae

Cattle, goat, sheep

Correlation modeling (Maxent)

Davie et al. 2014

Mongolia

Asia

Ikh Nart

Nature

Reserve,

Dornogobi

Aimag

Wolf (Canis lupus)

Canidae

Camel (Camelus bactrianus),

cattle (Bos tarus), goat (Capra

aegragus), horse (Equus ferus

caballus), sheep (Ovis aries)

Spatial association (Mahalanobis D2 (k)

analysis)

Edge et al. 2011

USA

North

America

Upper

Peninsula,

Michigan

Wolf (Canis lupus)

Canidae

Cattle, sheep, poultry, rabbit

Correlation modeling (single-sample

discriminant-function analysis)

Kaartinen et al.

2009

Finland

Europe

southern and

central

Wolf (Canis lupus)

Canidae

Sheep

Correlation modeling (general additive

models)

Karanth et al.

2012

India

Asia

Kanha Tiger

Reserve,

Madhya

Pradesh

Leopard (Panthera

pardus), tiger

(Panthera tigris),

wild dog (Cuon

alpinus)

Canidae,

Felidae

Buffalo, cattle, goat

Correlation modeling (logistic regression) to

identify landscape attributes associated with

risk; spatial interpolation to map risk

probabilities

Karanth et al.

2013

India

Asia

Western

Ghats,

Karnataka

Leopard (Panthera

pardus), tiger

(Panthera

tigris),wild dog

(Cuon alpinus)

Canidae,

Felidae

Buffalo, cattle, goat

Correlation modeling (logistic regression) to

identify landscape attributes associated with

risk; spatial interpolation to map risk

probabilities

Kissling et al.

2009

Argentina

South

America

Patagonia

Puma (Puma

concolor)

Felidae

Sheep (Ovis aries)

Correlation modeling (generalized linear

models)

Marucco and

McIntire 2010

Italy

Europe

western,

central and

eastern Alps

Wolf (Canis lupus)

Canidae

Sheep

Spatially explicit, individual-based model*

Meena et al.

2014

India

Asia

Gir Protected

Area, Gujarat

Asiatic lion

(Panthera leo

persica)

Felidae

Cattle, buffalo, goat, sheep

Spatial association (Getis-Ord G clustering)

Miller et al. 2015

India

Asia

Kanha Tiger

Reserve,

Madhya

Pradesh

Leopard (Panthera

pardus), tiger

(Panthera tigris)

Felidae

Cattle, buffalo, goat, pig

Correlation modeling (resource selection

function using logistic regression)

Shrader et al.

2008

South

Africa

Riemvasmaak,

Northern Cape

Caracal (Caracal

caracal)

Felidae

Goat

Spatial interpolation (giving-up density)

Soh et al. 2014

China

Asia

Hunchun, Jilin

Tiger (Panthera

tigris)

Felidae

Cattle

Correlation modeling (negative binomial zero-

inflated)

Treves et al.

2004

USA

North

America

Wisconsin,

Minnesota

Wolf (Canis lupus)

Canidae

Bison, cattle, poultry, sheep

Correlation modeling (single-sample

discriminant-function analysis)

Treves et al.

2011

USA

North

America

Wisconsin

Wolf (Canis lupus)

Canidae

Cattle

Correlation modeling (logistic regression)

Wilson et al.

2006

USA

North

America

Montana

Grizzly bear (Ursus

arctos)

Ursus

Cattle, sheep

Correlation modeling (logistic regression)

Zarco-González

et al. 2013

Mexico

North

America

Jaguar (Panthera

onca), puma (Puma

concolor)

Felidae

Cattle, goat, sheep

Correlation modeling (ensemble modeling

using Artificial Neural Network,

Environmental Distance, Genetic Algorithm

for Rule-set Production, Support Vector

Machines, Ecological Niche Factor Analysis

and Maxent

*Note: This study is not discussed in the review of modeling methods because it specifically addresses the recolonization of carnivores. To estimate

livestock depredation hotspots, the study multiplied the probability of wolf pack presence by the number of available livestock pastures, which is less

precise than modeling carnivore-livestock interactions more explicitly.

