draft...draft 3 48 resume 49 les individus font face à un compromis entre acquérir des ressources...

Draft

Black-tailed deer (Odocoileus hemionus sitkensis) adjust

habitat selection and activity rhythm to the absence of predators

Journal: Canadian Journal of Zoology

Manuscript ID cjz-2015-0227.R1

Manuscript Type: Article

Date Submitted by the Author: 04-Feb-2016

Complete List of Authors: Bonnot, Nadège; INRA, CEFS - Behaviour and Ecology of Wild Mammals ; CNRS, CEFE Morellet, Nicolas; INRA, CEFS Hewison, Aidan J Mark; INRA, CEFS Martin, Jean-Louis; CEFE CNRS - Montpellier, Benhamou, Simon; CEFE CNRS - Montpellier Chamaillé-Jammes, Simon; CEFE-CNRS UMR 5175,

Keyword: anti-predator behavior, crepuscular activity peaks, day-night contrast, GPS, <i>Odocoileus hemionus sitkensis</i>, relaxed selection, Sitka black-tailed deer

https://mc06.manuscriptcentral.com/cjz-pubs

Canadian Journal of Zoology

Draft

1

Title: 1

Black-tailed deer (Odocoileus hemionus sitkensis) adjust habitat selection and activity rhythm to 2

the absence of predators 3

4

Authors: 5

Bonnot Nadège1,2, Morellet Nicolas1, Hewison A. J. Mark1, Martin Jean-Louis2, Benhamou Simon2 & 6

Chamaillé-Jammes Simon2 7

8

Affiliations: 9

1 INRA, UR35 Comportement et Ecologie de la Faune Sauvage, CS 52627, 31326 Castanet-Tolosan, France 10

2 Centre d’Écologie Fonctionnelle et Évolutive UMR 5175, CNRS – Université de Montpellier – Université Paul 11

Valéry Montpellier – EPHE, 1919 route de Mende, 34293 Montpellier Cedex 5, France 12

13

E-mail addresses: 14

1. Bonnot Nadège (corresponding author): [email protected] 15

Telephone number : +33 (0)5 61 28 51 25 ; Fax number : +33 (0)5 61 28 55 00 16

2. Morellet Nicolas: [email protected] 17

3. Hewison A.J. Mark: [email protected] 18

4. Martin Jean-Louis: [email protected] 19

5. Benhamou Simon: [email protected] 20

6. Chamaillé-Jammes Simon: [email protected] 21

22

23

Page 1 of 33



Draft

2

Black-tailed deer (Odocoileus hemionus sitkensis) adjust habitat selection and activity rhythm to 24

the absence of predators 25

by Bonnot N, Morellet N, Hewison AJM, Martin J-L, Benhamou S & Chamaillé-Jammes S 26

27

ABSTRACT 28

Although individuals must generally trade-off acquisition of high-quality resources against 29

predation risk avoidance, removal of top predators by humans has resulted in many large herbivores 30

experiencing novel conditions where their natural predators are absent. Anti-predator behaviors should 31

be attenuated or lost in such a context of relaxed predation pressure. To test this prediction, we 32

analyzed daily and seasonal habitat selection and activity rhythm (both commonly linked to predation 33

risk) of GPS-collared Sitka black-tailed deer (Odocoileus hemionus sitkensis Merriam, 1898) on 34

predator-free islands (British Columbia, Canada). In marked contrast to the behavioral patterns 35

commonly observed in populations subject to predation risk, we documented a very low day-night 36

contrast in habitat selection. Moreover, we observed higher activity during daytime than nighttime, as 37

expected for non-hunted populations. We also showed that resource selection was primarily driven by 38

seasonal variations in resource availability. These results are consistent with the expected attenuation of 39

anti-predator behaviors in predation-free environments. However, we also observed marked 40

crepuscular activity peaks which are commonly interpreted as an anti-predator response in ungulates. 41

Our study indicates that large herbivores are able to adjust certain anti-predator behaviors under relaxed 42

selection, notably habitat selection and activity rhythm, while others persist despite the long-term 43

absence of predators. 44

45

Key words: anti-predator behavior; crepuscular activity peaks; day-night contrast; GPS; Odocoileus 46

hemionus sitkensis; relaxed selection; Sitka black-tailed deer 47

Page 2 of 33



Draft

3

RESUME 48

Les individus font face à un compromis entre acquérir des ressources de qualité et éviter leurs 49

prédateurs. De nombreuses populations de grands herbivores évoluent dans des environnements où 50

leurs prédateurs naturels sont maintenant absents. Dans ces conditions, les comportements anti-51

prédateurs des individus devraient être amoindris, voire supprimés. Nous avons testé cette hypothèse en 52

analysant les patrons journaliers et saisonniers de sélection des habitats et d’activité de cerfs à queue-53

noire (Odocoileus hemionus sitkensis) vivant sur des îles exemptes de tout risque de prédation en 54

Colombie Britannique (Canada). A la différence de ce qui est communément observé dans des 55

populations soumises au risque de prédation, le contraste jour-nuit dans la sélection des habitats est peu 56

marqué. L'activité est plus importante de jour que de nuit, comme attendu pour des populations non-57

chassées. De plus, la sélection des ressources sont principalement influencés par les variations 58

saisonnières et spatiales de leur disponibilité. Ces résultats sont cohérents avec une atténuation des 59

comportements anti-prédateurs attendue dans des environnements exempts de risque de prédation. 60

Cependant, des pics d’activité crépusculaires marqués persistent, qui sont pourtant communément 61

interprétés comme des réponses anti-prédatrices. Notre étude montre donc qu’en l’absence de risque de 62

prédation, certains comportements anti-prédateurs peuvent être ajustés, d'autres maintenus. 63

64

Mots-clés : comportement anti-prédateur ; pics d’activité crépusculaire ; contraste jour/nuit ; GPS ; 65

sélection relâchée ; Odocoileus hemionus sitkensis ; cerf Sitka à queue-noire 66

Page 3 of 33



Draft

4

INTRODUCTION 67

In a rapidly changing world, individuals must cope with novel conditions which potentially modify 68

the selection pressures acting on them (Sih et al. 2011). Predation pressure represents a major selective 69

force shaping the phenotypes of prey (Lima and Dill 1990; Caro 2005). Currently, individuals of 70

certain prey populations must adjust their behavioral and physiological responses to rapid changes in 71

the landscape of risk due to the return of their predators. For example, the reintroduction of wolves 72

(Canis lupus L., 1758) to the Greater Yellowstone Ecosystem was followed by changes in habitat 73

selection, vigilance level, feeding behavior, nutritional condition and mean progesterone concentrations 74

of elk, (Cervus elaphus L., 1758), leading to a decline in calf production and, thus, directly and 75

indirectly impacting population dynamics (Creel et al. 2007; Christianson and Creel 2010). However, 76

while there is currently a strong focus on how individuals respond to novel selection pressures, less is 77

known about their behavioral responses when a strong selection pressure is removed, i.e. relaxed 78

selection (Coss 1999; Lahti et al. 2009). Specifically, although the majority of 79

wild ungulate populations are behaviorally constrained by the avoidance of predation risk/disturbance, 80

there is a need to better understand how they adjust their behavior to the absence of predation risk 81

(Blumstein 2002, Ripple et al. 2014). Our aim here is to test the general hypothesis that, in predator-82

free environments, costly anti-predator behaviors should be attenuated or lost (e.g. Coss et al. 1999; 83

Blumstein and Daniel 2005), so that the behavior of prey should be solely driven by spatio-temporal 84

variation in resource availability and energetic needs. 85

Under relaxed predation pressure, most anti-predator responses should be attenuated, or lost, 86

because of their costs and associated negative impact on individual fitness, e.g. reduced foraging 87

efficacy, growth, reproduction or survival (Lima 1998; Raymond et al. 2001; Brown and Kotler 2004; 88

