FORACINC DECISIONS AT MULTIPLE SPATIAL

AND TEMPORAL SCALES: A BISON PERSPECTIVE

A Thesis

Presented to

The Faculty of Graduate Studies

of

The University of Guelph

by

DANIEL FORTiN

In partial fiilfilment of requirements

for the degree of

Doctor of Philosophy

December, 2000

O Daniel Fortin, 2000

National Library I*! oi cana, Bibliothèque nationale du Canada

Acquisitions and Acquisitions et Bibliographie Services services bibliographiques 395 Wellington Street 395, me Weilington OtlawaON KlAON4 OltawaûN K l A O N l Canada Canade

The author has granted a non- exclusive licence allowing the National Library of Canada to reproduce, loan, distribute or sel1 copies of this thesis in rnicroform, paper or electronic formats.

The author retains ownership of the copyright in this thesis. Neither the thesis nor substantial extracts fiom it may be printed or otherwise reproduced without the author's permission.

L'auteur a accordé une licence non exclusive permettant à la Bibliothèque nationale du Canada de reproduire, prêter, distribuer ou vendre des copies de cette thèse sous la fome de microfiche/fïim, de reproduction sur papier ou sur format électronique.

L'auteur conserve la propriété du droit d'auteur qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation.

ABSTRACT

FORAGING DECISIONS AT MULTIPLE SPATIAL

AND TEMPORAL SCALES: A BISON PERSPECTIVE

Daniel Fortin University of Guelph, 2000

Advisor: Dr. John M. Fryxell

This thesis is an investigation of the behavioural response of fiee-ranging bison

(Bison bison) to resource distribution and abundance across spatio-temporal scales. From

1996 to 1999.1 exarnined the searching behaviour. diet selection and habitat use of bison

in Prince Albert National Park, Saskatchewan.

Analysis of winter searchinglforaging paths reveaied that bison used area-

restricted search to find food underneath the snow. Bison perception of resource quality

varied with their short-term sampling experience. Computer simulations based on bison

behaviour and habitat characteristics indicated that searching efficiency should increase

as more sarnpling information is used to m e s s resource quality. but this increase should

rapidly level off. Simulations further suggest that bison can normally maximize their

searching efficiency by considering the information gathered within a foraging bout. This

result provides a potential explanation for the flexibility in bison assessment of resource

quality observed under field conditions.

Despite such flexibility in perception of resource quaiity, the diet choice of bison

was limited to only few plant species. Contingency models, which 1 developed based on

the maximization of short- and long-term gains, revealed dietary choices more consistent

with short-term goals. in surnmer, bison diet was similar in al1 meadows. in winter? diet

choice was still consistent with short-term goals, but bison exhibited frequency-

dependent selection for the two plant species providing the highest short-term

profitability. in contrast with theoretical predictions.

Meadow selection by bison was not directly related to the distribution of the plant

they most ofien consumed, suggesting scale-sensitivity in selection criteria. In sumrner.

the presence of nearby meadows and water areas increased the probability of use of a

given meadow. In winter, snow depth within meadows was a dominant factor related to

the probability of use. Larger bison herds were more likely to be o b s e ~ e d in meadows

close to other meadows. 1 conclude that prediction of resource and habitat selection at a

given spatial and temporal scale cannot be readily inferred from knowledge at other

scales. due to trade-offs among individual requirements of bison.

ACKNOWLEDGMENTS

Funding for this study was provided by research gants from Parks Canada and

the Natural Sciences and Engineering Research Council of Canada, and by scholarship

frorn the "Fonds pour la formation de Chercheurs et d'Aide à la Recherche" and the

Ontario Graduate Scholarships.

Many people deserve credit for the realization of this study. 1 thank my advisor,

Dr. John Fryxell, for his insighthl advice and guidance throughout my research. The

rnernbers of my advisory committee, Drs. Pablo Colucci, Bart Nolet and Tom Nudds,

ot'ten took time away from their busy schedule to discuss research issues and critically

review rny work. The positive attitude and encouragement of Dr. Nudds provided moral

support that 1 greatly appreciated.

1 thank the staff of Prince Albert National Park for providing logistical assistance

during this study. 1 am especially grateful to Dan Frandsen who always insured that the

field cornponent of the study was running smoothly. The field experience and logistical

support of Lloyd O'Brodovich largely contributed to the success of the fieldwork. 1 am

also thankful to Adam Pidwerbeski for giving me the opportunity to share my findings

with park visitors through different forums. Benveen my numerous field seasons. Ms.

Weider always took good care of the "bison*' truck. Wes Olson, Park warden in Elk

Island National Park. shared his knowledge of bison morphology. which became key

information for age and sex determination during my field seasons.

Régis Pilote. Mark Andruskiw and Sergio luirez managed to survive through

laborious field seasons. Régis even asked for more by coming back for a second fieezing

Acknowledgemenrs

winter. Mark also did not have enough! Once back in Guelph, he pursued his contribution

to my research through laboratory work, data enby and the proofkeading of manuscripts.

Lynn Wardle, Cheryl Campbell and many volunteers also provided assistance during

laboratory analyses.

All graduate students of the many waves that passed through the Nudds-Fryxell

laboratory over the past four years were very helpful. Glenn Benoy, John Wilmshurst,

Tristan Long, Karen McCoy, Tracy Hillis. John McKenzie. Kara Vlasman. Jen Simard,

Mark Drever, Yolanda Wiersma, Emily Gonzales, Alastair Wilson. Carneron McDonald

and Carita Bergman helped me improve my English and writing skills by providing

judicious comments on my work. Carita also gave me the opportunity to collaborate in

her research on wood bison, and to use son:e of her raw data for rny own study.

All the friends 1 made over the past four years made my life in Guelph most

enjoyable. Among others, Alastair, Chris M. and Chris J. made sure that. before 1 lefi. 1

acquired a good appreciation of Guelph's nightlife.

Finalement. je tiens à remercier ma famille, ainsi que MarieClaude, pour le

support inconditio~el dont ils ont fait preuve au cours de ma recherche et pour toujours

croire en mes capacités peu importe les circonstances.

TABLE OF CONTENTS

ACKNOWLEDGMENTS ................................................................................................. i ...

TABLE OF CONTENTS ................................................................................................. 111

... LIST OF TABLES ......................................................................................................... VILI

LIST OF FIGURES ........................................................................................................ x

PROLOGUE ................................................................................................................. 1

SEARCH STRATEGY ................................................................................................ 2

FORAGING DECISIONS ............................................................................................. 3

INTEGRATION OF INFORMATION ACROSS SPATIO-TEMPORAL SCALES ..................... 5

.......................... THE BISON OF PRINCE ALBERT NATIONAL PARK ... .......... 6 THESIS OUTLINE .................................. .. ......... 7

LITERATURE CITED ................................................................................................. 8

CHAPTER 1 : Searching Behaviour and Use of Sampling Information

by Free-Ranging Bison ........................................................................... 15

........................................................................................................... SUMMARY 16

INTRODUCTION ..................................................................................................... 17

............................................................................................................. METHODS 19

............................................................... Searching path characteristics 19

.......................... ..........*............................. Snow characteristics .... 2 0

........................................................................... Forage charactrristics 2 1

....................................................................................... Biomass 21

............................................................. Dry marrer intake rare 21

Digestibiliry .............................................................................. 22

Energy inrake raie ....................................................................... 23

. ....................................................... Field estimale ofrejérence windows 23

Data analysis .......................................................................................... 24

...................... Correlated random walk betweenjieding craters 24

Inter-crater resotrrce qualis, ...................................................... 2 5

iii

Table of Contents

Pairs of craters wifh similar resources ....................................... 25

RESULTS .............................................................................................................. 26

............................................................... Signs of area-restricted search 26

.................................. Correlated random walk behveenfeeding craters 27

................................................................... Inter-crater resource qualiiy 27

Pairs of craters with similar resources .................................................. 2 9

......................................................................................................... DISCU~SION 30

...................................................... Navigation during searching activiiy 30

.............................................. The use of resource sampling information 32

ACKNOWLEDGEMENTS ........................................................................................ 34

LITERATURE CITED .............................................................................................. 34

................................................................................................. FIGURE LEGENDS 42

...... CHAPTER 2: Optimal Searching Behaviour: the Importance of Habitat Sampling 46

...................................................... ......................*............................ SUMMARY 47

INTRODUCTION ..................................................................................................... 48

............................ MATERIALS AND METHODS .. .............................................. 50

................................................................................................ The mode1 50

................................................. Field estimation of mode1 parameters 5 3

Data analyses .......................................................................................... 55

RESULTS ......................................................................................................... 56

Reference windows size and searching eflciency .................................. 56

A reference point for omniscient animals ............................................... 57

DISCUSSION ...................................................................................................... 58

.................................... Reference window size and searching eflciency 58

A rejèrence point for omniscient animais ............................................... 62

ACKNOWLEDGEMENTS ..................................................................................... 64

LITERATURE CLTED .............................................................................................. 64

................................................................................................. FIGURE LEGENDS 70

Table of Contents

CHAPTER 3: An Adjustment of the Extended Contingency Mode1 of Farnsworth

and Illius (1998) ...................................................................................... 73

ACKNOWLEDGMENTS .......................................................................................... 77

LITERATURE CITED ....................................................................................... 7 7

.............................. CHAPTER 4: The Temporal Scale of Foraging Decisions in Bison 79

SUMMARY ............................................................................................................ 80

INTRODUCTION .................................................................................................... 81

CONTINGENCY MODELS ..................................................................................... 84

Short-term mode1 .................................................................................... 84

Long-term mode1 ................................................................................ 84

............................................................................... Situation [Il 8 6

Situation [II] ............................................................................... 86

............................................................................. Situation [III] 8 8

.............................................................................. Situation [IV 90

............................................................................................................. METHODS 91

Prey encounter rate fi) ........................................................................... 92

Handling time (h) and proportion of handling time e-rclusive

to searching lime (rl) ............................................................................ 95

Dry matter digestibility (d) and digestible energv (e) ............................ 97

Observed diet ........................................................................................ 98

RESULTS ....................................................................................................... 9 8

................................................................................. Mode1 parameters 9 8

Mode1 predictions and observed dier .................................................... 1 O1

DISCUSSION ........................................................................................................ 103

ACKNOWLEDGMENTS ......................................................................................... 107

LITERATURE CITED ........................................................................................... 108

............................................................................................... FIGURE LEGENDS 1 17

........ CHAPTER 5: A multiscale investigation of bison distribution and resource use 125

SUMMARY .......................................................................................................... 126

Table of Contents

................................................................................................. INTRODUCT~ON 128

METHODS ...................................... ... ................................................................. 130

Siudy urra ............................................................................................ 130

PIant and snow swrvey .......................................................................... 131

.................................................................................. Animal Locarions 132

GPS-collars ............................................................................... 132

............................................................................ Bison surveys 133

........................................................... Geographic information Systcm 133 ...................................................................................... Data analysis 133

Habitat use ............................................................................. 133

Meadow selection ...................................................................... 134

Group size ................................................................................. 135

........................................................................... Plant selection 136

............................................................................................................ DSULTS 138

............................................................................................ Habitat use 138

............................................................................. Daily dispiucements 138

.................................................................................. Meadow selecrion 139

Group ske .......................................................................................... 140

....................................................................................... Plant selecrion 141

....................................................................................................... DISCUSSION 142

............................................................................................ Habita1 use 142

Meadow selecrion .................................................................................. 143

....................................................................................... Plunt selecrion 145

....................................... ........................ Integrution ocross scales .. 147

....................................................................................... ACKNOWLEDGEMENTS 149

L~TERATURE C ITED ............................................................................................ 149

FIGURE LEGENDS ............................................................................................... 163

EPILOGUE ......................................................................................................... 167

............................................................................................ SEARCH STRATEGY 167

......................................................................................... FORAGING DECISIONS 169

Table of Contents

INTEGRATION OF INFORMATION ACROSS SPATIO-TEMPORAL SCA

THE NEXT STEP ................................................................................................... LITERATURE CITED ............................................................................................

vii

LIST OF TABLES

CHAPTER 1

Table 1. Surnmary of 2-way ANOVAs (factors: individual paths and crater size) testing whether the difference in size of pairs of snow crater comprised of similar resource was not systernatically influenced by the resource quality as

............................................................ perceived using various reference windows 4 1

CHAPTER 2

Table 1. Surnmary of ANOVAs testing for differences in the average potential energy intake rate of the resource found during 800-1000 foraging bouts when assessing resource quality with various reference windows in simulated habitats differing in spatial auto-correlation in resource distribution (a) andfi vaIues. For a given habitat type, reference windows with the same letter did not provide significantly differences in searching eficiency according to Tukey's posr hoc tests (a = 0.05). Average searching eficiency is displayed in figure 2 for each reference window and habitat type ........................................................... 69

CHAPTER 4

Table 1. Foraging parameters and characteristics used to determine the optimal diet of bison during 6 sampling penods of 1998 in Prince Albert National Park. Acronyms of the plant species considered are agro: Agropyron spp.. caaq: Carex aqiïatilis, caat: C. arherodes, cain: Calamagrostis inexpansa, hoju: Hordeilm jubarum, juba: Juncus balricus and scfe: Scolochloajèsrucacea. Sarnpling periods correspond to (1) 5 January to 16 February. (2) 17 February to 4 April, (3) 23 May to 19 June, (4) 20 June to 12 July. (5) 13 July to 7 August. and (6) 8 August to 3 September ........................................................... 1 14

Table 2. Ivlev's electivity index indicating preference or avoidance in plant species grazed by bison in Prince Albert National Park. Preference for Carex arherodes and Scolochloafesrucacea is suggested by positive indices. Species narne acronyms and dates of the investigation periods are indicated in Table 1 ... 1 16

CHAPTER 5

Table 1. Geornetric mean selection ratio among habitat types of Prince Albert National Park by female bison as detemined frorn GPS-locations during 1996- 1999. Water shore and water interior were pooled in winter. For a given

List of Tubles

season, the use of habitat variables with the same letter did not differ significantly (P < 0.05) ........................................................................................ 157

Table 2. Percent of fixes within 100 for pairs of female bison equipped with GPS-collars in sumrner (white m a ) and winter (gmy am). Pairwise cornparisons were restricted to animals tracked simultaneously. Sample size for each cornparison ranged between 228-1 245 pairs of fixes, with the exception of

............ animal B vs, animal E in winter where sample size was 34 pairs of fixes 158

Table 3. Variables influencing the probability of use of 23 meadows in Prince Albert National Park during the surnmer of ! 997-1998. as determined from stepwise logistic regressions. n i e independent variables considered in the analyses included: area of the meadow, area of water in the meadow, perimeter of water in the meadow, absence of water in the meadow (dicotomic variable). biomass of Agropyron spp.. Carex aquutilis (Biomass of Caaq). Carex acherodes (Biornass of Caat). Calamagrostis inexpnsa. Jirncus. balficus and Scolochloufestucacea (Biomass of Scfe), and percentage of an area of 2-km radius and 1-km radius covered by meadows, water areas and agricultural lands. Final models included only variables significant at P < 0.05 ...................... 159

Table 4. Variables inffuencing the probability of use of 23 meadows in Prince Alben National Park in winter, as determined from stepwise logistic regressions. The independent variables considered in the analyses included: area of the meadow, snow depth. snow density, snow softness. presence of cmsts in the snow column (dicotomic variable). biomass of Agropyron spp.. Curex uqziuarilis (Biomass of Caaq), Curex atherodes (Biomass of Caat), Calamagrostis inexpansa. Juncus balricirs and Scolochloafestircaceu (Biomass of Scfe). and percentage of an area of 2-km radius and 1 -km radius covered by meadows, water areas and agricultural lands. Final models included only variables significant at P < 0.05 ...................................................................... 160

Table 5. Variables influencing the average size of bison herds found in a total of 15 meadows of Prince Albert National Park. independent variables submitted to the stepwise multiple regressions are indicated in Table 2. The final models included only variables significant at P < 0.05 .................................. 161

Table 6. Bison selection of plant types in relation to the relative biomass of other plant types found in 25 meadows of Prince Albert National Park. Acronyms for plant types are Caat: Carex atherodes, Scfe: Scolochloa fèsiucacea and Other: other species altogether. Non-signifiant intercept (b log V) suggests that when the two contrasted plant types had equal availability. there was no selection by bison. Non-signifiant slope (b) suggests fiequency- independent selection arnong the contrasted plant srpes ....................................... 162

List of Figures

LIST OF FIGURES

CHAPTER 1

Figure 1. Relationship between the predicted and observed average net squared displacement, and the number of consecutive inter-crater movements. The average and 90-95% confidence limits for the expected net squared displacement were calculated fiom 1000 simulation of 6-10 average bison searching paths, depending on the number of observed paths for each inter- crater movement ................................................................................................... 43

Figure 2. Relationship between the predicted and observed average net squared displacement and the number of consecutive inter-crater movements for each observed path. The average and 90-95% confidence limits for the expected net squared displacement were calculated fiom 10000 simulations of expected searching paths. Paths for which at least half of their inter-crater movements were outside of the 90 or 95% confidence limits are indicated with * and **,

.............................................................................................................. respectively 44

Figure 3. Average variation in snow and forage characteristics (X + SE) for pairs of successive craters along a searching path and the area separating them (inter-crater). as calculated for al1 sites and only for sites that were not old crater areas covered by fresh snow .......................................................................... 45

CHAPTER 2

Figure 1. Sampling design of the eight adjacent and eight distant quadrats surveyed to determinef-values expected under field conditions. Note that the two most distant of the adjacent quadrats were also considered as distant ones ..... 71

Figure 2. Average expected energy intake rate of the resource found using area-restricted search during 800- 1000 foraging bouts (depending on reference window size). The sirnulated anirnals assessed resource quality using different reference windows in habitats varying in a- andf-values. High values of a reflect hi@ autocorrelation between sampling units, whereas highf-values suggest that the animal has relatively low eficiency in finding similar resources when in an intensive searching mode compare to when in an extensive searching mode. W 4 1 p G represents the searching efficiency of animals assessing resource quality based on the median energy intake rate available in the entire environment (i.e. 195 kJ / min). Wru,vD corresponds to a random

List of Figures

search and provided the same average for al1 values of a, thus only a = 14 was displayed .................................................................................................................. 72

CHAPTER 4

Figure 1. Fraction of plant biomass consumed by bison as a function of their height. (A) The fraction grazed declines significantly with height of three species, whereas (B) there was no relationship for the last four focal species. and their averages are displayed. Note that Carex uqzrarilis and Junctrs balriczrs both averaged 3 1 % of biomass grazed .................................................................. 1 19

Figure 2. Relation between bison intake rate and sward biomass in winter and during the growing season (late-spring and sumrner). Intake rate was calculated from the product of bite rate and bite rnass. which came frorn behavioral observations and from the literature, respectively ................................................. 120

Figure 3. Relation between cropping rate of bison and sward biomass in winter and during the growing season in Prince Albert National Park ............................ 121

Figure 4. Change in short-term (A) and daily (B) energy intake rate with an increase in diet breadth. as predicted h m contingency models for six periods of 1998. Prey species are ranked according to their short-term (kllmin) or daily (MJIday) profitability fiom lefi to right, and sequentially added to the bison's diet (i.e. the diet includes one prey type to seven prey types). A direct decline in intake rate indicates that a specialized diet woutd be optimal. Species name acronyms and dates of the investigation periods are indicated in Table 1 ............ 122

Figure 5. Cornparison between the observed diet of bison in Prince Albert National Park and the optimal diet predicted for short-term and daily intake rate mavimization during six periods of 1998. Species name acronyms and dates of the investigation periods are indicated in Table 1 ................................... 124

CHAPTER 5

Figure 1 . Relative occurrence (total # of individuals for a given species / total # of individuals for al1 species x 100) of bison. moose, white-tailed deer and elk recorded during ground surveys of25 meadows in Prince Albert National Park

.......................................................................................... during 1 997 and 1998 164

tisi of Figures

Figure 2. Relationship between average herd size of bison o b s e ~ e d in 15 meadows of Prince Albert National Park in 1997-1998, and the area (log- transfomed) of these meadows. Regression line is displayed for significant relationships (P < 0.05) ......................................................................................... 165

Figure 3. Selection ratio of different plant types recorded in 12 meadows (Le. n = 12 for al1 plant species) of Prince Albert National Park in 1998. A selection ratio of 1 indicates that resource was used in same proportion than it was available. A broken horizontal line displays this threshold. For a given season, selection of plant types with the sarne letter did no! differ significantly following pairwise cornparison (P < 0.05). Acronyms for plant types are Agro: Agropyron spp., Brci: Bromus ciliatus, Caat: Carex atherodes, Cain: Calamagrostis inexpansa, Ciar: Cirsium arvense, Juba: Juncus balticus, Scfe: Scolochloafes~ucacea, Soar: Sonchus urvensis and Other: other species

............................................................................................................... altogether 166

xii

PROLOGUE

On a Monday of August 1996,I entered one of the nurnerous meadows of Prince

Albert National Park, Saskatchewan. A bison (Bison bison) herd was dispersed

throughout the area. Most individuals were grazing, some were lying down and a few

calves were playing. Approximately 30 minutes later. the herd started moving towards

the forest. where they soon disappeared. Where were they going? Why were they there in

the first place?

Animal distribution and movement patterns in heterogeneous landscapes result

frorn decisions made in the context of the dynarniç interactions between the animal and

its biotic and abiotic environment. These decisions are embedded in a nested hierarchy of

different spatial and temporal scales of reference (Senft et al. 1987). A large herbivore

has to select. for example. a generai geographic range in which to settle for a certain

period of time. Then follows the selection of a general habitat in which to live. Once in

this habitat. the animal has to decide which resource patches are worth exploiting. which

resources to utilize. and how much to consume in each patch before moving on.

Food distribution appears to play a prominent role in shaping animal distribution

(McNaughton 1988, 1990, Krasiiiska and Krasinski 1995, Keane and M O ~ S O ~ 1999) that

can Vary across scales (Bergman 2000, Schaefer and Messier 1995~. Wallace et al.

1995). In this thesis, I am interested in capturing some of the subtle interplay between

plains bison (Bison bison bison) and the plant assemblages of Prince Albert National

Park. My investigation of foraging behaviour cornes through studies of search

Prologue

strategy and diet selection. 1 also investigate whether this interaction is responsible for the

distribution patterns of bison observed at different spatio-temporal scales.

SEARCH STRATEGY

The foraging eficiency of an animal can be heavily influenced by its search

strategy (Focardi et al. 1996). Nonetheless, it is ofien assurned that predators encounter

prey in direct proportion to their availability. and thus that searching is simply a random

process (e.g. Charnov 1976. Fryxell and Doucet 1993). Besides random search, animals

can also use area-restricted search to find prey. During area-restricted search. searching

effort is intensified in the vicinity of areas where suitable resources have k e n previously

found (Cézilly and Benhamou 1996). This behaviour can lead to higher searching

efficiency than random search (Benhamou 1992), because animals can take advantage of

the heterogeneous resource distribution in their environment. The use of environmental

cues (Benhamou 1989) and spatial memory (Edwards et al. 1996. Laca 1998) can further

increase the ability to find suitable resources. But what is a "suitable" resource?

The concept of resource quality holds a pivotal role in both searching and

foraging behaviours because animai decisions are directly based on their perception of

the value of resources (Cowie 1977, Krakauer and Rodriguez-Gironés 1995). The

reference point used to m e s s food quality is presumably dictated by past sampling

experience (Cowie 1977, McNarnara and Houston 1985). A suitable resource would then

simply be one of higher quality (e.g. higher profitability, as measured in Hlmin) than

alternate resources.

Prologue

Although classical foraging models most conunonly assume that animals are

"omniscient" with regard to resource distribution and abundance (Stephens and Krebs

1986). empirical studies have demonstrated that individuals often have incomplete

information about the value of resources (Valone 1991, Alonso et al. 1995, Olsson et al.

1999). Given that the foraging eficiency of omniscient animals should be superior

(McNamara and Houston 1985), it remains unctear why natural selection has not

favoured the use of large amounts of sampling information to assess resource quality for

al1 species. This paradox can be partly explained by the stochastic nature of resource

quality and distribution ofien observed in the environment (McNamara and Houston

1985). Because of this variability, it may be impossible to gather enough information to

behave as an omniscient animal before resource characteristics change. For species

foraging socially, sampling information may indeed be rapidly outdated due to food

depression (Bernstein et al. 1988). The value of sampling infornation during area-

restricted search in heterogeneous landscapes remains, however, largely unexplored.

Foraging behaviours of animals rarely occur at random, even for generalist

herbivores (Schaefer and Messier 19956). Optimal diet theory provides a framework to

generate hypotheses to explain observed foraging patterns. These mathematical models

generally predict a diet that would mavimize a given currency (cg. long-term rate of

energy intake), given various constraints faced by animals (Stephens and Krebs 1986).

Such constraints may be imposed by biotic or abiotic components of the environment. as

Prologue

well as by the animal's ability to utilize resources. Resource distribution and abundance,

the time required to handle a prey type, the possibility of searching while handling prey,

the digestible energy that can be extracted fiom a given prey type, the time required to

digest prey. the maximum amount of resource that can be consumed during a day, the

proportion of time that can be devoted to foraging activity and the state of the animal are

al1 factors that have been used in optimal foraging models (MacArthur and Pianka 1966.

Belovsky 1978. Stephens and Krebs 1986. Fryxell 1991, Illius and Gordon 1991,

Edwards et al. 1994, Farnsworth and Illius 1998, Bergman et al. 2001). The generally

good correspondence between predicted and observed diets suggests that we might have

partly begun to understand the mechanisms underlying dietary choice.

Classical foraging theory assumes that animals maximize their long-term gains

(Stephens and Krebs 1986). The temporal scale over which gain is actuily maximized

has been, however, largely overlooked in both theoretical and empiricai studies (see

Bergman et al. 2001). The optimal diet of herbivores can become temporal scale-

dependent because of imperfect correspondence benveen, for example. the instantaneous

rate of energy intake offered by a prey species and the total amount that can be ingested

during a day. In other words. a prey species that can provide rapid energetic gains may

not necessarily maximize the long-term rate of energy intake because of rapid saturation

of its digestive tract. A forager optimizing long-term gains should instead choose an

alternative prey species that has less saturation effect on digestion.

In addition to temporal scales, spatial scales should be taken into account when

predicting optimal foraging strategies. McNamara et al. (1993) demonstrated that, in

contrast to previous claims (Lucas and Schmid-Hempel 1988). optimal diet should be

Prologue

constant within and among the patches of a heterogeneous landscape. Variation in prey

abundance should only lead to an adjustment in patch residence time, and not to a change

in diet selection. In other words, fiequency-dependent response in diet selection would

require global changes in prey availability at the landscape level, and not only local

variation in abundance. Several previous studies have detected positive frequency-

dependent selection (see Gendron 1987). but these were not directly designed to test

McNamara et al.'s (1993) model. Further studies are needed to darifi this aspect of

foraging decisions.

[NTEGRATION OF MFORMATION ACROSS SPATIO-TEMPORAL SCALES

The success of foraging models to predict dietary choice of individuals does not

necessarily imply that animal distribution can be readily inferred across the landscape

based on food distribution. The tirne available to an animal must be used to satisfy

multiple objectives (Senft et al. 1987). Hence, animal decisions may reflect trade-offs

associated with non-dietary goals that may also be important in fitness mavimization

(Senft et al. 1987. Barton et al. 1992, Mysterud et al. 1999. Conradt et al. 2000). For

exarnple. the potential importance of water in animal distribution has long been

recognized (McHugh 1958). The establishment of water holes has even been used as a

management strategy leading to a more homogeneous use of resources in a landscape

(McHugh 1958. Bailey et al. 1996). The longer the animal has to travel to access water,

the less time is Iefl for foraging, a conflict that can lead directly to changes in dietary

choice (Arnold 1985). Likewise, the need for favourable microclimates, dety fiom

Prologue

predators, and relief fiom insects can influence foraging decisions (Melton et al. 1989,

Barton et al. 1992, Schaefer and Messier 1995a, Conradt et ai. 2000), complicating

predictions h m optirnality principles.

Understanding the effects of spatial scale also requires infonnation on the proper

temporal scale (Wiens 1986). Animals may respond to changes in plant phenology by

changes in diet (Albon and Langvatn 1992). Behavioural responses may also be induced

by seasonal patterns in temperatures. For exarnple, the presence of snow cover in winter

creates variation in travel costs and in water distribution over the landscape. Habitat

selection and resource use may be expected to simultaneously change (Schaefer and

Messier 1995a).

THE BISON OF PRINCE ALBERT NATIONAL PARK

The bison of Prince Albert National Pack, Saskatchewan. constitute good

candidates for studying plant-herbivore interactions and habitat selection by large

grazers. Green et al. (1989) noted the need for additional observations on ke-ranging

bison herds. The bison of Prince Albert National Park are free-ranging, which should

prevent some of the problems that can arise when attempting to relate findings on captive

animals to traits under natural conditions.

Resource distribution in the bison range is heterogeneous at multiple scales.

Discrete meadows are interspersed in forested areas. The meadows include mosaics of

plant comrnunities that differ in species composition and abundance. Water areas are also

unevenly distributed over the landscape. These levels of heterogeneity thus allow the

Prologue

investigation of multi-scale patterns in habitat selection and use.

