Impacts of deer management practices on the spatial dynamics of the tick Ixodes ricinus: A scenario analysis

Ecological Modelling 276 (2014) 1–13

Contents lists available at ScienceDirect

Ecological Modelling journal homepage: www.elsevier.com/locate/ecolmodel

Impacts of deer management practices on the spatial dynamics of the tick Ixodes ricinus: A scenario analysis Sen Li a,∗ , Sophie O. Vanwambeke a , Alain M. Licoppe b , Niko Speybroeck c a

Georges Lemaître Centre for Earth and Climate Research, Earth and Life Institute, Université catholique de Louvain, Louvain-la-Neuve, Belgium Department of Natural and Agricultural Environment Studies, Operational Directorate-General for Agriculture, Natural Resources and the Environment (DGARNE), Gembloux, Belgium c Institute of Health and Society (IRSS), Université catholique de Louvain, Brussels, Belgium b

a r t i c l e

i n f o

Article history: Received 13 June 2013 Received in revised form 12 December 2013 Accepted 21 December 2013 Available online 23 January 2014 Keywords: Cellular automata Deer management Spatial heterogeneity Tick population ecology Tick control

a b s t r a c t Deer, for example roe deer, red deer and fallow deer, are the common reproduction host types for European Ixodes ricinus ticks. Understanding the consequences of deer management on the spatial dynamics of ticks may advise the risk management of tick-borne diseases, and thus be of public health importance. We present a scenario analysis to understand such consequences by integrating multi-disciplinary knowledge within a predictive modelling framework. A spatial tick population model was adopted to explore how changes in the host population may affect woodland patch- and landscape-level tick dynamics. Scenarios on current and foreseen European deer management strategies were built based on expert knowledge. These scenarios were then tested through the described model for their potential effectiveness as tick control strategies. Our models indicate that: (i) reducing local deer densities could not effectively reduce tick abundance if woodland patches are well-connected, allowing deer population exchanges, (ii) controlling deer grazing intensity in grassland may not be an effective tick control strategy, (iii) local extinction of deer could decrease tick abundance considerably but deer reintroduction could lead to fast tick upsurge, and (iv) controlling human disturbances may reduce the tick density at landscape-level, as well as tick “hotspots” (i.e., areas with unusually high tick density) at woodland patch-level. Our results can instruct policy-makers on the potential impact on public health of wildlife management strategies, and suggest empirical investigations of disease risks. For optimising such simulation studies on disease risks, however, a better understanding of how biodiversity may influence the ecology of tick and pathogen transmission is required. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Ixodes ricinus (Acari: Ixodidae) is the most abundant and widely distributed tick species in Europe. Over the past two decades, the reported incidence of tick-borne diseases, such as Lyme borreliosis and tick-borne encephalitis, has increased in Europe (Randolph, 2008). This has made tick-borne zoonoses the most prevalent vector-borne diseases in Europe and hence kindled public health and scientific attention (Morens et al., 2004). Host availability is fundamental for the maintenance of ticks and pathogen populations (Ruiz-Fons and Gilbert, 2010). Success

∗ Corresponding author at: 457, Place Louis Pasteur 3, B1348 Louvain-la-Neuve, Belgium. Tel.: +32 0 10472870; fax: +32 0 10472877. E-mail addresses: [email protected], [email protected] (S. Li), [email protected] (S.O. Vanwambeke), [email protected] (A.M. Licoppe), [email protected] (N. Speybroeck). 0304-3800/$ – see front matter © 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.ecolmodel.2013.12.023

in managing tick-borne diseases depends on understanding tick dynamics, including the way in which tick feeding success is affected by varying host population densities and distributions. Ixodes ticks feed on a broad range of hosts. I. ricinus larvae primarily feed on small-sized animals such as rodents and insectivores (Gray et al., 1994; Jaenson and Talleklint, 1996), while hosts for nymphs are less specific, including birds, and small, medium, and ˜ et al., 2005). Adult ticks often large-sized mammals (Estrada-Pena have a narrower host range, with medium- and large-sized hosts, for example deer (Rizzoli et al., 2009; Tagliapietra et al., 2011). This three-host tick life cycle can be impacted by a diapause occurring at different life stages such that its duration may extend over two years (Gray, 1981). Ixodes ticks can be transported across the landscape as hosts move. Landscape effects, e.g., woodland fragmentation, shaping host distribution and limiting host movement at different spatial scales, can yield different influences on tick populations in different life stages (Li et al., 2012a). Deer and other large mammals are the target of direct management practices (e.g., fencing, hunting, translocation, etc.). In previous empirical studies,

2

S. Li et al. / Ecological Modelling 276 (2014) 1–13

deer density has been found to be positively related to tick densities and the tick infestations on rodents (Cagnacci et al., 2012; Killilea et al., 2008). There is a need to analyse the potential consequences of these management practices on the spatial dynamics of ticks, so that both wildlife managers and the public would be better informed about potential changes in disease risks (Luckhart et al., 2010). Predicting the impact of deer management on the spatial dynamics of I. ricinus requires a thorough understanding of (i) deer management practices, (ii) the spatially explicit response of deer movement patterns to such management practices and (iii) the spatial ecology of tick-deer relations. Agent-based modelling and cellular automata have the capacity to integrate these aspects (Grimm et al., 2005). To date, only a few studies have reported on biological process-based models modelling tick populations in a spatially explicit way (Killilea et al., 2008). These models have been used mainly to explore how spatial dynamics of tick densities and/or infections can be influenced by landscape heterogeneity and host movements. Some recent examples are: a multi-habitat model (Hoch et al., 2010) for I. ricinus tick, an agent-based model (Wang et al., 2012) for Amblyomma americanum, a multi-patch model (Watts et al., 2009) and a reaction diffusion model (Jones et al., 2011) for louping-ill virus, and a cellular automata model for Lyme disease (Li et al., 2012a). In general, these studies support either a single host following a random exploratory movement or assume a certain number of hosts continuously travelling between woodland patches. However, in the field, animals can exhibit a set of statistically distinguishable movement patterns as a function of their internal state, motion capacity, navigation capacity, environmental conditions and the management strategies impacting them (Nathan et al., 2008). Therefore, a multiphasic hypothesis for the host movement pattern (i.e., considering more than one host movement phase, e.g., both home-ranging phase, in which animals randomly move within their home range, and displacement phase, in which animals tend to move over longer distances for new habitats) may be an interesting extension to the current spatial tick modelling context. This study aimed to investigate the consequences of deer management practices on the spatial dynamics of ticks. We firstly described a spatial tick model adapted from a previous study (Li et al., 2012a). Then, based on expert knowledge, four scenarios of current and foreseen deer management practices were built, namely: (i) reducing local deer density, (ii) controlling deer grazing intensity in grassland, (iii) translocation of deer species, and (iv) controlling human disturbance and deer displacement between woodland patches. The influence on the spatial dynamics of ticks under these scenarios was tested through the model. Finally, we discussed these results with a focus on the effectiveness of these practices as tick controlling strategies.