Fig S1 Maps compiled to examine global potential livestock depredation in Fig. 2. Top: Livestock density from

the Gridded Livestock of the World v.2.0 (Robinson et al., 2007). Bottom: The world’s primary carnivore

species responsible for depredating livestock (species list in Table S1 adapted from Breitenmoser et al. 2005;

species range maps from the IUCN Red List of Threatened Species v.2014.2 and Letnic et al. 2012)

References

Abade L, Macdonald DW, Dickman AJ (2014) Assessing the relative importance of landscape and husbandry

factors in determining large carnivore depredation risk in Tanzania’s Ruaha landscape. Biol Conserv

180:241–248

Baruch-Mordo S, Breck SW, Wilson KR, Theobald DM (2008) Spatiotemporal distribution of black bear–

human conflicts in Colorado, USA. J Wildl Manage 72:1853–1862

Breitenmoser U, Angst C, Landry J-M, et al (2005) In: Wodroffe, R, Thirgood, S, Rabinowitz, A (eds) People

and wildlife: Conflict or Coexistence? Cambridge University Press, Cambridge, pp 49–71

Davie HS, Murdoch JD, Lhagvasuren A, Reading RP (2014) Measuring and mapping the influence of landscape

factors on livestock predation by wolves in Mongolia. J Arid Environ 103:85–91

Edge JL, Beyer DE, Belant JL, et al (2011) Adapting a predictive spatial model for wolf Canis spp. predation on

livestock in the Upper Peninsula, Michigan, USA. Wildlife Biol 17:1–10

Kaartinen S, Luoto M, Kojola I (2009) Carnivore-livestock conflicts: determinants of wolf (Canis lupus)

depredation on sheep farms in Finland. Biodivers Conserv 18:3503–3517

Karanth KK, Gopalaswamy AM, DeFries RS, Ballal N (2012) Assessing patterns of human-wildlife conflicts

and compensation around a central Indian protected area. PLoS One 7:e50433

Karanth KK, Gopalaswamy AM, Prasad PK, Dasgupta S (2013) Patterns of human–wildlife conflicts and

compensation: insights from Western Ghats protected areas. Biol Conserv 166:175–185

Kissling WD, Fernández N, Paruelo JM (2009) Spatial risk assessment of livestock exposure to pumas in

Patagonia, Argentina. Ecography 32:807–817

Letnic M, Ritchie EG, Dickman CR (2012) Top predators as biodiversity regulators: the dingo Canis lupus

dingo as a case study. Biol Rev Camb Philos Soc 87:390–413

Marucco F, McIntire EJB (2010) Predicting spatio-temporal recolonization of large carnivore populations and

livestock depredation risk: Wolves in the Italian Alps. J Appl Ecol 47:789–798

Meena V, Macdonald DW, Montgomery RA (2014) Managing success: Asiatic lion conservation, interface

problems and peoples’ perceptions in the Gir Protected Area. Biol Conserv 174:120–126

Miller JRB, Jhala Y V, Jena J, Schmitz OJ (2015) Landscape-scale accessibility of livestock to tigers:

Implications of spatial grain for modeling predation risk to mitigate human-carnivore conflict. Ecol Evol

5:1354–1367

Robinson TP, Franceschini G, Wint W (2007) The food and agriculture organization’s gridded livestock of the

world. Vet Ital 43:745–751

Shrader AM, Brown JS, Kerley GIH, Kotler BP (2008) Do free-ranging domestic goats show “landscapes of

fear”? Patch use in response to habitat features and predator cues. J Arid Environ 72:1811–1819

Soh YH, Carrasco LR, Miquelle DG, et al (2014) Spatial correlates of livestock depredation by Amur tigers in

Hunchun, China: relevance of prey density and implications for protected area management. Biol Conserv

169:117–127

Treves A, Martin KA, Wydeven AP, Wiedenhoeft JE (2011) Forecasting environmental hazards and the

application of risk maps to predator attacks on livestock. Bioscience 61:451–458

Treves A, Naughton-treves L, Harper EK, et al (2004) Predicting human-carnivore conflict: a spatial model

derived from 25 years of data on wolf predation on livestock. Conserv Biol 18:114–125