THESIS OUTLiNE

In this study, 1 use theoretical models to generate predictions that are then related

to field observations of bison in Prince Albert National Park. My thesis is comprised of

three major themes: 1 first investigate the winter searching behaviour of bison, then their

dietary choice. and finally 1 look at bison distribution and resource use at different spatial

and temporal scales.

More specifically. in Chapter 1 , I investigate the winter searching behaviour of

bison. based on the snow Crater paths they leave behind following their foraging activity.

Such "fingerprints" of searching activity allows investigation of whether short-term

samphg experience influences assessment of resource quality. To explain the patterns

observed in the first chapter, 1 create a simple model of area-restricted search that fonns

the basis of Chapter 2. This model shows how searching eficiency changes with the

amount of sampling information considered to assess resource quality during area-

restricted search in heterogeneous environments.

Chapter 3 is preparatory to Chapter 4 because it corrects the existing theory of

optimal diet for large herbivores developed by Farnsworth and Illius (1998). 1 first

provide new rules to determine whether handling or searching tirne lirnits the intake rate

of given prey. As s h o w by Farnsworth and Illius (1998), if more than one prey is

required to make a diet handling-lirnited, the last prey selected should often be consumed

at a rate lower than the encounter rate. 1 provide an equation that can be used to

Prologue

determine the rate at which the last prey should be accepted to make the diet exactly

handling-limited. 1 finally develop an equation to calculate the energy intake rate of a

multi-species diet during handling-limited foraging.

In Chapter 4,I use contingency models to predict the optimal diet maximizing

either the short-term or the daily rate of energy intake by bison. 1 then test which

temporal modeling scale is most consistent with the observed diet of bison, globally. over

their range.

Finally. in chapter 5,1 explore resource selection by bison at different spatio-

temporal scales. First, 1 look at the diflerentid use of different habitat types comprised in

the bison range. 1 then investigate the biotic and abiotic variables that relate most closely

to meadow selection. 1 pursue this by examining which plant species are selected in

meadows. and whether this selection is frequency-independent. The adequacy of

principles that relate to optimal fomging theory to explain animal distribution and

resource use across spatio-temporal scales remains ambiguous because of contrasting

findings (e.g. Serût et al. 1987, Orians and Wittenberger 1991, Bergin 1992. Schaefer and

Messier 1995~. Wallace et al. 1995. Bergman 2000, Ward and Saltz 1994. Mysterud et al.

1999). Hoping to shed some light on this problem. 1 also test whether optimal dietary

principles can explain bison distribution and their grazing patterns across multiple scales.

Albon. S. D. and Langvam. R. 1992. Plant phenology and the benefits of migration in a

temperate ungulate. Oikos 65: 502-5 13.

Alonso. J. C., Alonso, J. A., Bautista, L. M. and Muiloz-Pulido, R, 1995. Patch use in

cranes: a field test of optimal foraging predictions. Animal Behaviour 49: 1367-

1379.

Arnold. G. W. 1985. Regulation of food intake. In: R. J. Hudson and R. G. White, (eds),

Bioenergetics of Wild Herbivores. CRC Press, Inc., Boca Raton, Florida, pp. 8 1-

101.

Bailey, D. W., Gross, J. E., Laca, E. A., Rittenhouse, L. R., Coughenour, M. B., Swift, D.

M.. and Sims. P. L. 1996. Mechanisms that result in large herbivore grazing

distribution patterns. Journal of Range Management 49: 386-400.

Barton, R. A.. Whiten, A. Strurn, S. C., Byme, R. W. and Simpson. A. J. 1992. Habitat

use and resource availability in baboons. Animal Behaviour 43: 83 1-844.

Belovsky, G. E. 1978. Diet optimization in a generalist herbivore: the moose. Theoretical

Population Biology 14: 105- 134.

Belovsky. G. E. 1986. Optimal foraging and community structure: implications for a

guild of generalist grassland herbivores. Oecologia 70: 35-52.

Benhamou, S. 1989. An olfactory orientation model for marnmals' movements in their

home ranges. Journal of Theoretical Biology 139: 379-388.

Benhamou. S. 1992. Eficiency of area-concentrated searching behaviour in a continuous

patchy environment. Journal of Theoretical Biology 159: 67-8 1.

Bergin, T. M. 1992. Habitat selection by the Western kingbird in Western Nebraska: a

hierarchical analysis. Condor 94: 903-91 1.

Bergman. C. M. 2000. Behavioural responses to resource availability by Northem

ungulates. Ph.D. thesis. University of Guelph.

Prologue

Bergman, C. M., Fryxell, J. M., Gates, C. C. and Fortin, D. 2001. Ungulate foraging

strategies: energy maximizing or time minimizing? Journal of Animal Ecology (in

press).

Bernstein. C., Kacelnik, A. and Krebs, J. R. 1988. Individual decisions and the

distribution of predators in a patchy environment. Journal of Animal Ecology 57:

1007- 1026.

CéziIly. F. and Benhamou, S. 1996. Les stratégies optimales d'approvisionnement. Revue

d'Écologie (Terre Vie) 5 1 : 43-86.

Conradt. L.. Clutton-Brock, T. H. and Guinness, F. E. 2000. Sex differences in weather

sensitivity can cause habitat segregation: red deer as an example. Animal

Behavour 59: 1049- 1060.

Cowie. R. J. 1977. Optimal foraging in great tits (Parus major). Nature 268: 137- 139.

Charnov. E. L. 1976. Optimal foraging, the marginal value theorem. Theoretical

Population Biology 9: 129- 136.

Edwards. G. R., Newman, J. A., Parsons, A. J. and Krebs, J. R. 1994. Effects of the scale

and spatid distribution of the food resource and animal state on diet selection: An

example with sheep. Journal of Animal Ecology 63: 816-826.

Edwards. G. R., Newman. J. A.. Parsons, A. J. and Krebs, J. R. 1996. The use of spatial

memory by grazing animals to locate food patches in spatially heterogeneous

environments: an example with sheep. Applied Animal Behaviour Science 50:

147-160.

Focardi. S. Marcellini, P. and Montanaro, P. 1996. Do ungulates exhibit a food density

threshold? A field study of optimal foraging and movement patterns. Journal of

Prologue

Animal Ecology 65: 606-620.

Fortin. D. and Frandsen, D. 1999. Some preliminary results on the ecological hierarchy of

foraging activities in bison. Research Links 7: 13-1 5.

Fryxell. J. M. 199 1. Forage quality and aggregation by large herbivores. Arnerican

Naturalist 13 8: 478-498.

Fryxell, J. M. and Doucet, C. M. 1993. Diet choice and the functional response of

beavers. Ecology 74: 1297- 1306.

Famswcrth. K. D. and Illius, A. W. 1998. Optimal diet choice for large herbivores: an

extended contingency model. Functional Ecology 12:74-8 1.

Gendron. R. P. 1 987. Models and mechanisms of frequency-dependent predation.

American Naturalist 130: 603-623.

Green. W. C. H., Griswold, J. G. and Rothstein, A. 1989. Post-weaning associations

among bison mothers and daughters. Animal Behaviour 38: 847-858.

Illius, A. W. and Gordon, i. J. 1991. Prediction of intake and digestion in ruminants by a

model of rumen kinetics integrating animal size and plant characteristics. Journal

of Agricultural Science 1 16: 145-157.

Keane, J. J. and Momson. M. L. 1999. Temporal variation in resource use by black-

throated Gray Warblers. Condor 101 : 67-75.

Krakauer, D. C. and Rodriguez-Gironés, M. A. 1995. Searching and learning in a random

environment. J o d of Theoretical Biology 177: 41 7429.

Krasinska M. and Krasihski, 2. A. 1995. Composition, group size, and spatial

distribution of European bison bulls in Bialowiea Forest. Acta Theriologica 40:

1-21.

Prologue

Laca. E. A. 1998. Spatial memory and food searching mechanisms of cattle. Journal of

Range Management 5 1 : 370-378.

Lucas. J. and Schmid-Hempel, P. 1988. Diet choice in patches: time-constraint and state-

space solutions. Journal of theoretical Biology 13 1 : 307-332.

MacArthur. R. H. and Pianka. E. R. 1966. On optimal use of a patchy environment.

Amencan Naturalist 100: 603-609.

McHugh, T. 1958. Social behaviour of the Arnerican Buffalo (Bison bison bison). Zwlogica

43: 1-40.

McNaughton. S. J. 1988. Minera1 nutrition and spatial concentrations of Afncan

ungulates. Nature 334: 343-344.

McNaughton, S. J. 1990. Minerai nutrition and seasonal movements of African migratory

ungulates. Nature 345 : 6 13-6 15.

McNarnara, J. M. and Houston, A. 1. 1985. Optimal foraging and learning. Journal of

Theoretical Biology 1 17: 23 1-249.

McNarnara, J, M., Houston. A. 1. and Weisser, W. W. 1993. Combining prey choice and

patch use - what does rate-maximizing predict? Journal of Theoretical Biology

164: 219-238.

Melton. D. A.. Larter. N. C.. Gates. C. C. and Virgl, J. A. 1989. The influence of rut and

environmental factors on the behavior of wood bison. Acta Theriologica 34: 179-

194.

Mystemd, A.. Larsen, P. K., Ims, R. A., and 0stbye E. 1999. Habitat selection by roe

deer and sheep: does habitat ranking refiect resource availability? Canadian

Journal of Zoology 77: 776-783.

Prologue

Newman, J. A., Parsons, A. J., Thornley, J. H. M., Penning, P. D. and Krebs, J. R, 1995.

Optimal diet selection by a generalist grazing herbivore. Functiod Ecology 9: 255-

268.

Olsson, O. Wiktander, U., Holmgren, N. M. A. and Nilsson, S. G. 1999. Gaining

ecological information about Bayesian foragers through their behaviour. II. A

field test with woodpeckers. Oikos 87: 264-276.

Orians. G. H. and Wittenberger, J. F. 1991. Spatial and temporal scales in habitat

selection. American Naturalist 137: S29-S49.

Schaefer, J. A. and Messier, F. 1995a. Habitat selection as a hierarchy: the spatial scales

of winter foraging by muskoxen. Ecography 18: 333-344.

Schaefer. J. A. and Messier. F. 19956. Winter foraging by muskoxen: a hierarchical

apporach to patch residence time and cratering behaviour. Oecologia 104: 39-44.

Srnft. R. L., Coughenour, M. B., Bailey, D. W., Rittenhouse. L. R., Sala. O. E. and Swift.

D. M. 1987. Large herbivore foraging and ecological hierarchies. BioScience 37:

798-799.

Stephens. D. W. and Krebs, J. R. 1986. Foraging theory. Princeton University Press.

Princeton, New Jersey.

Valone. T. J. 1991. Bayesian and prescient assessment: foraging with pre-harvest

information. Animal Behaviour 41 : 569-577.

Wallace. L. L.. Turner, M.G., Romme W. H. O'Neill. R. V. and Wu. Y. 1995. Scale of

heterogeneity of forage production and winter foraging by elk and bison. Landscape

Ecology 10: 75-83.

Ward, D. and Sala D. 1994. Foraging at different spatial scales: Dorcas gazelles

foraging for Mies in the Negev desert. Ecology 75: 48-58.

Wiens. J. A. 1986. Spatial scale and temporal variation in studies of shrub-steppe birds.

In: Diamond, J. and Case, T. J . (eds), Community ecology. Harper and Row. New

York. pp. 154-1 72.

CHAPTER I

Searching Behaviour and Use of Sampling Information

by Free-Ranging Bison

DANIEL FORTIN

Searching behaviour used by predators can have important consequences for

trophic-ievel interactions and can affect the utility of classical foraging theories. 1

examined the searching behaviour of ke-ranging bison (Bison bison), and determined

whether they gsess resource quaiity on the basis of short- or long-term sampling

information acquired during search.

Bison used area-restricted search during their winter foraging activity. Their

movements between areas of suitabte resources were also influenced by local

environmental conditions, k ing sometirnes less sinuous and other times more sinuous

than expected from a correlated random walk model. Bison systematically avoided areas

offering a low expectation of energy intake. These signs of navigation should improve the

searching performance of area-restricted search and increase the minimum amount of

past sarnpling experience that should be considered to assess resource quality in order to

maximize searching eficiency. Nonetheless, the perception of resource quality by bison

varied during a foraging bout. After controlling for resource quality, 1 found that large

feeding craters were preceded by prior samples of poor resource quality.

1 conclude that bison take advantage of the structural characteristics of their

environment during searching activity and base their decisions on local rather than global

availability. contrary to the assumptions of classical optimal foraging theory.

The searching strategy of predators may importantly influence trophic-level

interaction because of its impact on foraging eficiency. Predictions of predator-prey

interactions oAen assume randorn search by predators, considering that prey are

encountered in proportion to their global availability (Owen-Smith and Novellie 1982,

Fryxell and Doucet 1993. Owen-Smith 1993, Farnsworth and Illius 1998). However,

animals can take advantage of spatial auto-correlation in resource distribution. such as

that inherent to most natural environrnents, by using area-restricted search to fmd prey

(Smith 1974. Haskell 1997). These foragers might increase their searching efficiency by

intensifying searching effort near areas where suitable resources have b e n previously

found (Benhamou 1992, Cézilly and Benhamou 1996). Hence. spatial bias in searching

effort should lead to resource encounter deviating from global availability . The expected

long-term benefits in an environment should differ depending on whether area-restricted

search occurs or not.

Searching efficiency can be increased by the use of environmental cues

(Benhamou 1989) and spatial rnemory (Edwards et al. 1996. Laca 1998) to help find

areas of suitable resource. This implies that animals have to be capable of discriminating

suitable from unsuitable cesources (Illius et ai. 1999). The perception of resource quality

by animals can thus have an important influence on area-restricted search patterns

(Krakauer and Rodriguez-Gironés 1995, Chapter 2). This has significant implications for

the application of optimal foraging theory. Most optimality models assume that animals

have perfect global knowledge of resource availability. Hence, foraging decisions should

be based on the long-term expectation for the entire environment (Chamov 1976,

S tephens and Krebs 1986). The good agreement between field observations and

quantitative predictions suggests that some animals do behave as fully-informed

individuals (Cowie 1977. DeBenedictis et al. 1978). in such cases, the optimal reference

point should not change fiom patch to patch (Charnov 1976. McNamara et ai. 1993), and

should thus remain constant over short periods of time. On the other hand. some field

observations suggest that foragers preferentially use recent information to adjust

subsequent foraging behaviour. For exampte, sheep increase selectivity afier having

recently consumed high quality food (Jung and Koong 1985). Additional empirical

studies have found that anirnals consider only limited environmental information. leading

them to overuse poor resource patches and undenise nch ones compared to predictions

based on optimality theory (Valone 1991. Alonso et al. 1995. Olsson et al. 1999). In surn.

it appears that there is variation among species on the use of short-term vs. long-term

sampling information. It is important to determine the amount of environmental

information considered by a species to better understand search behaviour, and to make

more reliable foraging predictions.

The concept of memory window can be used to investigate the amount of past

sampling experience used for resource quality assessment (Cowie 1977). This concept

suggests that. as time passes, new information is added to the animal's memory window

while an equivalent amount of the oldest information is discarded. Rather then

considering the arnount of information gathered during certain periods of time (Cowie

1977. Valone 1992, Mackney and Hughes 1995). here I considered the information

acquired during the sampling of given spatial amounts of habitat (referred to as reference

windows). Reference windows allow the investigation of whether short- or long-term

expectation is being used by the animal to assess resource quality during search activity.

In this study, 1 investigated the winter searching behaviour of fk-ranging bison

(Bison bison). More specifically, 1 examined whether bison find their food using area-

restricted search. whether bison make non-random movements to find areas of suitable

resources. and whether searching individuals decide resource suitability on the bais of

short- or Iong-term sampling. In winter, bison obtain food by digging through the snow

with sweeping motions of the head. This creates successive snow craters, allowing the

detemination of their searching intensity in a given area as well as their sampling

experience prior to each crater.

Searching path characteristics

The field study took place in Prince Albert National Park (PANP. Saskatchewan.

Canada), where approximately 220 free-ranging bison spend most of their time feeding in

grass and sedge meadows interspersed in a boreal forest.

Bison searching activity was studied in the winter of 1998 by recording

information on 10 snow crater paths until(1) the focal bison lefi the meadow. (2) the

bison lied down (i.e. end of the foraging bout). or (3) the pattern was obscured by tracks

of other individuals. Paths were found in six meadows located as far as 20 km apart, and.

from my frequent surveys, 1 estimated that they were studied when iess than 2 d old.

Searching paths were first described by recording the angle and the number of steps

between successive craters, as well as the average distance between steps. Foot print size

and the distance between steps suggested that adult individuals made al1 paths studied.

The average path comprised 15 consecutive cmers (range: 1 1-1 7), covering an area of 3

rn' of disturbed snow (range: 4 -1 7 mZ, n = 203). Thc area of each feeding crater was

estimated assuming the simpiest geometric shape associated with the crater, a rectangle in

most cases. Changes in crater size indicated variation in searching intensity, and thus

reflected changes in animals' perception of resource quality. with larger craters king

perceived as indicating higher quality.

Snoiv characreristics

Snow depth. density and softness were evaluated approximately 30 cm next to

each crater in an area of undisturbed snow. Snow characteristics were averaged from 1-4

locations for each crater, depending on crater area. Measurements were assumed to

reflect conditions at the time the foraging activities occuned. Snow density (@cm3) was

estimated by weighing with a spring scale a sample of the snow column collected usine a

metal tube, and dividing that mass by the volume of snow gathered. Snow sohess was

measured as the sinking depth of a bottle (300 g. 8.5 cm of diameter) dropped 50 cm

above the snow surface (Murray and Boutin 1991).

Chapier 1

Forage characreristics

Biomass

Forage biomass was estimated 30 cm fiom each crater in 0.25 m2 quadrats, on a

0-9 visual scale, where the plant communiîy was similar to that found in the crater. Most

of the individual plants in the craters were generally ungtazed (approximately 70 % from

visual estimation). allowing the corroboration of forage biomass estimates. The 0-9

categorical scale was calibrated by clipping above-ground vegetation in 39 random

quadrats (0.25 ml), and by weighing the sarnples d e r 60 hours of drying at 50°C (Dry

7 379e0.29(visuaI esiimaicl biomass [dm'] = 8,.,, . R' = 0.91).

Dry matter intake rate

The potential forage intake rate was estimated for adult bison from the

relationships between each of cropping rate and bite size with forage biomass. Factors

influencing cropping rate were established fiom behavioural observations conducted

during the winters of 1997 and 1998. Head pulling movements were considered as bites.

which were counted during 5-min observation bouts using a spotting scope. Actively

feeding animals within 100 m (most ofien within 50 m) of an observer were followed for

a maximum of four times per day. During the observation period, ! recorded snow

conditions, forage biomass and species cover in 3-5 quadrats (0.25 m') interspersed over

the foraging area. Average values were used in the subsequent statistical analyses.

Chapter 1

Stepwise multiple regression indicated that cropping rate decreased with increasing

biomass (RI = 0.07, Fi,isl = 13.6, P < 0.001), but was not infiuenced by snow conditions

(P > 0.14 for all snow variables). Consequently, snow conditions were not considered

when quantiQing the expected forage intake rate. The mean cropping rate during each

observation bout was multiplied by the expected bite size predicted by Bergman et al.'s

(2000) nonlinear relationship. The resulting dry matter forage intake rate (g/min)

expected at different biomass was iteratively fit to a Michelis-Menten equation, using

procedure NLIN of SAS (SAS [nstitute 1990). The resulting type i? bctional response

(Forage intake rate = E40.90 biomass] / [252.29 + biomass]) was used to convert the plant

biomass along searching paths to the expected rate of forage intake.

Digest ibiliiy

Dry matter digestibility of forages was determined ftom 481 plant samples of

known species composition collected in 25 meadows during the winter of 1997. Sarnples

of the tissue grazed were taken following the procedure of Hudson and Frank (1987), and

the percent cover of each species present was visually estimated. Samples were dned at

50°C for 60 hours. and their in vitro digestibility was estimated using Tilley and Terry's

(1963) method with cattle rumen fluid. Forage digestibility was then related to species

cover using stepwise multiple regression, with separate analyses for samples of species

growing in dry and wet areas. The regression mode1 for dry area forage explained 36.3 %

of the variation in forage digestibility on the basis of percent cover of six plant species

(F6.7~ = 6.07. P < 0.0001). For wet area forage, eight plant species explained 36.6 % of

the variation in forage digestibility (F8.411 = 29.08, P < 0.0001). These equations based on

cattle rumen fluid were converted to reflect bison digestibility following Bergman (2000).

Field estimates of plant cover were transformed to forage digestibility for bison using

these two corrected regression models.

Energy inrake rate

1 determined the digestible energy content (kJ/g) for a given species tissue from

the product of dry matter digestibility and the gross energy content of 18.5096 kJ/g

(National Research Council 1996). The expected energy intake rate (kJ/min) of forage

encountered in the field was calculated from the product of the digestible energy (kJ/g)

and the expected forage intake rate (g/min).

Forage characteristics were evaluated in 1-4 quadrats for each crater and average

for subsequent analysis. Within-crater measurements varied little because a crater usually

included a single plant assemblage.

Field esrimate of reference windoivs

From the sequence of cratets dong a search path, it was possible to investigate

different reference windows that might be used to m e s s resource quality during search.

Reference windows represent the environmental conditions averaged over a given area of

previous searching activity:

where WL is the expected energy intake rate averaged over the reference window L m' of

past sampiing experience. A is the area of crater i, with i = 1 being the first previous

crater. n is the number of previous craters needed to reach L, and E is the expected energy

intake rate in craters i to n. Using equation ( l ) , 1 estimated W2 to W,2 for each of the 10

searching paths for which 1 had detailed field information.

Data analysis

Correlated random walk betweenjéeding craters

The net squared displacement ( R I ) was calculated for each of the paths. and

cornpared to the R' predicted from a correlated random walk as calculated fiom equation

2 of Kareiva and Shigesada (1983). A path is comprised of several sequential inter-crater

movernents (S). and expected and observed R' were calculated for each path for up to 15

S (the maximum recorded in the field for most paths). A bootstrapped test of significance

was used to determine whether the difference between the observed and expected squared

displacements was statistically significant. 1 first generated a distribution of inter-crater

angles (0) and distances (0 based on the 10 observed bison paths. The expected R' was

calculated using cos 8 and 1 averaged over 15 values (most of the observed paths were

cornposed of 15 S) randomly taken fiom their respective distributions. For each S. mean

and confidence intervals were estimated fiom 10000 bootstrap simulations of Ï? . Any

observed 2' failing outside of this interval was considered significantly different than

expected from a correlated randorn walk. 1 also tested whether the average of R' for al1

10 paths was significantly different from that expected fiom a correlated random walk.

90-95% confidence limits were calculated fiom one thousand averages made of 6-10 3'

values that were randomly selected frorn the 10000 R' simulated. The number of Ï?

values retlected the number of observed paths with that total number of S (cf. Turchin

1998). More specifically, al1 10 paths were cornprised of 10 S, 9 paths had 1 1 S. 8 had 14

S and 6 were cornprised of 1 5 S.

Inter-crarer resortrce qliality

Snow and forage characteristics comprising the area ktween two successive

feeding craters were determined from a set of measurements done in a quadrat (0.25 m2)

half way along the path connecting the two craters. For each path. an average of 1 1 inter-

crater areas were recurded (range: 5-16). The difference in resource characteristics

between inter-crater areas and the two adjacent craters was compared using 2-way

repeated-measures ANOVAs for each environmental condition. with individual path and

sequence of visited areas (first crater, inter-crater area and second crater) as factors.

Pairs of craiers with similar resources

I tested whether bison considered their long- or short-term sarnpling experience to

25

Chaprer I

evaluate resource quality. The long-term expectation of plant quality should not change

from patch to patch over short time pends. In contrast, animals that use short-term

information to evaluate resource quality should display changing perception depending

on recent patterns of experience.

To test this prediction, 1 identified pairs of craters that offered identical expected

rates of energy intake (i.e. sarne plant cornmunity and same visual estimate of plant

biomass). When more than two craters had identical rates of gain. al1 possible pairs were

formed. 1 then performed a 3-way repeated-measures ANOVA with path, reference

window and crater size as factors. As a p s t hoc test. 2-way repeated-measures ANOVAs

(with path and crater size as factors) were done for each of the reference windows.

Signs of area-restricred search

The discrete nature and the various sizes of snow craters obsented along searching

paths provide evidence that area-restricted search was used by bison in winter. The size

of the smallest craters dong paths suggests that bison had to disturb at least 0.25 m' of

snow from head sweeping motions to sample the forage available. The large size of most

craters (3 m' on average) implies that bison included several adjacent "sarnpling units".

and. hence. that an intensive searching mode was used in atternpt to find similar resource

types. The segregation of snow craters indicates that an extensive searching mode was

used such that areas of poor qudity reçource were avoided.

Correlated random walk betweenfeeding craters

Movement between discrete snow craters did not occur randomly for al1 paths.

The Ï? averaged over al1 bison paths tended to be higher than expected from the

prediction of a correlated random waik (Fig. 1). Half of the inter-crater movements

differed significantly fiom what could be expected for a correlated random walk. under

95% confidence limits. This trend was largely influenced by path 5 (Fig. 2). Indeed. there

was little consistency among individual paths. Three of the 10 paths were comprised of

inter-crater movements that differed significantly from what could be expected for a

correlated random walk over half the time (under 95% confidence limits). Two of those

paths were over-estimated and one was under-estimated by a correlated random walk.

Under 90% confidence limits, 5 of the 10 path had at least half their inter-crater

movements differing from the correlated random walk's predictions, with 3 paths k ing

over-estimated and 2 under-estimated. These results provide evidence that bison

rnovement were flexible. k ing sometimes random and othet times less sinuous or more

sinuous than expected randomly.

Inter-crater resource quality

Other evidence of non-randomness during extensive searching mode is given by

my comparison of adjacent cratered areas. Bison were able to avoid poor feeding areas by

adjusting their searching path to the local conditions (Fig. 3). For ail sites. snow softness

associated with successive snow craters did not differ fiom measurements taken in the

space between them (ANOVA: Frsa = 0.9 1, P = 0.4 I ), however, differences were

observed in snow depth (ANOVA: Fz,% = 3.78, P = 0.03) and density (ANOVA: F2.96 =

3.48. P = 0.03). Snow was shallower (ANOVA, Post hoc test: Fi.97 > 4.45, P < 0.04 for

both comparisons, i.e. previous crater vs. inter-crater and following crater vs. inter-crater)

and denser in the area between craters (ANOVA, Post hoc test: F1.97 > 4.89. P < 0.03 for

both comparisons). For al1 sites (Fig. 3, ail sites). successive snow craters were separated

by an area with lower plant biomass than that associated with either crater (ANOVA.

Post hoc test: > 23.70. P < 0.0001 for both comparisons). Forage digestibility did not

vas, among sites (ANOVA: F2.96 = 1.21. P = 0.30). but the difference in biomass skould

lead to lower gains of energy between craters than within craters (ANOVA. Post hoc test:

> 28.83. P < 0.0001 for both comparisons). These differences could simply reflect

prior foraging. Indeed, some of the areas between craters had been used before they were

covered by fresh snow. 1 therefore extended the analysis by excluding craters separated

by previous foraging experience.

When old cratered areas were excluded (Fig. 3. see omitting old crater areas), no

significant difference was found in snow sohess (ANOVA: Fzsr = 0.29. P = 0.75). depth

(ANOVA: Frsz = 0.62. P = 0.54) or density (ANOVA: F2.8- = 1 .O 1. P = 0.37) between the

first craters. inter-crater areas and second craters. The same trends remained. however.

for forage characteristics. Biomass was still lower between successive craters than within

them (ANOVA. Post hoc test: Fim3 > 19.17. P < 0.0001 for both comparisons). and

therefore. despite the absence of irnpottant differences in forage digestibility (ANOVA:

F2.g2 = 2.27, P = 0. I I ) the potential energy gain was lower between craters than within

Chapter I

craters (ANOVA, Posr hoc test: Fi,s3 > 17.95, P < 0.0001 for both comparisons). This

suggests that the distance between craters was neither random nor simply dictated by

obvious visual cues to past foraging activity.

Pairs ofcraiers wirh similar resources

Bison searching behaviour was further influenced by their shon-term past

sampling experience. 1 found that resource encountered shortly before arriving at an area

influenced crater size (ANOVA: F1.176 = 5.90. P = 0.03), and thus searching intensity.

This general trend was largely influenced by W6 to W!a and particularly by W6 (Table 1).

H f I ? was not significant, but it could be sirnply caused by reduced power to detect

ditferences. due to reduced sample size (Table 1). For the significant teference windows.

the larger crater of a pair having similar resource quality was the one preceded by poorer

resource quality. For these reference windows. the trend was similar among paths

(ANOVA, interaction beiween path and crater size: P = 0.36).

It was not surprising that bison appeared responsive to more than one reference

window because larger windows include dl the information of smaller ones. My analysis

indicates that short-term environmental information iduenced the perception of resource

quality and that the reference window should be rather small. It was not possible.

however, to statistically infer which reference window was preferentially used by free-

I ranging bison.

The observed difference in crater six pairs was not iduenced by the distance

covered before amving to these craters (ANOVA: = 0.14, P = 0.71). In other words.