2. Materials and methods 2.1. Model profile The cellular automata model for the spatial ecology of ticks was adapted from Li et al. (2012a) and included significant modifications compared to the original model. Firstly, the tick feeding pattern was more detailed to better represent the reality by enabling low probabilities for adults to feed on small mammals (Cagnacci et al., 2012) and larvae to feed on deer (Kiffner et al., 2010). The simplified method based on fixed tick attachment rates in the original model was replaced by a more detailed one based on host-finding probabilities of ticks and controlled by the feeding capacities of ticks on hosts. This adaption further allows to simulate the effects of deer density, as a tick amplifier, on tick infestation

levels (Cagnacci et al., 2012). Secondly, rules for deer movement were developed to distinguish behaviours in home-ranging and displacement phases, allowing to represent more realistic deer movement patterns. Finally, pathogen transmission functions were not included as the study at hand focused on tick populations. The modifications aimed at improving the model’s specificity, as the focus in this study was on the effects of the movements of the major host types on the abundance and distribution of ticks. The model was coded into Repast Simphony (North et al., 2006). The space used was two-dimensional and organised as a lattice of cells. Each cell in the lattice had three layers: a tick population, a host population and a landscape layer. The lattice was programmed to evolve in discrete time steps following a set of transition rules to update the cell states simultaneously. The present model adopts a cell size of 1 ha and a time step of 1 week. 2.1.1. States and parameters The three layers of cell states were: Tick population layer: A “larva”–“nymph”–“adult” life stage structure was used (Fig. 1). In each stage, ticks could be in questing, feeding or interstadial development phases. When encountering hosts, questing ticks attached for blood meals, then dropped and developed into the next life stages. Female adult ticks (assuming half of the emerged adult ticks were females) also needed blood meals to produce (i.e., lay eggs that hatch into) larval ticks. Host population layer: Two generalised host types were used: small mammals (including rodents and lagomorphs) and deer. Each life stage of the ticks was assumed capable of feeding on both host types. The overall number of the two host types was fixed but the host distributions could vary between time steps as a result of movements. Indeed, host movements resulted in ticks being transported from one place to another. Landscape layer: The cell states used were “woodland”– “grassland”–“non-vegetated areas”. Deer were considered to inhabit woodland mostly and to spend a proportion of time in grassland. Ticks could inhabit grassland but woodland was considered more suitable. Small mammals could inhabit both woodland and grassland. Mortality rates of ticks, densities of hosts and movement patterns of hosts varied with land cover types. Hosts could enter non-vegetated areas, but would not stay, meaning that no ticks would drop off in non-vegetated area. The landscape layer was set static and could be based on either real or theoretical landscape maps. Parameters in the model were estimated from field and laboratory observations found in the literature (Table 1). 2.1.2. Transition rules 2.1.2.1. Rules for modelling the tick population development. For each cell at time step t, we modelled the questing tick populations as (see Table 1 for the values used and the sources of parameters): qLt = (1 − mqL ) · qLt−1 + ˇ · 0.5 · (1 − mAL )

dAL

D S ·(1 − mfL ) · (fAD + fASt−dAL −1 ) − fLt−1 − fLt−1 t−dAL −1

qNt = (1 − mqN ) · qNt−1 + (1 − mLN )

dLN

· (1 − mfN )

D S D S ·(fLt−d LN −1 + fLt−dLN −1 ) − fNt−1 − fNt−1

qAt = (1 − mqA ) · qAt−1 + (1 − mNA )

dNA

(1)

(2)

· (1 − mfA )

D S D S ·(fNt−d NA −1 + fNt−dNA −1 ) − fAt−1 − fAt−1

(3)

S. Li et al. / Ecological Modelling 276 (2014) 1–13

3

Fig. 1. Tick life cycle in the model Modified from Li et al. (2012a).

where qL, qN and qA indicate the questing populations of larval, nymphal and adult ticks; mqL , mqN and mqA the weekly tick mortality survival rates in questing phases; mfL , mfN and mfA the weekly tick mortality survival rates in feeding phases; dAL , dLN and dNA are development periods between feeding and moulting into the next life stages and mAL , mLN and mNA are weekly tick mortality rates in those periods. All m and d parameters are constant. All mortality rates are considered four times higher in grassland than in woodland (Sf), in relation to the drier conditions in grassland. ˇ indicates the number of eggs per adult. The factor of 0.5 in Eq. (1) indicates that half of the adult ticks are assumed to be eggs-producing females.fL, fN and fA indicate whether the feeding populations were larval, nymphal and adult ticks. Superscripts D and S indicate the tick population is feeding on deer and small mammals. The number of weekly feeding ticks in each cell was estimated by fTH = qTt ·pTH , where qT refers to the questing tick population of a life stage (L, N, A) and pTH , the host-finding probabilities, refers to the probability of ticks in the life stage (T) to attach on a host type (D, S). The estimated feeding ticks per host were assumed not to exceed the feeding capacity on the host (C), or the maximum number of tick attachments on one host in one week. Such capacities were estimated based on field observations of the ranges of tick attachments per host and tick feeding durations (Table 1). For example, given that the duration of the larva attachment can be as short as two days and the number of larvae on one small mammal can be as many as 30 (Cagnacci et al., 2012; Gray, 2002), the weekly feeding capacity of small mammals for larvae (CLS ) was estimated as (30 larvae/2 days) × 7 days = 105 larvae. Feeding durations were assumed to be less than one week for larvae (approximately three days, c.f. Macleod (1932)), and one week for nymphs and adults (Hancock et al., 2011).

within their home range, while (ii) in displacement phase, animals tend to move over longer distances for new habitats. Such a multiphasic hypothesis has been utilised in both theoretical models (Bartumeus et al., 2005; Skalski and Gilliam, 2003; Zhang et al., 2007) and empirical studies on movement ecology (Fryxell et al., 2008; Morales et al., 2004). Deer, for example roe deer Capreolus capreolus, red deer Cervus elaphus, and fallow deer Dama dama, are capable of moving over long distances. In the model, deer populations were initially considered in home-ranging phase but a proportion were allowed to switch to a displacement phase (pDis) when competition or human disturbance increased (Nathan et al., 2008). During the home-ranging phase, deer were allowed to explore within and around woodland. They could also venture into grassland for short stays (Putman, 1986), i.e., a proportion of a time step (pG). It has been reported that, the home ranges of roe deer, red deer and fallow deer can be as large as 1 km2 , 7.5 km2 or 20 km2 (Putman, 1988). During the displacement phase, deer were allowed to move over long distances (i.e., up to 5 km for roe deer and 120 km for red deer per week has been reported) looking for new woodland habitats that are not over-populated (or below the limit of the carrying capacity, K, of the woodland cell) (Georgii and Schroder, 1983; Mysterud, 1999; Wahlstrom and Liberg, 1995). In this phase, deer were assumed to travel quickly through grassland and the time spent in grassland was considered negligible. An extended Moore neighbourhood (Fig. 2) was considered for host movement. The host movement capacity (MC) denotes the minimum distance travel from the centre to the boundary of the neighbourhood. Thus the neighbourhood sizes were different for different host types in different movement phases. Two sets of rules were designed, respectively, for host movement patterns in home ranging and displacement phases.

2.1.2.2. Rules for modelling the host movement and tick transportation in a cellular space. A multi-phasic movement pattern was considered: (i) in home-ranging phase, animals randomly move

Movement rules in home ranging phase: The extended neighbourhood in this phase represents the home range of the host type concerned. By assuming that hosts had an equal probability to

4

S. Li et al. / Ecological Modelling 276 (2014) 1–13

Table 1 Model parameters. Parameter

Symbol

Wherea

Value

Source

Parameters in relation to tick life cycle Average no. eggs per adult

ˇ

W/G

2000

Randolph and Craine (1995)

Mortality rate in woodland (week−1 ) Questing larvae Questing nymphs Questing adult ticks Developing from engorged larvae into questing nymphs

mqL mqN mqA mLN

W W W W

0.03 0.03 0.02 0.03

Developing from engorged nymphs into questing adults

mNA

W

0.01

Developing from engorged adults into questing larvae

mAL

W

0.02

Scaling factor for mortality rates of questing and developing ticks

Sf

G

4

Daniel et al. (1976) Daniel et al. (1976) Daniel et al. (1976) Daniel et al. (1976), Gray (1981) Daniel et al. (1976), Gray (1981) Daniel et al. (1976), Gray (1981) Mount et al. (1997), Talleklint and Jaenson (1997)

Feeding mortality rate (week−1 ) Larvae

mfL

H

0.5

mfN

H

0.5

fA

m

H

0.35

dLN dNA dAL

W/G W/G W/G

45 53 45

Macleod (1932) Macleod (1932) Macleod (1932)

pLS pNS pAS

W/G W/G W/G

0.3 0.2 0.01

Hancock et al. (2011) Hancock et al. (2011) Assumption

pLD pND pAD

W/G W/G W/G

0.001 0.1 0.3

Assumption Hancock et al. (2011) Hancock et al. (2011)