Biro et al. 2006). How fast a particular anti-predator response will attenuate or be lost under relaxed 89

predation pressure is difficult to predict and will depend on the costs associated with the expression of 90

Page 4 of 33



Draft

5

the trait, mutation accumulation rates and pleiotropic effects (Lahti et al. 2009). However, some studies 91

suggest that behavioral traits may persist longer than morphological or physiological traits (Coss 1999; 92

Lahti 2009). 93

Because access to high quality resources is often linked to higher predation risk (Benhaiem et al. 94

2008), changes in both the level/type of predation risk or in resource availability often result in 95

behavioral adjustments among prey (Sih 1980; Fraser and Huntingford 1986; Lima 1998). One 96

common way for prey to trade off acquisition of high quality resources with the avoidance of predation 97

risk or disturbance is through alteration of habitat selection and activity rhythms (Lima and Dill 1990; 98

Kronfeld-Schor and Dayan 2003; Caro 2005). Although activity rhythms can be constrained by 99

endogenous traits, many studies suggest that patterns of habitat selection and activity are highly 100

flexible behaviors that can be adjusted to variations in resource availability and/or predation risk. 101

Indeed, prey often minimize their exposure to predation risk by resting in safer habitats during riskier 102

periods and by concentrating their foraging activities in risky forage-rich habitats when risk is lower 103

(Lima and Bednekoff 1999; Fortin et al. 2015). For example, many hunted ungulate populations are 104

more active and prefer forage-rich habitats during nighttime, i.e. when risk and disturbance linked to 105

human activity are lower (e.g. Godvik et al. 2009; Lone et al. 2014; Bonnot et al. 2015). Moreover, the 106

circadian activity peaks at sunrise and sunset of many wild ungulates are often considered to be an anti-107

predator response to temporal variation in predation risk (Caro 2005; Kamler et al. 2007; Loe et al. 108

2007; Long et al. 2013). Although the underlying mechanism is unclear, crepuscular activity peaks 109

were found to be either absent or not marked in predation-free deer populations (e.g. Loe et al. 2007; 110

Long et al. 2013), whereas marked activity peaks occurred around sunrise and sunset in deer faced with 111

a significant predation risk (e.g. Godvik et al. 2009; Ensing et al. 2014; Pagon et al. 2013). 112

To test our general hypothesis that anti-predator behaviors should be attenuated or lost under 113

relaxed selection, we studied the activity rhythms and habitat selection of wild Sitka black-tailed deer 114

Page 5 of 33



Draft

6

(Odocoileus hemionus sitkensis) living on two predator-free islands of British Columbia (Canada). 115

Although previous studies on the same study site suggest that anti-predator vigilance has persisted over 116

time despite the absence of predators (Chamaillé-Jammes et al. 2014; Le Saout et al. 2015), some anti-117

predator traits may persist longer than others (Blumstein 2002; Lahti et al. 2009). Assuming that the 118

typical day-night contrast in activity and habitat selection patterns of large herbivores is induced by 119

variation in the level of predation risk, we predicted that deer in this predator-free environment (1) 120

would be equally active during daytime and during nighttime with no marked crepuscular activity 121

peaks, and (2) that selection among the main habitat types would not differ markedly between day and 122

night. At a finer spatial scale (i.e. in the home-range), we predicted (3) that the selection of feeding 123

resources within forest habitat (the most widely used habitat) would be driven by seasonal variation in 124

resource availability rather than predation risk avoidance (i.e. we assumed that this lack of predation 125

risk negated the risk-resource trade-off so that spatial behaviour would be driven solely by feeding 126

constraints). Because environmental conditions, risk level, browsing pressure and resource availability 127

are quite similar on both islands (Pojar et al. 1980; Le Saout et al. 2014a), we considered them as two 128

replicate sites so that we expected no marked difference in behavioral patterns of the deer between 129

islands. 130

131

MATERIALS AND METHODS 132

1. Study area 133

Our study area consists of two islands on the eastern coast of Haida Gwaii archipelago (Fig. 1) 134

which is made up of more than 350 islands around 80km off the northern coast of British Columbia 135

(Canada): East Limestone (hereafter “ELI”; WGS84 coordinates: 52.91N 131.61W, 41ha) and Kunga 136

(WGS84 coordinates: 52.77N 131.57W, 395ha). The climate is temperate oceanic, with a mean annual 137

Page 6 of 33



Draft

7

temperature of 8.6°C and frequent precipitation, averaging 1400mm per year (data from the weather 138

station at Sandspit airport located 40km further north). 139

At the end of the nineteenth century, Sitka black-tailed deer were introduced on the largest islands 140

of the Haida Gwaii archipelago. The vast majority of the islands are uninhabited and the archipelago is 141

devoid of the main natural predators of deer (i.e. wolf and cougar (Puma concolor L., 1771)) so that it 142

was rapidly colonized, with deer arriving on Kunga and ELI more than 60 years ago (Vila et al. 2004). 143

Black bears are present on the largest islands of the archipelago, but are not established on the study 144

islands. They are not considered to threaten adult deer, but may be an opportunistic predator of fawns. 145

Hence, we assumed that the deer we monitored would not express anti-predator behaviors in relation to 146

predation risk by bear. The association of the mild climate and the absence of significant predation risk 147

resulted in a rapid increase in deer densities across the archipelago (Sharpe 1999). Current densities are 148

now markedly higher than those found in the natural range of Sitka black-tailed deer, estimated at 43 149

deer/km² and 88 deer/km² for Kunga and ELI, respectively (Le Saout et al. 2014a). No hunting 150

occurred on ELI and Kunga except for a “hunting for fear” experiment conducted in May 2012 on 151

Kunga: the experiment had no noticeable effect on individuals that were captured and collared for use 152

in this study (Le Saout et al. 2014b). We thus considered ELI and Kunga as predation risk free islands. 153

154

2. Habitat description 155

The study islands are covered by temperate forests consisting mainly of coniferous species, notably 156

Western Hemlock (Tsuga heterophylla (Raf.) Sarg.), Western Red Cedar (Thuja plicata Donn ex D. 157

Don) and Sitka Spruce (Picea sitchensis (Bong.) Carriere) (Pojar et al. 1980). The forest understory has 158

been heavily impacted by the high density of deer (Martin et al. 2010; Le Saout et al. 2014a) which 159

browse on all groups of vascular plants (Stockton et al. 2005), resulting in mainly bare or moss-covered 160

ground. The remaining vascular plants were mostly <50cm in height (Martin et al. 2010) and consisted 161

Page 7 of 33



Draft

8

mainly of coniferous regeneration, bryophytes and ferns (not consumed by deer), grasses (notably red 162

fescue, Festuca rubra L. and nootka reedgrass, Calamagrostis nutkaensis (J. Presl) Steud.) and shrubs 163

(notably red huckleberry, Vaccinium parvifolium Sm. and salal, Gaultheria shallon Pursh) (Pojar et al. 164

1980; Chollet et al. 2013b). In the winter 2010‐2011, a severe storm created large windfall areas with 165

no, or very little, canopy cover. For deer, these newly opened areas had high feeding potential. Around 166

the perimeter of the islands, the intertidal zone consisted mainly of shoreline land. We mapped these 167

habitats on a pre-determined 50x50m grid square map in April-May 2012 by exhaustively surveying 168

the whole of both islands on foot with the help of a hand-held GPS recorder. We assigned each pre-169

determined 50x50m grid square to one of three habitat type categories: windfall, shoreline or forest 170