Chapter I

a longer travel distance did not precede the larger Crater of the pair, as it could have been

predicted.

Free-ranging bison in Prince Albert National Park used area-restricted search

during their winter foraging activity. Their searching strategy suggests that, like

European thrushes (Turdus merula and T. philomelos, Smith 1974) and ferrets (Mus~ela

putorilrsjrro, Haskell 19971, bison take advantage of spatial auto-correlation in resource

distribution by adjusting their movements to habitat structure to find suitable food items.

Nuvigc~rion h r i n g seorching activiiy

Bovet and Benhamou (1 99 1 ) demonstrated that the efficiency of area-restricted

search can be increased by adjusting the search path to local conditions. Departure from a

correlated random walk is a sign of non-random movement in the landscape (Root and

Kareiva 1984), and can be used to investigate whether environmental cues influence

searching activity. For bison. extensive searching mode during area-restricted search (i.e.

when moving between craters) appeared flexible to local environmental conditions. k ing

sometimes less sinuous or sometimes more sinuous than random expectations. Fine-

tuning of path sinuosity according to local resource distribution has similarly been

observed in ferrets (Haskell 1997).

Bison adjustment of search path appeared partly dictated by the spatial

distribution of poor resources. Bison systematically avoided low quality resource when in

an extensive searching mode. Individuah were able to avoid areas where they would have

experience low energy intake by simply walking on without even sampling the resource

by digging through the snow. Spatial memory can be used to help avoid or locate food

patches in heterogeneous landscapes (Gillingham and Bunnell 1989. Edwards et al. 1996.

Dumont and Petit 1998, Laca 1998). Large herbivores can remember the locations and

relative availability of food for at least 20 days (Bailey et al. 1996). Many cues can also

be used to navigate in the environment. The seed heads and stems of some plant species

remain above the snow cover for most of the Saskatchewan winter. offering a visual

indication of the resource lying underneath the snow (Fortin and Frandsen 1999).

Animals can also adjust their searching path to olfactive cues (Benhamou 1989).

Reindeer could distinguish good and poor lichen sources by smell though 91 cm of sol?

snow (Helle 1984). The use of such environmental cues would not only increase the

searching eficiency, but it would also tend to increase the minimum arnount of sampling

information required to optimize searching activity (Chapter 2). Sarnpling information

becomes more valuable because of the increased predictability of resource distribution.

The optimal amount of sampling information should be, in most cases, less than the area

that bison normally visited during a foraging bout (Chapter 2). Such rather small optimal

reference windows could perhaps help explain the finding that the perception of resource

quality by bison was influenced by short-term sampling information.

The use of resource sampling information

1 found that bison searched areas of similar =source quality with different

intensities dependent on their short-term past sampling experience. This variation in

perception of resource quality was not influenced by the distance between craters. This

observation contrasts with those of Cuthill et al. (1990. 1994) on starlings (Slurnus

wlguris) and Todd and Kacelnik (1993) on pigeons (Columba iivia), who found that the

last travel distance influenced patch exploitation. The perception of a given resource type

by bison was simply responsive to the sequence of the resource characteristics

encountered during a foraging bout, a sign of incomplete knowledge of resource

distribution. Animals with incomplete information are ofien categorized into prescient

and Bayesian animals depending on the information they possess (Valone 1991).

Prescient fongers have sensory information or memory of patches that are spatially and

ternporaliy predictable in quality. As omniscient animals, they have unbiased estimates of

patch quality. Bayesian animals possess incomplete information of patch quality. which

leads them to overuse poor quality patches and underuse rich patches compared to

utilisation by omniscient foragers. The use of infomation during the foraging activity of

Inca doves (Colrïmbina inca) appears flexible to the predictability of patch quality: they

display presient behaviour when patch quality is temporally predictable and Bayesian

behaviour when unpredictable (Valone 199 1). Signs of Bayesian behaviour have also

been reported in other animal species. such as conunon cranes (Grus grus. Alonso et al.

1995) and lesser spotted woodpeckers (Dendrocopos minor. Olsson et al. 1999). Bison

appeared to be learning about resource quality distribution as they search for food, thus

displaying at least some characteristics associated with Bayesian animals.

The obsewed plasticity in perception of resource quality by bison during a

searching bout suggests quality assessment through reference windows of fairly small

size. Similarly, bumblebees (Bombus apposifus, Delphinium nelsonii) were found to use

an intermediate arnount of habitat sampling information to adjust their foraging activity

(Pleasants 1989). These observations contrast with the assumption of most optimality

models. The marginal value theorem and the contingency mode1 of optimal diet assume

that animals assess resource (or patch) quality relative to the long-term expectation for

the entire environrnent (Charnov 1976, Stephens and Krebs 1986, McNamara et al.

1993). hlany factors could cause divergence fiom this theoretical assumption. According

to McNamara and Houston (1985), the tirne required to leam the long-term benefits an

environrnent offers depends on the characteristics of the habitat. Such omniscience might

not be attainable in environrnents offering highly variable resource. The benefits of area-

restricted search increases in a decelerating fashion with the expansion of reference

windows and rapidly levels off (Chapter 2). Because of the costs involved with storing

environmental information (Dall and Cuthill 1997), the optimal reference window should

as small as possible. and thus. at whatever size that plateau begins.

Most foraging theories assume that animals base expectations on long-term

benefits. in assessing local resource quality (Olsson and Iiolmgren 1999). According to

Begm et al. (1996), the main source of imperfection of foraging models would be the

transgression of the assurnption of omniscience. Depending on factors such as resource

distribution and the consumer's foraging efficiency. violation of this assumption can lead

to divergence in predictions (Bernstein et al. 1989, 1991). Theories based on ecologically

unrealistic assurnptions obviously have limited applicability in understanding real

ecosystem dynamics (Bernstein et al. 199 1). My finding that bison use area-restricted

search during which they display signs of incomplete information, emphasizes the

importance of considering the searching behaviour of predators together with their use of

environmental information when studying predator-prey interaction.

The hnding of this study was provided by Prince Albert National Park,

University of Guelph. and scholorships fiom FCAR and OGS to DF. 1 would like to

thank Régis Pilote for his help in the field, Mark Andruskiw for assisting in the

laboratory work, and Park Wardens Lloyd O'Brodovich and Dan Frandsen for their

logistical support. 1 also thank Bart Nolet, Tom Nudds. Carey Bergman and Alastair

Wilson for their valuable comments on an early version of this paper.

Alonso. J. C., Alonso. J. A., Bautista, L. M. and Mufioz-Pulido, R. 1995. Patch use in

cranes: a field test of optimal foraging predictions. Animal Behaviour 49: 1367-

1379.

Bailey. D. W., Gross. J. E., Laça, E. A., Rittenhouse, L. R.. Coughenour, M. B.. Swift, D.

M. and Sims, P. L. 1996. Mechanisms that result in large herbivore grazing

distribution patterns. Journal of Range Management 49: 386-400.

Begon, M, Harper, J. L. and Townsend, C. R. 1996. Ecology individuals, Populations and

communities. Third edition. Blackwell Scientific Ltd, Cambridge, Massachusetts.

Benhamou, S. 1989. An olfactory orientation mode1 for mammals' movements in their

home ranges. Journal of Theoretical Biology 139: 379-388.

Benhamou, S. 1992. Efficiency of area-concentrated searching behaviour in a continuous

patchy environrnent. Journal of Theoretical Biology 159: 67-8 1.

Bemstein,C.. Kacelnik, A. and Krebs, J. R. 1988. Individual decisions and the

distribution of predators in a patchy environment. Journal of Animal Ecology 57:

1 007- 1026.

Bernstein.C.. Kacelnik, A. and Krebs, J. R. 1991. Individus! decisions and the

distribution of predators in a patchy environment. II. The influence of travel costs

and the structure of the environrnent. Journal of Animal Ecology 60: 205-225.

Bergman. C. M. 2000. Behavioural responses to resource availability by Northem

ungulates. Ph.D. thesis. University of Guelph.

Bergman. C. M., Fryxell. J. M. and Gates. C. C. 2000. The effect of tissue complexity

and sward height on the functional response of Wood Bison. Functional Ecology

14: 61-69.

Bovet, P. and Benharnou, S. 1991. Optimal sinuosity in central place foraging

movements. Animal Behaviow 42: 57-62.

Cézilly. F. and Benhamou, S. 1996. Les stratégies optimales d'approvisionnement. Revue

d'Écologie (Terre Vie) 5 1 : 43-86,

Charnov, E. L. 1976. Optimal foraging, the marginal value theorem. Theoreticai

Population Biology 9: 129-136.

Cowie, R. J. 1977. Optimal foraging in great tits (Parus major). Nature 268: 137-139.

Cuthill. 1. C., Kacelnik, A. Krebs, S. R., Haccou, P. and lwasa, Y. 1990. Starlings

exploiting patches: the effect of recent experience on foraging decisions. Animal

Behaviour 40: 625-640.

Cuthill, 1. C., Haccou, P. and Kacelnik, A. 1994. Starlings (Sturntrs vulgaris) exploiting

patches: response to long-tenn changes in travel time. Behavioural Ecology 5: 8 1 -

90.

Dall. S. R. X. and Cuthill, 1. C. 1997. The information costs of generalism. Oikos 80:

197-202.

DeBenedictis, P. A.. Gill, F. B.. Hainsworth. F. R., Pyke, G. and Wolf, L. L. 1978.

Optimal meal size in hummingbirds. American Naturalist 1 12: 30 1-3 16.

Dumont. B. and Petit, M. 1998. Spatial memory of sheep at Pasture. Applied Animal

Behaviour Science 60: 43-53.

Edwards. G. R.. Newman, J. A., Parsons. A. J. and Krebs. J. R. 1996. The use of spatial

memory by grazing animals to Iocate food patches in spatially heterogeneous

environments: an example with sheep. Applied Animal Behaviour Science 50:

147- 160.

Farnsworth, K. D. and Illius, A. W. 1998. Optimal diet choice for large herbivores: an

extended contingency model. Functiod Ecology 12: 74-8 1.

Fortin. D. and Frandsen, D. 1999. Some ptelirninary results on the ecological hierarchy of

foraging activities in bison. Research Links 7: 13-15.

F y e l l . J. M. and Doucet, C. M. 1993. Diet choice and the functional response of

beavers. Ecology 74: 1297- 1306.

Gillingham. M. P. and Bunnell, F. L. 1989. Effects of leaming on food selection and

searching behaviour of deer. Canadian Journal of Zoology 67: 24-32.

Haskell. D. G. 1997. Experiments and a model examining leaming in the area-restricted

search behaviour of ferrets (Mustela putorius furo). Behavioural Ecology 8: 448-

455.

Helle, T. 1984. Foraging behavior of the semi-domestic reindeer (Rang* tarandus) in

relation to snow in Finnish Lapland. Annales Universitatis Turkuensis Series A II

Biologica-Geographica-Geologica 68: 35-48.

Hudson. R. J. and Frank. S. 1987. Foraging ecology of bison in aspen boreal habitats.

Journal of Range Management 40: 71-75.

Illius. A. W.. Gordon, 1. J., Elston, D. A. and Milne, J. D. 1999. Diet selection in goats: a

test of intake-rate maximization. Ecology 80: 1008- 10 18.

Jung. H. G. and Koong, L. J. 1985. Effects of hunger satiation on diet quality by grazing

sheep. Journal of Range Management 38: 302-305.

Kareiva. P. M. and Shigesada, N. 1983. Analyzing insect movement as a correlated

random walk. Oecologia 56: 234-238.

Krakauer. D. C. and Rodnguez-Gironés, M. A. 1995. Searching and learning in a random

environment. Journal of Theoretical Biology 177: 4 17-429.

Laca E. A. 1998. Spatial memory and food searching mechanisms of cattle. Journal of

Range Management 5 1 : 3 70-3 78.

Littell, R. C.. Freund, R. J. and Spector, P. C. 1991. SAS" System for linear models.

Third edition. SAS bstitute Inc, Cary, North Carolina.

Mackney, P. A. and Hughes, R. N. 1995. Foraging behaviour and memory window in

sticklebacks. Behaviour t 32: 1241 -1253.

McNamara, J. M. and Houston, A. 1. 1985. Optimal foraging and learning. Journal of

Theoretical Biology 1 17: 23 1-249.

McNarnara. J.M.. Houston, A.!. and Weisser, W.W. 1993. Combining prey choice and

patch use - what does rate-maximizing predict? Journal of Theoretical Biology

164: 2 19-238.

Murray. D. L. and Boutin, S. 1991. The idluence of snow on lynx and coyote movements:

does rnorphology affect behaviour? Oecologia 88: 463-469.

National Research Council {subcommittee on beef cattle nutrition). 1996. Nutrient

requirements of beef cattle. National Academy Press, Washington.

Olsson. 0. and Holmgren. N. M. A. 1999. Gaining ecological information about

Bayesian foragers through their behaviour. 1. Models with predictions. Oikos 87:

25 1-263.

Olsson. O. Wiktander. U., Holmgren, N. M. A. and Nilsson, S. G. 1999. Gaining

ecological information about Bayesian foragers through their behaviour. 11. A

field test with woodpeckers. Oikos 87: 264-276.

Owen-Smith. N. 1993. Evaluating optimal diet models for an Afncan browsing ruminant.

the kudu: how constraining are the assumed constraints? Evolutionary Ecology 7:

499-524.

Chapter 1

Owen-Smith, N and Novellie, P. 1982. What should a clever ungulate eat? American

Naturalist 1 19: 15 1-1 78.

Pleasants, J. M. 1989. Optimal foraging by nectarivores: a test of the marginal-value

theorem. Arnerican Naturalist 134: 5 1-7 1.

Root, R. B. and Kareiva, P. M. 1984. The search for resources by cabbqe butterflies

(Pieris rapae): ecological consequences and adaptive significance of markovian

movements in a patchy environment. Ecology 65: 147-165.

SAS Institute. 1990. SA SIS TA^ User's Guide. Version 6. Fowth edition. SAS Institute

Inc, Cary. North Carolina.

Smith. J. N. M. 1974. The food searching behaviow of two European thrushes. 11. The

adaptiveness of the search patterns. Behaviow 49: 1-6 1.

Stephens, D. W. and Krebs, J. R. 1986. Foraging theory. Princeton University Press,

Princeton, New Jersey.

Tilley. J. M. A. and Terry, R. A. 1963. A two-stage technique for the in vitro digestion of

forage crops. Journal of the British Grassland Society 18: 104- 1 1 1.

Todd. 1. A. and Kacelnik, A. 1993. Psychological mechanisms and the Marginal Value

Theorem: dynarnics of scalar memory for travel time. Animal Behaviour 46: 765-

775.

Turchin. P. 1998. Quantitative analysis of movement: measuring and modeling

population redistribution in animals and plants. Sinauer Associates, Sunderland,

Massachusetts.

Valone. T. J. 199 1. Bayesian and prescient assessment: foraging with pre-harvest

information. Animal Behaviour 41: 569-577.

Valone, T. J. 1992. Patch estimation via memory windows and the effect of travel time.

Journal of Theoretical Biology 157: 243-25 1 .

Chapter 1

Table 1 . Summary of 2-way ANOVAs (factors: individual paths and crater size) testing

whether the difference in size of pairs of snow crater comprised of similar resources was

not systematically influenced by the resource quality as perceived using various reference

windo ws

Re ference F P Difference in N windows (rn') crater area (m2)

2 2.00 0.1 7 1.8 46 4 1.12 0.30 1.8 44 6 5.09 0.03 2.0 3 7 8 2.89 0.10 2.0 27 1 O 3.65 0.08 2.1 2 1 12 0.15 0.71 2.5 16

Figure 1. Relationship between the predicted and observed average net squared

displacement. and the number of consecutive inter-crater movements. The average and

90-95% confidence limits for the expected net squared displacement were calculated

from 1000 simulation of 6-1 0 average bison searching paths. depending on the number of

observed paths for each inter-crater movement.

Figure 2. Relationship between the predicted and observed average net squared

displacement and the number of consecutive inter-crater movements for each observed

path. The average and 90-95% confidence limits for the expected net squared

displacement were calculated from 10000 simulations of expected searching paths. Paths

for which at least half of their inter-crater movements were outside of the 90 or 95%

confidence limits are indicated with * and **. respectively.

Figure 3. Average variation in snow and forage characteristics (X + SE) for pain of

successive craters along a searching path and the area separating hem (inter-crater). as

calculated for al1 sites and only for sites that were not old Crater areas covered by fiesh

snow.

Chapter 1: Figure 1

O 2 4 6 8 :O 12 14 16

Inter-crater movements

Chapter 1: Figure 2

3000 . Paul 4'

2500 .

2000 . 1500 . R'

Path 5 *' W h 6

Expided diplacement Ohenred dimplacsmsnt - 95% confidence limt

500 . 90°h confidence limt

Inter-crater movements

Chupier 1: Figure 3

Snow characteristics

= 12.0

Forage characteristics

700 - first crater

h - inter-crater

N- - second crater

~ 1 1 Old crater areas ornnted

sites Old crater areas omitted

CHAPTER 2

Optimal Searching Behaviour: the Importance of Habitat Sampling

DANIEL FORTiN

Chaprer 2

Predictions of resource use theories largely depend on animal searching

efficiency, which in turn depends on the assumed searching strategy. 1 investigated how

efficiency of area-restricted search is infiuenced by the reference window size used by

large grazers, i.e. the extent of past sampling experience considered to assess resource

quality.

Searching eftlciency increased in a decelerating fashion as the reference window

increased. and leveled off sooner with p a t e r environmental stochasticity. An increase in

the ability of finding similar resource types when in an intensive searching mode (ISM)

and different resources when in an extensive searching mode (ESM) led to higher

searching efficiency, and to a plateau starting with larger reference windows. In most

cases. the increase in searching eficiency leveled off with the consideration of more than

110 units sampling units, representing only 30 rn' of p s t experience for bison (Bison

bison). The assurnption of random search when foragers use area-restricted search can

lead to biases from conventional foraging theories particularly in environments with high

spatial auto-correlation and for animals highly efficient in switching between ESM and

ISM.

My study emphasizes that reliable predictions about resource use theories require

considering the interaction between searching behaviour and habitat structure. together

uith the reference window used by the animal to assess resource quaiity.

Searching eficiency is an integral part of most foraging models. According to the

marginal value theorem (Chamov 1976, see dso Arditi and Dacorogna 1988), resource

consumption rate is maximized by leaving a patch when its prey density reaches a

threshold that reflects the long-term expected intake rate in the whole environment.

Calculation of this marginal value requires estimation of the forager's searching

efficiency (Olsson and Holmgren 1999). Similady, the long-term energy gain predicted

from a contingency model of optimal diet requires an estimate of the prey encounter rate

when searching (Stephens and Krebs 1986). Although "predictions of optimal foraging

models depend critically upon the assurned search strategy" (Focardi et al. 1996)' it is

often assumed that prey are simply encountered in proportion to their availability. and

thus that search occur randomly (e.g. Chamov 1976, Fryxell and Doucet 1993). For

example. the searching component of Owen-Smith and Novellie's (1982) contingency

model included the effective high reach of the animal. the effective path width scanned

and the walking rate while searching. The product of these three constants and the density

of the different prey species present in the landscape provides an estimate of the long-

term encounter rate of prey (in grams per second). In this classic form of the contingency

model. the nutrient intake rate of their "clever ungulate" was calculated assurning a

random search pattern. While such an assumption might be appropriate under some

circumstances, resource types in natural environments are often auto-correlated in space.

Animals might thus have some expectation of the prey species and density that may be

found locally. permitting the possibility that they adjust searching behaviour in response

to habitat structure.

Animals commody respond to resource clumping by using area-restricted search

to find their prey (Smith 1974, Haskell 1997). They adopt an intensive searching mode

(ISM) at the vicinity of successful locations by increasing their path sinuosity and

reducing traveling speed (Cézilly and Benhamou 1996). Subsequent resource depletion

triggers the adoption of an extensive searching mode (ESM), characterized by a searching

path of low sinuosity and high velocity (Cézilly and Benhamou 1996). This tactic can

increase searching eficiency (Benhamou 1992).

A general assumption of area-restricted search models is that prey are identical

and Vary only in abundance (Andersson 1978, Getty and Pulliam 1993, Benhamou 1994,

Haskell 1997). Howeve- a more general form of area-restricted search considers that

path sinuosity and speed is also adjusted to prey quality (Krakauer and Rodriguez-

Gironés 1995, Walsh 1996). The animal's perception of quality becomes a critical

element of area-restricted search because it directly dictates the switch between [SM and

ESM.

Based on past experience, animals expect a certain quality of resource, which then

influences their perception of what a "high quality" prey is. For example, sheep increase

selectivity d e r having recently consurned hi& quality food (Jung and Koong 1985).

Animals could even gain sufficient information to behave as omniscient foragers

(McNarnara and Houston 1985), Le. as if they had complete information about the

profitability and encounter rates of al1 prey found in their environment (Cowie 1977,

DeBenedictis et al. 1978). The influence that the extent of past sampling experience has

on searching efficiency remains largely unexpiored.

Studying area-restricted search of selective animals foraging on several ptey

species requires an adequate appraisal of the animal's perception of resource quality.

Leaming about habitat quality is usually incorporated into foraging models with two

approaches. The first uses a temporal weighting rule of the prey encountered, with more

weight given to the most recent experience (McNarnara and Houston 1987, Devenport

and Devenport 1993). The second approach considers memory windows for which equal

weight is given to al1 past foraging events that occurred during a fixed penod of time

(Cowie 1977. Valone 1992). As new information is gathered. an equal amount of the

oldest information is discarded. Instead of using the information gathered over a given

period of time, here 1 consider information collected over given areas of habitat sampled

(hereafter referred to as reference windows).

A simple area-restricted searc h mode1 based on the behaviour of large grazers and

on their habitat features was used to 1) investigate how searching eficiency changes with

reference window size, and 2) demonstrate how the long-term expectation assigned to a

simulated omniscient animal would influence its searching performance. The latter

objective should provide an indication of the potential biases that could be expected in

various types of habitat from the predictions of classical theories.

The mode1

Distribution of vegetation is often nearly continuous, with the expected digestible

energy intake rate of grazers utilizing such a resource varying between and among plant

species due to variation in quality and abundance. Resource availability in natural

environments is thus expected to follow a more or less smooth gradient of quality that

could resemble. for example, a fiactal landscape (Tyler and Hargrove 1997). My intent

was not, however, to directly model an animal moving in a simulated landscape, but

rather to model the consequences for searching efficiency of considering various

reference windows before deciding adopt a given searchiny mode (ESM or [SM) in

landscapes with auto-correlated resowce distribution.

The spatial auto-correlation in resource types observed in natural habitats makes

animals likely to encounter successive prey similar to each other. Accordingly, my model

considered the quality of a given sampling unit as a hc t ion of the previous one.

where ES is the expected energy intake rate (kJ / min) of resource for the sampling unit S,

6 is a random value taken fiom the normal distribution

for which the average is O and the standard deviation a.

Compared to ESM, ISM should lead to a higher likelihood of finding similar

successive prey. For example, fiefd measurernents for bison (Bison bison) suggest that the

difference between adjacent sampling units could be on average approximately twice as

variable when in an ESM than in an ISM (see Field esfimution of model parameters).

This increase in spatial autoçorrelation is reflected in my model by mdtiplying Svalues

by f (where f < 1) whenever the animal uses an ISM to tind the next sampling unit:

A habitat offers a restricteci range of prey types, and thus Es-, is bounded between

E, t f l .~ and where E,J~,,v is the lowest and E,tus the highest available E of any prey

types that can be found where the feeding bout takes place. To respect this condition.

whenever Es-, < E , i f , , ~ Es-, becomes equal to E.v,,v + 14. Similady, whenever Es./ >

E,\,..i.y. Es.! becornes E.I~.~;L\. - ]a f: In this latter case. f is used because when the animal is

offered a sampling unit greater than E J , ~ an ISM would be u d .

Following each sampling unit, the animal exhibits an ISM to search for the next

prey item when the resource is perceiveci as k i n g of'good quality. and thus whenever:

where W (kl I min) represents a reference window for which the expected energy intake

rate of the prey encountered over the last L unit of area (ml) is avenged. Omniscient

animals are assumed to use a fixed reference point to assess resource quality that is based

on the whoIe enviroment (McNamara et al. 1993). I thus dso investigated the searching

performance that would result if the animal used a constant reference window: the

average energy intake rate for the resource found over the entire home range (w41.G).

Chapter 2

When the current resource is of poorer quality han the reference window value (Le. when

equation 3 is false). the resource is perceived as poor quality. and the animal switches to

an ESM until the next sampling unit (thus equation 1 applies).

Field estimation of mode1 parameters

To ensure the use of parameter values that reflect natural situations, simulations

were pararneterized based on 10 searching paths recorded during winter 1998 together

with general information on the bison habitat of Prince Albert National Park (PANP),

Saskatchewan (see Chapter 1). Bison are grazers throughout the year. and in winter they

access vegetation by digging through the snow wiîh sweeping motions of the head. The

switch between [SM and ESM is directly reflected by the difference in size and the

discrete nature of the craters of disturbed snow found in meadows (Chapter 1).

For all simulations. I fixed the sampling unit at 0.25 m2. because this Crater size

was arnong the smallest obsemed in the field (n = 152). A simulated foraging bout

included 350 sample units, and thus coveced 87.5 m'. close to the largest area of

disturbed snow recorded for a foraging bout. The exact area covered during a feeding

bout was not critical here because I simply needed an area ttiat was large enough to

minimize the variance in searching efficiency between foraging bout replicates.

Simulations were performed with differentjlvalues to reflect variation in the

efficiency of switching h m an ISM to an ESM. Potentialf-values were estimated in the

field a s the variation in the degree of similarity in prey profitability between adjacent

sampling units (representing ISM) compared to distant ones (ESM). On 6-7 September

Chapter 2

1999. five locations were randomly selected in areas where winter feeding paths were

surveyed in 1998 (Chapter l), as well as in five other random locations in the sarne three

rneadows. At those locations, eight 0.25 m\uadrats adjacent to each other were recorded

aloog a random bearing (figure 1). The two rnost distant of these adjacent quadrats were 3

m apart. Starting at those two quadrats and along the sarne bearing, 1 recorded six

additional quadrats also separated by 3 m. This distance was chosen to reflect the average

distance of 2.65 I2.98 m (mean i SD, n = 145) observed ùetween sequentiai feeding

craters of bison. For each of the eight adjacent and eight distant quadrats, percent cover

of plant species was visually estimated and the biomass determined using a cdibrated

disk [biomass (g 1 m') = 81.958 + 10.004 x Disk height (cm), R' = 0.48, P < 0.0001, n =

1801, as described by Varha and Matches (1977). Using biomass estimates. and

considering the winter estimates of a type II functional response as weti as the digestible

energy of the recorded plants, E (kJ / min) was calculated as specified in Chapter 1. The

difference in E of sequential quadrats was calculated for both sets of adjacent and distant

sampling units. Values off were equal to the ratio of standard deviation for adjacent

quadrats ( 6 ~ ~ ~ ~ ) and for distant ones (a). For the 10 areas surveyed, f averaged 0.53 f

0.3 1 (range: O. 14-1.00). Based upon this information. the effect of variation in f was

investigated by performing the simulations with average (0.53) and low (0.25)f-values.

Based on plant surveys in 25 meadows interspersed in the bison range of PANP. 1

assigned a value of 150 kJ / min to E,tfLv and 240 kJ 1 min to E.tLu to sirnulate the range of

resource quality found in a typical winter habitat. Tbe general effect of spatial auto-

correiation in resource distribution was investigated by changing variability in overall

prcy profitability found in the habitat. Based on the 10 bison foraging paths (Chapter 11,

Chapter 2

different habitat types were simulated by using low (14), intermediate (24) and high (45)

a-values. A high a-value led, for example, to low auto-correlation between sampling

units.

For each simulated environment, 1000 foraging bouts were perforrned for W ~ I G

Wk4.vD and other reference windows up to Wj0, whereas 800 foragiiig bouts were

simulated for larger reference windows. Simulated anirnals were assumed to begin a

foraging bout with a reference window including values encountered in a similar habitat

type, My simulations did not include enetgy expenses, and thus assumed that those costs

do not influence searching strategy because, for instance, there is no relationship between

energy expended to access the resowce and its quality.

ANOVAs (with Tukey's post hoc tests, a = 0.05) were used to compare the

searching eficiency obtained using the various reference windows investigated. When

viewed for their statistical sense, the statistical significance of F- and P-values would be

arbitrary for my simulations because the numkr of replicates was subjectiveIy decided

(Tyler and Hargrove 1997). These results would then simpEy provide some indication

about the rate of change in searching efficiency that followed the expansion of the

reference windows. However, these analyses can also be considered for their biological

implication. For example, the absence of statisticai differences in the searching efficiency

obtained by using different reference windows during the 1000 simulated foraging bouts

would suggest that. during more than the entire winter (assuming 7 foraging bouts per

day, Plumb and Dodd 1993), searching efficiency was too variable for the animal to

expect that any systematic improvement would be achieved by the use ofa particular

reference window.