CLS

H

105

Maximum nymph attachments on one small mammal

CNS

H

6

Maximum adult attachments on one small mammal

CAS

H

0.1

LD

H

800

Maximum nymph attachments on one deer

ND

C

H

300

Maximum adult attachments on one deer

CAD

H

150

Cagnacci et al. (2012), Gray (2002) Oliver (1989), Cagnacci et al. (2012) Cagnacci et al. (2012), Oliver (1989) Gray (2002), Kiffner et al. (2010) Oliver (1989), Kiffner et al. (2010) Oliver (1989), Kiffner et al. (2010)

Parameters in relation to host movement patterns Movement capacity (m week−1 ) Small mammal in home ranging phase Deer in home ranging phase Deer in displacement phase

MCS MCD H MCD D

W/G W W

100 500 2000

Proportion of time step spent in grassland for deer (%) Proportion of deer population in displacement phase (%)

pG pDis

G W

35 0.1

Parameters in relation to landscape heterogeneity Density of small mammal (ha−1 )

dS

W/G

80

D

W W/G W

0.12 120 0.3

Nymphs Adults Duration of developmental phases (week) Developing from engorged larvae into questing nymphs Developing from engorged nymphs into questing adults Developing from engorged adults into questing larvae Small mammals finding probability Questing larvae Questing nymphs Questing adults Deer finding probability Questing larvae Questing nymphs Questing adults Host feeding capacity (week−1 ) Maximum larva attachments on one small mammal

Maximum larva attachments on one deer

−1

Density of deer (ha ) Carrying capacity of small mammal (ha−1 ) Carrying capacity of deer (ha−1 ) a

C

d KS KD

Gray (1981), Hancock et al. (2011) Gray (1981), Hancock et al. (2011) Gray (1981), Hancock et al. (2011)

Kikkawa (1964) Putman (1988) Georgii and Schroder (1983), Mysterud (1999), Wahlstrom and Liberg (1995) Putman (1986) Assumption Escutenaire et al. (2000) Delbeuck (2008), OFFH Assumption Assumption

This column indicates where the parameter can be applied: W, in woodland; G, in grassland, H, on host animals.

move in each direction, the home ranging population of a host type in the original cell was evenly divided into eight parts to be considered for movement. In each time step, the following two sub-steps (i) and (ii), respectively, for the trigger and completion of movement were repeated eight times, once for each direction. In sub-step (i), a location of the destination was randomly generated

by using a modified Gaussian model (a detailed description can be found in Appendix S2 in Li et al. (2012a)) giving the concerned home ranging host population (i.e., the one eighth part) a higher probability to move at shorter distances. A movement trigger value of 100 m (size of one sell) was adopted, representing the required distance to reach the adjacent cells. Thus, the hosts would stay in

S. Li et al. / Ecological Modelling 276 (2014) 1–13

5

for less than one week. When a proportion of hosts had moved between habitats, the same proportion of feeding ticks was transported. When deer had only spent a proportion of a time step in grassland, the same proportion of ticks feeding on them dropped off, and the same proportion of questing ticks were picked up. 2.1.3. Process overview and scheduling At each successive time step, transition rules for tick development and host movement patterns were applied. For each life stage, questing ticks that emerged (moulted in the previous time step from a previous life stage) were added in. Questing ticks that died in the previous time step were removed. Then, ticks attached on hosts in the same cell and host movements were considered. Attached ticks on the out-moving animals were transported and all dropped off at the end of the time step. Cell states were updated simultaneously at each time step. Fig. 2. Extended Moore neighbourhood. There are eight directions: east, northeast, north, northwest, west, southwest, south, and southeast.

the original cell if the distance to the randomly-generated destination was less than the movement trigger value. If the distance was higher than the hosts’ movement capacity (with a low probability), the destination would be adjusted to the closest boundary cell of the extended neighbourhood, so that the movements beyond home ranges were disallowed. In sub-step (ii), there were two different situations when considering the completion of movement. First, completion of movement was possible when the destination was a suitable habitat for the concerned host type. The hosts could then only complete the movement if the host population was below the carrying capacity of its host type in the destination cell. Second, completion of movement was not possible when the destination was a less suitable habitat for the concerned host type (i.e., when considering the deer moving to grassland). The hosts would then spend a proportion of time in the cell and then return to the original cell. In all other situations, the movements in this direction were considered to have failed, with as a result hosts remaining in the original cell. Movement rules in displacement phase: Host populations in displacement phase were considered to move together from a cell. The extended neighbourhood in this phase represents the search window of the host. Displacing populations would start to search from the nearest eight neighbouring cells to the next 16 adjacent cells and then reach over the extended neighbourhood. The searching activity would be finished when a habitat cell was found in a different patch and the host population there was below the host carrying capacity of the cell. If no such habitat cell was found, the hosts remained in the original cell at the end of the time step. As we focused on deer management, small mammals were assumed to perform home ranging behaviours only. It has been reported that small mammals have relatively small home range areas (e.g., about 0.2 ha for the bank vole Myodes glareolus and the wood mice Apodemus sylvaticus (Kikkawa, 1964; Kozakiewicz et al., 2007). In the model, small mammals were assumed to inhabit both woodland and grassland and their movement capacity (MCS ) was assumed to be 100 m. Given a value of 100 m to trigger the movement between adjacent cells, it was therefore modelled that most small mammal movements could be completed within each cell, except for a small proportion, amongst those inhabiting areas close to cell boundaries. Moving hosts transported ticks. As the time step of this study was one week, nymphs and adults, assumed to take one week to feed, could be transported between cells. The transport of larvae, however, was considered negligible, as larvae were assumed to feed

2.1.4. Inputs, outputs and initialisation The model required the input of a land cover map and outputted the questing population of ticks in different life stages. The land cover map could be a real world map or an artificial map designed for theoretical investigations. In the present study, all simulations were initialised with a tick density of 50 000 adults per ha in woodland areas, according to a range of 0–1620 adults sampled per ha in Belgium (Li et al., 2012b) and by assuming a 5–9% sampling efficiency (Daniels et al., 2000). Hosts were assumed homogenously distributed in habitats at the initial time step. Then, as the model ran and movement rules were applied, their populations would become heterogeneous and varying between time steps. 2.2. Model evaluation 2.2.1. Comparison with field observations The model performance was examined by testing its ability to reflect relative differences in observed tick abundance in different sites in south Belgium as reported in Li et al. (2012b). Tick sampling data for comparison were selected (in total 125 samples in 51 sites) based on the following criteria: (i) in Wallonia (deer data); (ii) in rural areas (to avoid potential strong effects of human disturbance on local deer density); and (iii) sampled in the same period of the same year (to avoid potential strong influence of climatic differences on tick population). Using these criteria, only four samples could be used (Thuin, Marche-en-Famenne, Wellin and Marcheles-Dames). Land cover maps (5000 m × 5000 m, resolution 100 m) included “woodland” (including broadleaf and coniferous forests), “grassland” (including pasture and moorland) and “non-vegetated area” (including built-up areas, agricultural land and forest clearings). Deer densities in the woodland were calculated based on the spring standing population of roe, red and fallow deer provided by the “Direction Générale des Ressources Naturelles et de l’Environnement” of the Walloon Region. 2.2.2. Sensitivity analysis The sensitivity of the model to all parameters was assessed with the land cover map of Thuin (i.e., the site with the greatest number of sampled ticks) as used in Section 2.2.1. The sensitivity index was calculated as: S = Log10 (Di /D0 )/Log10 (Pi /P0 ), with D0 the density index at equilibrium when using the default values for all parameters (values in Table 1) and Di the density index when the value of a parameter (P) is increased from the default P0 to Pi by 5%. It has been demonstrated in a number of biological models (Keeling and Gilligan, 2000; Ogden et al., 2005) that S is a robust indicator of the sensitivity. A higher absolute value of S means a stronger effect of the parameter change on the tick abundance. S values of 1 and −1 indicate positive and negative linear effects.