Wilson SM, Madel M, Mattson D, et al (2006) Landscape conditions predisposing grizzly bears to conflicts on

private agricultural lands in the western USA. Biol Conserv 130:47–59

Zarco-González MM, Monroy-Vilchis O, Alaníz J (2013) Spatial model of livestock predation by jaguar and

puma in Mexico: conservation planning. Biol Conserv 159:80–87

Supplementary material 1

Data

November 2015

Jennifer R. B. Miller

Input Behavior of Farmer Production Factors in the Range of Asian Elephant Distribution: Survey Data from 1264 Households in Yunnan Province, China

Article

Full-text available

Nov 2023
Diversity

This article, based on the sustainable livelihood framework and survey data from 1264 households in Xishuangbanna Dai Autonomous Prefecture, Puer City, and Lincang City in Yunnan Province, China, analyzes the impact mechanism of livelihood capital on the production input behavior of farmers affected by Asian elephant damage and the moderating effect of Asian elephant damage on this process using ordinary least squares (OLS) models. The study finds the following: (1) Asian elephant damage has a significant negative effect on farmers’ production input, meaning that as the severity of Asian elephant damage increases, farmers reduce their input into agricultural production factors. (2) Livelihood capital has a significant positive effect on farmers’ production input, and both the increment and stock of livelihood capital promote an increase in farmers’ production input. (3) Asian elephant damage strengthens the influence of livelihood capital on farmers’ inputs of agricultural production factors. Based on these findings, four recommendations are proposed: emphasizing the cultivation and enhancement of farmers’ livelihood capital, improving strategies for managing and preventing wildlife damage, optimizing the economic compensation mechanism for human–wildlife conflicts, and adhering to sustainable development and resource allocation. These recommendations aim to enhance wildlife conservation and management policies, strengthen farmers’ risk-coping capabilities, and ensure the sustainability of agricultural production and livelihoods.

Robust inference and errors in studies of wildlife control

Preprint

Full-text available

Oct 2023

Randomized, controlled trials (RCT) are seen as the strongest basis for causal inference, but their strengths of inference and error rates relative to other study have never been quantified in wildlife control and rarely in other ecological fields. We simulate common study designs from simple correlation to RCT with crossover design. We report rates of false positive, false negative, and over-estimation of treatment effects for five common study designs under various confounding interactions and effect sizes. We find non-randomized study designs mostly unreliable and that randomized designs with suitable safeguards against biases have much lower error rates. One implication is that virtually all studies of lethal predator control interventions appear unreliable. Generally, applied fields can benefit from more robust designs against the common confounding effects we simulated.

Assessing human–carnivore conflict and the identification of hotspot areas to prioritize mitigation efforts

Article

Sep 2023
Hum Dimens Wildl

An ensemble modeling approach to predict spatial risk patches of the Persian leopard-livestock conflicts in Lorestan Province, Iran

Article

Full-text available

Jul 2023
ENVIRON SCI POLLUT R

This study was conducted in the Lorestan Province in the west of Iran with two objectives of identifying major environmental variables in spatial risk modeling and identifying spatial risk patches of livestock predation by the Persian leopard. An ensemble approach of three models of maximum entropy (MaxEnt), generalized boosting model (GBM), and random forest (RF) were applied for spatial risk modeling. Our results revealed that livestock density, distance to villages, forest density, and human population density were the most important variables in spatial risk modeling of livestock predation by the leopard. The center of the study area had the highest probability of livestock predation by the leopard. Ten spatial risk patches of livestock predation by the leopard were identified in the study area. In order to mitigate the revenge killing of the leopards, the findings of this study highlight the imperative of implementing strategies by the Department of Environment (DoE) to effectively accompany the herds entering the wildlife habitats with shepherds and a manageable number of guarding dogs. Accordingly, the identified risk patches in this study deserve considerable attention, especially three primary patches found in the center and southeast of Lorestan Province.