Reference windows size and searching eficiency

Setting f to 1 corresponds to a random search (WR4,~rD)r which resulted in the

discovery of prey averaging 195 kJ 1 min in al1 habitat types (with the variability among

foraging bouts being inversely proportional to the environmental stochasticity). This

value corresponds to the median quality of the prey available in the environment. Thus.

the area-restricted search with the use of any reference window in a spatially auto-

correlated environment increased searching efficiency by a rate of at least 7 kJ / min

compared to WR4.vD (figure 2). Considering that bison can spend 10.7 h / day p i n g

(Hudson and Frank 1987). this increase in eficiency would lead to an average daily

increase in resource discovery of 4494 kJ.

Searching eficiency increased in a deceleraring fashion following the expansion

of the animal's reference windows. and rapidly reached a plateau (figure 2). The

steepness of this increase varied with the habitat variability (a), with the consequence

that the reference windows at which searching eficiency started leveling off also varied

with o (figure 2. Table 1) . in habitats with average f and low a. simulated animals using a

W60 were as eficient in their searching behaviour as those considering three times more

sampling information (Le. WIBO). In habitats shulated with average f and intermediate a,

animals using a W18 had similar searching eficiency than those using a WI8o (Table 1).

The size of the reference window had even less influence for animals searching resources

in habitats where high variability was expected between feeding stations (average f and

high a, figure 2). In this case, the searching eficiency of animals using W6 did not

significantly differ fiom those using w180 (Table 1 ). Overail, iittle was gained in any

habitat by using a reference window larger than W18 when f was set to 0.53 (figure 2).

However. searching eficiency showed strong dependence tof-values. A low

value off reflected a high eficiency in finding similar prey species using an ISM

compared to an ESM. Lowering f led to an overall higher searching eficiency (figure 2),

and to an increase in the importance of considering an adequate reference window. For

example. when searching in habitats with a low a, simulated animals that considered a

W120 rather than W? increased their expected energy intake rate by 6 kJ 1 min when f was

0.53 and by 13 kJ / min when f was set to 0.25. Mien f was 0.25 and a 14, the searching

efficiency provided by WlaO was not significantly different to the one provided by W120;

rhus providing yet no evidence of a prolonged plateau (figure 2, Table 1). In contrast, the

searching efficiency in habitats where o w& 45 and 24 leveled off with the consideration

of and W30. respectively (Table 1).

A refirence point for omniscient animals

In some landscapes, omniscient animals are assumed to use the average resource

qudity for the environment as a reference point to assess local resource quality (Pleasants

1989. Howell and Hart1 1980). An animal that would use its perfect knowledge of the

average resource available in the environment (which equals W R I , ~ , 195 kJ / min) to

assess quality during its area-restricted search in an auto-correlated landscape would

increase its searching eficiency by up to 8.6 kJ / min when f is set to 0.53, and by up to

24.0 kJ / min when f is 0.25 compared to a random search (see W ~ I G VS. WK~,VD in figure

2). However, using complete information about the average resource available did not

mâuimize searching eficiency (figure 2, Table 1). In al1 habitat types, animals can

perform as well or even improve their searching efficiency by the consideration of

sarnpling information gathered during less than a foraging bout (i.e. with a WL 5 Wa)

compared to the use of a fixed reference window based on the average resource quality

for the entire environment (W.,I-~;) .

Rejèrence window size and searching eflciency

The efficiency of area-restricted search is dependent on the size of the reference

uindow used by the animal to assess resource quality. Searching efficiency increased in a

decelerating fashion following an expansion of the reference windows. The low

performance of animais using smail reference windows has k e n previously reporteci by

houe (1983) and Valone (1992). However. their studies had important differences with

this one. houe and Valone were both interested in foraging behaviour, and not directly in

searching behaviour. These two studies considered environrnents in which discrete

patches of resource were randomly distributed, preventing the animal h m adjusting their

Chaprer ?

searching behaviour with the habitat structure. in contrast, my model considered animals

that can take advantage of the spatial auto-correlation found in natural landscapes by

means of area-restricted search. Walsh (1996) showed that the use of area-restricted

search in environments comprised of well-delimited patches of resource interspersed in a

resource-fiee mamx c m strongly influence patct: quality discrimination and animal

distribution. Such environments would reflect the resource distribution in houe (1983)

and Valone ( 1992). My focus on large grazers led to the consideration of environments

with continuous resource distribution instead of well-delimited patches (Arditi and

Dacorogna 1988). At some spatial scales, such continuous habitats would be nonetheless

characterized by species aggregation because of the "sedentary habit" of plant and the

spatial heterogeneity of the environment (Schemske et al. 1994). Spatial variation in

factors that affect plant demography, such as soi1 characteristics. creates ecotones where

plant distribution cannot only Vary in species composition, but also in state of rnaturity.

For example. a gradient in soi1 moisture could affect plant species development, leading

to intra-specific variation in the digestible energy available to the gazer (Van Soest

1994). Therefore, as assurned in my model, relatively smooth gradients in prey

profitabiliiy might be expected in natural envirorments fiom this intra- and inter-specific

variation in plant distribution and characteristics. Despite the important differences in

mode1 designs, my results corroborate those of Valone (1992) and houe (1983),

indicating that the use of small, compared to large. reference windows should also be

avoided during searching activity because the perception of resource quality they provide

would be overly affected by the stochastic nature of the environments.

Valone (1992) and [noue's (1983) findings that a reference window of

intermediate size can be optimal contrast with my results. Intermediate reference

windows maximize performance when old information does not reflect current reality for

reasons such as resource depletion (Valone 1992). My mode1 does not assume the

absence of resource dep!etion within the sampting units, but like Krakauer and

Rodriguez-Gironés (1995). it assumed that the entire spectrum of prey types is available

throughout the foraging bout. My personal observations performed at the end of foraging

bouts confirm this possibility for bison in PANP, where grazing intensity is rather low. In

such cases. information on resource distribution and quality previously gathered still

reflects the resource that can be encountered at the end of the foraging bout.

Other constraints c m influence the size of teference windows considered by the

animal when assessing resource quality. According to Dall and Cuthill(1997). there are

costs associated with storing environmental information. These costs are related to the

allocation of limited nervous system resources between the different conflicting demands

made on an animal, and may be subject to natural selection (Real 1991). Animals might

have a selective advantage by considering a minimal amount of environmental

information. Given that the increase in searching efliciency that follows the reference

window's expansion reaches a plateau from which Litle efficiency is gained from an

increase in the arnount of environmental information considered, the animal should select

the reference window corresponding to the start of this plateau. Therefore. although

longer reference windows will not lead to lower searching efficiency, they are likely to

diminish the animal's fitness due to trade-offs? and should tend to be avoided.

The reference window where the plateau begins varied according to the

stochasticity in resource distribution. In my simulation, stochasticity was manipulated by

Chapter 2

chmging the overall variability between successive sampling stations (a in equaiion 1

and 2) . This could reflect a change in the grain of habitat heterogeneity, where a high

variability would reflect a habitat comprised of a mosaic of small aggregations of

resources of similar nature. 1 found that the beginning of the plateau reflected the grain of

the habitat patches, with, for example, small reference windows already adequately

appraising resource quality in habitat with high variation (i.e. high a) between sampling

units. Resource distribution in such variable environrnents is less predictable From one

feeding station to the next. Consideration of a small number of sampling stations thus

provides as good as knowledge about the local resource as possible in highiy variable

environments. Hirvonen et al. (1999) also found that the consideration of distant past

experience becomes increasingly important in environments with smaller variance in

patch quality. In such environrnents, short reference windows would tend to build up an

"inaccurate picture of the world as a whole" (Krebs and Inrnan 1993).

The searching efficiency and the beginning of this plateau were influenced byf: ln

natural environments. f may vary either because of habitat structure or because of animal

searching ability. Fine-tuning path sinuosity according to resource distribution can be

leamed and has been shown to increase the efficiency of the area-restrkted search (Bovet

and Benhamou 199 1. Haskell 1997). Spatial memory (Gillingham and Bumeil 1989.

Laca 1998, Dumont and Petit l998), together with visual (Fortin and Frandsen 1999,

Zollner and Lima 1999) and olfactive (Benhamou 1989) cues can be used to navigate in

the environment and to help find resources. Such cues would lower the prevailingf;

augment the importance of considering an adequate reference window. and increase the

advantage of considering Iarger reference windows.

Despite the large variation in parameter values 1 used, the rapidity with which the

increase in searching efficiency leveled off with the expansion of the reference windows

came out as an interesting generality of my model. Results suggest that only 30 m' of

habitat sarnpling information would be sufficient for bison to maximize their searching

behaviour under most field situations.

A reference point for omniscient animals

An accurate estimate of the long-term expectation of resource quality is required

to make adequate predictions from most classical foraging theories. Tome ( 1988) and

1-ivoreil and Giraldeau (1997) estimated the expected value based on behaviourai

observations. This approach is possible in such laboratory experiments where the

relationship between animal behaviour and resource distribution and quality can be

accurately quantified. However, it would be a challenging process to conduct for wild

animals in their natural environments. Howell and Hart1 (1980) and Pleasants (I989)

suggested the use of average prey density available in the habitat. This convenient

approach, however. leaves out the consideration of the interaction between animal

behaviour and habitat structure, potentially leading to biases in some environments.

Indeed. 1 showed that the average quality of the resource available in the environment

would only be an adequate representation of the long-term expectation when animals do

not adjust their behaviour to the structural component of the habitat (Le. f =l).

Given that an animal can make effective use of its area-restricted search to find

resources, any reference windows would provide a higher searching esciency than what

could be expected by a simple random search (WRIIvD), which would thus lead to higher

long-tenn expectations. Even an animal that uses a perfect knowledge of the average

resource quality in the environment (i.e. 195 kJ I min) to assess resource quality would

end up with a largely increased searching performance ( W ~ K >> W,U,VD), and thus higher

long-term expectation. ïhis indicates that Wm:~rD cannot represent a stable solution under

most natural situations.

The optimal searching strategy should include the consideration of a reference

window large enough so that any increase in reference window size does not lead to

higher searching efficiency. The searching eficiency reached by this optimal reference

window would then represent to the long-term expectation for an omniscient animal. My

findings thus indicate that the bias that would result from considering the expected long-

term resource quality given by WR4SD rather than by the optimal reference window in

foraging models would increase with the degree spatial auto-correlation of the resource

found in the habitat (figure 2). The use of WAA,vD rather than optimal reference window to

pararneterize the long-term expectation would lead to biases such as predicting longer

patch residence time and broader diet fiom the classical theories. With a broader diet

comes the prediction of higher stability in the plant-herbivore system (Fryxell and

Lundberg 1997). Failure to consider the adjustment between searching behaviour and

habitat structure thus makes it bazardous to forecast ecosystem dynarnics. My study

underlines the importance of considering the interaction between animal searching

efficiency and habitat structure, together with the animal's reference window, when

predicting resource utilization.

Chapter 2

The funding of this study was provided by Prince Albert National Park,

University of Guelph, and scholorships fiom FCAR and OGS. 1 would like to thank

Régis Pilote for his help in the field, Mark Andmskiw for assisting in the laboratory

work. and Park Wardens Lloyd O'Brodovich and Dan Frandsen for their logistical

support. 1 thank Carey Bergman, Tristan Long, Tom Nudds, Glenn Benoy and John

Fryxell for their valuable comrnents on a previous version of this paper. 1 am also grateful

to Jemifer Simard and Kara Vlasman for their discussion on different parts of this work.

Andersson, M. 1978. Optimal foraging area: size and allocation of search effort.

Theoretical Population Biology 13: 397-409.

Arditi. R. and Dacorogna B. 1988. Optimal foraging on arbitrary food distributions and

the definition of habitat patches. American Naturalist 13 1 : 837-846.

Benhamou. S. 1989. An olfactory orientation model for mammals' movements in their

home ranges. Journal of Theoretical Biology 139: 379-388.

Benhamou. S. 1992. Efficiency of area-concentrated searching behaviow in a continuous

patchy environment. Journal of Theoretical Biology 159: 67-81.

Benhamou. S. 1994. Spatial memory and searching eficiency. Animal Behavior 47:

1423-1433.

Bovet, P. and Benhamou, S. 1991. Optimal sinuosity in central place foraging

movements. Animal Behavior 42: 57-62.

Cézilly, F. and Benhamou, S. 1996. Les stratégies optimales d'approvisionnement. Revue

d'Écologie (Terre Vie) 51: 43-86.

Chamov, E. L. 1976. Optimal foraging, the marginal value theorem. Theoretical

Population Biology 9: 129- 136.

Cowie. R. J. 1977. Optimal foraging in great tits (Parus major). Nature 268: 137-139.

Dall. S. R. X. and Cuthill, 1. C. 1997. The information costs of generalism. Oikos 80:

197-202.

DeBenedictis. P. A., Gill, F. B.. Hainsworth, F. R., Pyke. G. and Wolf, L. L. 1978.

Optimal meal size in humrningbirds. American Natwalist 1 12: 30 1-3 16.

Devenport. J. A. and Devenport, L. D. 1993. Time-dependent decisions in dogs (Canis

familiaris). Journal of Comparative Psychology 1 07: 1 69- 1 73.

Dumont, B. and Petit, M. 1998. Spatial mernory of sheep at Pasture. Applied Animal

Behaviour Science 60: 43-53.

Focardi. S. Marcellini, P. and Montanam, P. 1996. Do ungulates exhibit a food density

threshold? A field study of optimai foraging and movement patterns. Journal of

Animal Ecology 65: 606-620.

Fortin, D. and Frandsen, D. 1999. Some preliminay results on the ecological hierarchy of

foraging activities in bison. Research Links 7: 13-1 5.

Fryxell. J. M. and Doucet, C. M. 1993. Diet choice and the functional response of

beavers. Ecology 74: 1297- 1306.

Fryxell, J. M. and Lundberg, P. 1997. Individual behavior and community dynarnics.

Population and community biology series 20. Chapman and Hall, New York. New

York.

Getty, T. and Pulliam, H. R. 1993. Search and prey detection by foraging sparrows.

Ecology 74: 734-742.

GiIlingharn. M. P. and Bunnell, F. L. 1989. Effects of leaming on food selection and

searching behaviour of deer. Canadian Journal of Zoology 67: 24-32.

Haskell. D. G. 1997. Experiments and a model examining learning in the area-restricted

search behavior of ferrets (Mustela putoritïsftïro). Behavioral Ecology 8: 448-

455.

Hirvonen. H., Ranta. E.. Rita, H. and Peuhkuri, N. 1999. Significance of memory

propenies in prey choice decisions. Ecological Modelling 1 15: 177- 189.

Howell, D. J. and Hartl, D. L. 1980. Optimal foraging in glossophagine bats: when to

give up. American Naturalist 1 15: 696-704.

Hudson. R. J. and Frank, S. 1987. Foraging ecology of bison in aspen boreal habitats.

Journal of range management 40: 7 1-75

[noue. T. 1983. Foraging strategy of a non-omniscient predator in a changing

environment (1) model with a data window and absolute criterion. Researches on

Population Ecology 25: 8 1-104.

Jung. H. G. and Koong, L. J. 1985. Effects of hunger satiation on diet quality by grazing

sheep. Journal of Range Management 38: 302-305.

Krakauer. D. C. and Rodriguez-Gironés. M. A. 1995. Searching and learning in a randorn

environment. Journal of Theoretical Biology 177: 41 7-429.

Krebs. J. R. and inman A. J. 1992. Learning and foraging: individuds, groups, and

populations. American Naturaiist 140: S63-S84.

Chapter 2

Laca, E. A. 1 998. Spatial memory and food searching mechanisms of cattle. Journal of

Range Management 5 1 : 370-378.

Livoreil, B. and Giraldeau, L.-A. 1997. Patch departure decisions by spice finches

foraging singly or in groups. Animal Behavior 54: 967-977.

IllcNamara, J. M. and Houston, A. 1. 1985. Optimal foraging and leaming. Journal of

Theoretical Biology 1 17: 23 1-249.

McNarnara, J. M. and Houston, A. 1. 1987. Memory and the efficient use of information.

Journal of Theoretical Biology 125: 385-395.

McNamara, J. M., Houston, A. 1. and Weisser, W. W. 1993. Combining prey choice and

patch use - what does rate-maximizing predict? Journal of Theoretical Biology

164: 219-238.

Olsson. 0. and Holmgren, N. M. A. 1999. Gaining ecological information about

Bayesian foragers through their behaviour. 1. Models with predictions. Oikos 87:

25 1-263.

Owen-Smith, N and Novellie, P. 1982. What should a clever ungulate eat? Amencan

Naturalist 1 19: 15 1-1 78.

Pleasants. J. M. 1989. Optimal foraging by nectarivores: a test of the marginal-value

theorem. Arnerican Naturalist 134: 5 1-7 1.

Plumb. G. E. and Dodd, J. L. 1993. Foraging ecology of bison and cattle on mixed

prairie: implication for natural area management. Ecological Applications 4: 631-

643.

Real. L. A. 199 1. Animal choice behavior and the evolution of cognitive architecture.

Science 253: 980-986.

Chapter 2

Schemske, D. W., Husband, B. C., Ruckelshaus, M. H. Goodwillie, C., Parker, 1. M. and

Bishop, J. Ci. 1994. Evaluating approaches to the conservation of rare and

endangered plants. Ecology 75: 584-606.

Smith. J. N. M. 1974. The food searching behaviour of two European thrushes. II. The

adaptiveness of the search patterns. Behaviour 49: 1-61.

Stephens. D. W. and Krebs, J. R. 1986. Foraging theory. Princeton University Press,

Princeton, New Jersey.

Tome, M. W. I988. Optimal foraging: food patch depletion by mddy ducks. Oecologia

76: 27-36.

Tyler. J. A. and Hargrove, W. W. 1997. Predicting spatial distribution of foragers over

large resocrce landscapes: a modeling analysis of the Ideal Free Disiribution.

Oikos 79: 376-386.

VaIone. T. J. 1992. Patch estimation via memory windows and the effect of travel time.

Journal of Theoretical Biology 157: 243-25 1.

Van Soest. P. J. 1994. Nutritional ecology of the ruminant. Second edition. Corne11

University Press, Ithaca, New York.

Vartha E. W. and Matches, A. G. 1977. Use of a weighted-disk measure as an aid in

sampling the herbage yieid on tall fescue pastures grazed by cade. Agronomy Journal

69: 888-890.

Walsh, P. D. 1996. Area-restricted search and the scale dependence of patch quality

discrkination. Journal of Theoretical Biology 183: 35 1-361.

Zollner, P. A. and Lima, S. L. 1999. Orientational data and perceptual range: real mice

aren't bhd . Oikos 84: 164-166.

Chapter 2

Table 1. Surnrnaxy of ANOVAs testing for differences in the average potential energy

intake rate of the resource found during 800-1000 foraging bouts when assessing resource

quality with various reference windows in simulated habitats differing in spatial auto-

correlation in resource distribution (a) andf-values. For a given habitat type, reference

windows with the same letter did not provide significantly differences in searching

efficiency according to Tukey's posr hoc tests (a = 0.05). Average searching efficiency is

displayed in figure 2 for each reference window and habitat type.

Tuke y Reference Tukey Reference Tukey Reference

Figure 1. Sampling design of the eight adjacent and eight distant quadrats surveyed to

determinefivalues expected under field conditions. Note that the two most distant of the

adjacent quadrats were also considered as distant ones.

Figure 2. Average expected energy intake rate of the resource found using area-restricted

search during 800- 1000 foraging bouts (depending on reference window size). The

simulated animais assessed resource quality using different reference windows in habitats

varying in a- andf-values. High values of a reflect low autocorrelation between sampling

units. whereas highf-values suggest that the animal has relatively low eficiency in

finding similar resowces when in an intensive searching mode compare to when in an

extensive searching mode. represents the searching efficiency of animais assessing

resource quality based on the median energy intake rate available in the entire

environment (i.e. 195 kJ 1 min). WM,vD corresponds to a random search and provided the

sarne average for al1 values of a, thus only a = 14 was displayed.

Chopter 2: Figure 1

Adjacent H , 3 m ,

E I 17r...1 I I I

t Distant l

Chapter 2: Figure 2

Reference windows (m *)

CHAPTER 3

An Adjustment of the Extended Contingency Mode1 of Farnsworth and Illius (1998)

DANIEL FORTiN

To be pub Iished in Funclional Ecology (2001)

Farnsworth and Illius (1 998) modified the classical contingency model (Stephens

and Krebs 1986) so that it could take into account the overlap between searching and

handling time observed in large herbivores (Spalinger and Hobbs 1992, Laca et al. 1994).

Their model predicts an optimal diet based on estimates of prey encounter rate (A.

preylmin), digestible energy (e, in kJ/prey), handling tirne (h, in midprey), and the

proportion of h exclusive from searching activity (q). Given that 1 > q > O. which will be

assumed throughout this paper, animals can search for the next prey while chewing the

last bite d u h g the portion (1-q)h of handling time, whereas they cannot search during

the portion qh of handling time devoted to cropping a food item. Famsworth and Illius

( 1998) also distinguished between foraging that is limited by encounter rate and by

handl ing time. Interestingly, they showed that if more than one prey is required to make a

diet handling-limited. the last prey should often be consumed at a rate lower than the

encounter rate. This contrasts with the 0-1 rule characteristic of the classical foraging

models (Stephens and Krebs 1986). The model of Farnsworth and Illius (1998)

illuminates our understanding of foraging decisions in large herbivores. but algebraic

errors in some of its equations lead to inaccurate predictions during handling-limited

foraging. My objective is to provide revised equations that indicate (1) when a diet is

handling-lirnited. (2) the rate at which the last type of a multi-prey type diet should be

accepted to make the diet just handling-limited, and (3) the energy intake rate provided

by a handling-limited diet based on the partial consumption of a given prey type.

Farnsworth and Illius (1998) stated on p. 76 that a prey type (i) is subject to

handling-limited foraging when "h, 2 IIA,", which could also be witten as (1-rt,)h, + q,h,

2 1 /A,. Because herbivores can search for new bites while they are masticating a previous

Chapier 3

bite. the proper conflict is benveen expected tirne to the next bite to be encountered IIR,

and the expected time to chew the last bite (l-qi)hi. When (1-q,)hi 1 1/;1,, the forager is

limited by handling, because bite rate = 1 /hi. In contrast, when (1-v,)h, < lIA,, then the

bite rate = A,/(! +A, hi). Extendhg this argument to multi-species foraging, and based on

equation 9 of Farnsworth and Illius (1996) and equation 2 of Farnsworth and Illius

(1 998). it can be derived that handling-limited foraging should rather occur when:

where m is the total number of prey types included in the diet. The optimal diet is

obtained by ranking prey by increasing profitability (dh) . and then expanding the diet

until foraging becomes handling-limited. If consumption of the most profitable prey type

is limited by handling tirne, the animal should specialize on this prey type. Hence. the

intake rate would simply correspond to profitability of the highest ranked prey (e, /hr).

When more than one prey type is required to reach handling limitation, the last type

should be partially accepted at a rate that relates to the encounter rate for the more highly

ranked items already in the diet. The acceptance rate of prey type m can be expressed as:

(?il4 1% . By definition. partial acceptance of prey type m entails that these plants are 1=I

accepted at a rate lower than their encounter rate. and thus that (yA ln < L. Because I m

i = I

n, represents the minimal amount of prey m that would make the diet handling-Iimited.

the following equality is implied:

With the rearrangement of equation 2 we can find n,,, from:

m-i m-l

2 A, x A

which corresponds to the time that would be spent searching after chewing (Le. during

"pure" search) if only m-1 prey types were accepted in the diet divided by the time that

can be spent searching while handling prey m. Because q, > O. equation 3 would provide

higher n, than equation 17 of Famsworth and Illius (1 998). However. their equation

reflected encounter-limited foraging rather than handling-limited foraging, as it should

have.

Given the acceptance of al1 the m-1 prey encountered and the partial acceptance

rate of prey m. the energy intake rate of the handling-limited diet becomes:

m-l

The diet would be optimum only if the handling-lirnited foraging provides a greater

intake rate than the rate for encounter-limited foraging without inclusion of the last prey

type:

m-l

Z 'iei

If inequality (5) does not hold. only m-1 prey types should be included in the diet.

1 believe that the study of Farnsworth and [llius (1998) constinites an important

contribution to foraging theory, and my revision of their equations now allows the

prediction of optimal diet also during the handling-limited foraging of large herbivores.

The tûnding for this study was provided by Parks Canada University of Guelph.

and scholarships frorn FCAR and OGS. 1 thank John Fryxell for his advice and

constructive comments on this paper. 1 am gratefbl to Keith Farnsworth and Andrew

[Ilius for encouraging me to pursue this work.

Famsworth. K. D. and illius, A. W. 1996. Large gazers back in the fold: generaliing the

prey mode[ to incorporate rnammalian herbivores. Functional Ecology 10: 678-

Chapier 3

680.

Farnsworth, K. D. and Illius, A. W. 1998. Optimal diet choice for large herbivores: an

extended contingency model. Functional Ecology 12: 74-81.

Laca. E.A.. Ungar, E.D. and Demment, M.W. 1994. Mechanisms of handling time and

intake rate of a large mammalian grazer. Applied Animal Behaviour Science 39:

3- 19.

Spalinger, D. E. and Hobbs, N. T. 1992. Mechanisms of foraging in mammalian

herbivores: new models of functiond response. American Naturalist 140: 325-

348.

Stephens. D. W. and Krebs. J. R. 1986. Foraging theory. Princeton University Press.

Princeton. New Jersey.

CHAPTER 4

The Temporal Scale of Foraging Decisions in Bison

DANIEL FORTIN,

JOHN M. FRYXELL and &GIS PiLOTE

Assessing the temporal d e under which gain is maximized is critical for

understanding diet choice by animals. Classical foraging theory assumes that animals

rnmimize long-terrn rates. Few studies have concurrently considered several temporal

scales, however. weakening tests of rate-maximizing models. We used contingency

models based on maximization of short-term vs. long-tenn energy intake by bison (Bison

bison). Mode1 predictions were tested against field observations conducted during six

periods of 1998: two penods in the winter, one in the spring and three in the summer.

During most of the year, foraging characteristics and plant attributes suggest that intake

rate of bison should be limited by handling time over short period of time, and by

digestive constraints over long period of time. Diet predictions varied across temporal

scaies for four of the six sampling periods. Selecting Agropyron spp., rather than Car-

arherodes. during these periods would result in an increase of daiIy energy intake by as

much as 15565 kJ, but would necessitate a longer daily foraging time. We observed.

instead. that bison prefemd C. a~herodes to Agropyron spp., suggesting that patterns of

diet selection by bison were more consistent with maximization of short-term than of

long-term energy intake. We provide evidence t h a ~ contrary to established principles of

classic optimality models, the foraging decisions of bison reduced potential long-term

gains by maximizing short-term gains.

Chopter 4

There is a growing awareness of the importance of choosing the rnost appropriate

tirne scale in foraging studies (Lucas 1983, 1990, Gass and Roberts 1992, Gross et al.

1993. Wallis de Vries and Daieboudt 1994, Wilmshurst and FryxeH 1995). According to

classical foraging theory, animais are expected to maximize the long-term rate of energy

intake (Stephens and Krebs 1986). There is no clear guideline. however, indicating which

temporal scales should be considered. Investigators usually use their biological intuition

to select temporal scales (Gass and Roberts 1992) and they commonly choose a single

one (Owen-Smith and Novellie 1982, Langvatn and Hanley 1993. Wilmshurst et al. 1995,

1999. van Wieren 1996).

Under some conditions, predictions are insensitive to time frarne. For example,

the original contingency mode1 predicts that animais would rnaximize both their short-

term and long-term rates of intake by making the same diet choice (Stephens and Krebs

1986). Likewise. time minimizers and energy maximizers should select an identical diet

(Lucas 1990). Consideration of other constraints, however, can lead to different optimal

diets at different temporal scales (Gass and Roberts 1992). This is the case, for example.

when black-capped chickadees (Parus atricapillus) encounter prey simultaneously rather

than sequentially (Barkan and Withiam 1989).

Large mammalian herbivores face constraints not considered by the classicak diet

model. Field evidence suggests that the rate of digestion is sometimes the predominant

factor limiting the daily intake rate by large herbivores (Belovsky 1978, Mould and

Robbins 1982. Wilmshurst et al. 1995). The need to spend t h e in other activities or to

rnaintain thermal balance can also constrain feeding time, setting an upper limit to the

daily food intake (Arnold 1985, Belovsky and Slade 1986, Memll 1991). Consideration

of these constraints c m lead to predictions of optimal diet that are sensitive to temporal

scale (Bergman et al. 200 1).

Farnswonh and Illius (1 996, 1998) indicated that, over short time scaies, intake

rate of large herbivores can be either restricted by the encounter rate with plants or. when

resources are highly abundant, by the time required to handle food. Over long time

scales, the consideration of digestive and time constraints leads to four foraging

situations:

[Il Resources do not allow animals to meet their daily voluntary intake during the

maximum time that can be dlocated to foraging activity (T[KT]). and

foraging is limited by encounter rate.

[II] Resources do not allow animais to meet their daily voluntary intake during

TL-icr]. and foraging is limited by handling time.

[III] Resources do allow animals to meet their daily voluntary intake during

T[.KT], and foraging is limited by encounter rate.

[IV] Resources do allow animais to meet their daily voluntary intake during

T[.-rcr]. and foraging is limited by handling time.