(Fig. 1). When a given grid square contained more than one habitat type, we assigned the square to the 171

dominant category. Based on this survey, the estimated proportions of the different habitat types 172

available on each island were 46% forest, 22% windfall and 32% shoreline on ELI, and 72% forest, 173

17% windfall and 11% shoreline on Kunga. For the forest habitat type only, we also assessed the 174

availability of food resources (i.e. all vascular plants potentially eaten by deer) by visually scoring the 175

cover of the three main plant species present in the understory stratum (<1.5m) of each forest grid 176

square (for more details, see Chollet et al. 2013a). 177

178

1. Deer capture, and GPS and activity data 179

From 2011 to 2014, we caught Sitka black-tailed deer using baited box-traps in March-April and 180

August-October of each year under the BC Wildlife Act Permit NA11‐68421 approved by Parks 181

Canada Animal Care Task Force. Each caught deer was ear-tagged, sexed, aged, weighed and equipped 182

with a GPS collar (Lotek S 7000 GPS, for deer heavier than 20kg only) programmed to obtain the 183

deer’s location with a minimum schedule of one GPS fix every 6 hours (at 04:00, 10:00, 16:00 and 184

22:00 h PST). For the analysis, we assigned each GPS fix to the nearest 50x50m grid square and we 185

Page 8 of 33



Draft

9

removed fixes that did not fall within any grid square. GPS collars were also equipped with 186

accelerometers which measured acceleration in the forward/backward (X-axis) and in the side-to-side 187

and rotary (Y-axis) axes. Activity is measured as the average difference in acceleration between 188

consecutive measurements on each axis, within a relative range between 0 and 255 per 4 minute 189

interval for each axis. The two values of activity measured respectively on the X and Y axes were 190

highly correlated across individuals (Pearson correlation coefficient r = 0.82), suggesting that they 191

provide comparable indices of activity (see also Stache et al. 2013 on roe deer Capreolus capreolus L., 192

1758; Coulombe et al. 2006 on white-tailed deer Odocoileus virginianus Zimmermann, 1780). For the 193

analyses of activity rhythms, we thus used the sum of the values for the X and Y axes as our measure of 194

activity per 4 minute interval, with values ranging from 0 (no activity) to 510 (high activity). It is 195

however worth noting that activity values are not directly proportional to actual activity intensity. 196

Based on direct observations and Receiver Operating Characteristic analyses, we estimated that activity 197

values lower than 18 indicate to an individual at full rest, and values greater than 36 indicate an active 198

animal (i.e. travelling or browsing), while intermediate values indicate a mixture of resting and active 199

behavior during the 4 min interval. 200

We obtained GPS data for 12 deer on Kunga (nine adult females and three adult males) and 12 deer 201

on ELI (eight adult females, three adult males and one yearling male). Activity data were available for 202

the same 12 deer on Kunga, but for 10 deer on ELI because of sensor failure on two collars (one adult 203

male and one adult female). Although we might expect some behavioral differences between sexes, 204

notably in habitat use during calving (e.g. Weckerly 1993), we did not formally analyze sex differences 205

because of the low sample sizes. Note, however, descriptive analysis suggested that, except possibly for 206

the calving season, there were few sex differences (see Supplementary Material Figs S1). 207

208

2. Daily activity rhythms 209

Page 9 of 33



Draft

10

Because we hypothesized that the behavioral responses of deer should vary among seasons, we 210

defined five seasons in relation to established within-year variation in climate, vegetation growth and 211

deer reproductive status: i) February to March (Feb-Mar), the coldest season on our study site, with a 212

mean temperature of 3°C, and the lowest level of resource availability, ii) April to the 15th June (Apr-213

miJun), when both resource quality and quantity are relatively high and mean temperature is 8°C, iii) 214

16th June to August (miJun-Aug), the birth/lactation season, when resource quality and quantity remain 215

high and the mean temperature is highest at 13°C, iv) September to October (Sep-Oct), the rutting 216

season, with a mean temperature of 10°C and a decrease in resource availability due to intensive 217

browsing, and v) November to January (Nov-Jan), the beginning of winter, with a mean temperature of 218

4°C and a relative low level of resource availability. Seasonal average temperatures were recorded 219

using 63 temperature recorders (i-Buttons) placed at 10 and 100 cm above the ground throughout both 220

islands between 2011 and 2012. Daily and seasonal variations in ambient temperature are presented in 221

Supplementary Fig. S2, as well as variations among the three main habitat types (i.e. forest, windfall 222

and shoreline), although no marked among-habitat differences in temperature were observed. 223

To describe spatio-temporal variation in activity rhythms, we constructed generalized additive 224

mixed models (GAMM) for investigating variation of activity level (continuous response variable 225

ranging from 0 to 510) in relation to time over a 24-hour cycle (continuous explanatory decimal 226

variable from 0 to 23.99). Because time was described as a 24-hour cycle, we used a cyclic cubic 227

regression spline (Wood 2006). Given that we had repeated measures of activity over time for each 228

individual, the deer’s identity was included in each model as a random effect. To test for differences in 229

activity rhythms among seasons for each island, we compared GAMMs including the season as a factor 230

in interaction with time (i.e. incorporating a separate smoothing spline for each season) with the 231

simplest GAMM model including time only (see Supplementary Table S1). We then compared these 232

pairs of GAMMs (one pair for each island) based on AIC values (Burnham and Anderson 1998). 233

Page 10 of 33



Draft

11

GAMMs were fitted using the ‘mgcv’ package (Wood 2011) implemented in R software version 3.1.3 234

(R Development Core Team 2015). 235

236

3. Habitat selection analyzes 237

5.1. Day-night contrast in habitat use within islands 238

Our hypothesis was that habitat selection patterns of deer in predation-free environments should 239

vary at the seasonal time scale only, tracking seasonal variation of resource availability, but not at the 240

daily scale because of the absence of variation in the level of predation risk between night and day. 241

Therefore, we analyzed variation in the daily habitat selection patterns of deer among seasons and 242

islands by contrasting daytime versus nighttime habitat use (see Bonnot et al. 2013). For this, we used 243

GPS locations in generalized linear mixed models, with the time of day (day versus night) as the 244

dependent variable and the habitat type (three modalities: shoreline, windfall and forest) as an 245

explanatory factor. These models allowed us to estimate the probability that each habitat type was used 246

during daytime in comparison to its use during nighttime (i.e. an estimated probability of 1 indicates a 247

habitat that is used only during daytime, while a probability of 0 indicates a habitat that is used only 248

during nighttime). While habitat selection is often analyzed using binomial logistic regression of the 249

use of a resource versus its availability at a given scale (e.g. resource selection functions, RSF Manly et 250

al. 2002; Boyce et al. 2002), our method provides a simple alternative which is not affected by habitat 251

availability, assuming that this does not change over the course of a day (see Bonnot et al. 2013, for 252

more details on this approach). Our expectation was that there should be no difference between daytime 253

and nighttime habitat use, hence, the probability of using each habitat during daytime should not differ 254

from 0.5. For valid comparison, the number of GPS fixes sampled during daytime must be similar to 255

that sampled during nighttime. Hence, we considered GPS fixes obtained at 10:00 and 22:00 PST (or 256

11:00 and 23:00 PDT in summer) to describe daytime and nighttime habitat use, respectively. We 257

Page 11 of 33



Draft

12

removed the miJun-Aug season from this analysis, because the difference in GPS sampling rate 258

between day and night was too large (as the majority of fixes recorded at 23:00 PDT occurred during 259

twilight during this season and so were removed from the nighttime habitat use estimates, resulting in 260

753 and 430 available fixes during day and night respectively; see Supplementary Table S2). 261