Where, why and how carnivores kill domestic animals in different parts of their ranges: An example of the Eurasian lynx

Article

Full-text available

Jul 2023

Protection of Farms from Wolf Predation: A Field Approach

Article

Full-text available

Jun 2023

Abstract: The livestock sector is facing serious challenges in combatting the increasing predation of domestic livestock. In this scenario, wild carnivores, especially wolves, represent key predators. To allow the coexistence of wild and domestic animals, defense methodologies consisting of multiple integrated antipredator strategies must be tested and implemented based on the geographical management context of each farm. This study investigated the potential of a novel antipredator method (PAN, Project Farmers-Nature in Italian) in protecting livestock (goats and horses) from wolves on a farm located in the Sibillini Mountains National Park, over a three-month period (June–September 2022). The PAN field approach involved two phases: (1) interviews with farmers and inspections of how the farm and pasture are structured and (2) monitoring predator abundance using camera traps and transects in order to understand the wildlife habits. Information on predator movement around the grazing area was shared with the farmer, who was actively involved in implementing strategies to protect livestock. The stable presence of one pair of wolves was confirmed in the grazing area, placing grazing livestock at risk. The farmer was advised to strengthen the existing antipredator strategy (herd protection dogs) by introducing two trained Maremma-Abruzzese sheepdog puppies to protect his animals. The implemented actions demonstrate how professional experts can serve as a strategic intermediary between livestock and wildlife conservation in the management of the current conflicts. Keywords: livestock farming; wolf; predations; monitoring; antipredator strategies

Human-Wildlife Conflict, SAFE Strategy

Book

Full-text available

Jun 2023

The Safe System approach of WWF, which is holistic and long-term, seeks to address the current shortcomings of HWC in Bhutan. The approach intends to create safe environment for both people and their assets and wildlife and their habitats in a landscape for their harmonious co-existence. The approach was piloted in nine Gewogs in four Dzongkhags. One of the activities of the pilot project was to conduct rapid assessment of HWC in these nine Gewogs. This report is based on the results of the rapid assessment.

Wildlife Conflict Area Demarcation—Special Reference to Human–Elephant Conflict

Chapter

Nov 2023

Human–elephant conflict (HEC) is presently a major issue in West Bengal. This chapter aims to demarcate conflict intensity zones based on selected landscape features, i.e. percentage of forest cover, multi-crop area density, built-up area density, forest habitat patch shape complexity road network buffering, canal buffering and water patch density. All of these data were obtained from individual forest range images (LISS-III, P6 Satellite image 2019). Individual thematic layers were prepared using kriging tool and reclassification done for five classes in ArcGIS 10.1 version software. The Analytic Hierarchy Process (AHP) method was used in order to elucidate weight and rank of selected individual parameters like forest cover, multi-crop area, built-up patch density, patch fragmentation, road and canal and water patch density. Weighted overlay analysis was applied to demarcate the wildlife conflict zones using selected parameter layers. This zonation will help the planners or policy makers for implementing preventive measures in controlling HEC effects.

Spatial Analysis of Attitudes Toward Timber Rattlesnakes ( Crotalus Horridus ) Where Encounters Occur Within an Exurban Landscape

Article

Oct 2023

Assessing spatio-temporal patterns of human-wildlife conflicts in a human-dominated landscape: a case study from Iran

Article

Full-text available

Jul 2023
BIODIVERS CONSERV

Human-wildlife conflict is considered one of the significant challenges that conservation authorities face in managing protected areas and the surrounding lands. This study aims to investigate the spatial and temporal trends and characteristics of conflicts between rural communities and wild animals in a protected area in northeastern Iran. We collected data on wild animals’ attacks on humans and livestock and damages to crops within the Ors-e-Sistan Protected Area (OSPA) and its 5 km buffer zone for 2010–2020 via official reports and questionnaire surveys (n = 373). We found that multiple species-specific and human-related factors determine conflict hotspots in space and time. Wild boars (Sus scrofa), wolves (Canis lupus) and leopards (Panthera pardus saxicolor) were perceived to be responsible for the highest number of attacks on humans and livestock. Wild boars were responsible for more than 90% of attacks on agricultural lands. Hotspots of attacks on livestock were located in the pastures around villages and pastures within the protected area. In contrast, hotspot areas of attacks on humans and damage to crops were located in the orchards and farms of villages on the protected area’s northern and southern fringes. Temporal patterns in the variability of conflict revealed that grazing seasons and species’ nocturnal behaviors caused more attacks and damages during warm seasons and nights. Conflict hotspots and their temporal clustering, identified in this study, can guide managers to focus mitigation activities in prioritized areas and allocate management resources that reduce conflicts between people and wild animals. The assessment of the type of species and their characteristics, land-use type, and distribution of human settlements is recommended when identifying locations and occasions of conflicts.