When the daily voluntary intake cannot be fulfilled during T[KT] (situation [Il

and [II]). the optimal diet is scaie-insensitive. This is because, at both temporal scales, the

optimal diet is determined by ranking prey types according to their short-term

profitability (i.e. digestible energy I handling time, in Hlrnin) and expanding the diet.

starting with the most profitable prey, until the intake rate reaches a mavimurn (algebraic

solutions for this optimum are given in Farnsworth and Illius [1998] and Fortin [2001]).

In other words, when the daily voluntary intake cannot be reached, the animal should

always ingest energy as quickly as possible. in contrast, when the digestive constraint is

reached within T[.-IcT] (situation [III] and [IV]), the diet can become scale-sensitive.

Under such foraging situations, the optimal diet over short time scales is determined by

considering short-tem profitability. whereas the diet that maximizes daily gains is found

by ranking prey types according to their daily profitability (daily profitability of species i

= digestible energy of species i x voluntary intake of an animal feeding on prey type i, in

kJ/day). The short- and long-tenn profitability c m then differ, leading to different

predictions of optimal diet.

For exarnple, taller (or higher biomass) plants ofien allow for larger bites and

higher instantaneous intake rate (Laca et al. 1992, Wallis de Vries and Daleboudt 1994.

Gordon et al. 1996, Bergman et al. 2000), and can be handled faster. At the same time.

taller (or higher biomass) plants generally provide less digestible energy because of

maturational changes in quality (Fryxell 1991. Van Soest 1994, Wilmshurst et al. 1999).

These opposite trends can create differences when ranking prey according to their long-

and short-ten profitability. and thus it can create an optimal diet that is time scale-

sensitive.

Here, we use contingency models based on short-ten vs. daily rates of energy

intake. We then test which temporal modeling scale is most consistent with the observed

diet of free-ranging bison (Bison bison) in Western Canada.

Short-term model

We determined the diet maxirnizing the short-term energy intake rate (kllmin) of

bison based on equations 1 through 17 of Farnsworth and Illius (1998), with some

modifications (Fortin 2001). Al1 parameters were estimated as a Fwiction of dry matter

mass rather than individual prey.

Long-term model

The diet ma,,imizing daily energy intake rate was detennined by including the

constraints imposed by time and daily voluntary intake to a contingency model that also

considers the overlap between searching and handling observed in large herbivores.

A daily feeding time of 642 min has k e n reported for bison (Hudson and Frank

1987). which we assurned to be the maximum ddly activity time (T[KT]). Bergman et al

200 1 ) have demonstrated that the daily voluntary intake of bison (G. in glday) shares a

positive-linear relationship with dry matter digestibility, as comrnonly o b s e ~ e d in other

species (Mould and Robbins 1982, Van Soest 1994, Wilmshurst et al. 1999). Based on

the common assurnption (Owen-Smith and Novelie 1982. Stephens and Krebs 1986.

Fryxell and Doucet 1993) that the prey types constituting the animal's diet are accepted

in proportion to their encountered rate (A. g/min), the linear relationship between G and

the dry matter digestibility (4 can be expressed as:

where i represents individual prey types in a diet including m species. The fitîed constants

y and 9 are, respectively, equal to 142.464 and 687 1.344 in bison (Bergman et al. 200 1 ),

assuming an animal of 636 kg (Belovsky 1986). When foraging is lirnited by prey

encounter rate. the tirne needed to reach the daily voluntary intake (T[EK~,, in mintday)

is given by T[E,\.T~, = G,H,,,. Hm (midg) is the expected tirne required to encounter and

to crop 1 g of the m prey types included in the diet, and is found by:

where h (midg) is handling time and q is the proportion of h that does not ovedap with

searching activity. During handling-limited foraging, the time needed to reach the daily

voluntary intake (T[H..I,vD],, in rnidday) is given by:

Chopter J

Considering the time constraint, the daily voluntary intake cannot be reached when

T[E.vc],,, > T[;IcT] or T[H..IND], > T ~ c T ] .

Sitrration [Il

When resource does not dlow the fulfillment of the daiIy voluntary intake during

T/KT] and foraging is limited by encounter rate (situation [Il), the daily energy intake

rate is simply given by equation 5 of Farnsworth and Illius (1998) multiplied by T[.-~cT]:

m-l z Alel

where e, is the digestible energy of species i (kllg of dry matter). To detemine the

optimal diet. prey types are ranked according to their short-term profitability (dh. in

kJImin), and the diet is expanded as long as the inequality 8 of Famswonh and Illius

(1998) is met.

Situation [II]

When resources do not allow animds to meet their daily voluntary intake during

TLKT]. and foraging is limited by handling time (situation [Li]). determination of the

optimal diet also requires ranking prey types acccrding to their short-term profitability

Chapter J

(elh). and adding thern to the diet until foraging becomes handling-limited, Following

Fortin (200 1 ). diet is handling-limited whenever: f A, - a )hi 2 1. i=l

When this requires more than one type of prey, the last type (m) should be

accepted at a proportion @,) that is relative to the m-l prey already included in the diet,

and that would just make the diet handling-limited (see equation 3 of Fortin 2001):

m-l

1 4

Intake rate under situation [II] is then given by:

("Z", ) ~ m e m - - '4e, + 1 = 1 C- 1 = i h,

I , = m - ~ hm T [A CT] C'[(l+ P,) r = l

Note that if only one prey is required to make the diet handling-limited, Pm equals zero,

and intake rate is simply given by: elh x T[JcT].

Situation [IIiJ

When the resource allows the fuifiIlment of the daily voluntary intake during less

than T[,rcr], and foraging is limited by encounter rate (situation [III]), determining the

optimal diet requires to rank the prey types according to their potential daily profitability

(eG, in kJlday), and then adding them sequentially to the diet until the daily voluntary

intake c m be achieved. If this requires more than one type of prey. the last type should be

accepted at a rate (#,,,) that is relative to the m-1 prey already included in the diet and that

would just fulfill the animal's daily voluntary intake during T[.~cT]. This implies that

T[~.vc-j, = T'[KT] = G',H',. The daily voluntary intake that includes the partial

acceptance of prey in (G',) is given by:

G ' m = '=' i= l m-l Y + O Z'j(l+#m) i= l

and the expected time required to encounter and crop 1 g of a diet that now includes

partial acceptance of prey type m (H,) follows:

1 + Y4 h, 1, + ("?A, Mm vmhm H ' m = l='

r = l m-l

XA,(l+#rn)

The equality T[.-rc~] = G', H', can be expanded and rearranged in the form of a

polynomial: O = a , k 2 + b,#,,, + cm, where

We can tlien find 4, as the lowest positive value given by:

The intake rate for a situation [[II] of foraging becomes:

nt-l

1 4e1 + cc4 )ke*

Chapter 4

Situation [IV]

Finally, when the resowce allows hlfillment of the daily voluntary intake during

T[..icr] and foraging is handling-limited (situation [IV]), determining the optimal diet

also requires the prey type to be ranked according to their potential daily profitability

(eo, and to be included in the diet. sequentially, until the daily voluntary intake can be

achieved. If the prey type with the highest daily profitability is handling-limited on its

own. the optimal diet would predict specialization on this type of prey. When more than

one prey type is required to make the diet handling-limited. the last type (m) should be

included in a proportion p,, which is determined by equation 5. The daily energy intake

rate in situation [IV] is found by equation 6, in which TLK'T] is substituted by the

T[~.\ . 'D] ',,, given by:

Whenever more than one prey is required for handling-limited foraging, T[F/.~,VD] ',,, has to

exactly equal T[..IcT] to be treated as situation [rV] foraging. Faster tùlfillment (Le.

T[HA.VD] c T[.-IcT]) indicates that the animal could have foraged for a longer period of

time. and thus that opportunity for foraging on the more digestible prey type had k e n

lost. This case should then be treated as a situation Dm, with a partial acceptance & of

the prey type m (where & cp,) .

Chapier 4

The field study took place in Prince Albert National Park (Saskatchewan,

Canada), which has a population of approximately 220 plains bison (Bison bison bison).

The optimal diet of bison was determined for six periods during 1998: (1) 5 January to 16

February. (2) 17 February to 4 April, (3) 23 May to 19 June, (4) 20 June to 12 July, (5) 13

July to 7 August and (6) 8 August to 3 September. Information on plant phenology was

collected in 25 meadows Iocated throughout the bison range. Although >170 plant

species were recorded, for logistic reasons, we focused on seven of the most abundant

kind of plants: Agropyron spp.. Carex atherodes, C. aquaf ilis, Calamagrosfis inexpansa,

Hordeum jubarum. Jirncus balticus and Scolochloafestucacea. Throughout the year,

these species represented 50 - 72 % of the total biomass available and 8 1 - 99 % of the

bison diet.

During the spring, summer and winter of 1997 and 1998, the foraging behavior of

bison (excluding yearlings and calves) was recorded using focal animal sarnpling

(Altmann 1974) during 5-min periods or until the animals walked out of sight. Using a

spotting scope. we followed any bison -400 m. We recorded every bite, displacement of

the Front feet, head raising and lowering movement, and pawing and head sweeping

motion for snow clearing into a tape recorder. In high vegetation or deep snow, head

pulling movements were recorded as bites. Data were later transcribed using a stopwatch

to measure the time between movements (following Wilmshmt et ai. 1999). th-

frequency of behaviors per 5-min period, and the nurnber of feeding stations. A feeding

station was defined as the area fiom which the animal fed without displacing its front feet

(Bailey et al. 1996).

Following an observation session, forage characteristics and snow conditions

were averaged over 3-5 puadrais s p ~ a d over the foraging area. Quadrats were 1 m2

during the spring and summer (referred to as the growing season), and 0.25 m2 in winter.

Sample size varied with the distance covered by the focal animais. During the growing

season. the total dry biomass (dm') was determined by the distance (in cm) that a plastic

disk settled from the ground (Vartha and Matches 1977), according to the regression: y =

8 1 .%8 + 10.004 x (Fi. 179 = I6Z.99, R' = 0.48, P < 0.000 1). In winter. forage biomass was

based on a 0-9 visual scale. according to y = 82.229e0.->' (FIJ8 = 366.27, R' = 0.91. P <

0.000 1 ). Evaluation of plant biomass was supplemented by estimation of the percent

cover and. during the growing season. the percent green biomass of each plant species.

Snow conditions consisted of density and depth. Snow density (@cm3) was determined

by weighing, with a spring scale, a sample of the snow colurnn collected with a metal

tube inserted vertically into the snow and dividing the mass by the volume of the snow

gathered.

Prey encounter rate fi)

Estimation of A requires knowledge of the average effective plant biomass.

stepping rate while searching, distance between steps and the width of the animal's

foraging path.

Plant characteristics were determined in 16-90 evenly spaced quadrats, depending

on the area of each of the 25 meadows surveyed. During winter, plant biomass was

visually estimated on a 0-5 scale, according to: y = 67.876e0.w'x = 295. 1 1, R~ =

0.88, P < 0.0001). This equation differed from the previous one because a different

observer performed the meadow surveys.

Grazers do not consume plants completely (Ungar and Noy-Meir 1988. Burlison

et al, 1991. Edwards et al. 1995, Bergman et al. 2000). suggesting that the vegetation

cannot be considered entirely available to the animal. Following Owen-Smith and

Novellie (1 982) and Owen-Smith (1 993), we assumed that the resource actually available

was the fraction of a given plant species consurned by the animal. Grazed depth was

determined for each focal species during the growing season of 1998. Sampling units

consisted of average height of 15 grazed and 15 adjacent ungrazed plants. Approximateiy

40 g of ungrazed individuai plants were collected. at the grazed depth of nearby plants.

Samples were weighed with a spring-scale following drying at 50°C for 72 h. We

established the relationship behveen the proportion of the plant mass consurned (P) and

plant hcight. Although estimated only during the growing season, we assumed that P

related similarly to species height in winter, For each of the six sarnpling periods, the

height of each plant species was also measured at randomly selected stations among the

25 meadows, with each sample consisting of 15 measurernents. From this information. P

was established for each sampling p e n d based on the tegression analyses. The effective

dry biomass of a species was calculated as effective biomass = total biomass x cover x

grazed depth, in winter, and effective green biomass = effective biomass x proportion of

green tissue, during the growing season.

Step rate was calculated by measwing the time required to accomplish 4-20 steps

when moving between feeding stations. This sequence excluded the animal's first step

because its initiation was often difficult to determine accurately. This omission could lead

to a slight ovetestirnate of semhing speed because it ignores the initial period of

acceleration (Shipley et al. 1996). In winter. the distawe between footprints in the snow

was directly measured and averaged over 5 replicates for each animal path. During the

growing season. distance between steps was established based on animals 15-35 m away

fiom a biind. by counting the total number of steps between landmarks divided by the

distance traveled. The search speed consisted of the p d u c t of step rate and step size.

We estimated the effective foraging path width (W) during the behavioral

observations. We considered that a foraging animal moves its head fiom side to side in a

semi-ci-rcle of diarneter W (m). We first estimated the total biomass grazed in the visited

area (dm2) by summing visual estimates of the proportion of cover of the species grazed

by bison. For each obsewed feeding bout. we then estimated the arnount of forage grazed

per station (g) from the product of the observed average number of bites per station and

the expected mas of these bites. Expected bite mass (S, in gtbite) was determined by

reanalyzing the data displayed in figure 4 of Bergman et al. (2000). As commonly

O bserved (Bradbq et al. 1996, Wilmshurst et al. 1999). the dry mass of bites increased

in a decelerating manner with increasing total dry biomass (V), according to:

On the basis of the estimated parameters and fiom the assumption that a feeding station

represents half a circie, W was estimated from:

8 x Biomass grazed per station

~c x Total biomass grazed

Finally, using al1 the estimated parameters, we determined the encounter rate (dmin) of

each species from: encounter rate = effective biomass x search speed x effective path

width. in winter. and from: encounter rate = effective green biomass x search speed x

effective path width, in spnng and summer.

Hundling time (h) and proportion of handling rime exclusive ro searching rime (ti)

The intake rate of dry matter in food concentrated-patches (process 3 foraging) is

lirnited by handling time (Spalinger and Hobbs 1992). Thetefore. we used functional

responses estimated in food-concentrated patches to determine the handling time of each

of the seven focal species. Following Bradbury et al. (1996). dry matter intake rate (F)

was determined by multiplying bite rate (D) observed in the field by bite mass (5'). Forage

intake rate was then related to forage biomass using the Michaelis-Menten (Michaelis and

Menten 19 13) form of the functional response:

where yr is the maximum feeding rate (g/min) and P i s the haif-saturation constant (g/m2)

(Wilmshurst et ai. 1999, Bergman et ai. 2000). Two functional responses were

detemined: one for spring and sumrner and one for winter.

Assessment of the expected intake rate for a given plant species requires field

estimates of biomass at a relevant scaie. According to Bergman et al. (2000). the area of a

bite represents the scale of relevance in foraging processes. in rnany instances, species

occurred in rnixed swards and their aggregation was srnailer than the m a of the disk (0.3

ml) used to estimate biomass during the growing season. For this reason. the relationship

between plant height and disk biomass was detennined during the growing season of

1997 in stands largely dorninated (cover >70%) by each of the focal species. Positive

relationships between height and biornass were obsewed for all of the seven focal species

(P < 0.005 in al1 cases). Using these regression relationships. we transformed the average

height of each species observed during each of the six sampling period into dry rnatter

biomass. As suggested by equation 16 of Farnsworth and Mus (1996), when foraging is

handling-limited (process 3 foraging, sensu Spalinger and Hobbs [1992]), the inverse of

intake rate is equal to handling tirne. Therefore, we estimated the average time required to

crop and chew 1 g of a given prey species (h in midg of dry matter) based on our

functionai responses and this inverse relationship.

Farnsworth and Illius (1996, 1998) indicated that handling time can be partitioned

into cropping time (qh) , during which the animal cannot search for food, and chewing

time ([1-VIA), during which the animal can finish handling the prey while searching for

the next bite. q thus represents the proportion of handling time spent cropping. Following

the modeling approach of Spalinger and Hobbs (1992), we calculated the time required to

crop a bite from the relationship between bite rate (D, bitdmin) and bite mass (S, glbite):

where R,, is the rate of food processing in the absence of cropping (gfmin) and h' is the

average time required to crop a bite (minhite). The distinct values of h; found during the

growing season and in winter were divided by the average bite mass of species i to

determine the time require to crop a gram of plant species i ( 5 , in midg). 7, was then

found from the ratio of h and h,.

Dy matter digesiibility (d) and digestible energy (e)

Dry matter digestibility and digestible energy of the seven focal plant species

were detemined by collecting samples of their above-ground tissue at random locations

throughout the bison range. The samples collected during each of the sarnpling pieds

were separated by tissue (stems and leaves) and dried at 50°C for 72 h. Dry matter

digestibility was determined following Tilley and Terry's (1963) method using cattle

rumen liquor. The resulting estimates of forage digestibility for cattle were converted to

dry matter digestibility for bison. following Bergman et al.3 (2000) procedure. We then

determined the digestible energy content (e,, Wg) for a given species tissue h m the

product of dry matter digestibility and the gross energy content of 18.4096 kl/g (National

Research Council 1996). Only the digestible energy of the tissue consumed by bison was

considered in the contingency models. Tissue selectivity was determined during the

assessrnent of grazing depth, as well as during other plant surveys.

Observed dier

Bison diet was determined by estimating the total biomass grazed per unit of area

in each quadrat, averaging over 25 meadows for each of the seven focal species, as well

as for non-target plant species. Although white-tailed deer (Odocoileus virginianus). elk

(Cenw elaphlrs) and moose (Alces alces) were occasionally obsewed grazing in the

meadows, their grazing intensity was likely small (see Chapter 5).

We first describe our field evaluation of the parameters used in the contingency

models. and then present our predictions of optimal diet for dit'ferent temporal scales and

compare these predictions to the obsewed diet of bison.

Mode1 paramerers

Grazed depth increased with plant height for most species (P < 0.002), with the

exception of Hordetrrn jtrbatum = 0.002, P = 0.97) and Carex aquatilis (Fi.9 = 1.18.

P = 0.3 1). The proportion of biomass grazed declined with increasing plant height for

.4gropyron spp.. Calamugrostis inexpansa. and Scolochloafesrucacea. ( P < 0.05. figure

1 A), whereas no significant relationship were observed in other species (P > 0.18, figure

1 B). Based on this information, and considering the average species height measured

during each of the sampling periods (Table l), we determined that, on average, bison

consumed iess than half of the plant biomass available during any period of the year

(Table 1). In late spring (period 3: 23 May - 19 June), the biomass of most plant species

was still low, and bison generally grazed stems together with leaves. Carex atherodes and

Scolochloafesrucacea were the only exception. with leaves only consumed. During the

other periods of the year, grazing activiîy was limited to leaves for Agropyron spp..

Carex arherodes. C. aqicatilis. Calamugrosris inexpansa and Scolochloafistucacea. Dry

mmer digestibility varied between 40-75% throughout the year. tending to be higher for

Rgropyron spp. and C. arherodes than the other species (Table 1 ).

Bison traveled between feeding stations at a stepping rate 58.9 k 6.2 steplmin

(average + sd. n = 21) in spring and summer, and 48.3 I 8.7 steplmin (n = 23) in winter.

whereas the distance between these steps was 0.55 + 0.09 m (n = 13) in spring and

summer and 0.34 10.06 m (n = 10) in winter. Hence. searching speed was. on average.

twice as fast dunng the growing season ihan in winter. Bison path width averzged 1.78 + 1 .O3 m (n = 101 ) during the growing season and 1.91 I 1.27 m (n = 2 10) in winter.

leading to a searched area of 57.9 m2/min and 3 1.7 m2imin, respectively. As observed in

cattle (Illius and O'Connor 2000). bison tended to graze onIy the green biomass during

the growing season. During winter, however. al1 the accumulated biomass was used. High

iiiornass during winter partly compensated for low searching speed, providing an

intemediate encounter rate in winter (Table 1). During any pend of the year. the higher

forage biomass of C. atherodes conferred to this species the highest encounter rate (Table

1 ).

Forage intake rate in mixed swards increased with biomass towards an asymptote

of 78.43 g/min during the growing season and 72.68 glmin in winter (figure 2). This

maximum intake rate was comparable to pior measurements by Hudson and Frank

(1986) and Bergman et ai. (2000), afler accounting for differences in body m a s ( s e

Bergman et al. 2000). The recorded functional responses likely represented process 3

foraging. First. behaviorai observations were performed over short periods of time (5

min), requiring the visit of only few feeding stations. Second, we found weak but

significant negative relationships between bite rate and biomass (winter: = 5.90. R'

= 0.03, P = 0.016: growing season: Fi.iiz = 14.55, R' = 0.12. P = 0.0002, figure 3). which

is characteristic only of process 3 foraging (Bradbuy et al. 1996). This suggests that the

observed functional responses reflect a situation in which intake rate was restricted by

cropping and chewing rather than by encounter rate (Spalinger and Hobbs 1992).

Using these functional responses and the average seasonal biomass. we calculated

that it took between 0.01 5 and 0.020 min to handle 1 g of forage during any period of the

year (Table 1 ). The iarger bites that c m be cropped in species with hi@ local abundance

reduce the handling time per unit mass because less time is spent cropping relative to

chewing. Reduction of cornpetition between cropping and chewing was responsibte for

the rapid handling time in tall species such as C. atherodes and Scolochloafesrucacea.

The proportion of handling time exclusive to searching time ( q ) varied between 0.37 - 0.68. being more important during winter (Table 1). Tall species like C. atherodes and

Scolochloafesrucacea tended to have more overlap between searching and handling,

because of the longer chewing time associated with the larger bites they offered.

Mode1 predicrions and observed diet

Due to high encounter rate with the highly profitable prey (Table l), both short-

and long-term models suggest that intake rate would be maximized by specializing on a

single plant species during winter or summer (figure 4). During these seasons, bison diet

should be handling-limited over short period of time, leading, over longer period of time,

to a diet that reflected a situation [IV] of foraging. For haif the sampling periods, the

optimal diet maxirnizing short-term and long-term rate of energy intake differed. because

of inconsistency between the ranking of plant types according to their short-terni and

Iong-term profitability. During three periods (period 2,4 and 5), specializing on C.

arherodes would maximize animais' short-term energy intake rate (idmin), because the

hi&-energy content and rapid handling tirne lead to high short-term profitability

(kJ/min). On the other hand, Agropyron spp. offered the highest dry matter digestibility.

which also increased the daily voluntary intake, hence offering the highest daily

profitability (kJIday) (situation [IV]). Consequently. during periods 2.4 and 5, a bison

specializing on Agropyron spp. rather than C. arherodes would forage 3 1 - 63 min longer.

ending up with an additionai 6677 - 15565 kJ at the end of the day.

The longer foraging period associated with the selection of Agro~yron spp.

translates to higher activity costs. If the additional foraging time is entirely spent walking

in snow of depth of 25 * 8 cm and density of 0.15 * 0.06 @cm2 (n = 906) as observed

during period 2, or in absence of snow as during period 4 and 5, we found fiom equations

1 1 and 12 of Turner et al. (1994) that activity costs wodd only increase by 21 2-274 kJ.

Specialization on Agropyron spp. would still result in a higher net daily energy intake

rate.

C. arherodes constituted most of the bison diet throughout the year (figure 5).

Positive IvIev electivity indices indicate use of this species disproportionate to its

availability (Table 2). In contrast, Agropyron spp. made up only a small amount of the

diet. tending to be avoided during most of the year. The preference for C. atherodes and

the avoidance of Agropyron spp. during the three periods of the year where the mode1

prediction differed suggests that bison foraging decisions lead to short-term maximization

of energy intake. Scolochloafesii~caceu was also an important part of bison's diet (figure

5). which was not predicted at any temporal scales.

A mixed diet was predicted during the late-spnng (period 3). Both models

suggested that Hordeum jubatum shouId be eaten whenever encountered, whereas only a

Faction of the Agropyron spp. encountered should be included in the diet. The important

difference in encounter rate ktween these two species (Table 1) led to the prediction that

for every 1 g of Hordeum jubalum consurned, 7.7 g (Le. p, = 7.7) of Agropyron spp.

should be accepted in the animals' diet to maximize the short-term intake rate. However,

with this handling-Iimited diet, the daily voluntary intake would be fùlfilled in only 3 17

min of the 642 min potentiaily available per day for foraging. Bison could thus increase

their daily energy intake by foraging more intensively on Hordeum jubatum. As

expIained in the methods (see case when a potential situation [IV] should be treated as a

situation [III] to maximize the daily intake rate), the daily energy intake rate wodd be

mavimized by reducing the consumption of Agropyron spp. (# 2 < p?). which would make

the diet encounter-limited. For every 1 g of Hordeum jubatirm, only 2.0 g (Le. Q 2 = 2.0)

of Agropyron spp. should be accepted in the diet. The consequent 325-min increase of

daily foraging yields only an extra 60 kJ at the end of the day. This is because Agropyron

spp. and Hordeum jubatum had almost similar digestible energy content whereas

Hordetrm jubatirm had a much lower encounter rate. In this case, following the short-term

optimal diet for the entire day rather than the diet that maximizes daily intake rate should

enhance the "net" daily energy intake rate. Regardless, bison did not follow any of these

spring predictions, but they rather kept foraging largely on C. atherodes as in other

penods of the year (figure 5, Table2).

Our study demonstrates the potential dependence of optimal diet predictions on

temporal scale. At both temporal scales we considered. optimal diet of bison should

usually comprise a single plant species. but the identity of that species differs between

scales. Several factors contribute to the prediction of nanow diets in bison. First. the

overlap between chewing and searching time reduces the proportion of time spent in

exclusive search, making the diet more likely to become handling-limited with inclusion

of fewer prey types. Second, bison are highly efficient in foraging on low biomass swards

(Bergman et al. 2000). This eficiency translates into fast handling times even at low

plant biomass. Third. t!e most profitable (in kllmin or kJ/day) plant species were highly

abundant in Prince Albert National Park, perhaps in part because the bison population is

still relatively small.

Our assessment of resource availability, however, made the prediction of a broad

diet more likely. Large grazers do not normaily consume the entire above-ground

biornass encountered (Ungar and Noy-Meir 1988, Burlison et al. 199 1, Edwards et ai.

1995, Bergman et al. 2000). We followed the convention of considering only the Fraction

of each plant species consumed by bison as available (Owen-Smith and Novellie 1982,

Owen-Smith 1993). By following this procedure, we effectively link tissue selectivity

with species selectivity. This approach should not jeopardize the validity of our findings,

however. for several reasons. First, we were interested in selectivity occurring at the

species level rather than at the tissue level. At this level, tissues not actually consumed by

bison should not reflect their perception of a given plant species (Bergman et al. 2000).

Second. the obsewed gnzed depth should represent the minimum available Fraction of

each plant species. Therefore. the consideration of the entire above-ground biomass

would increase the expected encounter rate with each species, which tends to shift

predictions towards even narrower diets. Given that our models already suggest a

specialized diet. taking a more liberai estimate of resource availability would result in

identical predictions during most of the year (with spring. period 3. as the only possible

exception).

The observed diet was most consistent with short-term rather than long-term

goals. For half of the sampling periods (Le. periods 2,4 and 9, short-terni gains occurred

at the expense of long-terni gains. Bison avoided Agropyron spp., which would have

enhanced daily intake, prefemng instead Carex arherodes. Such findings are important

because they constitute direct evidence that foraging decisions by bison reduce their

potential long-term energy gain, contrary to established principles of classic optimality

models (Barkan and Withiam 1989). The foraging behavior of black-capped chickadees

has also been investigated under multi-temporal scales, with Barkan and Withiam (1989)

concluding that chickadee foraging was more consistent with short-term than long-term

goals. The authors argued that the results of Green et al. (1 98 1) about pigeons could also

be explained by short-term energy maximization.

It appears that the relevant temporal scale could vary across species of large

herbivores. Wallis de Vries and Daleboudt (1994) observed that foraging decisions by

cattle were more consistent with long- than short-term energy maximization. In contrast.

Illius et al. (1 999) observed that goats (Capra hircus) selected a diet maximizing short-

t e m forage intake, which also coincided with short-tem energy intake. Maxirnization of

daily energy gains was unlikely in Illius et al.'s (1999) study because of poor covariance

between short-term energy intake rate and forage digestibility arnong plant species.

Probably the most striking support for our interpretation coma from the study by

Bergman et al. (300 1). Their study focused on a single plant species. Carex atherodes,

the same species preferred by bison in Prince Albert National Park. Optimality modeling,

based on patch choice rather than diet choice, was used to detemine whether foraging

decisions led to the maximization of either short- or long-term energy gains in wood

bison (Bison bison arhabascae). It was found that wood bison displayed patch

preferences most consistent with the maximization of short-term goals. Our combined

findings suggest that bison make consistent foraging decisions at different hierarchical

levels: first ma..imizing short-term energy intake rate by choosing Carex arherodes. and

then improving gains even further by selecting patches yielding the highest short-tem

gains.

Several reasons could explain short-term maximization by bison. First, bison rnay

need to get relief fiom insect harassrnent, to scan for predators, or to maintain thermal

balance or social statu (Bergman et al. 200 1). The time saved by selecting a diet that

maximizes short-term intake would be less than 63 min (range: 3 1 - 63 min) in our study.