Therefore, we used generalized linear mixed models with a binomial distribution, a logit link and 262

with the deer’s identity as a random effect given that we had repeated measures over time for each 263

individual. To test for spatio-temporal variation in daytime versus nighttime habitat use, we included 264

the effect of season (four modalities, see Supplementary Table S2) and island (two modalities: Kunga 265

and ELI) in a three-way interaction with habitat type in the most complex model. We compared this full 266

model with three simpler nested models including the two-way interactions between habitat type and 267

island and between habitat type and season, and the simple main effect of habitat type. Finally, we 268

compared the above models with the constant model. We used the Akaike’s Information Criterion 269

(Burnham and Anderson 1998) for model selection (see the summary table in Supplementary Table S1). 270

The model with the lowest AIC value reflects the best compromise between precision and complexity. 271

All generalized linear mixed models were fitted using the ‘lmer’ function in the library ‘lme4’ (Bates et 272

al. 2015) implemented in R software 273

274

5.2. Patterns of resource selection within forest habitat 275

To describe spatio-temporal variation in the selection of resources within the home-range, we 276

focused on forest habitat only, the most frequently used habitat on both islands. Because we had no 277

clear prediction concerning the day-night contrast in the selection of food resources, we used a classic 278

use-availability design by contrasting the use of resources in the understory at observed deer locations 279

within forest habitat with the resources available at random locations. The use of resources in the 280

understory by a given individual was computed from observed locations within forest habitat whereas 281

Page 12 of 33



Draft

13

their availability was computed from locations randomly drawn within the forest part of that 282

individual’s home range (delimited by a 100% minimum convex polygon of used locations), with a 283

similar number of used and available locations per individual for each island and each season. 284

Then, we used resource selection functions (RSFs, Manly et al. 2002) to explore how the season 285

and the presence or absence of the main feeding resources (described below) contributed to resource 286

selection patterns. Previous studies on the feeding behavior of Sitka black-tailed deer in May-August 287

have shown that deer of ELI and Kunga mostly browse on grasses, notably red fescue and small-288

flowered Wood-rush (Luzula parviflora (Ehrh.) Desv.), and on Sitka Spruce (ca 90% of their average 289

combined browsing time; Le Saout 2009; Chollet et al. 2013a). To a lesser extent, shoots of other trees 290

and shrubs (notably Western Red Cedar and Red Huckleberry) were also browsed (ca 10% of browsing 291

time). 292

In view of the above, we determined the presence/absence of three main resource types in each 293

50x50m grid square: i) grasses and shrubs (i.e. Red Huckleberry) grouped under the “GRH” acronym, 294

ii) Sitka Spruce (“SSP”) and iii) Western Red Cedar (“WRC”). Given that more than one resource type 295

may occur in a given pixel at the same time, we defined seven exclusive categories in relation to the 296

presence or absence of each resource type in a given grid square: the three mono-specific modalities 297

“WRC”, “SSP” and “GRH” (for grid squares containing only one resource type) and the four multi-298

specific modalities “SSP-GRH”, “SSP-WRC”, “WRC-GRH”, “WRC-SSP-GRH” for grid squares 299

containing either two or three of the resource types. We described spatial variation in patterns of 300

resource selection by using generalized linear mixed models with a binomial distribution contrasting 301

the use (use=1) versus the availability (use=0) of each resource type across seasons. The most complex 302

model included the three-way interaction between resource type (7 modalities), season (5 modalities: 303

Nov-Jan, Feb-Mar, Apr-miJun, miJun-Aug and Sep-Oct) and island (two modalities: ELI and Kunga). 304

We compared this full model with the 7 simpler nested models including the two-way interactions 305

Page 13 of 33



Draft

14

between resource type and season and between resource type and island, with models including 306

resource type, season and island as main effects and with the constant model using the Akaike’s 307

Information Criterion (Burnham and Anderson 1998; see Supplementary Table S1). 308

309

RESULTS 310

1. Daily activity rhythms 311

On both islands, and irrespective of the season, deer were markedly more active during daytime 312

than nighttime (see Fig. 2 and Supplementary Fig. S3). Although these activity rhythms were 313

remarkably consistent, the difference between daytime and nighttime activity levels varied slightly 314

among seasons. Indeed, the best models for both islands included the interaction between time over the 315

24-hour cycle and season (ELI: ∆AIC = 28259 and Kunga: ∆AIC = 69353 with the models including 316

time only, see Supplementary Table S1). The day-night contrast in activity level was particularly 317

marked during spring and summer (April-August) on ELI compared to the other seasons, particularly 318

fall (September-October), but was most pronounced during early winter (November-January) on 319

Kunga. Moreover, deer exhibited marked peaks in activity clearly closely linked to the crepuscular 320

periods (just after sunrise and just before sunset). These peaks in activity were remarkably similar 321

between islands in both their amplitude and timing. 322

323

2. Habitat selection patterns 324

2.1. Day-night contrast in habitat use within islands 325

Overall, we found no marked day-night contrast in habitat selection patterns for deer living on these 326

predator-free islands. The model with the highest support (with a much lower AIC value than any 327

simpler model tested, i.e. with ∆AIC ≥ 32; see Supplementary Table S1) included the three-way 328

interaction between habitat, season and island, indicating some variation in the day-night contrast in 329

Page 14 of 33



Draft

15

habitat use among seasons and islands. However, inspection of parameter estimates and their 330

confidence intervals (Fig. 3) indicates that the contrast was not pronounced (i.e. close to 0.5 threshold 331

which indicates identical day-night use). In detail, deer on ELI used windfall slightly more during 332

daytime than nighttime (all estimated probabilities were greater than 0.55), whereas there was no 333

marked day-night contrast in windfall use on Kunga (estimated probabilities ranged from 0.43 to 0.52). 334

The estimates for the day-night contrast in the use of shoreline habitat were more variable among 335

seasons (ranging from 0.32 to 0.61; see Fig. 3), but less precise (wider confidence intervals), probably 336

partly because of the relatively low use of this habitat type (around 6% of the total number of fixes 337

were located in shoreline habitat, see Supplementary Fig. S4 and Table S3). Hence, there was no 338

significant day-night contrast in the use of shoreline habitat in most seasons on either island 339

(confidence intervals included 0.5). However, shoreline habitat was used more during nighttime than 340

daytime from November to January on both islands, but more during daytime than nighttime from 341

February to March on Kunga only. Finally, for forest habitat, the most frequently used habitat on both 342

islands (see Supplementary Fig. S4 and Table S3), there was no marked day-night contrast in use on 343

either island, with all estimated probabilities ranging between 0.45 and 0.52. 344

345

2.2. Resource selection within forest habitat 346

The model with the most support for explaining variation in resource selection within forest habitat 347

(i.e. use versus availability of each resource type) included the three-way interaction between resource 348

type, season and island (∆AIC > 490 with the nearest nested model, see Supplementary Table S1). 349

Firstly, deer tended to select locations with higher specific richness (Fig. 4). That is, mono-specific 350

patches were generally more avoided than richer patches providing two or three resource types, 351

although mono-specific patches providing the highly sought after Western Red Cedar (“WRC”) were 352

also highly selected on ELI. On Kunga, deer particularly selected patches that simultaneously provided 353

Page 15 of 33



Draft

16

the three resource types (“WRC-GRH-SSP”). Deer also exhibited seasonal variation in resource 354

selection. Notably, from April to October they selected patches that provided both Sitka Spruce and 355

grasses and/or Red Huckleberry (“SSP-GRH”). Selection for patches providing only grasses and/or 356