6

S. Li et al. / Ecological Modelling 276 (2014) 1–13

Fig. 3. Configuration of the artificial landscape. The spatial extension of the landscape is 50 × 50 cells (cell size = 1 ha). Grey area indicates three woodland patches: (A) size = 47 ha, (B) size = 177 ha, and (C) size = 226. The minimum movement capacities required for deer to move between A and B is 800 m, between B and C is 500 m, and between A and C is 500 m. White areas indicate grassland.

2.3. Scenarios Four scenarios relevant to current and foreseen deer management practices in Europe were developed. Specifically, the scenarios address the potential effects of wildlife management practices on deer patterns. The model predicted their consequences on the spatial pattern of ticks using an artificial landscape (Fig. 3). It further allows assessing whether I. ricinus can be managed through controlling its major reproduction hosts. The design of the theoretical landscape was based on the research questions to be explored. As the main target of the present study was the spatial relationship between tick and deer populations, the structure of the theoretical landscape was kept simple to limit the effects of landscape configuration on the results. The woodland area was assumed to be fragmented, reflecting a common field situation. Grassland and woodland covers were included, as these are the major vegetated land cover types in many European countries. Such design allowed testing the effects of deer movement between-land cover types and between woodland patches. 2.3.1. Scenario 1 – reducing local deer density Reducing the deer density, via hunting, can sometimes benefit conservational plans for the protection of woodland resources and for a better recovering and quality of deer population. In the simulation, we intended to examine whether a failed deer density reduction in one woodland patch will influence the tick density over the landscape. Deer densities observed by the “Direction Générale des Ressources Naturelles et de l’Environnement” in Belgium were used to parameterise the model. We assumed the three woodland patches had a deer density (dD ) of 0.18 heads/ha (i.e., close to the highest deer density of 0.19 heads/ha in Büllingen in 2008) and the controlling target was 0.08 heads/ha (i.e., the averaged deer density in South Belgium in 2008). Then, four situations were considered: (i) the control was successful in all woodland patches, and a situation for which the controlling failed (i.e., deer density did not change) (ii) in woodland A, (iii) in woodland B or (iv) in woodland C. For each situation, three sets of simulations were compared with movement capacity of deer (MCD H ) = 800 m (i.e., deer can move between woodlands A, B and C), 500 m (i.e., deer can move between woodlands A and C and between woodlands B and C but not between woodlands A and B) and 300 m (i.e., deer cannot reach to other woodland patches). Our hypothesis was that, if deer can move between woodland patches to transport ticks, the failed reduction of deer in one woodland patch may increase the density in both local and nearby woodland patches, resulting in a general increase in tick density at landscape level. 2.3.2. Scenario 2 – controlling deer grazing intensity in grassland Grassland grazing by deer, especially by red and fallow deer, has been reported to cause significant damage on agricultural grassland production in Europe (Licoppe, 2006; Putman and Moore, 1998;

Trdan and Vidrih, 2008). Roe deer, the most abundant and selective browser of European ungulate species, is often underestimated in its impact on grassland, pasture and ecotone composites of trees and shrubs (Putman and Moore, 1998; Reimoser and Putman, 2011). The prevention of such damage can be reached via exclusive fencing. As complete removal of deer grazing in grassland by fencing is considered undesirable in many natural forests, there is an essential need to understand the impact of grazing intensities on woodland dynamics for potential deer management policies (Hester et al., 2000). In reality, managing the grazing intensity of wildlife is not easy, but via hunting the numbers can be regulated, allowing for density changes in specific areas (Hester et al., 2000). Moreover, a cline of grassland usage may also reflect the situations in which enclosures are incomplete. Thus, a proportion of deer are free-ranging deer that have not been enclosed, or the deer fenced inside the woodland can jump over the low fence and still graze in grassland. In the model, grazing intensity was represented in two ways: the grazing range and time spent in grassland. In the simulation, the impact of reduced deer grazing intensities on tick dynamics was tested by decreasing movement capacities of deer (MCD H = 400 m, 300 m, 200 m and 100 m) and the proportion of time spent in grassland (pG = 75%, 50%, 25% and 0% of a week). Thus, in all simulations, deer populations could not exchange between woodland patches. Our hypothesis was that controlled deer grazing intensity could lead deer to spend more time in woodland and thus increase the tick density in woodland. 2.3.3. Scenario 3 – translocation of deer species Translocation of deer species has been widely practised in Europe during the recent past for meeting the local hunting demands and rehabilitating habitats where local deer species were extinct. Roe deer have been reintroduced in some of their historic regions in France (Calenge et al., 2005), Italy (Dupanloup et al., 2002) and Portugal (Vernesi et al., 2002). The current distribution of a western red deer lineage in eastern Europe was suggested to be mainly artificial and a result of translocation (Niedziałkowska et al., 2010). In the model, we tested the effects of different time periods between the local extinction and reintroduction of deer species. Deer populations in all woodlands were removed after 10 years of simulation. Then reintroductions were considered after 1, 2, 3, 5 and 10 years, respectively. The aim was to explore to what extent the deer extirpation could reduce the local tick population and how long it would take to restore the local tick population after deer reintroduction. Our hypothesis was that the extinction of deer species would not eradicate ticks (as tick population could still be maintained by small mammals), but would greatly reduce the tick population (as only limited numbers of adult ticks could feed on small mammals). 2.3.4. Scenario 4 – controlling human disturbance and deer displacement between woodland patches Human disturbance (e.g. hunting, tourism, management practice etc.) may result in deer displacement and a net deer population exchange between woodland patches. In the displacement phase, red deer may migrate to other woodland patches, hide for several days and then return (Sunde et al., 2009). Roe deer, though preferring to hide in ground vegetation, may also perform displacement when being released after capture for fitting tracking devices (Morellet et al., 2009). Besides, immigration of other deer species, for example via natal dispersal (Prévot and Licoppe, 2013), may result in the competitive displacement of roe deer (Dolman and Wäber, 2008; Latham, 1999). In the model, controlling human disturbance was assumed to reduce deer displacement rates. We assumed up to 50% of deer population to be displaced (pDis) by disturbance (Sunde et al., 2009) and a lower movement capacity (MCD H = 300 m) for home-ranging deer to avoid disturbance

S. Li et al. / Ecological Modelling 276 (2014) 1–13

7

Table 2 Validation results. Site

Deer density (ha−1 )

No. of sampled nymphs

Estimated nymph density in field (ha−1 )a

Simulated nymph density (ha−1 )

Thuin Marche-en-Famenne Wellin Marche-les-Dames

0.13 0.14 0.11 0.07

125 108 88 42

2.78–5.00 E4 2.40–4.32 E4 1.96–3.52 E4 0.93–1.68 E4

2.95 3.00 2.41 1.52

a

± ± ± ±

0.01 E4 0.01 E4 0.01 E4 0.00 E4

Assuming 5–9% sampling efficiency (Daniels et al., 2000).

(Jayakody et al., 2011). Finally, we simulated a set of decreased deer displacement rates (pDis, from 50% to 0% of deer population in displacement phase, with a successive decrease of 5%) for testing their impact on tick population dynamics. Our hypothesis was that a reduced displacement rate of deer would reduce the tick density, as it could relatively increase the home-ranging population of deer (so that more ticks would be transported to grassland in which tick mortalities are higher). 3. Results Equilibrium levels (with a weekly change of questing tick densities <0.5%) were achieved in all simulations in five to eight simulated years after initialisation. Hence, to have a safe confirmation of the equilibrium, outcomes were recorded from the 10th to 30th simulated years (521st–1560th weeks) for model evaluation, sensitivity analysis and scenario analysis. Different initial values of tick densities in different life stages were tested as well and did not affect the results. 3.1. Model evaluation 3.1.1. Comparison with field observations The results in Table 2 show that, the simulated output largely reflects the differences of field sampled data in the four sites (comparing field sampled data and simulated data: Pearson’s chisquared = 6.05, p-value = 0.1). The unexplained difference (e.g., the sample from Thuin shows the highest numbers of ticks while in Marche-en-Famenne numbers are highest with the model) may be explained by different reasons. Microclimatic conditions, for example, may have been different on the sampling occasions and local small mammal densities can vary. The local deer compositions may also have influenced the results, as, based on the field data, Thuin and Marche-les-Dames were frequented by roe deer only, while Marche-en-Famenne and Wellin were frequented by both roe and red deer. 3.1.2. Sensitivity analysis In general, tick densities in different post-egg life stages were sensitive to different sets of parameters (Fig. 4). Larval tick densities were sensitive to average numbers of eggs per adult, the weekly mortality rate of questing larvae, the duration of and the weekly mortality rate in the developmental phase from engorged adult ticks to questing larvae, the density of deer, and the weekly maximum number of adult attachments on one deer. Nymphal tick densities were sensitive to many parameters. High sensitivities were observed for the mortality in developing phases from engorged larvae to questing nymphs and from engorged adults to questing larvae, the mortality rate of feeding larvae, the mortality rate of questing nymphs, the weekly maximum number of larvae and nymphs attachments on one small mammal, and the density of deer and small mammals. Adult tick densities show high sensitivities to developing phase mortality from engorged nymphs to questing adults. They were also sensitive to the weekly mortality rates in questing adult phase, the weekly maximum nymph attachment on small mammals and small mammal densities.