This time saving appears rather small, but we have no idea of its potential fitness

importance. Kagel et al. (1986) mentioned that sources of uncertainty that may intempt

an animal's foraging activity can lead to discounting of future rewards. The gregarious

nature of bison exposes them to cornpetition via hth exploitation and interference,

perhaps making resource availability uncertain. Herd movements are determined by a

small number of individuals (McHugh 1 %8), making feeding opportunities somewhat

unpredictable for other individuals. Disturbance by predators or humans can prematurely

terminate foraging bouts. We suggest that the uncertainty of foraging time or resource

availability could increase the value of rapid energy acquisition at the expense of long-

term gains.

Despite good agreement during most of the year, the observed diet during spring

did not correspond to predictions given by either of ow models. The conswnption of

Hordeum jubatirm and Agropyron spp. would have maximized both long- and short-term

energy gain. but bison avoided them in deference to Carex atherodes. The importance of

Carex atherodes in the diet of bison has k e n reported in other landscapes with similar

plant availability (Reynolds et al. 1978, Larter and Gates 1991, Bergman et al. 2001).

However. as in these studies, we observed that bison did not specialize solely on Carex

atherodes. The bison of Prince Albert National Park consurned few other species.

including a noticeable arnount of Scolochloafesrucacea. This species often had the

Chapter 4

second highest short-term profitability (Utmin). A broader diet than predicted is

commonly observed in large herbivores (Belovsky 1978, Owen-Smith and Novelie 1982,

Owen-Smith 1993, Wilmshurst et al. 1995, Illius et al. 1999, Bergman et al. 2001).

Factors such as animal variation, discrimination errors, and sampling update information

can be used to explain such partial preferences (Wilmshurst et ai. 1995, Illius et al. 1999).

Because of incomplete information about resource distribution, the sequence of resources

encountered can influence the likelihood of inclusion in the diet. The non-omniscience of

bison in their perception of diet quality can lead to diet expansion (Chapter 1). Finaily, an

animal's optimal diet could change during the course of a day (Newman et ai. 1995) or as

a function of animal state (Edwards et al. 1994).

The generally good agreement provided by our approach suggests that the

foraging decisions of bison can be largely explained by simple rules of energy-

maximization and that bison preferred rapid energy acquisition, which may come at the

expense of long-term gains.

Funding for this work was provided by Parks Canada, University of Guelph. and

scholarships from the Fonds FCAR and OGS to DF. 1 thank Mark Andruskiw and Sergio

Jujrez. as well as many volunteers for their help in the field and the laboratory. Park

Wardens Lloyd O'Brodovich, Dan Frandsen and Norman Stolle made the fieldwork

possible from their logisticai support. John Wilmshurst, Tom Nudds and Bart Nolet

provided constructive comments that helped to improve this paper.

Altmann, J. 1 974. Observational study of behavior: sarnpling methods. Behaviour 49:

227-267.

Arnold. G. W. 1985. Regulation of food intake. in: R. J. Hudson and R. G. White, (eds),

Bioenergetics of Wild Herbivores. CRC Press, Inc., Boca Raton, Florida, pp. 8 1 - 101.

Bailey, D. W., Gross, J. E., Laca, E. A.. Rittenhouse, L. R., Coughenour, M. B., Swift. D.

M. and Sims, P. L. 1996. Mechanisms that result in large herbivore grazing

distribution patterns. Journal of Range Management 49: 386-400.

Barkan. C. P. L. and Withiam. M. L. 1989. Profitability, rate maximization, and reward

delay: a test of the simultaneous-encounter mode1 of prey choice with Parus

atricapillus. American Naturalist 134: 254-272.

Bélovsky. G. E. 1978. Diet optimization in a generalist herbivore: the moose. Theoretical

Population Biology 14: 105- 134.

Belovsky. G. E. 1986. Optimal foraging and community structure: implications for a

guild of generalist grassland herbivores. Oecologia 70: 35-52.

Belovsky, G. E. and Slade. J. B. 1986. The importance of body size and thermal

environment in the time budgets of some grassland herbivores. Oecologia 70: 53-

62,

Bergman, C. M., Fryxell, J. M. and Gates, C. C. 2000. The effect of tissue complexity

and sward height on the functional response of Wood Bison. Functional Ecology

14: 6 1-69.

Chapter 4

Bergman, C. M., Fryxell, J. M., Gates, C. C. and Fortin, D. 2001. Ungulate foraging

strategies: energy maximizing or time minimizing? Journal of Animal Ecology (in

press).

Bradbury, J. W.. Vehrencamp, S. L., Clifion, K. E. and Clifion, L. M. 1996. The

relationship between bite rate and local forage abundance in wild Thomson's

gazelles. Ecology 77: 2237-2255.

Burlison. A. J., Hodgson, J. and Illius, A. W. 1991. Sward canopy structure and the bite

dimensions and bite weight of grazing sheep. Grass and Forage Science 46: 29-

38.

Edwards. G. R.. Newman, J. A.. Parsons, A. J. and Krebs. J. R. 1994. Effects of the scale

and spatial distribution of the food resource and animal state on diet selection: An

example with sheep. Journal of Animai Ecology 63: 8 16-826.

Edwards. G. R.. Parsons. A. J.. Penning. P. D. and Newman, J. A. 1995. Relationship

between vegetation state and bite dimensions of sheep grazing contrasting plant

species and its implications for intake rate and diet selection. Grass and Forage

Science 50: 378-388.

Farnsworth. K. D. and Illius. A. W. 1996. Large grazers back in the fold: generalizing the

prey model to incorporate mammalian herbivores. Functional Ecology 10: 678-

680.

Farnsworth. K. D. and Illius, A. W. 1998. Optimal diet choice for large herbivores: an

extended contingency model. Functional Ecology 12: 74-8 1.

Fortin. D. 2001. An adjustment of the extended contingency model of Farnsworth and

Illius ( 1 998). Functional Ecology 15: (in press).

Chaprer 4

Fryxell, J. M. 1991. Forage quality and aggregation by large herbivores. American

Naturalist 138: 478-498.

Fryxell, J. M. and Doucet, C. M. 1993. Diet choice and the functional response of

beavers. Ecology 74: 1297- 1306.

Gass. C. L. and Roberts, W. M. 1992. The problem of temporal scale in optimization:

three contrasting views of hummingbird visits to flowers. American Naturalist

140: 829-853.

Gordon. 1. J.. Illius, A. W. and Milne, J. D. 1996. Sources of variation in the foraging

eficiency of grazing ruminants. Functionai Ecology 10: 2 19-226.

Green. L.. Fisher Jr., E. B., Perlow, S. and Sherman, L. 1981. Preference reversal and self

control: choice as a hnction of reward arnount and delay. Behaviour Analysis

Letters 1 : 43-5 1.

Gross. J. E.. Shipley, L. A., Hobbs, N. T., Spaiinger, D. E. and Wunder, B. A. 1993.

Functional response of herbivores in food-concentrated patches: tests of a

mechanistic mode]. Ecology 74: 778-791.

Hudson, R. J. and Frank, S. 1987. Foraging ecology of bison in aspen boreal habitats.

Journal of Range Management 40: 71 -75.

Illius. A. W.. Gordon. 1. J., Elston, D. A. and Milne, J. D. 1999. Diet selection in goats: a

test of intake-rate maximization. Ecology 80: 1008-1 0 18.

Illius, A. W. and O'Connor. T. G. 2000. Resource heterogeneity and ungulate population

dynarnics. Oikos 89: 283-294.

Kagel. J. H.. Green. L. and Caraco, T. 1986. When foragers discount the future:

constraint or adaptation? Animal Behaviour 34: 271-283.

Laca, E. A., Ungar, E. D., Seligman, N. and Demment, M. W. 1992. Effects of sward

height and bulk density on bite dimensions of cattle grazing homogeneous swards.

Grass and Forage Science 47: 91 -1 02.

Langvatn R. and Hanley, T. A. 1993. Feeding-patch choice by red deer in relation to

foraging eficiency: an experiment. Oecologia 95: 164-1 70.

Larter. N. C. and Gates, C. C. 1991. Diet and habitat selection of wood bison in relation

to seasonal changes in forage quantity and quality. Canadian Journal of Zoology

692677-2685.

Lucas. J. R. 1983. The role of foraging time constraints and variable prey encounter in

optimal diet choice. American Naturalist 122: 191 -209.

Lucas. J. R. 1990. Time scale and diet decisions. In: R. N. Hughes, (ed). Behavioural

mechanisms of food selection. NATO AS1 series G, vol 20, Springer, Berlin.

Heidelberg. New York, pp. 165-1 84

McHugh. T. 1958. Social behaviour of the American Buffalo (Bison bison bison).

Zoologica 43: 1-40.

MerriIl. E. H. 1991. Thermal constraints on use of cover types and activity time of elk.

Applied Animal Behavior Science 29: 25 1-267.

Michaelis. L. and Menten, M. L. 1913. Kinetics of invertase action. Biochemische

Zeigschrift 49: 333-369.

Mould. E.D. and Robbins, C. T. 1982. Digestive capabilities in elk compared to white-

tailed deer. Journal of Wildlife Management 46: 22-29.

National Research Council (subcommittee on beef cattle nutrition). 1996. Nutrient

requirements of beef cattle. National Academy Press, Washington.

Newman, J. A., Parsons, A. J., Thornley, J. H. M., Penning, P. D. and Krebs, J. R. 1995.

Optimal diet selection by a generalist grazing herbivore. Functional Ecology 9:

255-268.

Owen-Smith. N. 1993. Evaiuating optimal diet models for an Afncan browsing ruminant,

the kudu: how constraining are the assumed constraints? Evolutionary Ecology 7:

499-524.

Owen-Smith, N. and Novellie, P. 1982. What should a clever ungulate eat? American

Naturalist 1 19: 15 1-178.

Reynolds. H. W.. Hansen. R. M. and Peden, D. G. 1978. Diets of the Slave River

lowland bison herd, Northwest Tenitories, Canada. Journal of Wildlife

Management 42: 58 1-590.

Shipiey, L. A.. Spalinger, D. E., Gross, J. E., Hobbs, N. T. and Wunder, B. A. 1996. The

dynarnics and scaling of foraging velocity and encounter rate in mammalian

herbivores. Functional Ecology 10: 234-244.

Spalinger, D. E. and Hobbs, N. T. 1992. Mechanisms of foraging in marnmalian

herbivores: new models of functional response. American Naturalist 140: 325-

348.

Stephens, D. W. and Krebs, J. R. 1986. Foraging theory. Princeton University Press.

Princeton, New Jersey.

Tilley. J. M. A. and Terry, R. A. 1963. A two-stage technique for the in vitro digestion of

forage crops. Journal of the British Grassland Society 18: 104-1 1 1.

Turner, M. G., Wu, Y., Wallace, L. L., Rornrne, W. H. and Brenkert, A. 1994. Simulating

winter interactions arnong ungulates, vegetation, and tire in Northem Yellowstone

Park. Ecological Applications 4: 472-496.

Ungar. E. D. and Noy-Meir, 1. 1988. Herbage intake in relation to availability and sward

structure: grazing processes and optimal foraging. Journal of Applied Ecology 25:

1045-1062.

Van Soest. P. J. 1994. Nutritional ecology of the ruminant. Second edition. Comell

University Press, Ithaca, New York.

van Wieren. S. E. 1996. Do large herbivores select a diet that rnaxirnizes short-term

energy intake rate? Forest Ecology and Management 88: 149-1 56.

Vartha. E. W. and Matches, A. G. 1977. Use of a weighteddisk rneasure as an aid in

sarnpling the herbage yieid on taIl fescue pastures grazed by cattie. Agronomy

Journal 69: 888-890.

Wallis de Vries. M. F. and Daleboudt, C. 1994. Foraging strategy of cattle in patchy

grassland. Oecologia 100: 98- 106.

Wilrnshurst. J. F. and Fryxell, J. M. 1995. Patch selection by red deer in relation to

energy and protein intake: a re-evaluation of Langvatn and Hanley's (1 993)

results. Oecologia 104: 297-300.

Wilmshurst. J. F.. Fryxell, J. M. and Colucci, P. E. 1999. What constrains daiIy intake in

Thomson's gazelles? Ecology 80: 2338-2347.

Wilmshurst. J. F.. Fryxell. J. M. and Hudson. R. J. 1995. Forage quality and patch choice

by wapiti (Cervus elaphus). Behavioural Ecology 6: 209-2 17.

Chaprer 4

Table 1. Foraging parameters and characteristics used to determine the optimal diet of

bison during 6 sampling periods of 1998 in Prince Albert National Park. Acronyms of the

plant species considered are agro: Agropyron spp., caaq: Carex aquatilis, caat: C.

atherodes, cain: Calamagrosris inexpansa, hoju: Hordeum jubatum, juba: Juncus balticus

and scfe: Scolochloafesiucacea. Sampling periods correspond to (1) 5 January to 16

February, (2) 17 February to 4 April, (3) 23 May to 19 June. (4) 20 June to 12 July. (5) 13

July to 7 August. and (6) 8 August to 3 September.

Sampling Plant species peciods apro caaq caat cain hoj u juba SC fe

A) Elongated height (cm), n = 8-59 1 33.74 74.21 101.30 63.26 32.76 49.66 104.04 2 34.01 68.89 96.19 57.03 30.69 51.06 100.40 3 23.34 38.67 35.52 27.1 1 16.74 27.94 34.54 3 28.23 48.1 1 47.95 35.73 27.58 44.22 49.36 5 40.40 65.41 81.39 45.17 31.94 45.63 74.35 6 48.3 1 72.00 87.06 48.89 30.54 52.4 1 8 1.83

B) Total biomass (g/m2), n = 752-964 1 16.00 21.01 65.42 54.76 1.1 1 8.54 88.95 3 - 15.17 23.19 64.94 53.34 1.10 8.28 86.85 3 6.75 5.08 17.41 13.68 0.5 1 7.25 11.02 4 12.03 6.86 66.72 27.26 1 .O0 8.08 34.58 5 19.23 19.76 121.65 46.34 3.93 5.34 63.71 6 36.64 23.27 135.52 44.61 7.07 4.67 80.59

C) Biomass gazed individual plant (%). n = 10-44 1 38.25 30.98 40.23 20.48 39.30 31.06 22.15 2 38.07 30.98 40.23 25.47 39.30 31.06 23.28 3 45.03 30.98 40.23 49.41 39.30 31.06 43.68 4 41.84 30.98 40.23 42.51 39.30 31.06 39.09 5 33.91 30.98 40.23 34.96 39.30 31.06 3 1.35 6 28.76 30.98 40.23 31.98 39.30 3 1 .O6 29.03

D) Encounter rate (glmin) 1 194.13 206.45 834.84 355.77 13.89 84.12 624.97 2 183.18 227.88 828.73 430.98 13.77 81.60 641.26

Chaprer 4

E) Handling time (minig) 1 0.0203 0.01 75 2 0.0203 0.0177 3 0.0181 0.0171 4 0.0178 0.0165 5 0.0172 0.0157 6 0.0169 0.0155

F) Proponion of handling tirne exclusive to searching activity 1 0.68 0.62 0.58 0.63 0.65 2 0.67 0.62 0.59 0.63 0.65 3 0.45 0.43 0.44 0.46 0.4 1 4 0.44 0.42 0.4 1 0.44 0.41 5 0.43 0.40 0.38 0.42 0.41 6 0.43 0.39 0.37 0.42 0.41

G ) Dry matter digestibility (%), n = 6-1 0. with the exception * n = 3. ** n = 5 1 54.40 40.17** 56.1 1 44.12 48.06 45.64 2 53.97 41.31 51.98 40.36 47.01 46.67 3 75.79 62.58 71.24 61.96 76.06 68.85* 4 68.56 58.09 65.30 53.06 55.38 53.76 5 65.96 54.83 64.53 54.15 60.21 56.22 6 56.55 52.66 60.16 43.74 58.88 53.83

H) Digestible energy content (kJ/g of dry matter) 1 10.0 1 7.39 10.33 8.12 8.85 8.40 2 9.94 7.60 9.57 7.43 8.65 8.59 3 13.95 1 1.52 13.12 11.41 14.00 12.67 4 12.62 10.69 12.02 9.77 10.20 9.90 5 12.14 10.09 11.88 9.97 1 t .O9 10.35

Table 2. Ivlev's electivity index indicating preference or avoidance in plant species

grazed by bison in Prince Albert National Park. Preference for Carex atherodes and

Scolochloafesfitcacea is suggested by positive indices. Species name acronyms and dates

of the investigation periods are indicated in Table 1.

Period Plant Species agro caaq caat cain hoju juba scfe others

1 0.4 1 -1.00 0.46 -0.53 -1.00 -1.00 -0.18 -0.87

Figure 1. Fraction of plant biomass consurned by bison as a function of their height. (A)

The fraction grazed declines significantly with height of three species, whereas (B) there

was no relationship for the last four focal species, and their averages are displayed. Note

that Carex aqiratilis and Juncus balticus both averaged 3 1% of biomass grazed.

Figure 2. Relation between bison intake rate and sward biomass in winter and during the

growing season (late-spring and summer). lntake rate was calculated from the product of

bite rate and bite mass. which came from behavioral observations and from the literature.

respectively.

Figure 3. Relation between cropping rate of bison and sward biomass in winter and

during the growing season in Prince Aibert National Park.

Figure 4. Change in short-terni (A) and daily (B) energy intake rate with an increase in

diet breadth. as predicted from contingency models for six periods of 1998. Prey species

are ranked according to their short-terni (kitmin) or daily (MJJday) profitability from left

to right. and sequentially added to the bison's diet (i.e. the diet includes one prey type to

seven prey types). A direct decline in intake rate indicates that a specialized diet would be

optimal. Species name acronyms and dates OP the investigation periods are indicated in

Table 1.

Chaprer 4

Figure 5. Comparison behveen the observed diet of bison in Prince Albert National Park

and the optimal diet predicted for short-term and daily intake rate maximization during

six periods of 1998. Species name acronyms and dates of the investigation perds are

indicated in Table 1.

Chapier 4: Figure 1

Height (cm)

O

+ Carex aquafilis . o.. Carex alherodes

O - - - Hnrdgurn jubafum , -, .,..,Js baificus

-t Agropyron spp. 4.. Calamagmstis inexpensa

- - . Swlochloa kstucacea

Height (cm)

I

Chupter 4: Figure 2

Dry biornass (V, dm2)

1 0- F = (78.43 Y) 1 (1 10.65 + V) Growing season O , i I i

O 200 400 600 800 1000

Chaprer 4: Figure 3

a 1 + Winter 1 . . .o. - - Growing O

400 600 a00

Dry biomass (g/m2)

Chapier 4: Figure 4 A

740 . Period 1 Period 4

720 - 710 -

700 -

h 560 ::: . \ 690

r .- E 540 --- 680 -2

3 caat scfe agm hoju Cain juba c a q caat scfe agro caaq hoju juba cain 5 al 600 -- c.

820 Period 2 Penod 5 E

Qi X (O CI

C ,- %

g 540 C al E 520 8 z 500 720 O x caat scfe agro hoju juba caaq Cain eaat scie agro hoju caaq juba Cain VI 900 760

Petitni 3 *. Penod 6 r----- 740 1 --•

i IO0 660

hoju agro caat scfe juba caaq Cain mat hop scfe caaq agro luba Cain

Oiet breadth

Chopter 4: Figure 3 B

125 . . . .. . . . . m a caat agro scfe hoju juba cain caaq

Y Peffod 2 1 4 5 . .

al ,, 120 .- CU agro caat scfe hoju jubacaaqcain

260 240 Period 3

* - - 220 .- -6-

a 200 .

130 .

160 140

180 - .

agro caat scfe caaq hoju juba Cain 200

Period 5

agro caat hoju scfe juba caaq cain t75

Period 6 170 . * -* ---*

80 150 hoju agro caat scfe jubacaaq cain caat hoju agro juba scfe caaq cain

Chapter 4: Figure j

Short-term intake rate Daily intake rate

agrocaaq caat cain hoju juba scfeother agrocaaq caat cain hoju juba scfeother

100 4 n

Plant species

fl Period 6

CHAPTER 5

A multiscale investigation of bison distribution and resource use

DANIEL FORTIN.

JOHN M. FRYXELL, LLOYD O'BRODOVICH AND DAN FRANDSEN

Predictions of animal distribution and resource use require multi-scale

consideration because animals can use different sets of selection criteria at different

scales. We investigated whether the observed patterns of distribution and resource use by

free-ranging bison can be attributed to rules of energy maximization that hold across

multiple scales. Optimality theory predicts specialization on Carex atherodes and

frequency-independent selection among plant species, that is, local variation in C.

orherodes biomass should not influence the diet but only the t h e spent in individual

patches.

Contrary to our predictions, the probability of meadow use and the average group

size observed in meadows were not related to the presence of C. atherodes. During

summer and winter, general landscape features within the daily radius of bison (2 km in

summer and 1.3 km in winter), together with abiotic characteristics of meadows,

appeared to have the major influence on the probability of meadow utilization. The

summer use of resources found within meadows kvas. however. closely related to energy

ma.~imization principles. Namely, C. atherodes dominated bison diet, this species was

seIected in al1 meadows, and diet composition was independent of frequency in the

environment. In winter, diet was still dominated by C. atherodes, but the selection of

Scolochloajëstttcacea and Cirsium arvense conflicted with theoretical predictions.

Additionally. the relative selection of C. atherodes and S. festucacea was positive

fiequency-de pendent.

Chapter 5

Our study suggests that bison distribution and resowce use are influenced by both

abiotic and biotic factors, which Vary in relative importance at different spatio-temporal

scales. We propose that the complex nature of selection reduces the utility of optimality

diet models based solely on energy maximization principles for predicting patterns of

grazing distribution across spatio-temporal scales.

Understanding the deteninants of the distribution and abundance of organisms is

fundamental to ecological studies (Krebs 1985). Adequate prediction of animai

distribution patterns require multi-scale consideration because diverse sets of selection

criteria can used at different scales (Se& et al. 1987, Orians and Wittenberger 199 1.

Bergin 1992, Schaefer and Messier 1995, Wallace et al. 1995). Congruence in selection

across scales has, however. also been reported for several animal species (Ward and Saltz

1994, Sedgwick and Knopf 1992. Carey et al. 1992. Hall and Mannan 1999). These

conflicting observations emphasize the importance of multiscale studies. Nonetheless,

such studies are still limited compared to investigations at a single scale (Ward and Saltz

1994. Bergman et al. 2001).

As suggested by Levin ( 1 992), global comprehension of ecologicai systems

requires understanding of how information is transferred across scales. Levin indicated

that we must attempt to simpliQ and retain only the most essential information of

cornplex systems. Modeling c m be used to detemine the minimal arnount of detail that

reproduces observed patterns, potentidly illuminating the essentiai nature of ecological

systems. Optimal foraging theory (Stephens and Krebs 1986) can provide such a

framework allowing investigation of whether animal distribution can be simply

explained. at various spatio-temporal scales, by the spatial patterns of resources that

maximize a given currency. The study of Ward and Saltz (1994) is encouraging in this

respect. because it reports field observations consistent with optimal foraging predictions

at different spatial scales.

Contingency models of optimal diet predict which prey types should be eaten, and

in what relative proportion (Stephens and Krebs 1986, Farnsworth and Illius 1998, Fortin

2001). 1 showed in Chapter 4 that, during summer and winter, bison would maximize

short-term gains by specializing on Carex atherodes, which was closely reflected by

obsewed grazing patterns. Predictions were tested, however, only at the landscape level.

Therefore. it remains unknown whether selection for C. atherodes can explain bison

distribution at different spatial scaies and throughout the year.

Regardless of whether short- or long-tem gains are maximized by an animal,

contingency models (Stephens and Krebs 1986. Farnsworth and Illius 1998, Fortin 2001,

Chapter 4) also suggest positive frequency-dependence in prey selection within an

animal's home range (Hubbard et al. 1982). As the availability of highly profitable prey

(either in kJ/min or kJ/day) decreases, their encounter rate should similarly decrease,

eventually leading to diet expansion. In other words, decline in optimal prey types should

reduce the attack rate by predators, inducing positively frequency-dependent selection

(Fryxell and Lundberg 1997).

This theoretical prediction should be. howevec dependent on spatial scale. In a

patchy environment. depletion of the best prey has to occur at the landscape level in order

for a positive frequency-dependence to occur. McNarnara et al. (1993) demonstrated that.

in heterogeneous environments' variation in prey availability among patches (or even

resource depletion within patches) should not influence prey selection but only patch

residence time. Therefore, prey selection among patches should be frequency-

independent.

We tested whether energy maximizsttion pinciples as predicted by optimal diet

Chapter 5

theory can explain bison distribution and resource use across scales. We started by

determining, at the landscape level, whether bison select and spend most of their t h e in

meadows compared to other types of habitat. We then predicted that bison should spend

more time in meadows offering high avaitability of C. atherodes. that larger herds should

occur in meadows offering high abundance of C. atherodes, and that selection for C.

a~herodes should occur in al1 meadows. This selection should be frequency-independent.

and thus insensitive to the inter-meadow variation in the biomass of C. atherodes.

Finally, these predictions should not Vary between surnmer and winter, because of the

consistency in optimal diet predictions.

Sttrdy area

We studied the free-ranging bison of Prince Albert National Park (53'44'N,

106°40'W) during the winters (1 January - 5 April) of 1997-1998 and summers (20 June -

6 September) of 1997-1 999. The population, estimated at 220 individuais (Fortin and

Frandsen 1999), is established in the Southwest corner of the park. This area of the park

is characterized by approximateIy 85% forest, 10% meadow and 5% water. The forest is

mainly deciduous in the southem part of the bison range and coniferous in the northem

part. Richness of plant species in individual meadows ranged From 16 to >62, and.

overall. more than 170 plant species were recorded.

Although a trail network allows public access to a mail portion of the bison

Chapter 5

range. relatively few visitors travel in the area. A small wolf (Canis lupus) population is

present but bison predation appears rare, presumably due to the abundance of alternative

prey such as white-tailed deer (Odocoileus virginianus).

Plant und snow survey

Plant phenology and snow conditions were followed in 25 meadows distributed

throughout the bison range. Plant characteristics were determined in 16-90 evenly spaced

qudrats. depending on meadow area. Quadrats were I m' during the summer and 0.25

ml in winter. Total dry biomass (dm2) of summer forage was determined by measuring

the height (in cm) that a calibrated disk settled fiom the ground (Vartha and Matches

1977): y = 8 1.958 + 10.004 x = 162.99, R' = 0.48. P < 0.0001). In winter, plant

0.64 lx biomass was estimated From a calibrated 0-5 visual scale: y = 67.876e (Fi+4z =

295.1 1, R' = 0.88, P < 0.0001). Evaluation of plant biomass was supplemented by

estimation of percent cover and, in surnmer, percent green biomass of each plant species.

Snow was characterized by its depth. density, sohess. and the presence of cmst.

Snow density (g/cm3) was detenined by weighing, with a spring scale, a sarnple of the

snow column collected with a metal tube inserted vertically into the snow and dividing

the mass by the volume of the snow gathered. Snow softness was indexed by the sinking

depth of a bottle (300 g, 8.5 cm in diameter) dropped 50 cm above the snow surface

(Murray and Boutin 1991).

Bison diet was determined in al1 quadrats by visually estimating the percent

grazed of each plant species. Although white-tailed deer, elk (Cervus elaphus) and moose

(Alces alces) were present, their occurrence in meadows was relatively infrequent (Figure

1). Hence, we expect tliat the observed grazing patterns should largely reflect bison

selection.

Animal Locations

During the entire study period (1 997- 1999), a total of 10 fernale bison were

equipped with Global Positioning System (GPS) collars (GPS 1000 collar fiom Lotek

Engineering, inc. Newmarket, Ontario, Canada). Some females were followed for more

than one season. In 1997, we followed two individuals during the surnmer and four in

winter: in 1998. a total of four females were tracked both in summer and in winter: an

additional four individuals were followed in the summer of 1999. In winter of 1997,

animal locations were recorded at 00:OO and every 2 hours fiom M:00 to 16:OO. During

the rest of the study period. animal locations were recorded every h e e hours, starting at

00:OO. Overall. 79 % of the potential fixes were successtiilly recorded. Differential

correction of these locations was successful at 98%. providing accuracy within 12 m 95%

of the time (Moen et al. 19%). The other 2% of locations were accurate within 40 m 50%

of the time and ~ i t h i n 100 m 95% of the time.

Bison surveys

In 1997 and 1998, the 25 meadows were also opportunisticaily surveyed for the

presence of bison herds. Each meadow was inspected a maximum of once s day. For each

season, the two years of observations were pooled before analysis, and only meadows

surveyed more than 10 times (range: 10-12 1 surveys per season) were considered. This

criterion reduced the sample size to 22 meadows in surnmer and 23 meadows in winter.

Geographic Information System

Aerial photographs (l:60,000) taken in 1995 and 1999 were digitized, geo-

referenced. and imported into a Geographic Information System (GIS). No photographs

were available for two of the 25 meadows surveyed for plant and snow characteristics.

These two meadows were. therefore, excluded fiom our large-scale analysis of habitat

selection. Using ArcView GIS (Version 3. l) , a buffer of 2 km in diameter was drawn

from the edge of the 23 meadows. Al1 distinguishable meadows, rivers, ponds and lakes

within this large-scale study area were drawn and included as GIS layers of information.