Red Huckleberry (“GRH”) was also higher from April to October on ELI, whereas there was no 357

marked seasonal variation for this resource type on Kunga. In winter (from November to March), deer 358

selected patches providing Western Red Cedar (“WRC”, “GRH-WRC” and “SSP-WRC”) more on both 359

islands and, to a lesser extent, during fall (Sept-Oct, on ELI only). 360

361

DISCUSSION 362

In a predator-free environment, we should expect prey to attenuate or lose their anti-predator 363

behavior due to relaxed selection pressure. This decline in anti-predator responses should be 364

particularly marked under conditions of resource limitation because poor body condition and increased 365

foraging competition may affect how individuals trade off food for security (e.g. Heithaus et al. 2007). 366

In this study of Sitka black-tailed deer behavior on two densely-populated islands without predators, 367

we found partial support for an attenuation of anti-predator behavior, as i/ activity levels were higher 368

during daytime than nighttime, which is coherent with earlier studies on predation-free ungulates (e.g. 369

Loe et al. 2007), but unusual among human-disturbed populations (e.g. Pagon et al. 2013); ii/ the day-370

night contrast in habitat use was lower than that frequently observed in populations of deer subject to 371

predation and/or disturbance (e.g. Godvik et al. 2009; Bonnot et al. 2015); and iii/ forest patch use was 372

linked to seasonal variation in key resource availability. However, peak activity did coincide with 373

sunrise/sunset, as commonly observed in deer populations subject to predation and/or disturbance (e.g. 374

Langbein et al. 1997; Pagon et al. 2013). 375

376

Higher activity during daytime than nighttime in a predator-free environment 377

Page 16 of 33



Draft

17

Published studies on the activity rhythms of large herbivores have consistently shown that 378

individuals faced with a significant threat of predation or disturbance during daytime concentrate their 379

foraging activities during nighttime or during crepuscular periods (e.g. Crosmary et al. 2012; Fortin et 380

al. 2015). In this study, we showed that Sitka black-tailed deer living in a predation-free environment 381

were more active during daytime than during nighttime (Fig. 2 and Supplementary Fig. S3). This result 382

differs from our initial hypothesis, but is consistent with previous studies on predation-free deer 383

populations which also showed higher activity levels during daytime (e.g. Loe et al. 2007; Massé and 384

Côté 2013; Ensing et al. 2014). Our result is also in marked contrast with the pattern observed in most 385

hunted populations where deer are more active during nighttime (e.g. Beier and McCullough 1990; 386

Langbein et al. 1997; Pagon et al. 2013), likely as a result of the higher level of human-induced 387

disturbance and/or hunting during daytime. This interpretation in terms of a behavioral adjustment to 388

the absence of predation risk is supported by a small sample of GPS monitored deer (N=5) on the 389

neighboring Reef island where deer have been exposed to regular hunting since 1997, consisting in an 390

initial reduction of over 70% in the deer population between 1997 and 2000, and then maintaining them 391

at around 10-15 deer/km²). These hunted deer are markedly nocturnal in contrast to those from ELI and 392

Kunga (but see the Nov-Jan season; Supplementary Fig. S5). Although this result does not support our 393

initial prediction of no differences in activity levels between daytime and nighttime under relaxed 394

predation pressure, it does imply that the deer on these islands have made a marked behavioral 395

adjustment to the absence of predation risk. 396

Crepuscular peaks in activity have been commonly interpreted as an anti-predator behavior (Caro 397

2005; Kamler et al. 2007; Monterroso et al. 2013; Swinnen et al. 2015). Indeed, many deer populations 398

exhibit bimodal activity rhythms that peak around sunrise and sunset (e.g. white-tailed deer: Beier and 399

McCullough 1990; Massé and Côté 2013, black-tailed deer: Long et al. 2013, roe deer: Pagon et al. 400

2013). Contrary to our predictions, deer in the predation-free environment of our study site also 401

Page 17 of 33



Draft

18

exhibited such peaks in activity around sunrise and sunset (Fig. 2). This contrasts with studies that 402

reported an absence of marked crepuscular peaks in activity levels of predation-free and food-limited 403

ungulates (Loe et al. 2007 on arctic reindeer (Rangifer tarandus L., 1758); Long et al. 2013 on Sitka 404

black-tailed deer). However, other studies of predation-free deer populations have shown persistence of 405

such peaks in activity during crepuscular periods (e.g. Massé and Coté 2013 on white-tailed deer). 406

Moreover, a recent study on the same islands of Haida Gwaii showed that avoidance of resources 407

tainted with olfactory wolf cues persisted despite the lack of exposure to this predator for more than 408

100 years (Chamaillé-Jammes et al. 2014). Explaining why certain behaviors that are interpreted as 409

anti-predator persist in predation-free environments whereas others are lost is difficult. Although 410

different biological processes may be involved, for example “buttressing pleiotropy” (i.e. when the 411

expression of traits is tightly linked so that selection for one trait can favor persistence of a second 412

linked trait despite relaxed selection for that trait), we stress the need for further studies that quantify 413

the costs of a given anti-predator behavior for individual fitness. 414

415

Limited day-night difference in habitat use in a predator-free environment 416

We observed only limited variation in deer habitat use between day and night on both islands (most 417

season-specific estimates for the probability of use of a given habitat during daytime fell within the 0.4-418

0.6 range, where 0.5 indicates identical day-night use). In particular, the use of forest, the habitat that 419

the deer used most intensively (Supplementary Fig. S4 and Table S3), was almost identical during 420

daytime and nighttime, with no seasonal variation (Fig. 3). There was also no day-night contrast in the 421

use of windfall habitat for deer on Kunga, however, deer used windfall slightly more during the day 422

than during nighttime on ELI. We speculatively suggest that the higher daytime use of windfall on ELI 423

was the consequence of deer actively seeking the higher temperatures (linked to greater luminosity) in 424

this habitat compared to closed canopy forest. Indeed, previous studies have shown that deer are able to 425

Page 18 of 33



Draft

19

adjust their habitat use patterns in response to changes in ambient temperature to avoid increased 426

thermoregulation costs (e.g. Beier and McCullough 1990; Mysterud 1996). This hypothesis is 427

consistent with the observation of higher ambient temperature in windfall than in forest during daytime 428

from April to October on ELI (see Supplementary Fig. S2). 429

The low day-night contrast in habitat selection that we observed markedly differs from what is 430

commonly observed in other ungulate populations subject to the risk of predation and/or disturbance, 431

where patterns of habitat selection are highly differentiated between daytime and nighttime (e.g. Fortin 432

et al. 2015; Tambling et al. 2015). In human-disturbed landscapes, deer exhibit strong selection for 433

habitats providing good cover during daytime (i.e. when human disturbance is greater) and for open 434

forage-rich habitats during nighttime (e.g. Beier and McCullough 1990; Godvik et al. 2009; Padié et al. 435

2015). For example, roe deer living in a human-disturbed landscape in South-West France had an 436

average probability of 0.8 of being located in forest during daytime, but a probability of more than 0.7 437

of being located in open habitats during nighttime (Bonnot et al. 2013). Lastly, although the day-night 438

contrast was also generally not marked in most seasons, during winter, for an as yet unknown reason 439

deer on both islands tended to use shorelines more at nighttime (Fig. 3). Note, however, that the 440

availability of this habitat was quite low, and it was rarely used (Supplementary Fig. S4 and Table S3), 441

so that model estimates were not precise. 442

As expected, at a smaller spatial scale and focusing on forest habitat which was the most widely 443

used across the year on both islands (Supplementary Fig. S4 and Table S3), we found that the Sitka 444

black-tailed deer of Haida Gwaii used forest patches in relation to the spatial distribution of the three 445

main food resource types (Chollet et al. 2013a, b; Le Saout et al. 2014a). Firstly, richer patches that 446

provided a mix of two or all three resource types were generally more selected than mono-specific 447

patches, although this pattern was more pronounced on Kunga (Fig. 4). This result suggests that these 448

food resources are to some degree complementary for offsetting the deer’s’ energetic and nutrient 449