Non-lethal techniques for reducing depredation

Article

Full-text available

Jan 2005

Collapse of the world's largest herbivores

Article

Full-text available

May 2015

Large wild herbivores are crucial to ecosystems and human societies. We highlight the 74 largest terrestrial herbi-vore species on Earth (body mass > – 100 kg), the threats they face, their important and often overlooked ecosystem effects, and the conservation efforts needed to save them and their predators from extinction. Large herbivores are generally facing dramatic population declines and range contractions, such that ~60% are threatened with extinction. Nearly all threatened species are in developing countries, where major threats include hunting, land-use change, and resource depression by livestock. Loss of large herbivores can have cascading effects on other species including large carnivores, scavengers, mesoherbivores, small mammals, and ecological processes involving vegetation , hydrology, nutrient cycling, and fire regimes. The rate of large herbivore decline suggests that ever-larger swaths of the world will soon lack many of the vital ecological services these animals provide, resulting in enormous ecological and social costs.

Human–Grizzly Bear Coexistence in the Blackfoot River Watershed, Montana

Chapter

May 2014

This chapter describes how a rural agricultural community in the Blackfoot River watershed in Montana used collaborative decision making to reduce conflicts with grizzly bears. Through effective communication, regular meetings, and partnerships, a host of on-the-ground projects were implemented through the Blackfoot Challenge, a grassroots watershed group, that significantly reduced conflicts with grizzly bears. The effort focused on changing people's practices and behaviors, not changing their value systems, and emphasized protecting human safety and livelihoods. The chapter explores the specific ways that collaborative decision making was carried out and offers numerous field-based lessons and recommendations that will be useful to natural resource managers in facilitating human-grizzly bear coexistence.

BEHAVIORALLY MEDIATED TROPHIC CASCADES: EFFECTS OF PREDATION RISK ON FOOD WEB INTERACTIONS

Article

Jul 1997

Spatial analysis: a guide for ecologists

Book

Jan 2005

Status and ecological effects of the world’s largest carnivores

Article

Jul 2014

Spatial analysis: A guide for ecologists, second edition

Book

Sep 2014

Nowadays, ecologists worldwide recognize the use of spatial analysis as essential. However, because of the fast-growing range of methods available, even an expert might occasionally find it challenging to choose the most appropriate one. Providing the ecological and statistical foundations needed to make the right decision, this second edition builds and expands upon the previous one by: • Encompassing the basic methods for spatial analysis, for both complete census and sample data • Investigating updated treatments of spatial autocorrelation and spatio-temporal analysis • Introducing detailed explanations of currently developing approaches, including spatial and spatio-temporal graph theory, scan statistics, fibre process analysis, and Hierarchical Bayesian analysis • Offering practical advice for specific circumstances, such as how to analyze forest Permanent Sample Plot data and how to proceed with transect data when portions of the data series are missing. Written for graduates, researchers and professionals, this book will be a valuable source of reference for years to come.

Spatial Analysis: A Guide for Ecologist

Article

Apr 2005

The number and variety of statistical techniques for spatial analysis of ecological data are burgeoning and many ecologists are unfamiliar with what is available and how the techniques should be used. This book provides an overview of the wide range of spatial statistics available to analyze ecological data, and provides advice and guidance for graduate students and practicing researchers who are either about to embark on spatial analysis in ecological studies or who have started but need guidance to proceed. © M.-J. Fortin and M.R.T. Dale 2005 and Cambridge University Press 2009.

The analysis of spatial association

Article