Data analysis

Habitat use

The large-scale study area was divided into six habitat types: agiculturai land,

meadow, meadow edge (forested area lwated I 25 m fiom meadows), forest interior (>

25 m ffom meadows), water shore (25 m from the shore line into the watered area) and

water interior (> 25 m fiom water shore). Ia winter, water shore and water interior were

frozen and bison use snow as a water source. Consequently, these two categories were

both considered as water interior. The proportion of GPS-fixes recorded in each habitat

type was determined for every animal.

Habitat use was investigated using compositional analysis (Aebischer et al. 1993).

Although this technique overcomes some problems faced by other types of analyses (see

Tufto et al. 1996), use or availability estimates of O cannot be directly included in the log-

ratio computation. Based on Aebischer et al. (1993) and Pendleton et al. (1998), unused

habitats were replaced by a usage of 0.00 1 %, which is a value lower than any observed

use.

!Meadow selection

Habitat features influencing the probability of meadow use were investigated. for

each season, with stepwise logistic regressions. Regression models included only

significant variables at P < 0.05. To avoid multicollinearity problems, we followed a rule-

of-thumb comrnonly used with multiple regression analysis that consists of considering

only one of two variables that are more closely related to each other (significantly) than

to the dependent variable (Hair et al. 1992). The independent variables considered

included meadow characteristics. as well as characteristics of the area covered by a 1-

and 2-km radius starting at the meadow edge (see Table 3 and 4 for a complete list of

independent variables). Two sets of analysis were performed considering different

dependent variables: the first one based on GPS-locations, and the second, on bison

surveys.

CPS-collurs. - GPS-fixes associated with the selection of a given meadow

included al1 fixes recorded in that meadow, as well as in the meadow edge because bison

were often observed lying in edges between feeding bouts. Meadows never visited were

assigned a value of O in the logistic regressions, whereas other meadows were assigned a

value of 1. Individual bison were considered as replicates. Years were pooled before

analysis of summer data, because not dl habitat information was available annually.

Bison sirrveys. - Meadows in which bison were observed during less than 10% of

the surveys were considered as rarely used, i.e. 4 out of 10 swveys, which was the

minimum number of surveys done for any meadow during any season. These rarely-used

meadows were assigned a value of O in the logistic regressions, whereas other meadows

had a value of 1.

Groirp size

Factors related to the average size of herds observed in meadows were

investigated using stepwise multiple regressions. Because group size was influenced by

meadow characteristics (see resulis). seasonal variation in herd size was tested using a

paired t-test.

Plant selection

Plant selection was investigated fkom the plant sweys of 1998. Three surveys

were done in summer and four in winter. Before analysis, biomass used and available was

averaged among surveys for each plant species. Compositional analysis was then

performed considering individual meadows as the sampling unit. Unused plant types

were replaced by a usage of 0.001%. In contrast to large-scale habitat use, where al1

habitat types were available for every animal. not ail plant types were available in every

meadow (Le. presence of O availabiiity). Therefore. to make resource use in meaciows

more comparable and to reduce the number of zeros, we restricted our analysis to the

eight most common plant types (other plant types were pooled in a ninth category). and to

the meadows frequented &y bison where these plants were most abundant (Le. 12

meadows in summer and 11 in winter). Overall, these eight plant species comprised 95%

of the bison diet in summer (range among meadows: 70-100%) and 96% in winter (range:

63 - 1 00%).

Althouph this approach minimized the number of meadows where some plant

types were unavailable. such cases remained. Unavailable resources resulted in missing

values in the matrix of "log-ratio transfonned difference value" (d-matrix, Pendleton et

al. 1998). Before calculating the likelihood ratio test. missing values were replaced by the

rnean of the nonmisshg values for that habitat class (see Aebischer et al. 1993 and

Pendleton et al. 1998 for details). We then determined the significance level of the

likelihood ratio test using the standard appmach based on a dimibution Although

plant species use as a denorninator in the analyses (Scolochloafistucacea) was always

Chapier j

present, the absence of other plant species in some meadows could have led to a test that

may not perfectly conform to the 2 distribution (Aebischer et al. 1993). Nonetheless,

given the magnitude of the p-values (p < 4.7 x in both cases. see results), we felt

confident to have avoided type 1 error. Posr hoc pairwise cornparison consisted of paired

t-tests perfonned on the untransformed d-matrix, and thus directly took into account the

presence of missing values.

C. atherodes and Scolochloafesiucacea have been shown to have high short-term

profitability (in kJ/min) (Chapter 4). Therefore, we performed pairwise cornparisons

between C. arherodes. S. festircacea and others plant types combined to test frequency-

dependent selection in bison. Following Elton and Greenwood (1 987) we evaluated

frequency-dependence using the equation:

where U,, = total biornass of plant type i grazed / total biornass of plant type j grazed. A,,

= total biomass of plant type i available 1 total biomass of plant type j available. b,, is the

dope of the relationship between log U,, and log A,,, for which difference from unity

indicates frequency-dependence. Finally, "b, log 6,'' is the intercept of the relationship

for which a 6 , is a selectivity coefficient. An intercept different from O indicates plant

selectivity. As for the compositional analysis, unused plant types were replaced by a

usage of 0.00 1 %. but unavailable resources corresponded to missing values. The

significance of the estimated parameters was determined by t-test following Greenwood

and Elton ( 1979). Because we considered three groupings of plants, three tests were done

to cover ail pairwise comparisons. Risk of type I error was reduced by setting the

significance level to 0.01 7 following Bonferroni adjustment (Sokal and Rohlf 1995).

Habitat use

Female bison equipped with GPS-radio collars did not use habitat types in

proportion to their availability during both seasons (summer: X' = 50.20,9 df, P <

0.000 1. winter: X2 = 27.69.5 df, P < 0.000 1). These bison spent 69 * IO% (mean * sd.

range: 43-78%. n = 10 individuals) of their time in meadows, although their total area

covered onIy 9% of the bison range. Forested areas at the edge of meadows (< 25m from

edge) were also used more than expected from their availability. whereas little use was

made of other habitat types (Table 1). There was no significant seasonal variation in the

use of these habitats (MANOVA on q-matrix: Factor "season", Wilks' lambda = 0.5 1.

exactF=2.19,d.f.=nurn4,den9.P=O.15).

Daily displacements

Daily displacements by collared bison varied seasonally (3-way ANOVA: Factor

"season". F1.107~ = 49.2, P < 0.0001). The daily displacement corresponded to distance

between locations recorded at 00:OO on successive days. Overall, daily displacements

averaged 1986 * 1636 m (n = 538) in summer and 1297 1236 m (n = 537) in winter, a

significant difference (Tukey's studentized range test: P < 0.0 1).

Meadow selecrion

Green (1 992) considered that individual bison remaining at distances p a t e r than

100 m beionged to different herds. Based on this criterion, we observed that meadow

selection was not independent arnong all female bison equipped with GPS-collars due to

the gregarious nature of these animals (Table 2). Logistic regressions were therefore

performed on individual animals. For each season. larger meadows were generally more

likely to be visited (Table 3 and 4). Meadow utilization was also related to the resource

found within a distance reflecting the daily displacement of bison. In surnmer. rneadows

that were largely surrounded (i.e. radius of 2 km) by water were more likely to be visited.

This pattern remained consistent arnong individuals spending M e time with each other.

Herd surveys, which considered al1 types of bison herds (e.g. bull or mixed herds),

confirmed the importance of water availability on the probability of meadow use during

the summer. In winter. meadows with deep snow were less likely to be visited by bison in

1997 by some individuals (Table 4). This pattern vanished in 1998, a year with shallower

snow depth (paired t-test, t = 15.9 P = 0.0001, n = 25 meadows). Snow depth averaged 38

I 4 cm (range: 3 1-46 cm) in 1997 and 27 f 4 cm in 1 998 (range: 1 7-33 cm).

Resource biomass recorded within meadows had little influence on meadow use

by bison during any period of the year (Table 3 and 4). Even the biomass of the species

Chapter 5

dominating bison diet, C. atherodes, did not display a significant positive relationship for

any of our multiple logistic regression models (Table 3 and 4). Individual logistic

regressions confirmed the lack of significant positive relationships between meadow

selection (0-1) and biomass of C. atherodes (range of PkSE: -0.00620.005, P = 0.26 to

-0.01+0.005. P = 0.04 in swnmer and -0.02k0.02, P = 0.15 to 0.005k0.007. P = 0.48 in

winter). For exarnple, if C. atherodes influenced the summer selection of meadows, the

eight meadows located in the northern part of the bison range should have k e n heavily

used because they offered higher biomass of C. atherodes (squared-root transformed)

than the other rneadows (ANOVA. Flu = 13.58, P = 0.001). Instead, the northern

meadows were virtually ignored in summer. Only a single animal was observed during al1

summer surveys cornpared to 10 herds during winter. Similarly, one GPS-location was

recorded in these meadows during the summer compared to 1 17 locations by four

animals in winter.

Overall. our results thus suggest that meadow selection was influenced more by

large-scale landscape features than by the food they offered. The two sampling methods

of meadow selection offered consistent results, although the herd surveys confirmed

mainly the importance of meadow area on bison selection (Table 3 and 4).

Group size

Not only did the area of meadows influence their probability of k ing used, but

also the average herd size encountered in summer (figure 2). Surnrner herd size in

meadows was also related negatively to the absence of water in meadows as well as to the

vicinity of agricultural lands (Table 5). in winter, average herd size was not related to

meadow area (figure 2), but covaried only with the concentration of meadows within a 1 -

km radius (Table 5). There was no significant relationship between average herd size in a

meadow and the average biomass of C. atherodes recorded during the sumrner (Fl,iz =

0.08, P = 0.79) or the winter (FlVl2 = 2.77, P = 0.12).

The average group size was larger in summer than in winter (paired t-test, t =

2.25. n = 1 1. P = 0.05). Considering every herd observed chroughout the bison range in

1997 and 1998, group s i x averaged 9.8 i 8.6 individuals (mean sd, range: 1-50, n =

288) in winter and 21 -4 * 27.7 individuais (range: 1-168, n = 239) in summer.

Plant selecrion

Bison displayed strong selection among the plant species these meadows had to

offer (summer: X 2 = 76.73,8 df. P < 0.0001, winter: X' = 49.70.8 df. P < 0.0001). During

both seasons. plant species such as Agropyron spp., Bromus ciliatus, Calamagrosris

inexpansa were generally used l e s than expected from their availability (Figure 3). The

degree of selection for C. atherodes during summer was higher than that of any other

species. A similar trend in summer selection was found with the analysis of fiequency-

dependent selection (Table 6). Overall, the selection for C. aiherodes was consistent

across meadows (Figure 3)' and independent of its relative abundance (Table 6). as

predicted.

In winter. however, there was no statistical difference between the degree of

seIectioii of C. atherodes. S. fëstucacea and Cirsium awense (Figure 3. TabIe 6). These

Chapter 5

species were sornetimes grazed more heavily than expected and other times more lightly

than expected fiom availability (Figure 3). Such variation can be partly exptained by

bison switching behavior in food selection (Table 6). hdeed, in contmt to the summer

patterns, the relative consumption of C. atherodes and S. fesrucacea in winter was

positively frequency-dependent. which differed ftom our prediction. Further seasonal

variation in species selection was evident h m the use of Cirsium arverise. C. arvense

possesses spiny-toothed leaves that becorne softer in winter. This species was unutilized

in summer. but in winter, its selection equded that of C. atherodes and S.fesrtrcacea

(Figure 3).

Although. foraging decisions of bison have k e n s h o w io more closely reflect the

muimization of their short-term intake rate (Chapter 4). our results suggest that the

success of optimal diet information to predict grazing patterns strongly depends on

spatio-temporal scales.

Habitat use

FemaIe bison spent most of their time in meadows. This pattern was consistent

witii predictions of optimal diet models, because C. atherodes grows in open areas.

Female bison spent a limited amount of tirne using water areas. which simply indicates

that the need for water can be rapidly met. Also as reported by Krueguer (1986), forest

interior appeared to be used by bison mainly when traveling between meadows. ïhis

observation contrasts with the important use of forested areas by European bison (Bison

bonasus) (Krasinski 1978, Krasinski and Krasinska 1992). Despite the rninor use of forest

interior by plains bison. forested areas at the edge of meadows were used more than

expected from availability. We frequently observed bison resting in the shade at the edge

of meadows during hot sumrner days, a behavior also common in other ungulate species

(Arnold 1985). Lying in the shade cm reduce radiative heat gain. potentially reducing

water loss (Parker and Robbins 1985, Schmidt-Nielsen 1990) and preventing risk of

hyperthermia. The use of meadow edges was also frequent during gusty winter days. a

pattern previously reported (Fuller 1966 in Van Camp 1975). The decrease to wind

exposure should reduce convective heat toss through bowidary layer effects (Parker and

Robbins 1985, Gebrernedhin 1987, Bakken 1992). Seeking cover could thus potentially

reduce themoregulatory costs, although the therrnoneutral zone is broad in bison

(Christopherson et al. 1978). The important use of meadow edges already suggests that

factors other than energy intake can have an important influence on bison distribution.

Meadow selecrion

The factors associateci with meadow selection conflicted with our predictions

based on optimal diet for energy maximization. Indeed. the probability of meadow use

was not positively related to the biomass of C. arherodes during any season. Energy

maximization niles were, therefore. unsuccessfùl in predicting bison distribution at a

large spatial scale, presurnabIy because ecological factors affecting survival and

reproduction were scaie-sensitive (Bergin 1992).

Our study reveals that factors infiuencing the use of particular meadows are less

closely related to the forage species that meadows have to offer than to the resowces and

habitat features in the vicinity of these meadows. As pointed out Pearson et al. (1995), the

pattern of animal distribution is, at a large spatial-scale, a reflection of resource

distribution and movement capability by the animal. It appears that bison did not only

adjust their movements to the presence of snow, but perhaps also their perception of the

landscape. We found, indeed, that daily displacement of fernale bison passed from 2 km

in summer to 1.3 km in winter. Concunently, environmental variables influencing the use

of a meadow and herd size observed in that meadow were associated with the habitat

characteristics found within a 2-km radius during the surnmer and within a 1-km radius in

winter. This trend exists despite that both sets of independent variables (i.e.

characteristics within a 1-km and 2-km radius) were submitted to the analyses for both

seasons.

In summer. meadows largely surrounded by water areas within a 2-km radius

were more likely to be visited by bison. McHugh (1958) indicated that bison require daily

access to surface water in summer. Water distribution has been considered as the primary

determinant of the grazing distribution patterns of m e r s at large spatial scales (Baitey et

al. 1996). In winter, meadows largely surrounded by other meadows within 1-km (rather

than 2-km in summer) were more likely to be visited by larger herds than more isolated

sites. Snow depth also had some influence on the probability of meadow use by bison. a

finding supported by observations in other bison populations (Luter and Gates 199 1,

Pearson et aI. 1995). Snow conditions can influence factors such as animal movement.

energy costs of locomotion, feeding ability, probability of swvival and susceptibility to

predation (Van Camp 1975, Telfer and Kelsail 1984, Turner et al. 1994). Bison are

particularly inclined to the potentially negative effects of snow conditions compared to

other North Arnerican ungulates (Telfer and Kelsall 1984).

According to the theory of Ideal Free Distribution, animals should be distributed

arnong habitats proportional to rewards (Fretwell and Lucas 1970). Accordingly, we

predicted that group size and the biomass of C. atherodes should be positively related.

but significant relationships were not obsewed during any season. We could also predict

a relationship between group size and meadow size. No significant relationship was.

however, detectable in winter. We suggest that the absence of a relationship could partly

reflect seasonal variation in herd size. As commonly reported (e-g. McHugh 1958,

Komers et al. 1993), we obsewed large aggregations of bison during the summer.

Perhaps such aggregations are more likely in large areas simply by chance. Alternatively,

perhaps the size of winter herds may be suficiently small relative to meadow size to

avoid critical levels of competition for resource or space, dlowing the use of small

meadows by al1 herds.

Plant selecrion

During both seasons, C, arherodes held the highest rank in term of plant selection

by bison. and it dominated their diet. The importance of C. atherodes in bison diet has

been observed in other landscapes offering similar plant types (Reynolds et al. 1978.

Larter and Gates 1991, Bergman 2000), and has already been reported in Prince Albert

National Park (Chapter 4). What was not previously reported, however, was the variation

in the degree of success with which energy maxirnization principles can predict resource

use arnong meadows and across seasons. Optimal foraging predictions held up best

during the summer. Selection for C. atherodes was consistently observed in al1 meadows

and it was independent fiom the relative abundance of this species. The fiequency-

independence of sumrner diet selection in bison was consistent with dietary observations

on moose (Danell and Ericson 1986, Lundberg et al. 1990). Ow findings were therefore

consistent with McNarnm et al.'s (1993) predictions: the sununer diet of bison remained

immutable in a patchy environment.

Our winter predictions were less consistent with optimality predictions. Bison diet

was still dominated by C. atherodes, but S. fesrucacea and C. arvense were also often

selectively used. Moreover. we observed positive fiequency-dependence in forage

selection by bison. Winter foraging decisions thus appearcd to lead to sub-optimal diet

for the maximization of their short-term rate of energy intake. Several factors may

contribute to the divergence fiom our prediction. Perhaps, for example. the cost of

digging a new Crater (kJ/min) surpasses the difference profitability (measured in kJ/min)

between plant species co-occurring at the feeding station. Bison would then minimize the

number of feeding craters by maximizing the use of each one, which could still be

optimal in tems of net energy gains and would lead to diet expansion. Consistently, it

has been showed that increase in search cost c m lead to a broadening of the diet (Lifjeld

and Slagsvold 1988). Also, cratering patterns of reindeer (Rangifer tarandus) appears to

be responsive to changes in snow conditions (Collins and Smith 1991).

Given that optimal diet based on net energy gains could predict the use of both C.

Chapter 5

alherodes and S. festircacea, the search strategy of bison might be responsible for the

frequency-dependent selection obsewed between these two plant species. Bison use area-

restricted search to find their food in winter (Chapter 1). This strategy Ieads to increased

search efforts in areas where suitable food types are found. Stronger search effort should

occur at the vicinity of the more abundant of these two forage types because this type is

the more likely to be encountered. The spatial auto-correlation in plant distribution

should thus result in the encounter of the most abundant plant species in a proportion

exceeding availability, whereas the least abundant of the two plant species should be

encountered less oAen than expected from overall availability. Atthough the plant species

rare in a specific meadow could still be accepted in the diet whenever encountered. its

proportionally low encounter rate could resutt in a usage lower than expected based on

availability. This speculative explanation for the observed frequency-dependent selection

between C. citherodes and S. fissrucacea across meadows remains to be verified.

Integrarion across scales

The distribution of several ungulates species has been related to the spatial

variation in forage quality (see Albon and Langvatn 1992). Krasinska and Krasinski

( 1995) indicated that food selection is the major factor causing concentration of European

bison in particular mas. Mysterud et al. (1999) success~lly predicted habitat selection

by domestic sheep (Ovis aries) across spatial d e s bas4 on food availability. Ward and

Saltz ( 1 994) observed consistency in the foraging behaviour of doms gazetles (Guzella

dorcas) across spatial scaIes. Despite these observations, it appears that information on

diet selection cannot always be readily transferred to higher scale to explain animal

distribution.

Our study revealed that rules of energy intake maximization were, indeed,

insufficient to explain the distribution and resource use by plains bison across spatio-

temporal scales. In her study in the Mackenzie Bison Sanctuary (Northwest Temtories.

Canada). Bergman (2000) also reported that the relation between habitat use by wood

bison (Bison bison athabascae) and the distribution of C. atherodes varied among spatial

scaIes. Likewise, Schafer and Messier (1995) reported reversals of selection arnong

forage species by muskoxen (Ovibos moschatzrs) at different spatial scales. The scale-

dependent selection observed in our study could be the reflection that maximization of

bison fitness relies on trade-offs associated with nondietary goals (Senft et al. 1987,

Barton et al. 1992, Mystemd et al. 1999). in such instances, the complex nature of

selection can reduce the utility of simple optimality principles in predicting habitat

selection (Barton et al. 1992).

Our findings confirm Senft et al.'s (1987) contention that the relative importance

of plant-herbivore interaction decline at larger scales. Abiotic factors appear to become

more important (Bailey et al. 1996). Bison distribution is not only influenced by the

characteristics of meadows, but it is also linked to the characteristics of adjacent areas, an

effect also reported by Pearson et al. (1995) in ungulates and Fortin and Arnold (1997) in

birds. The exact nature of relationships between resource characteristics and animal

distribution is somewhat obscure (Barton et al. 1992). and our study underlines the

existence of important spatio-temporal vanations in the relative influence of abiotic and

biotic factors on bison distribution and resource use. Appropriate protocols of

management should therefore consider the hierarchical nature of bison selection.

The funding for this study was provided by Parks Canada, University of Guelph,

and scholarships from FCAR and OGS. We thank Mark Andruskiw, Régis Pilote and

Sergio Juiirez, as well as many volunteers for their help in the field and the laboratory.

Mark Andruskiw. Tom Nudds and Bart Nolet provided constructive comments on a

previous version of this manuscript. We are grateful to John McKenzie and Tracy Hillis

for sharing their experience with ArcView. We acknowledge the Southem Caiifornia

Integrated GPS Network and its sponsors. the W.M. Keck Foundation. NASA, NSF.

USGS. SCEC. for providing some of the data used in this study to perform the

differential correction of bison locations.

Aebischer, N. J.. Robertson, P. A. and Kenward, R. E. 1993. Compositional analysis of

habitat use from animai radio-tracking data. Ecology 74: 13 13-1325.

Albon. S. D. and Langvatn. R. 1992. Plant phenology and the benefits of migration in a

temperate ungulate. Oikos 65: 502-5 13.

Arnold. G. W. 1985. Regulation of food intake. In: R. J. Hudson and R. G. White, (eds.),

Bioenergetics of Wild Herbivores. CRC Press, Inc., Boca Raton, Florida pp. 8 1 - 101.

Bailey, D. W., Gross, J. E., Laca, E. A., Riîtenhouse, L. R., Coughenour, M. B., Swift, D.

M., and Sims, P. L. 19%. Mechanisms that result in large herbivore grazing

distribution patterns. Journal of Range Management, 49,386-400.

Bakken, G. S. 1992. Measurement and application of operative and standard operative

temperatures in ecology. American Zoologist 32: 194-2 16.

Barton, R. A., Whiten, A. Strum, S. C., Byrne. R. W. and Simpson, A. J. 1992. Habitat

use and resource availability in baboons. Animal Behaviour 43: 83 1-844.

Bergin. T. M. 1992. Habitat seiection by the Western kingbird in Western Nebraska: a

hierarchical analysis. Condor 94: 903-9 1 1.

Bergman, C. M. 2000. Behaviourai responses to resource availability by Northem

ungulates. Ph.D. thesis. University of Guelph.

Bergman. C. M., J. M. Fryxell. C. C. Gates, and D. Fortin. 2001. Ungulate foraging

strategies: energy maximizing or t h e minimizing? Journal of Animal Ecology (in

press).

Carey. A. B., Horton. S. P. and Biswell. B. L. 1992. Northern spotted owls: influence of

prey base and landscape charmer. Ecological Monographs 62: 223-250,

Christopherson, R. J., Hudson. R. J. and Richmond, R. J. 1978. Comparative winter

energetics of Amencan bison, yak, Scottish Highland and hereford calves. Acta

Theriologica 23: 49-54.

Collins. W. B. and Smith, T. S. 1991. Effecrs of wind-hardened snow on foraging by

reindeer (Rangifer larandus). Arctic 44: 21 7-222.

Danell. K. and Ericson, L. 1986. Foraging of moose on two species of birch when these

occur in different proportions. Holarctic Ecology 9: 79-84.

Elton, R. A. and Greenwood, J. J. D. 1987. Frequency-dependent selection by predators:

cornparison of parameter estimates. Oikos 48: 268-272.

Farnsworth, K. D. and Illius, A. W. 1998. Optimal diet choice for large herbivores: an

extended contingency model. Functional Ecology l2:74-8 1.

Fortin. D. 200 1. An adjustment of the extended contingency model of Farnsworth and

Illius (1998). Functional Ecology 15: in press.

Fortin. D. and Arnold. G. W. 1997. The influence of road verges on the use of nearby

small shrubland remnants by birds in the central wheatbelt of Western Australia.

Wildlife Research 24: 679-689.

Fortin, D. and Frandsen, D. 1999. Some preliminary results on the ecological hierarchy of

foraging activities in bison. Research Links 7: 13-15.

FretweII. S. D. and Lucas, H. L. 1970. On territorial behaviour and other factors

influencing habitat distribution in birds. Acta Biotheoretica 19: 16-36.

Fryxell. J. M. and Lundberg, P. 1997. Individual behavior and community dynamics.

Population and comrnunity biology series 20. Chapman and Hall, New York. New

York.

Fuller, W. A. 1966. The biology and management of the bison of Wood Buffalo National

Park. Canadian Wildlife Service, Wildlife Management Bulletin Series i . No. 16.

Gebremedhin, K. G. 1987. Effect of animal orientation with respect to wind direction on

convective heat loss. Agricultural and Forest Meteorology 40: 199-206.

Green. W. C. H. 1992. Social influences on contact maintenance interactions of bison

rnothers and calves: group size and nearest-neighbour distance. Animal Behaviour

43: 775-785.

Greenwood, J. J. D. and Elton, R. A. 1979. Analysing experiments of frequency-

dependent selection by predators. Journal of Animal Ecology 48: 721-737.

Hair. Jr.. J. F., Anderson, R. E., Tatharn, R. L. and Black, W. C. 1992. Multivariate Data

Analysis with readings. Macmillan Publishing Company, New York, New York.

Hall, L. S. and Mannan, R. W. 1999. Multiscaled habitat selection by elegant trogons in

Southeastern Arizona. Journal of Wildlife Management 63: 45 1-461.

Hubbard. S. F., Cook. R. M., Glover, J. G. and Greenwood, J. J. D. 1982. Apostatic

selection as an optimal foraging strategy. Journal of Animal Ecology 5 1: 625-633.

Komers. P. E., Messier. F. and Gates, C. C. 1993. Group structure in wood bison:

nutritional and reproductive determinants. Canadian Journal of Zoology 71 : 1367-

1371.

Krasinski. Z. A. 1978. Dynarnics and structure of the European bison population in the

Bialowiea Primeval Forest. Acta Theriologica 23: 13-48.

Krasiiiska, M. and Krasinski, 2. A. 1995. Composition, group size. and spatial

distribution of European bison bulls in Bialowieh Forest. Acta ï'heriologica 40:

1 -2 1.

Krasiiiski, 2. A. and Krasinska, M. 1992. Free ranging European bison in Borecka Forest.

Acta Tleriologica 37: 30 1-3 17.

Krebs. C. J. 1985. Ecology: The Experimental Analysis of Distribution and Abundance.

Third Edition. Harper and Row, Publishers, Inc., New York. New York.

Knreger. K. 1986. Feeding relationships among bison, pronghorn, and prairie dogs: an

experimentai anaiysis. Ecology 67: 760-770.

Larter. N. C., and Gates, C. C. 1991. Diet and habitat selection of wood bison in relation

to seasonal changes in forage quantity and quality. Canadian Journal of Zootogy

69: 2677-2685.

Levin, S. A. 1992. The problem of pattern and scale in ecology. Ecology 73: 1943-1967.

Lifjeld, J. T. and Slagsvold, R. 1988. Efk t s of energy costs on the optimal diet: An

experiment with pied fiycatchers Ficedital hypoleuca feeding nestlings. Omis

Scandinavica 19: 1 1 1 - 1 18.

Lundberg. P., Astrom. M. and Danell, K. 1990. An experimental test of frequency-

dependent f d selection: winter browsing by moose. Holarctic Ecology 13: 177-

182.

McHugh, T. 1958. Social behaviour of the American Buffalo (Bison bison bison). Zmlogica

43: 1-40.

McNarnara, J. M.. Houston. A. I. and Weisser, W. W. 1993. Combining prey choice and

patch use - what does rate-maximizing predict? Journal of Theoretical Biology

164: 2 19-73 8.

Moen. R.. Pastor, J.. Cohen. Y. and Schwartz, C. C. 1996. Effects of moose movement

and habitat use on GPS colla performance. Journal of Wildlife Management 60:

659-668.

Murray. D. L. and Boutin. S. 1991. The influence of'snow on lynx and coyote movements:

does morphology affect behaviour? Oecologia 88.463-469.

Mystentd. A.. Larsen, P. K.. Ims. R. A. and 0stbye E. 1999. Habitat selection by me deer

and sheep: does habitat ranking refiect resource availability? Canadian Journal of

Zoology 77: 776-783.

Orians. G. H. and Wittenberger. J. F. 1991. Spatial and temporal scales in habitat

selection. American Naturalist 137: S29449.

Parker, K. L. and Robbins, C. T. 1985. Thermoregulation in ungulates. ln: R. J. Hudson

and R. G. White, (eds.), Bioenergetics of Wild Herbivores. CRC Press, Inc.. Boca

Raton. Florida, pp. 161-1 82.

Pearson, S. M., Turner, M. G., Wallace. L. L. and Romme, W. H. 1995. Winter habitat

use by large ungulates following fire in Northem Yellowstone National Park.

Ecological Applications 5: 744-755.

Pendleton, G. W.. Titus, K.. DeGayner. E., Flatten, C. J. and Lowell, R. E. 1998.

Compositional analysis and GIS for study of habitat selection by goshawks in

Southeast Alaska. Joumal of Agicultural, Biological, and Environmental

Statistics 3: 280-295.