Page 19 of 33



Draft

20

demands, and/or that the richest patches may offer the best opportunities to buffer against uncertainty 450

in resource availability. Although we supposed that the two islands could be considered as replicates for 451

our study, and while this was largely the case, we did observe some differences in resource selection 452

patterns, with deer on ELI preferring patches providing Western Red Cedar, whereas deer on Kunga 453

preferred patches providing Sitka Spruce and grasses and/or Red Huckleberry in combination (Fig. 4). 454

Secondly, we found marked seasonality in the patterns of resource selection that can be readily 455

explained by seasonality in resource availability. The increased winter use of Western Red Cedar, an 456

evergreen coniferous species which is palatable and sought after throughout the year, likely reflects its 457

critical importance as an alternative feeding resource during winter, when more nutritious and preferred 458

species (i.e. grasses and shrubs) are scarce and when the availability of fallen cedar leaves increases 459

due to higher wind frequency and strength. This interpretation is consistent with earlier studies on the 460

diet and feeding behavior of Sitka black-tailed deer on these same islands. They showed that conifer 461

consumption increased during winter, whereas deer mainly consumed grasses, shrubs and spruce buds 462

during spring and summer (Chollet et al. 2013a; Poilvé 2013). The spring/summer preference observed 463

for Sitka spruce and Huckleberry is equally consistent with the phenology of these two resources. Deer 464

almost exclusively consume young tender spruce buds that are only available in spring and huckleberry 465

shoots and leaves that also emerge in spring and further develop during the course of summer. 466

467

Conclusion 468

Our study suggests that large herbivores can adjust their habitat selection pattern and daytime 469

activity rhythm to the absence of predation risk in a context of limiting resources. However, these 470

observations contrast with the observed persistence of other anti-predator behaviors in this predator-471

free environment, notably crepuscular activity peaks documented in this study, but also behavioral 472

responses to predator cues and high levels of vigilance during feeding (see Chamaillé-Jammes et al. 473

Page 20 of 33



Draft

21

2014; Le Saout et al. 2015, respectively). Further studies are needed to explore how and why some 474

anti-predator behaviors persist in the absence of predation risk, while others attenuate rapidly, and may 475

disappear altogether (Coss 1999; Blumstein 2002; Blumstein and Daniel 2005; Lahti et al. 2009). 476

477

ACKNOWLEDGMENTS 478

N. Bonnot benefited from a post-doctoral grant from ANR BAMBI (2010 BLAN 1718 01). 479

Funding for the field work was provided by the ANR BAMBI, and by Forest Renewal BC, the South 480

Moresby Forest Replacement Account, the International Research Group (GRDI) Dynamics of 481

Biodiversity and Life-History traits and the French Ecological Society (SFE). The Canadian Wildlife 482

Service of Environment Canada and the Laskeek Bay Conservation Society (LBCS) provided logistical 483

support. The Archipelago Joint Management Board provided the necessary permits to work in Gwaii 484

Haanas and deer captures were done under Parks Canada Animal Care Protocol and research Permit 485

number 9059 and under Wildlife Act permit NA11-68421 issued by the Ministry of Natural Resource 486

Operations of the Province of British Columbia, Canada. Special thanks to Malcolm Hyatt, Flore Saint-487

André, Lukas Ostermann and to Barb Rowsell, our project coordinator, for help in the field. We thank 488

two anonymous referees for constructive comments on a previous version of this manuscript. 489

490

491

Page 21 of 33



Draft

22

REFERENCES 492

Bates, D., Maechler, M., Bolker, B., and Walker, S. 2015. lme4: Linear mixed-effects models using Eigen 493

and S4. R package version 1.1-8 [online]. Available from http://CRAN.R-project.org/package=lme4 494

Biro, P.A., Abrahams, M.V., Post, J.R., and Parkinson, E.A. 2006. Behavioural trade-offs between growth 495

and mortality explain evolution of submaximal growth rates. J. Anim. Ecol. 75: 1165-1171. J. Anim. Ecol 496

Beier, P., and McCullough, DR. 1990. Factors influencing white-tailed deer activity patterns and habitat use. 497

Wildl. Monogr. 109: 3-51. 498

Benhaiem, S., Delon, M., Lourtet, B., Cargnelutti, B., Aulagnier, S., Hewison, A.J.M., Morellet, N., and 499

Verheyden, H. 2008. Hunting increases vigilance levels in roe deer and modifies feeding site selection. Anim. 500

Behav. 76:611-618. 501

Blumstein, D.T. 2002. Moving to suburbia: ontogenetic and evolutionary consequences of life on predator-502

free islands. J. Biogeogr. 29: 685–692. 503

Blumstein, D.T., and Daniel, J.C. 2005. The loss of anti-predator behaviour following isolation on 504

islands. Proc. R. Soc. Lond. B Biol. Sci. 272(1573): 1663-1668. 505

Bonnot, N., Morellet, N., Verheyden, H., Cargnelutti, B., Lourtet, B., Klein, F., and Hewison, A.J.M. 2013. 506

Habitat use under predation risk: hunting, roads and human dwellings influence the spatial behaviour of roe deer. 507

Eur. J. Wildl. Res. 59: 185-193. 508

Bonnot, N., Verheyden, H., Blanchard, P., Cote, J., Debeffe, L., Cargnelutti, B., Klein, F., Hewison, A.J.M, 509

and Morellet, N. 2015. Interindividual variability in habitat use: evidence for a risk management syndrome in roe 510

deer? Behav. Ecol. 26: 105-114. 511

Boyce, M., Vernier, P.R., Nielsen, S.E., and Schmiegelow, F.K.A. 2002. Evaluating resource selection 512

functions. Ecol. Model. 157: 281-300. 513

Brown, J. S., and Kotler, B. P. 2004. Hazardous duty pay and the foraging cost of predation. Ecol. Lett. 7: 514

999-1014. 515

Burnham, K.P., and Anderson, D.R. 1998. Model selection and inference: A practical information-theoretic 516

approach. Springer-Verlag, New-York, New-York, USA. 517

Page 22 of 33



Draft

23

Caro, T. 2005. Antipredator defenses in birds and mammals. University of Chicago Press, Chicago, London, 518

USA. 519

Chamaillé-Jammes, S., Malcuit, H., Le Saout, S., Martin, J.-L. 2014. Innate threat-sensitive foraging: Sitka 520

black-tailed deer remain more fearful of wolf than of the less dangerous black bear even after 100 years of wolf 521

absence. Oecologia, 174: 1151-1158. 522

Chollet, S., Baltzinger, C., Le Saout, S., and Martin, J.-L. 2013a. A better world for bryophytes? A rare and 523

overlooked case of positive community-wide effects of browsing by overabundant deer. Ecoscience, 20: 352-524

360. 525

Chollet, S., Baltzinger, C., Ostermann, L., Saint-André, F., and Martin, J.-L. 2013b. Importance for forest 526

plant communities of refuges protecting from deer browsing. For. Ecol. Manage. 289: 470-477. 527

Christianson, D., and Creel, S. 2010. A nutritionally mediated risk effect of wolves on elk. Ecology, 91: 528

1184-1191. 529

Coss, R.G. 1999. Effects of relaxed natural selection on the evolution of behavior. In Geographic variation 530

in behavior: Perspectives on evolutionary mechanisms. Edited by S.A. Foster and J.A. Endler. Oxford University 531