3.2. Scenario analysis 3.2.1. Scenario 1 – reduction of deer population may not be an effective method to control the tick population if deer can move between neighbouring woodland patches A successful deer population reduction from 0.18 to 0.08 heads/ha in all woodland patches contributed to a reduction of approximately 50.1% of the tick population at landscape level (Fig. 5a and e). When deer were able to move between woodland patches (movement capacity = 800 m and 500 m), failing to reduce deer density in one woodland patch would increase the questing tick density in all other woodland patches (Fig. 5b–d). There was no influence on tick densities in other woodlands, when deer could not move to neighbouring patches (movement capacity = 300 m). Overall questing tick densities were highest when the reduction of deer population failed in woodland C. Woodland C, being relatively larger, would hold a larger deer population for the same deer density. No major changes in tick abundance in grassland were observed. 3.2.2. Scenario 2 – controlled deer grazing intensity can reduce the tick abundance in grassland, but increase it in woodland As deer grazing intensity was controlled (or deer movement capacity and the proportion of time steps deer spent in grassland decreased) in the simulations, questing tick abundance decreased in grassland but increased in woodland (Fig. 6). Extensive deer browsing control in grassland (by decreasing the movement capacity from 400 to 100 m per week and the proportion of time steps spent in grassland from 75% to 0%) could reduce the questing tick abundance in grassland by 87.1% and increase it by 10.7% in woodland, resulting in a combined overall landscape tick increase effect of 10.5% (the simulated ratio of tick density in grassland to that in woodland was approximately 1:105 ). 3.2.3. Scenario 3 – translocation of deer: local extinction of deer could decrease tick abundance, but could not eradicate ticks Local deer extinction could decrease the questing tick density substantially, namely by 76% (failing to find hosts before dying) (Fig. 7). There were two equilibrium levels: (i) a high level of 3.5 × 105 questing ticks per ha with both deer and small mammal populations and (ii) a low level of 8.5 × 104 questing ticks per ha when only small mammals were present. In all situations, the local questing tick abundance started to decrease after 45 weeks of deer extirpation and to increase 45 weeks after deer reintroduction. This is associated with the model parameters on the duration of developmental phases (Table 1). When tick started to decrease after deer extirpation, it took about 140 weeks to reduce to a lower equilibrium level. When tick started to increase after deer reintroduction, it took about 210 weeks to recover the tick density to a higher equilibrium level. 3.2.4. Scenario 4 – decreased displacement of deer could decrease the tick density in woodland, and also avoid the formation of areas with extremely high tick density Decreased deer displacement decreased average questing tick densities in woodland patches, as well as the highest tick density

8

S. Li et al. / Ecological Modelling 276 (2014) 1–13

Fig. 4. Sensitivity of model variables. Sensitivity indices were calculated for woodland larval tick density (Sl), nymphal tick density (Sn) and adult tick density (Sa).

in woodland cells (Fig. 8). A control of displacement proportion of deer population from 50% to 0% reduced the overall questing tick density from 3.9 × 105 to 3.6 × 105 ticks per ha and reduced the highest questing tick density from 7.4 × 106 to 4.6 × 105 ticks per ha. Along with an increased displacement rate, fewer ticks were noticed in grassland (not shown).

4. Discussion This study aimed to explore possible consequences of wildlife management practices on tick spatial dynamics. A populationbased, multi-host and multi-land cover cellular automata model for the I. ricinus population ecology was adapted from a previous

S. Li et al. / Ecological Modelling 276 (2014) 1–13

9

Fig. 5. Simulated questing tick densities under different deer population reduction scenarios. MC = movement capacity in home ranging phase. Failed control means the deer population remained at 0.18 heads/ha. The control target was 0.08 heads/ha.

study (Li et al., 2012a) so that it allowed simulating how various deer movement patterns influence the spatial distribution and abundance of ticks in heterogeneous landscapes. A set of scenarios on current and foreseen deer management practices and the corresponding changes of deer movement patterns in Europe were established. Those scenarios were then tested for their potential impacts on the spatial dynamics of ticks through the model using an artificial landscape, comprised of three woodland patches surrounded by grassland. We showed how complex ecological models for disease vectors can be adapted for exploring consequences on tick densities of various management strategies, allowing policy makers to assess the potential effects of managing deer population on public health. The study firstly predicted that controlling the deer population would not always be effective to control ticks. In isolated areas, reducing the deer population can reduce the local tick abundance considerably even if the deer reduction attempts failed in other areas. However, when target areas were well-connected with other areas for deer population exchange, the reduction of local tick abundance would not be effective unless deer populations were reduced in all connected areas. In line with a number of field studies ˜ 2002, 2003), this prediction emphasises the effect (Estrada-Pena, of interactions between landscape connectivity and spatial heterogeneity of host populations. Moreover, the relative location of woodland patches in the network appeared to be influential on the spatial tick population patterns. In scenario 1, while deer could not directly move between woodland A and B (movement capacity = 500 m), the failed control of deer population in woodland A still

increased the tick population in woodland B via woodland C. This may suggest a direction for future empirical work towards examining the potential of geographical accessibility of woodland patches (the degree to which the woodland is available to as many deer as possible) as a local risk factor for ticks and tick-borne diseases. To help preventing tick-borne diseases, hunting, for example, may be planned as a means to control deer population in woodland of great geographical accessibility. A second prediction was that decreasing deer grazing in grassland adjacent to forest moderately reduces the tick abundance in grassland and increases the tick abundance in woodland. This is in accordance with previous modelling studies (Hoch et al., 2010; Jones et al., 2011), suggesting a higher tick density in enclosed forests (Gilbert et al., 2012). Ticks increased at the landscape level, as the exposure of ticks to the higher mortality rates in grassland decreased. Moreover, ticks feeding on rodents may be increased where deer presence is strictly controlled (e.g., via fencing), which may result in an increase in disease transmission (Perkins et al., 2006; Rosà and Pugliese, 2007). Hence, controlling deer browsing between connecting and contrasting land covers may not be an effective way to reduce tick abundance and the risk of exposure to pathogens. Besides, along with previous findings from scenario 1, the two game-harvesting strategies in Europe could be compared: hunting free ranging animals or animals in enclosed forest parks. Hunting in parks is prohibited in Belgium but is allowed in some countries such as France, Spain and Portugal. The access to pastures or other woodlands is impossible for park game animal and the tick abundance could be higher in parks than in free ranging

10

S. Li et al. / Ecological Modelling 276 (2014) 1–13

Fig. 8. Effects of deer displacement on the questing tick densities in woodland. Averaged density () and maximum density () were shown.