Reynolds. H. W.. Hansen, R. M. and Peden. D. G. 1978. Diets of the Slave River lowland

bison herd. Northwest Temtories, Canada. Journal of Wildlife Management 42: 58 1 -

590.

Schaefer. J. A. and Messier, F. 1995. Habitat selection as a hierarchy: the spatial xales of

winter foraging by muskoxen. Ecography 18: 333-344.

Schmidt-Nielsen, K. 1990. Animal physiology: adaptation and environment. Fourth

Edition. Cambridge University Press, New York, New York.

Sedgwick. J. A. and Knopf, F. L. 1992. Describing willow flycatcher habitats: scale

perspectives and gender differences. Condor 94: 720-733.

Senft. R. L.. Coughenour. M. B.. Bailey, D. W., Rittenhouse, L. R., Sala, O. E. and Swifi,

D. M. 1987. Large herbivore foraging and ecological hierarchies. BioScience 37:

798-799.

i Chaprer 5

Sokal, R. R. and Rohlf, F. J. 1995. Biometry: the principles and practice of statistics in

biological research. Third edition. W. H. Freeman and Company, New York, New

York.

Stephens, D. W., and J. R. Krebs. 1986. Fotaging theory. Princeton University Press,

Princeton, New Jersey.

Tel fer, E. S. and Kelsall, J. P. 1984. Adaptation of some large North American mamrnals

for suwival in snow. Ecology 65: 1828-1834.

Tufto, J,, Andersen, R. and Linnell, J. 1996. Habitat use and ecological correlates of

home range size in a small cervid: the roe deer. Journat of Animal Ecology 65:

7 15-724.

Turner, M. G., Wu, Y., Wallace, L. L., Ronune. W. H. and Brenkert. A. 1994. Simulating

winter interactions among ungulates. vegetation. and fire in Northern Yellowstone

Park. Ecotogicai Applications 4: 472-496.

Van Camp. J. 1975. Snow conditions and the winter feeding behaviour of Bison bison in

Elk Island National Park. Can. Wildl. Serv. Rep. C WS-53-75.

Vartha E. W. and Matches, A. G. 1977. Use of a weighted-disk measure as an aid in

sampling the herbage yieId on taIl fescue pastures grazed by cartle. Agronomy

Journal 69: 888-890.

Wallace. L. L.. Turner, M. G., Romrne W. H. O'Neill, R. V. and W u Y. 1995. Scale of

heterogeneity of forage production and winter foraging by elk and bison. Landscape

Ecology 10: 75-83.

Chapter 5

Ward. D. and Saltz, D. 1994. Foraging at difietent spatial scales: Dorcas gazelles

foraging for Mies in the Negev desert. Ecology 75: 48-58.

Table 1. Geometric mean selection ratio among habitat types of Prince Albert National

Park by female bison as determined from GPS-locations during 1996- 1999. Water shore

and water interior were pooled in winter. For a given season. the use of habitat variables

with the sarne letter did not differ significantly (P < 0.05).

Summary Meadow Meadow Water Forest Water Agricultural edge shore interior interior land

Summer Mean 7.13 1 .48 0.27 0.25 0.08 O Range 4.70-8.32 1.19-2.03 0-0.92 0.16-0.52 0-0.30 0-0

Com~arison a b bc cd d e Winter Mean 8.05 1.20 - 0.16 0.57 O Range 7.35-9.01 1.00-1 .30 - 0.08-0.28 0.4 1-0.84 0-0

Comparison a ab - c bc d

Chaprer 5

Table 2. Percent of fixes within 100 for pairs of female bison equipped with GPS-collars

in summer (white area) and winter (gray ma) . Pairwise cornparisons were restricted to

animals tracked simultaneously. Sample size for each cornparison ranged between 228-

1245 pairs of fixes, with the exception of animal B vs. animal E in winter where sample

size was 34 pairs of fixes.

Animal id B C D E F G 9D 5D CF

C D

5 1 18 15 1 O 3 O 1 1 13 10 6

Table 3. Variables influencing the probability of use of 23 meadows in Prince Albert

National Park during the summer of 1997-1998, as determined fiom stepwise logistic

regressions. The independent variables considered in the analyses included: area of the

meadow. area of water in the meadow, perimeter of water in the meadow, absence of

water in the meadow (dicotomic variable), biomass of Agropyron spp., Carex aquatilis

(Biomass of Caaq), Carex atherodes (Biomass of Caat), Calamagrosris inexpansa.

Jrrncirs buliicus and Scolochloafestucacea (Biomass of Scfe), and percentage of an area

of 2-km radius and 1-km radius covered by meadows, water areas and agricultural lands.

Final models included only variables significant at P < 0.05.

Survey Significani independeni variable: PkSE technique Intercept Area (ha) Water in a 2- Biornass of

km radius (%) hoju (glmL) GPS-collars (Animal id)

C -3.81&1.58* 0.56&0.22* D -2.32+1.04* 0.3tk0.1 S* E -5.52&2.35* 0 . 1 9 H . 1 1 ~ 0.35kO. 15* F - 1.70~.90@ 0.33&0.14* G -3.84k 1.47* * 0.29H. 12*

CF -7 .2~2 .81" 0.81k0.40* 12 -2.54&0.95** 0.1%.08* 5D -4.3%1.98* 0.21Hl.12~ 1.45H.67* 9D -2.54&0.95** O. 19*0.08*

Herd surveys -1.94H.88* O. 19fO.09*

"<O. 10. *P<0.05. **P<O.Ol

Chaprer 5

Table 4. Variables intluencing the probability of use of 23 meadows in Prince Albert

National Park in wintei as determined h m stepwise logistic regressions. The

independent variables considered in the analyses included: area of the meadow, snow

depth. snow density. snow sohess. presence of cmts in the snow colurnn (dicotomic

variable). biomass of Agropyron spp.. Carex aquatilis (Biomass of Caaq). Carex

arherodes (Biomass of Caat). Calamagrostis inexpansa. Jlrncus balticus and Scolochloa

fesrlmwa (Biomass of Scfe). and percentage of an area of 2-km radius and 1 -km radius

covered by meadows. water areas and agricultural lands. Final models included only

variables significant at P < 0.05.

Survey Significant independent variable: FISE technique lntercept Area (ha) Snow depth Biomass Biomass Biomass

{cm) of scfe of caat of a g o (dm') ce/m2, ( g f d

GPS-collars (Animal id)

Herd surveys 1997- 1998

-3.43f 1.16** 0.73+0.09* 'P<O. 10, *P<O.OS, **P<0.01. NNon-significant intercept

Chapter 5

Table 5. Variables influencing the average size of bison herds found in a total of 15

meadows of Prince Albert National Park. Independent variables submiîîed to the stepwise

multiple regressions are indicated in Table 2. The finai models included only variables

signifiant at P < 0.05.

Season Predictive mode1 of group size R~ F df P Summer 9.57 + 7.46 (log Ares)** 0.90 26.97 3.12 0.0001

- 1.36 (% of 2-km radius covered by agricultural lands)** * - 1 1.81 (absence of water in meadow)*

Winter - 1 1.69 + 8.17 (% of 1-km radius covered 0.57 14.38 1.12 0.003 by rneadows)* *

*P<O.Oj. **P<O.Ol. ***P<O.Ool

Table 6. Bison selection of plant types in relation to the relative biomass of other plant

types found in 25 meadows of Prince Albert National Park. Acronyms for plant types are

Caat: Carex atherodes, Scfe: Scolochloajèstucacea and Other: other species altogether.

Non-significant intercept ( b log V) suggests that when the two contrasted plant types had

equal availability, there was no selection by bison. Non-significant slope (b) suggests

frequency-independent selection among the contrasted plant types.

Comparison Regression Ho: Intercept = O Ho: Slope = 1 df R' P b log V t P b t P

Summer Caat-Scfe 10 0.51 0.009 9 2.09 0.06 9 1.21 0.25

Caat-Other 10 0.29 0.07 5.44 3.28 0.008 2.01 1.01 0.33 Scfe-Other 10 0.26 0.09 5.47 1.88 0.09 2.86 1.21 0.25

Winter Caat-Scfe 16 0.60 0.0002 1.47 1.68 0.11 2.34 2.79 0.013

Caat-Other 16 0.25 0.04 6.69 2.69 0.02 2.95 1.52 0.15 Scfe-Other 16 0.28 0.02 5.44 2.34 0.03 3.18 1.73 0.10

Figure 1. Relative occurrence (total # of individuals for a given species / total # of

individuals for al1 species x 100) of bison, moose, white-tailed deer and elk recorded

during ground swveys of 25 meadows in Prince Albert National Park during 1997 and

1998.

Figure 2. Relationship between average herd size of bison obsewed in 15 meadows of

Prince Albert National Park in 1997-1998, and the area (log-transforrned) of these

meadows. Regession the is displayed for significant relationships (P < 0.05).

Figure 3. Selection ratio of different plant types recorded in 12 meadows (Le. n = 12 for

al! plant species) of Prince Albert National Park in 1998. A selection ratio of I indicates

that resource was used in same proportion than it was available. A broken horizontal tine

displays this threshold. For a given season, selection of plant types with the same letter

did not di ffer signi ficantl y following pairwise cornparison (P < 0.05). Acronyms for plant

types are A p : Agropyron spp., Brci: Brornus ciliatus, Caat: Carex atherodes, Cain:

Calamagrostis inexpansa. Ciar: Cirsium arvense, Juba: Juncus balticus. Scfe: Scoiochloa

festucacea, Soar: Sonchus arvensis and ûther: other species altogether.

Chapter 5: Figure 1

Summer

bison moose deer elk 1 O0 -

Winter

bison moose deer elk

U ng ulate species

Chapier 5: Figure 2

Surnmer 40

Winter 40 -

Ln (Area, rd)

Chap~er 5: Figure 3

Pairwise cornparison c c a bc d bc b bc c

Summer

Agro Brci Caat Cain Ciar Juba Scfe Soar Other 1

i b b a b a b a b b

Winter

Agro Brci Caat Cain Ciar Juba Scfe Soar Other

Plant types

This study offers insights on the foraging behaviour and habitat selection of bison

that can help answer the simple questions raised in the thesis prologue: where are bison

going after leaving a meadow and why are they in particular locations? Bison travel from

meadow to meadow in search of resources such as water and plant species. Selection of

meadows by bison is reiated to general landscape features, among which snow depth and

the proportion of water or meadow within their daily radius appear to be of particular

importance. This thesis Further outlines the importance of spatio-temporal pattern of

resource distribution on habitat selection and resource use by bison.

SEARCH STRATEGY

1 first investigated plant-herbivore interaction by examining bison searching

activity in winter. Rather than searching randomly, bison used area-restricted search in

response to the spatial heterogeneity of resource distribution. Area-restricted search c m

be charactenzed by an intensive searching mode (ISM) during which animals travel with

high sinuosity and low speed as they remain in the area where they found high quality

resource. and by an extensive searching mode (ESM) during which animals travel with

low sinuosity and hi& speed because they are moving away fiom a patch of poor quality

(Cézilly and Benhamou 1996). 1 found that most snow craters were comprised of several

adjacent feeding stations. in these cases, anirnals presurnably considered that the resource

was of suficient quality to keep searchingtforaging in the are% thus reflecting an ISM.

Hence, the discrete nahue of snow craters dong a path nicely reflects the shifi between

ISM and ESM. This behavioural response to habitat heterogeneity provides evidence that

bison take advantage of spatial auto-correlation in resource distribution during their

searching activity. 1 found that not only do bison shifi from ISM to ESM in response to

their perception of resource quality. but they also fine-tune their ESM io the resource

distribution by avoiding areas of poor resowe quality without even sampling by digging

in the snow. Spatial memory or environmental cues can be used to lwate or avoid certain

patches of resource (Gillingham and Bunnell 1989, Edwards et al. 1996, Dumont and

Petit 1998. Laca 1998). The adjustment of search behaviour to the local resource

distribution should increase searching efficiency (Bovet and Benharnou 199 1).

My field observations suggest that as bison sample their habitat. they continually

readjust their perception of patch quality. Their perception of a given resource even

displays flexibility during a foraging bout: the larger crater of pairs encountered during a

foraging bout. offering exactly the same potential rate of energy intake, tended to be

preceded by sarnples of poor resource quality. Given that large herbivores should be able

to remember resource information during periods much longer than the time required to

complete a foraging bout (Bailey et al. 1996). why are individuals using only short-term

sampling information?

The cornputer simulations performed in chapter 2 suggest that the value of

sarnpling information for the maximization of searching performance increases in a

decelerating fashion. and eventually reaches a plateau. 1 argue that. because of the costs

that could be involved with storing environmental information (Dali and Cuthill 1997).

animals should assess resource quality by considering only the minimum amount of

sampling information that would maximize their searching efficiency (Le. the b e g i ~ i n g

of the plateau, Chapter 2, figure 2).

This arnount of environmental information varies with the spatial auto-correlation

in resource distribution (Chapter 2). 1 found that sampling information is less valuable to

assess resource quality in environments with low spatial auto-correlation, because

resource distribution is less predictable fiom one feeding station to the next.

Consideration of a small nurnber of sampling stations thus provides as good as

knowledge about the local resource as it is possible to gain in highly variable

environments.

Despite the investigation of a wide range of spatial auto-correlation in resource

distribution. the increase in searching eficiency always levelcd off rapidly with the

amount of past sampling experience considered by simulated animals. In most

environments encountered by bison, simulations suggest that the amount of information

gathered during part of a foraging bout should be suficient to adequately mess resource

quality. This prediction seems to explain the flexibility in bison assessrnent of resource

quality observed under field conditions.

FORAGING DECISIONS

Despite the flexibility in perception of resource quaiity indicated in Chapter 1,

diet choice of bison was limited mainIy to a single sedge species: Carex atherodes. 1 used

optimal diet models to better understand what might be the cause of this selection

(Chapter 4).

The contingency mode1 is the most frequently used mode1 of optimal diet

(Farnsworth and Illius 1998). The classical form of this mode1 does not, however,

adequately reflect the foraging behaviour of most herbivores, because it assumes no

overlap between searching and handling activity (see Stephens and Krebs 1986).

Moreover. it considers that forage intake is limited by the rate at which selected prey are

encountered. Consequently, intake rate cannot be limited strictly by the time required to

handle prey. The recent contribution of Farnsworth and Illius (1996, 1998) to the

foraging theory of large herbivores is valuable because it acknowledges these previously-

ignored possibilities. Unfortunately, their model included some algebraic errors that lead

to inaccurate predictions. In Chapter 3 , l corrected their mode1 by providing new sets of

rules and equations. With this correction. their model can be used to estimate the rate of

energy intake by herbivores, but only over short periods of time. Over extended periods

of time, however. foraging rnodels should also consider the digestive and time constraints

faced by herbivores because of the potential limit these constraints can set on rates of

daily intake (Belovsky 1978. Mould and Robbins 1982. Langvatn and Hanley 1993.

Wilmshurst et al. 1995. 1 999).

Based on the work of Verlinden and Wiley (1989), Farnsworth and Illius (1998)

created an addi tional contingency model that included the overlap between searching and

handling activity together with digestive and tirne constraints. Verlinden and Wiley's

model (1 989) has been recently criticized, however, because of its underlying assumption

that the turnover time of al1 food types is one day (Hirakawa 1997a). Some of the

parameters required in Veriinden and Wiley's (1989) model (or the version revised by

Hirakawa 1997a . 1997b). such as the turnover time of each prey, the capacity of the

digestive tract and time required to fil1 the digestive tract with the highest quality prey.

are rarely measured for any species of large ruminant. An alternative way to represent the

constraint irnposed by digestive processes is the relationship between daily voluntary

intake and plant digestibility (Van Soest 1994, Wilmshurt et ai. 1995, 1999, Bergman et

al. 2001). Based on this relationship and considering time constraints, 1 developed a

contingency mode1 allowing prediction of the optimal diet based on the mmimization of

daily gains in large herbivores (Chapter 4).

Assessing the temporal scale under which foraging gains are maximized is critical

for the understanding of diet choice by animals. Classical foraging theory assumes that

animals maximize long-term rates (Stephens and Krebs 1986). Lack of multi-scale

investigations (Bergman et al. 2001) and conflicting empirical results (Barkan and

Withiarn 1989, Wallis de Vries and Daleboudt 1994, Lllius et al. 1999. Bergman et al.

2001). cast doubt on this assumption. 1 investigated the temporal s a l e of foraging

decisions in bison using the short-term and daily models of optimal diet developed in

Chapter 3 and 4. These contingency models were pararneterized based on plant surveys

and behavioual observations of bison in Prince Albert National Park.

My analyses clearly showed that predictions of optimal diet strongly depend on

temporal scale. Diet predictions varied between temporal scales for four of the six

sampling periods. In general, bison made dietary choices that were more consistent with

short-tem than long-term objectives. This trend provides evidence that, c o n t r q to the

assumption of classic optimality models. foraging decisions by bison oflen "sacrifice"

long-tem gains by rnaximizing the short-term ones. Despite some connadictory results

(Wallis de Vries and Daleboudt 1994). my findings join a slowly increasing body of

evidence arnong the multi-sale studies (Illius et ai. 1999, Bergman et al. 2001)

suggesting that rnaximization of long-term energy gains may not constitute an adequate

base for predicting the foraging behaviour of many species of large herbivores.

The relative success of the contingency model in predicting the dietary choices of

bison during rnost of the year suggests that this model could constitute an adequate tool

to forecast plant-bison interactions in Prince Albert National Park. For exarnple, should

the abundance of Carex atherodes decline with increasing bison abundance, the rules of

maximization of short-term gains suggests that grazing pressure should concurrently

increase on Scolochloafistucacea and then on Agropyron spp. Dpamic plant-herbivore

interactions could be anticipated, to a certain degree, using modeling approaches based

on diet selection (Fry'tell and Lundberg 1997).

Additional factors could, however, M e r improve diet predictions (Chapter 5) .

and shouid therefore also be taken into account when predicting the diet of bison and

their impact on plant communities. Consideration of principles derived from optimal diet

and marginal value theorem (Charnov 1976) suggests that diet selection should be

frequency-independent in a patchy environment (McNamara et al. 1993). Bison

adequately followed this prediction in summer, but not in winter. In winter, animal

selection varied among meadows depending on the relative abundance of their preferred

food, C. a~herodes. Bison increased their consumption of S.festucacea in meadows in

which the abundance of C. atherodes was low. Optimality theory would predict an earlier

departure from the meadow rather than a switch in diet (McNamara et al. 1993). Such

deviation from the theory can help explain why a broader die! than expected from

contingency models was observed in the field. Frequencydependent selection observed

during winter should be considered when predicting plant-herbivore interactions and

population dynamics.

[NTEGRATION OF INFORMATION ACROSS SPATIO-TEMPORAL SCALES

The ability of optimal diet mode1 to predict diet choice of animals does not

necessariIy irnply that their distribution can be anticipated from the suggestions of

contingency models. The plant species that bison consume the most are found in

meadows. which is where bison spend most of their time (Chapter 5). The meadows most

likely to be visited were not, however, those offering greater quantities of food plants

preferred by bison. Despite the tight link that has been observed in empirical studies

(McNaughton 1988. 1990. Krasinski and Krasinska 1995, Keane and Morrison 1999). my

findings suggest that it could be risky to predict animal distribution based solely on food

distribution.

Variation in the role that food can have on animal distribution is well reflected in

the study of Mystemd et al. (1999). They found that habitat selection by free-ranging

domestic sheep (Ovis aries) was directly related to food availability. whereas, for roe

deer (Cupreolus caprolus), habitat selection was related to adequate cover against

predation. Animais should select habitats that satise simultaneously several of their

fitness needs. Trying to satis& simultaneously several needs may result in trade-offs,

which rnay Vary at different spatio-temporal scales (Senft et ai. 1987. Barton et ai. 1992.

Mysterud et al. 1999. Conradt et al. 2000). Scaie-dependence in trade-offs has k e n

suggested by Senft et al. (1987) and Bailey et al. (1996), and observed in empirical

studies such as these of Bergman (2000), Wallace et ai. (1995) and Schaefer and Messier

(1 995).

1 found that bison distribution and resource use were infiuenced by both abiotic

and biotic factors that varied in relative importance at different spatio-temporal scaies.

The strength of plant-herbivore interaction is known to decline at larger scales (Senfi et

al. 1987. Bailey et al. 1996). Consistently, the factors influencing the use of particular

meadows were more closely related to regional features outside the meadow than to

forage species found in the meadows. In summer, the presence of other meadows and

water availability within a 2-km radius of a given rneadow increased the likelihood of

being used by bison. In winter. bison adjusted their movernent to the presence of snow

cover. reducing their daily displacement fiom 2 km in summer to 1.3 km in winter. Snow

depth within meadows rose as a dominant factor related to their probability of meadow

use. Large winter herds of bison were most likely to use meadows sunounded (within 1

km) by other meadows. These results suggest that management strategies should consider

characteristics of the rnatrix in which habitat patches are found (Pearson et al. 1995.

Fortin and Arnold 1997).

THE NEXT STEP

In this thesis. 1 investigated resource use by bison at different spatial and temporal

scales. Although contingency models based on the maximization of energy intake rate

appears to have a limited power in explainhg animal distribution across multiple

temporal and spatial scales, my results suggest that broader physiological models could

be more successful. From a physiological standpoint, the preference for meadows largely

surrounded by other meadows and for meadows with shallow snow couM reflect an

attempt by animals to reduce the energy expended in locomotion. The importance of

water distribution on meadow selection during the swnmer could reflect an attempt to

ensure daily access to water supply, and thus to maintain water balance. Within

meadows. bison maximized their short-term energy intake rate by selecting plant species

having high short-term profitability. The adjustment of search paths to resource

distribution by the use of area-restricted search should increase the probability of finding

of high quality resource with the shortest possible displacement. This search strategy

should diminish the energy expended in locomotion during foraging activity. From these

observations, 1 suggest that future work should concentrate on net energy gains together

with water balance. Such a mode1 might better explain the distribution of the bison of

Prince Albert National Park than energetic models that consider only rates of energy

intake.

For now, my thesis provides evidence that prediction of resource and habitat

selection at a given spatial and temporal scale cannot lx readily inferred based solely on

knowledge at other scales. presumably due to trade-offs among individual requirements

of bison.

Bailey. D. W., Gross, J. E., Laca, E. A., Rittenhouse, L. R., Coughenour, M. B., Swift, D.

M. and Sims, P. L. 1996. Mechanisms that result in large herbivore grazing

distribution patterns. Journal of Range Management 49: 386-400.

Barkan. C. P. L. and Withiam, M. L. 1989. Profitability, rate maximization, and reward

delay: a test of the simultaneous-encounter mode1 of prey choice with Parus

africapillirs. American Naturaiist 134954-272.

Barton, R. A., Whiten. A.. Stnrm. S. C., Byrne, R. W. and Simpson. A. J. 1992. Habitat

use and resource availability in baboons. Animal Behaviour 43: 83 1-844.

Belovsky. G. E. 1978. Diet optimization in a generalist herbivore: the moose. Theoretical

Population Biology 14: 105- 134.

Bergman, C. M. 2000. Behavioural responses to resource availability by Northem

ungulates. Ph.D. thesis, University of Guelph.

Bergman. C. M.. Fryxell. J. M.. Gates. C. C. and Fortin. D. 2001. Ungulate foraging

strategies: energy maximizing or time minimizing? Journal of Animal Ecology (in

press).

Bovet. P. and Benhamoli. S. 1991. Optimal sinuosity in central place foraging

movements. Animal Behaviour 42: 57-62.

Cézilly. F. and Benhamou, S. 1996. Les stratégies optimales d'approvisionnement. Revue

d'Écologie (Terre Vie) 5 1: 43-86.

Charnov. E. L. 1976. Optimal foraging, the marginal value theorem. Theoretical

Population Biology 9: 129-136.

Conradt, L., Clutton-Brock, T. H. and Guinness, F. E. 2000. Sex differences in weather

sensitivity can cause habitat segregation: red deer as an example. Animal

Behavour 59: 1049- 1060.

Dall. S. R. X. and Cuthill. 1. C. 1997. The information costs of generalism. Oikos 80:

197-202.

Dumont. B. and Petit, M. 1998. Spatial memory of sheep at Pasture. Applied Animal

Behaviour Science 60: 43-53.

Edwards, G. R.. Newman, J. A., Parsons. A. I. and Krebs. J. R. 1996. The use of spatial

memory by gazing animals to locate food patches in spatially heterogeneous

environments: an example with sheep. Applied Animal Behaviour Science 50:

147-160.

Farnsworth. K. D. and Illius. A. W. 1996. Large gazers back in the fold: generalizing the

prey model to incorporate mammalian herbivores. Functional Ecology 10: 678-

680.

Farnsworth. K. D. and Illius, A. W. 1998. Optimal diet choice for large herbivores: an

extended contingency model. Functional Ecology 12: 74-81.

Fortin. D. and Arnold. G. W. 1997. The influence of road verges on the use of nearby

small shrubland remnants by birds in the central wheatbelt of Western Australia.

Wildlife Research 24: 679-689.

Fyell . J.M. and Lundberg. P. 1997. Individual behavior and comrnunity dynarnics.

Population and community biology series 20. Chapman and Hall, New York. New

York.

Gillingham, M. P. and Bumeil, F. L. 1989. Effects of learning on food selection and

searching behaviour of deer. Canadian Journal of Zoology 67: 24-32.

Hirakawa, H. 1997a. How important is digestive quality? A correction of Verlinden and

Wiley's digestive rate model. Evolutionary Ecolog~ 1 1 : 249-252.

Hirakawa. H. 19976. Digestion-constrained optimal foraging in generalist mammalian

herbivores. Oikos 78: 37-47.

Himonen. H.. Ranta E.. Rita, H. and Peuhkuri, N. 1999. Significance of memory

properties in prey choice decisions. Ecological Modelling 1 15: 177-189.

IIlius. A. W.. Gordon. 1. J.. Elston, D. A. and Milne J. D. 1999. Diet selection in goats: a

test of intake-rate maximization. Ecology 80: 1008-10 18.

Keane. J. J. and Momson. M. L. 1999. Temporal variation in resource use by black-

throated Gray Warblers. Condor 10 1 : 67-75.

Krasinska. M. and Krasinski, Z. A. 1995. Composition. group size. and spatial

distribution of European bison bulls in Bialowieia Forest. Acta Therioiogica 40:

1-21.

Laca. E. A. 1998. Spatial memory and food searcliing mechanisms of cattle. Journal of

Range Management 5 1 : 370-378.

Langvatn. R. and Hanley. T. A. 1993. Feeding-patch choice by red deer in relation to

foraging eficiency: an expriment. Oecologia 95: 164- 170.

McNamara. J. M.. Houston. A. 1. and Weisser. W. W. 1993. Combining prey choice and

patch use - what does rate-maximizing predict? Journal of Theoretical Biology

164: 219-238.

McNaughton, S. J. 1988. Mineral nutrition and spatial concentrations of African

ungulates. Nature 334: 343-344.

McNaughton. S. J. 1990. Mineral nutrition and seasonal movements of Afiican migratory

ungulates. Nature 345: 61 3-61 5.

Mould. E. D. and Robbins, C. T. 1982. Digestive capabilities in elk compared to white-

tailed deer. Journal of Wildlife Management 46: 22-29.

Mystemd. A.. Larsen. P. K., [ms, R. A. and Bstbye E. 1999. Habitat selection by roe deer

and sheep: does habitat ranking reflect resource availability? Canadian Journal of

Zoology 77: 776-783.

Pearson. S. M.. Turner. M. G., Wallace, L. L. and Romme. W. H. 1995. Winter habitat

use by large ungulates following fire in Northern Yellowstone National Park.

Ecological Applications 5: 744-755.

Schaefer, J. A. and Messier. F. 1995. Habitat selection as a hierarchy: the spatial scales of

winter foraging by muskoxen. Ecography 18: 333-344.

Senft. R. L.. Coughenour, M. B., Bailey, D. W.. Rittenhouse. L. R.. Sala. O. E. and Swift.

D. M. 1987. Large herbivore foraging and ecological hierarchies. BioScience 37:

798-799.

Stephens, D. W. and Krebs, J. R. 1986. Foraging theory. Princeton University Press,

Princeton. New Jersey.

Verlinden. C. and Wiley, R. H. 1989. The constraints of digestive rate: an alternative

model of diet selection. Evolutionary Ecology 3: 264-272.

Van Soest. P. J. 1994. Nutritional ecology of the ruminant. Second edition. Cornell

University Press. Ithaca, New York.

Wallace, L. L., Turner, M.G., Romme, W. H., O'Neill, R. V. and Wu, Y. 1995. Scale of

heterogeneity of forage production and winter foraging by elk and bison.

Landscape Ecology 10: 75-83.

Wallis de Vries, M. F. and Daleboudt, C. 1994. Foraging strategy of cattle in patchy

grassland. Oecologia 100: 98- 106.

Wilmshurst, J. F., Fryxell, J. M. and Colucci. P. E. 1999. What constrains daily intake in

Thomson's gazelles? Ecology 80: 2338-2347.

Wilmshurst, J. F., Fryxell, J. M. and Hudson, R. J. 1995. Forage quality and patch choice

by wapiti (Cervus elaphus). Behavioural Ecology 6: 209-21 7.