Press, Oxford, UK. Pp. 180-208. 532

Coulombe, M.-L., Massé, A., and Côté, S.D. 2006. Quantification and accuracy of activity data measured 533

with VHF and GPS telemetry. Wildl. Soc. Bull. 34: 81-92. 534

Creel, S., Christianson, D., Liley, S., and Winnie, Jr J.A. 2007. Predation risk affects reproductive 535

physiology and demography of elk. Science, 315: 960. 536

Crosmary, W.-G., Valeix, M., Fritz, H., Madzikanda, H., Côté, S.D. 2012. African ungulates and their 537

drinking problems: hunting and predation risks constrain access to water. Anim. Behav. 83:145-153. 538

Ensing, E.P., Ciuti, S., de Wijs, F.A., Lentferink, D.H., Ten Hoedt, A., Boyce, M.S., and Hut, R.A. 2014. 539

GPS based daily activity patterns in European red deer and North American elk (Cervus elaphus): indication for 540

a weak circadian clock in ungulates. PLoS ONE, 9: e106997. 541

Fraser, D.F., and Huntingford, F. 1986. Feeding and avoiding predation hazard: the behavioral response of 542

the prey. Ethology, 73: 56-68. 543

Page 23 of 33



Draft

24

Godvik, I., Loe, L.E., Vik, J., Veiberg, V., Langvatn, R., and Mysterud, A. 2009. Temporal scales, trade-offs, 544

and functional responses in red deer habitat selection. Ecology, 90: 699-710. 545

Heithaus, M.R., Frid, A., Wirsing, A.J., Dill, L.M., Fourqurean, J.W., Burkholder, D., Thomson, J. and 546

Bejder, L. 2007. State‐dependent risk‐taking by green sea turtles mediates top‐down effects of tiger shark 547

intimidation in a marine ecosystem. J. Anim. Ecol. 76: 837-844. 548

Kamler, J.F., Jędrzejewska, B., and Jędrzejewski, W. 2007. Activity patterns of red deer in Białowieża 549

National Park, Poland. J. Mamm. 88: 508-514. 550

Kronfeld-Schor, N., and Dayan, T. 2003. Partitioning of time as an ecological resource. Annu. Rev. Ecol. 551

Evol. Syst. 34: 153-181. 552

Lahti, D.C., Johnson, N.A., Ajie, B.C., Otto, S.P., Hendry, A.P., Blumstein, D.T., Coss, R.G., Donohue, K., 553

and Foster, S.A. 2009. Relaxed selection in the wild. Trends Ecol. Evol. 24: 487-496. 554

Langbein, J., Scheibe, K.M., and Eichhorn, K. 1997. Seasonal changes in the circadian behaviour patterns 555

in European mouflons (Ovis ammon musimon Pallas, 1811). 1st Intern. Symp. on Physiology and Ethology of 556

Wild and Zoo Animals, Berlin, Germany, Sept. 18–21, 1996. Z. Säugetierkd. 62: 117-123. 557

Le Saout, S. 2009. Study of Sitka black-tailed-deer browsing behaviour in relation to habitat 558

impoverishment on Haida Gwaii archipelago (BC, Canada). Sci. Rep. AgroParisTech, Paris and CEFE, CNRS 559

Montpellier (France). 560

Le Saout, S., Chollet, S., Chamaillé-Jammes, S., Blanc, L., Padié, S., Verchere, T., Gaston, A.J., Gillingham, 561

M.P., Gimenez, O., Parker, K.L., Picot, D., Verheyden, H., and Martin, J.-L. 2014a. Understanding the paradox 562

of deer persisting at high abundance in heavily browsed habitats. Wildl. Biol. 20: 122-135. 563

Le Saout, S., Padié, S., Chamaillé-Jammes, S., Chollet, S., Côté, S., Morellet, N., Pattison, J., Harris, E., 564

and Martin, J.-L. 2014b. Short-term effects of hunting on naïve Sitka black-tailed deer (Odocoileus hemionus 565

sitkensis): behavioural response and consequences on vegetation growth. Can. J. Zool. 92: 915-925. 566

Le Saout, S., Martin, J.-L., Blanchard, P., Cèbe, N., Hewison, A.J.M., Rames, J.-L., Chamaillé-Jammes, S. 567

2015. Seeing a ghost? Vigilance and its drivers in a predator-free world. Ethology, 121: 651–660. 568

Lima, S.L. 1998. Nonlethal effects in the ecology of predator-prey interactions. Bioscience, 25-34. 569

Page 24 of 33



Draft

25

Lima, S. L., and Bednekoff, P. A. 1999. Temporal variation in danger drives antipredator behavior: the 570

predation risk allocation hypothesis. Am. Nat. 153(6): 649-659. 571

Lima, S.L., and Dill, L.M. 1990. Behavioral decisions made under the risk of predation: a review and 572

prospectus. Can. J. Zool. 68: 619-640. 573

Loe, L.E., Bonenfant, C., Mysterud, A., Severinsen, T., Øritsland, N.A., Langvatn, R., Stien, A., Irvine, R.J., 574

and Stenseth, N.C. 2007. Activity pattern of arctic reindeer in a predator-free environment: no need to keep a 575

daily rhythm. Oecologia, 152: 617-624. 576

Lone, K., Loe, L.E., Gobakken, T., Linnell, J.D., Odden, J., Remmen. Jr, Mysterud, A. 2014. Living and 577

dying in a multi-predator landscape of fear: roe deer are squeezed by contrasting pattern of predation risk 578

imposed by lynx and humans. Oikos, 123: 641-651. 579

Long, E.S., Jacobsen, T.C., Nelson, B.J., and Steensma, K.M.M. 2013. Conditional daily and seasonal 580

movement strategies of male Columbia black-tailed deer (Odocoileus hemionus columbianus). Can. J. Zool. 91: 581

679-688. 582

Manly, B.F.J., McDonald, L.L., Thomas, D.L., McDonald, T.L., and Erickson, W.P. 2002. Resource 583

selection by animals: statistical design and analysis for field studies. 2nd edn. Kluwer Press, New-York, New-584

York, USA. 585

Martin, J.-L., Stockton, S.A., Allombert, S., and Gaston, A.J. 2010. Top-down and bottom-up consequences 586

of unchecked ungulate browsing on plant and animal diversity in temperate forests: lessons from a deer 587

introduction. Biol. Invasions, 12: 353-371. 588

Massé, A., and Côté, S.D. 2013. Spatiotemporal variations in resources affect activity and movement 589

patterns of white-tailed deer (Odocoileus virginianus) at high density. Can. J. Zool. 91: 252-263. 590

Monterroso, P., Alves, P. C., and Ferreras, P. 2013. Catch me if you can: diel activity patterns of mammalian 591

prey and predators. Ethology, 119: 1044-1056. 592

Mysterud, A. 1996. Bed-site selection by adult roe deer Capreolus capreolus in southern Norway during 593

summer. Wildl. Biol. 2(2): 101-106. 594

Page 25 of 33



Draft

26

Padié, S., Morellet, N., Hewison, A.J.M., Martin, J.-L., Bonnot, N., Cargnelutti, B., Chamaillé-Jammes, S. 595

2015. Roe deer at risk: teasing apart habitat selection and landscape constraints in risk exposure at multiple 596

scales. Oikos, 124: 1536-1546. 597

Pagon, N., Grignolio, S., Pipia, A., Bongi, P., Bertolucci, C., and Apollonio, M. 2013. Seasonal variation of 598

activity patterns in roe deer in a temperate forested area. Chronobiol. Int. 30: 772-785. 599