Fig. 6. Relative abundance of questing ticks in grassland (a) and in woodland (b) under different deer browsing controlling scenarios. Deer grazing scenarios were set up by using all possible combinations of deer movement capacity (400 m, 300 m, 200 m and 100 m) and proportion of time deer spent in grassland (75%, 50%, 25% and 0%). The relative tick abundance was calculated by dividing the tick abundance simulated for each scenario by the default tick abundance simulated (i.e., when both movement capacity and proportion of time spent in grassland were the highest).

areas. Promoting empirical tick investigations in forest parks may be necessary for better acknowledging potential contacts between ticks and humans. A third prediction was that the local extinction of deer species could lead to an approximately 76% control of tick populations in

woodland. This is slightly lower than the empirical 86–96% field control of ticks in Scottish forests and moorland by excluding deer locally, through fencing (Gilbert et al., 2012). The observed difference may be due to stochasticity but also factors such as a higher tick mortality rates (Macleod, 1932) or a relative lower small mammal abundance (Gilbert et al., 2000) found in moorland may have played a role here. In our simulations, the tick population was maintained by an unchanged number of smaller mammals across the landscape. However, there is evidence that the exclusion of deer species in woodland can result in an increase of small mammals while the introduction of deer species may not affect local small mammal communities (Smit et al., 2001). Moreover, the sensitivity analysis indicated that an increased small mammal population can increase the nymphal and adult tick abundance (Fig. 4). Reintroducing deer may raise tick abundance to a higher level than before. Hence, the reintroduction of deer species into tick-infested areas should be designed carefully, especially concerning the location of the release sites. A fourth prediction was that a decreased deer displacement between forest patches in response to a control of human disturbance could decrease the possibility of tick “hotspots” with particular high tick densities, as well as the overall tick densities. Controlling human disturbance of animals thus seemed to be effective in controlling ticks. In scenario 4, deer exchange between woodland patches resulted from deer displacement only, as homeranging deer could not reach other woodland patches. Increased deer displacement could reduce the proportion of home-ranging

Fig. 7. Evolutions of questing tick densities in woodland under different deer translocation scenarios. Deer were extirpated in the simulated week of 521 (10 years). 1 year, 2 years, 3 years, 5 years and 10 years indicate deer reintroductions after 1, 2, 3, 5 and 10 years of deer extirpation.

S. Li et al. / Ecological Modelling 276 (2014) 1–13

deer, leading to fewer ticks being transported from woodland to adjacent grassland where tick mortalities were high. Hence, the effect of deer displacement on tick abundance was positive at landscape-level. The positive local-level effect on the maximum tick density was due to the simulated clustered distribution of deer in woodland, which is in accordance with field observations under high human disturbance (Jayakody et al., 2008). In the model, all deer followed the same rules to select destination woodland cells for displacement. Some cells were selected constantly by moving deer as they are closer (and thus easier) to travel to. This resulted in deer population clusters and therefore more drop-offs of fed ticks in and around these cells. As a result, “hotspots” of ticks were formed. In general, deer displacement seemed to not play an important role at landscape level, which is in line with the findings in the sensitivity analysis. However, its effects on the local tick distribution could be vast and deems further exploration. Such findings could hardly be gained from traditional non-spatial models, highlighting the utility of spatial ecological models in the present study. Finally, there is a public health reason to create “quiet zones” (i.e., area excluding human disturbance) in a relatively isolated location to reduce local deer movement. Such management practice can lower the contact probability between animals and humans. In any case, the spatial heterogeneity found in and between results underlines the need to know those places effectively used by humans, where the latter will be exposed to tick bites. The model could be extended in different ways, allowing a better understanding e.g., of the impact of local biodiversity. Firstly, the host layer could be adapted in a species-specific fashion. In the present study, a generalised layer mixing several common deer species was used to cope with current data limitations with respect to the composition of local deer populations. However, roe, red and fallow deer are different in size and in their capacities for hosting ticks (Matuschka et al., 1993; Vor et al., 2010), and can establish different movement patterns across multiple spatio-temporal scales. In Wallonia, the authorities aim to reduce the red deer and wild boar densities (see legal rule M.B.08.05.1993 amended in 2008 by Decree of the Walloon Government), which could have an indirect positive effect on roe deer populations. It may be interesting to study how such modification of deer composition could influence tick spatial dynamics. Secondly, the risk of tick-borne diseases, such as Lyme disease and tick-borne encephalitis, could be modelled by including a disease transmission function allowing exploring the tick infection prevalence. I. ricinus ticks can feed on a wide range of host species. However, some of these hosts are incompetent for pathogen transmission, such as deer, whilst most small mammals, especially rodents and squirrels, are competent vectors. Therefore, deer are commonly hypothesised to have a “dilution effect”, thus an increase in deer population can decrease the probability of tick feeding on small mammals and hereby decrease the tick infection prevalence. However, such a hypothesis has been reported to be more complex than explained and scale-dependent (Cagnacci et al., 2012). Future spatial modelling studies may allow exploring whether or not the “dilution” effects depend on landscape configurations. Improved understanding of the spatio-temporal dynamics of tick feeding patterns on hosts is also desirable. Still, such developments of the model rely heavily on interdisciplinary investigations, for example between ecology, geography and computer sciences, on the local composition and movement patterns of wildlife species. Thirdly, future modelling attempts may focus on finer spatial and temporal scales. A more detailed tick biology model could be daily-based, so that the feeding durations and the tick transportation during feeding can be represented in a better way. On a finer spatial scale, more landscape effects can be explored, for example the presence and characteristics of ecotones (i.e., woodland-grassland interfaces) that are often favourable to tick survival and development. Finally, it can be promising to

11

include the phenology of tick. The seasonal population dynamics of ticks and their host animals have been studied extensively (Killilea et al., 2008; Tagliapietra et al., 2011). A number of tick behaviours have been reported to be highly sensitive to season, for example diapause, maturation, and host-finding activity. The used modelling approach has some shortcomings resulting from the simplifying assumptions made, for example about the tick biology and climatic influence. This has also been reflected in the sensitivity analysis: the most sensitive variables concerned tick mortality in different phases (i.e., questing, feeding and developing between life stages) and feeding tick numbers on hosts. Similar model sensitivities have been found for existing models, e.g., the I. scapularis model of Ogden et al. (2005) and I. ricinus model of Hoch et al. (2010). Both models were highly sensitive to the feeding mortality of larvae and feeding numbers of immature ticks. For a better understanding of the spatial tick dynamics, a better representation of tick biology in natural conditions and an improved function for the development of tick life stages may be needed. To our knowledge, there is no clear evidence to claim that density dependence acts on I. ricinus numbers. The use of this factor in models can perfectly balance the populations but this can be equally well done through climate-related mortalities only. Andrewartha and Birch (1954) dismiss the existence of density dependence in the population dynamics of I. ricinus. In our models we used density dependence, but future modelling may need to include climate to investigate whether or not climate is sufficient to explain the field observations. Further empirical investigations could help to increase the scientific knowledge about the aforementioned aspects. However, a model is always a purposeful representation of reality. Including additional unknowns may increase the model complexity by creating too many degrees of freedom and thus can be harmful. Hence, there is a need in the design of ecological models to maximise the specificity while minimising the complexity. 5. Conclusions In summary, we used a spatial modelling approach to integrate existing interdisciplinary knowledge for predicting consequences of different wildlife management strategies. Prior information and expert knowledge on deer movement patterns and their possible responses to the management strategies were combined to improve the understanding of the spatial dynamics of deer, the tick natural hosts. Then, by using a cellular automata model, the effects of several potential wildlife management practices were evaluated such as the control of the deer population, grazing, translocation and displacement on the spatial dynamics of tick at woodland patch- and landscape-levels. The modelling approach presented here can be used for further experiments about the adaptive management of wildlife and tick-borne diseases. Acknowledgements Sen Li is a research fellow (Aspirant FNRS) at the National Fund for Scientific Research, Belgium. The authors wish to thank the anonymous reviewers for their valuable comments that have enhanced the work. References Andrewartha, H.G., Birch, L.C., 1954. The Distribution and Abundance of Animals. University of Chicago Press. Bartumeus, F., Da Luz, M.G.E., Viswanathan, G.M., Catalan, J., 2005. Animal search strategies: a quantitative random-walk analysis. Ecology 86, 3078–3087. Cagnacci, F., Bolzoni, L., Rosà, R., Carpi, G., Hauffe, H.C., Valent, M., Tagliapietra, V., Kazimirova, M., Koci, J., Stanko, M., Lukan, M., Henttonen, H., Rizzoli, A., 2012. Effects of deer density on tick infestation of rodents and the hazard of tick-borne encephalitis, I: empirical assessment. Int. J. Parasitol. 42, 365–372.