Poilvé, E. 2013. Etude du régime alimentaire de populations de cerfs à queue-noire (Odocoileus 600

hemionus sitkensis) de l’archipel Haïda Gwaii. M.Sc. thesis, Université de Bourgogne, Dijon, France. 601

Pojar J. 1980. Relationships between introduced Sitka black-tailed deer and the plant life of the Queen 602

Charlotte Islands, Bristish Columbia. Research Branch, Ministry of Forests, Smithers, BC. 603

R Core Team. 2015. R: A language and environment for statistical computing. R Foundation for Statistical 604

Computing, Vienna, Austria. [online] Available from http://www.R-project.org/. 605

Raymond, M., Berticat, C., Weill, M., Pasteur, N., and Chevillon, C. 2001. Insecticide resistance in the 606

mosquito Culex pipiens: what have we learned about adaptation? Genetica, 112-113: 287–296. 607

Ripple, W.J., Estes, J.A., Beschta, R.L., Wilmers, C.C., Ritchie, E.G., Hebblewhite, M., Berger, J., 608

Elmhagen, B., Letnic, M., Nelson, M.P., Schmitz, O.J., Smith, D.W., Wallach, A.D., and Wirsing, A.J. 2014. 609

Status and ecological effects of the world’s largest carnivores. Science, 343(6167): 1241484. 610

Sharpe, S. 1999. Management of deer on the Queen Charlotte Islands: Biology of the species. In 611

Proceedings of the Cedar Symposium: growing western redcedar and yellow-cypress on the Queen Charlotte 612

Islands/Haida Gwaii. Edited by G.G. Wiggins. B.C. Ministry of Forests/Canada–British Columbia South 613

Moresby Forest Replacement Account, Victoria, B.C. pp 118–124. 614

Sih, A. 1980. Optimal behavior: can foragers balance two conflicting demands? Science, 210: 1041-1043. 615

Sih, A., Ferrari, M.C., and Harris, D.J. 2011. Evolution and behavioural responses to human-induced rapid 616

environmental change. Evol. Appl. 4: 367-387. 617

Stache, et al. 2013 Stache, A., Heller, E., Hothorn, T., and Heurich, M. 2013. Activity patterns of European 618

roe deer (Capreolus capreolus) are strongly influenced by individual behaviour. Folia Zool. 62(1): 67-75. 619

Stockton, S.A., Allombert, S., Gaston, A.J., and Martin, J.-L. 2005. A natural experiment on the effects of 620

high deer densities on the native flora of coastal temperate rain forests. Biol. Conserv. 126: 118–128. 621

Page 26 of 33



Draft

27

Swinnen, K.R.R., Hughes, N.K., and Leirs, H. 2015. Beaver (Castor fiber) activity patterns in a predator-622

free landscape. What is keeping them in the dark? Mamm. Biol. 80: 477-483. 623

Tambling, C.J., Minnie, L., Meyer, J., Freeman, E.W., Santymire, R.M., Adendorff, J., and Kerley, G.I. 624

2015. Temporal shifts in activity of prey following large predator reintroductions. Behav. Ecol. Sociobiol. 69(7): 625

1-9. 626

Vila, B., Guibal, F., Torre, F., and Martin, J.-L. 2004. Assessing spatial variation in browsing history by 627

means of fraying scars. J. Biogeogr. 31: 987-995. 628

Weckerly, F.W. 1993. Intersexual resource partitioning in black-tailed deer: a test of the body size 629

hypothesis. J. Wildl. Manage. 57: 475-494. 630

Wood, S.N. 2006. Generalized Additive Models: An Introduction with R. Chapman and Hall/CRC Press. 631

Wood, S.N. 2011. Fast stable restricted maximum likelihood and marginal likelihood estimation of 632

semiparametric generalized linear models. J. R. Stat. Soc. B. 73: 3-36. 633

634

635

Page 27 of 33



Draft

28

FIGURE LEGEND 636

Fig. 1 Map of the study area in the Haida Gwaii archipelago (British Columbia), with the location of the study 637

islands (East-Limestone (ELI) and Kunga) and the neighbouring islands Reef and Louise, one of the main 638

islands of the archipelago (map courtesy of the Gowgaia Institute). The right panel presents detailed maps of ELI 639

and Kunga with the main habitat types (adapted from Le Saout 2009). 640

641

Fig. 2 Profiles of Sitka black-tailed deer (Odocoileus hemionus sitkensis) activity levels (black lines, see text for 642

construction of the activity index) over the 24-hour cycle across seasons and for the two islands (ELI and 643

Kunga). The average activity levels for nighttime and daytime are represented by the dark grey and light grey 644

dotted lines, respectively. The grey shading represents sunrise and sunset (i.e. incorporating variation in the 645

timing of sunrise and sunset over a given season). 646

647

Fig. 3 Estimated daytime versus nighttime habitat use probabilities of Sitka black-tailed deer (Odocoileus 648

hemionus sitkensis) among the four seasons and the three habitat types for the two islands (ELI and Kunga) as 649

generated by the best model. The dotted lines represent a level of use that is identical between daytime and 650

nighttime (i.e. an estimated probability of daytime use = 0.5) and bars represent the 95% confidence intervals 651

around the probability estimates. Values above (below) the dotted line indicate that the given habitat type is used 652

more (less) during the day than during the night for that season. 653

654

Fig. 4 Selection estimates with their 95% confidence intervals as predicted by the best model describing patterns 655

of resource selection of Sitka black-tailed deer (Odocoileus hemionus sitkensis) among the five seasons for each 656

island (A. ELI and B. Kunga). A value different from 1 (i.e. with confidence intervals that do not overlap the 657

black horizontal line) indicates that the use of a given resource type is not proportional to its availability, but 658

rather is selected (selection estimates > 1) or avoided (selection estimates < 1). The percentages given above 659

tick-marks indicate the proportion of each resource type available on each island. Resource types: “SSP”: 660

presence of Sitka Spruce only; “GRH”: presence of grasses and/or Red Huckleberry only; “WRC”: presence of 661

Western Red Cedar only; “SSP-GRH”: presence of both Sitka Spruce and grasses and/or Red Huckleberry; 662

Page 28 of 33



Draft

29

“SSP-WRC”: presence of both Sitka Spruce and Western Red Cedar; “GRH-WRC”: presence of both grasses 663

and/or Red Huckleberry and Western Red Cedar; “GRH-SSP-WRC”: presence of all three resource types. 664

Page 29 of 33



Draft

181x105mm (300 x 300 DPI)

Page 30 of 33



Draft

Profiles of Sitka black-tailed deer (Odocoileus hemionus sitkensis) activity levels (black lines, see text for construction of the activity index) over the 24-hour cycle across seasons and for the two islands (ELI and Kunga). The average activity levels for nighttime and daytime are represented by the dark grey and light

grey dotted lines, respectively. The grey shading represents sunrise and sunset (i.e. incorporating variation in the timing of sunrise and sunset over a given season).

171x236mm (300 x 300 DPI)

Page 31 of 33



Draft

Estimated daytime versus nighttime habitat use probabilities of Sitka black-tailed deer (Odocoileus hemionus sitkensis) among the four seasons and the three habitat types for the two islands (ELI and Kunga) as generated by the best model. The dotted lines represent a level of use that is identical between daytime

and nighttime (i.e. an estimated probability of daytime use = 0.5) and bars represent the 95% confidence intervals around the probability estimates. Values above (below) the dotted line indicate that the given

habitat type is used more (less) during the day than during the night for that season. 112x90mm (300 x 300 DPI)

Page 32 of 33



Draft

182x76mm (300 x 300 DPI)

Page 33 of 33



draft...draft 3 48 resume 49 les individus font face à un compromis entre acquérir des ressources...

Documents