12

S. Li et al. / Ecological Modelling 276 (2014) 1–13

Calenge, C., Maillard, D., Invernia, N., Gaudin, J.C., 2005. Reintroduction of roe deer Capreolus capreolus into a Mediterranean habitat: female mortality and dispersion. Wildl. Biol. 11, 153–161. Daniel, M., Cerny, V., Dusbabek, F., Honzakova, E., Olejnicek, J., 1976. Influence of microclimate on the life cycle of the common tick Ixodes ricinus (L.) in thermophilic oak forest. Folia Parasitol. 23, 327–342. Daniels, T.J., Falco, R.C., Fish, D., 2000. Estimating population size and drag sampling efficiency for the blacklegged tick (Acari: Ixodidae). J. Med. Entomol. 37, 357–363. Delbeuck, C., 2008. Cellule etat de l’environnement Wallon: Environmental Outlook for Wallonia 2008. Service Public de Wallonie. Dolman, P.M., Wäber, K., 2008. Ecosystem and competition impacts of introduced deer. Wildl. Res. 35, 202–214. Dupanloup, I., Schneider, S., Excoffier, L., 2002. A simulated annealing approach to define the genetic structure of populations. Mol. Ecol. 11, 2571–2581. Escutenaire, S., Chalon, P., Verhagen, R., Heyman, P., Thomas, I., Karelle-Bui, L., AvsicZupanc, T., Lundkvist, A., Plyusnin, A., Pastoret, P.P., 2000. Spatial and temporal dynamics of Puumala hantavirus infection in red bank vole (Clethrionomys glareolus) populations in Belgium. Virus Res., 91–107. ˜ A., 2002. Understanding the relationships between landscape connecEstrada-Pena, tivity and abundance of Ixodes ricinus Ticks. Exp. Appl. Acarol. 28, 239–248. ˜ A., 2003. The relationships between habitat topology, critical scales Estrada-Pena, of connectivity and tick abundance Ixodes ricinus in a heterogeneous landscape in northern Spain. Ecography 26, 661–671. ˜ A., Osacar, J.J., Pichon, B., Gray, J.S., 2005. Hosts and pathogen detection Estrada-Pena, for immature stages of Ixodes ricinus (Acari: Ixodidae) in North-Central Spain. Exp. Appl. Acarol. 37, 257–268. Fryxell, J.M., Hazell, M., Borger, L., Dalziel, B.D., Haydon, D.T., Morales, J.M., McIntosh, T., Rosatte, R.C., 2008. Multiple movement modes by large herbivores at multiple spatiotemporal scales. Proc. Natl. Acad. Sci. U.S.A. 105, 19114–19119. Georgii, B., Schroder, W., 1983. Home range and activity patterns of male red deer (Cervus elaphus L.) in the Alps. Oecologia 58, 238–248. Gilbert, L., Jones, L.D., Hudson, P.J., Gould, E.A., Reid, H.W., 2000. Role of small mammals in the persistence of Louping-ill virus: field survey and tick co-feeding studies. Med. Vet. Entomol. 14, 277–282. Gilbert, L., Maffey, G.L., Ramsay, S.L., Hester, A.J., 2012. The effect of deer management on the abundance of Ixodes ricinus in Scotland. Ecol. Appl. 22, 658–667. Gray, J.S., 1981. The fecundity of Ixodes ricinus (L.) (Acarina, Ixodidae) and the mortality of its developmental stages under field conditions. Bull. Entomol. Res. 71, 533–542. Gray, J.S., 2002. Biology of Ixodes species ticks in relation to tick-borne zoonoses. Wien. Klin. Wochenschr. 114, 473–478. Gray, J.S., Kahl, O., Janetzkimittman, C., Stein, J., Guy, E., 1994. Acquisition of Borrelia burgdorferi by Ixodes ricinus ticks fed on the European hedgehog, Erinaceus europaeus L. Exp. Appl. Acarol. 18, 485–491. Grimm, V., Revilla, E., Berger, U., Jeltsch, F., Mooij, W.M., Railsback, S.F., Thulke, H.H., Weiner, J., Wiegand, T., DeAngelis, D.L., 2005. Pattern-oriented modeling of agent-based complex systems: lessons from ecology. Science 310, 987–991. Hancock, P.A., Brackley, R., Palmer, S.C., 2011. Modelling the effect of temperature variation on the seasonal dynamics of Ixodes ricinus tick populations. Int. J. Parasitol. 41, 513–522. Hester, A.J., Edenius, L., Buttenschøn, R.M., Kuiters, A.T., 2000. Interactions between forests and herbivores: the role of controlled grazing experiments. Forestry 73, 381–391. Hoch, T., Monnet, Y., Agoulon, A., 2010. Influence of host migration between woodland and pasture on the population dynamics of the tick Ixodes ricinus: a modelling approach. Ecol. Model. 221, 1798–1806. Jaenson, T.G.T., Talleklint, L., 1996. Lyme borreliosis spirochetes in Ixodes ricinus (Acari: Ixodidae) and the varying hare on isolated islands in the Baltic Sea. J. Med. Entomol. 33, 339–343. Jayakody, S., Sibbald, A.M., Gordon, I.J., Lambin, X., 2008. Red deer Cervus elephus vigilance behaviour differs with habitat and type of human disturbance. Wildl. Biol. 14, 81–91. Jayakody, S., Sibbald, A.M., Mayes, R.W., Hooper, R.J., Gordon, I.J., Lambin, X., 2011. Effects of human disturbance on the diet composition of wild red deer (Cervus elaphus). Eur. J. Wildl. Res. 57, 939–948. Jones, E., Webb, S., Ruiz-Fons, F., Albon, S., Gilbert, L., 2011. The effect of landscape heterogeneity and host movement on a tick-borne pathogen. Theor. Ecol. 4, 435–448. Keeling, M.J., Gilligan, C.A., 2000. Bubonic plague: a metapopulation model of a zoonosis. Proc. R. Soc. Lond. B: Biol. 267, 2219–2230. Kiffner, C., Lödige, C., Alings, M., Vor, T., Rühe, F., 2010. Abundance estimation of Ixodes ticks (Acari: Ixodidae) on roe deer (Capreolus capreolus). Exp. Appl. Acarol. 52, 73–84. Kikkawa, J., 1964. Movement, activity and distribution of the small rodents Clethrionomys glareolus and Apodemus sylvaticus in woodland. J. Anim. Ecol. 33, 259–299. Killilea, M., Swei, A., Lane, R., Briggs, C., Ostfeld, R., 2008. Spatial dynamics of Lyme disease: a review. Ecohealth 5, 167–195. Kozakiewicz, M., Choluj, A., Kozakiewicz, A., 2007. Long-distance movements of individuals in a free-living bank vole population: an important element of male breeding strategy. Acta Theriol. 52, 339–348. Latham, J., 1999. Interspecific interactions of ungulates in European forests: an overview. For. Ecol. Manag. 120, 13–21. Li, S., Hartemink, N., Speybroeck, N., Vanwambeke, S.O., 2012a. Consequences of landscape fragmentation on Lyme disease risk: a cellular automata approach. PLoS ONE 7, e39612.

Li, S., Heyman, P., Cochez, C., Simons, L., Vanwambeke, S., 2012b. A multi-level analysis of the relationship between environmental factors and questing Ixodes ricinus dynamics in Belgium. Parasit. Vectors 5, 149. Licoppe, A., 2006. The diurnal habitat used by red deer (Cervus elaphus L.) in the Haute Ardenne. Eur. J. Wildl. Res. 52, 164–170. Luckhart, S., Lindsay, S.W., James, A.A., Scott, T.W., 2010. Reframing critical needs in vector biology and management of vector-borne disease. PLoS Negl. Trop. Dis. 4, e566. Macleod, J., 1932. The bionomics of Ixodes ricinus L. the sheep tick of Scotland. Parasitology 24, 382–400. Matuschka, F.R., Heiler, M., Eiffert, H., Fischer, P., Lotter, H., Spielman, A., 1993. Diversionary role of hoofed game in the transmission of Lyme disease spirochetes. Am. J. Trop. Med. Hyg. 48, 693–699. Morales, J.M., Haydon, D.T., Frair, J., Holsiner, K.E., Fryxell, J.M., 2004. Extracting more out of relocation data: building movement models as mixtures of random walks. Ecology 85, 2436–2445. Morellet, N., Verheyden, H., Angibault, J.-M., Cargnelutti, B., Lourtet, B., Hewison, M.A.J., 2009. The effect of capture on ranging behaviour and activity of the European roe deer Capreolus capreolus. Wildl. Biol. 15, 278–287. Morens, D.M., Folkers, G.K., Fauci, A.S., 2004. The challenge of emerging and reemerging infectious diseases. Nature 430, 242–249. Mount, G.A., Haile, D.G., Daniels, E., 1997. Simulation of blacklegged tick (Acari: Ixodidae) population dynamics and transmission of Borrelia burgdorferi. J. Med. Entomol., 461–484. Mysterud, A., 1999. Seasonal migration pattern and home range of roe deer (Capreolus capreolus) in an altitudinal gradient in southern Norway. J. Zool. 247, 479–486. Nathan, R., Getz, W.M., Revilla, E., Holyoak, M., Kadmon, R., Saltz, D., Smouse, P.E., 2008. A movement ecology paradigm for unifying organismal movement research. Proc. Natl. Acad. Sci. U.S.A. 105, 19052–19059. Niedziałkowska, M., Jedrzejewska, B., Honnen, A.C., Otto, T., Sidorovich, V.E., Perzanowski, K., Skog, A., Hartl, G.B., Borowik, T., Bunevich, A.N., Lang, J., Zachos, F.E., 2010. Molecular biogeography of red deer Cervus elaphus from eastern Europe: insights from mitochondrial DNA sequences. Acta Theriol., 1–12. North, M.J., Collier, N.T., Vos, J.R., 2006. Experiences creating three implementations of the repast agent modeling toolkit. ACM Trans. Model. Comput. Simul. 16, 1–25. OFFH, L’Observatoire de la faune, de la flore et des habitats en Wallonie Ministère de la Région Wallonne, DGRNE. Ogden, N.H., Bigras-Poulin, M., O’Callaghan, C.J., Barker, I.K., Lindsay, L.R., Maarouf, A., Smoyer-Tomic, K.E., Waltner-Toews, D., Charron, D., 2005. A dynamic population model to investigate effects of climate on geographic range and seasonality of the tick Ixodes scapularis. Int. J. Parasitol., 375–389. Oliver Jr., J.H., 1989. Biology and systematics of ticks (Acari: Ixodida). Annu. Rev. Ecol. Syst. 20, 397–430. Perkins, S.E., Cattadori, I.M., Tagliapietra, V., Rizzoli, A.P., Hudson, P.J., 2006. Localized deer absence leads to tick amplification. Ecology 87, 1981–1986. Prévot, C., Licoppe, A., 2013. Comparing red deer (Cervus elaphus L.) and wild boar (Sus scrofa L.) dispersal patterns in southern Belgium. Eur. J. Wildl. Res., 1–9. Putman, R.J., 1986. Foraging by roe deer in agricultural areas and impact on arable crops. J. Appl. Ecol 23, 91–99. Putman, R.J., 1988. The Natural History of Deer. Christopher Helm, London. Putman, R.J., Moore, N.P., 1998. Impact of deer in lowland Britain on agriculture, forestry and conservation habitats. Mamm. Rev. 28, 141–164. Randolph, S.E., 2008. Tick-borne encephalitis incidence in Central and Eastern Europe: consequences of political transition. Microbes Infect. 10, 209–216. Randolph, S.E., Craine, N.G., 1995. General framework for comparative quantitative studies on transmission of tick-borne diseases using Lyme borreliosis in Europe as an example. J. Med. Entomol. 32, 765–777. Reimoser, F., Putman, R., 2011. Impacts of wild ungulates on vegetation: costs and benefits. In: Putman, R., Apollonio, M., Andersen, R. (Eds.), Ungulate Management in Europe: Problems and Practices. Cambridge University Press, Cambridge, UK/New York, pp. 144–191. Rizzoli, A., Hauffe, H.C., Tagliapietra, V., Neteler, M., Rosà, R., 2009. Forest structure and roe deer abundance predict tick-borne encephalitis risk in Italy. PLoS ONE 4, e4336. Rosà, R., Pugliese, A., 2007. Effects of tick population dynamics and host densities on the persistence of tick-borne infections. Math. Biosci. 208, 216–240. Ruiz-Fons, F., Gilbert, L., 2010. The role of deer as vehicles to move ticks, Ixodes ricinus, between contrasting habitats. Int. J. Parasitol. 40, 1013–1020. Skalski, G.T., Gilliam, J.F., 2003. A diffusion-based theory of organism dispersal in heterogeneous populations. Am. Nat. 161, 441–458. Smit, R., Bokdam, J., den Ouden, J., Olff, H., Schot-Opschoor, H., Schrijvers, M., 2001. Effects of introduction and exclusion of large herbivores on small rodent communities. Plant Ecol. 155, 119–127. Sunde, P., Olesen, C.R., Madsen, T.L., Haugaard, L., 2009. Behavioural responses of GPS-collared female red deer Cervus elaphus to driven hunts. Wildl. Biol. 15, 454–460. Tagliapietra, V., Rosà, R., Arnoldi, D., Cagnacci, F., Capelli, G., Montarsi, F., Hauffe, H.C., Rizzoli, A., 2011. Saturation deficit and deer density affect questing activity and local abundance of Ixodes ricinus (Acari, Ixodidae) in Italy. Vet. Parasitol. 183, 114–124. Talleklint, L., Jaenson, T.G.T., 1997. Infestation of mammals by Ixodes ricinus ticks (Acari: Ixodidae) in south-central Sweden. Exp. Appl. Acarol. 21, 755–771. Trdan, S., Vidrih, M., 2008. Quantifying the damage of red deer (Cervus elaphus) grazing on grassland production in southeastern Slovenia. Eur. J. Wildl. Res. 54, 138–141.

S. Li et al. / Ecological Modelling 276 (2014) 1–13 Vernesi, C., Pecchioli, E., Caramelli, D., Tiedemann, R., Randi, E., Bertorelle, G., 2002. The genetic structure of natural and reintroduced roe deer (Capreolus capreolus) populations in the Alps and central Italy, with reference to the mitochondrial DNA phylogeography of Europe. Mol. Ecol. 11, 1285–1297. Vor, T., Kiffner, C., Hagedorn, P., Niedrig, M., Ruhe, F., 2010. Tick burden on European roe deer (Capreolus capreolus). Exp. Appl. Acarol. 51, 405–417. Wahlstrom, L.K., Liberg, O., 1995. Patterns of dispersal and seasonal migration in roe deer (Capreolus capreolus). J. Zool. 235, 455–467.

13

Wang, H.-H., Grant, W.E., Teel, P.D., 2012. Simulation of climate–host–parasite–landscape interactions: a spatially explicit model for ticks (Acari: Ixodidae). Ecol. Model. 243, 42–62. Watts, E.J., Palmer, S.C.F., Bowman, A.S., Irvine, R.J., Smith, A., Travis, J.M.J., 2009. The effect of host movement on viral transmission dynamics in a vector-borne disease system. Parasitology 136, 1221–1234. Zhang, X.X., Johnson, S.N., Crawford, J.W., Gregory, P.J., Young, I.M., 2007. A general random walk model for the leptokurtic distribution of organism movement: theory and application. Ecol. Model. 200, 79–88.