THE PREDATORY SEQUENCE AND THE INFLUENCE OF INJURY RISK ON HUNTING BEHAVIOR IN THE WOLF A THESIS SUBMITTED TO THE FACULTY OF THE GRADUATE SCHOOL OF THE UNIVERSITY OF MINNESOTA BY DANIEL ROBERT MACNULTY IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE May 2002 © Daniel Robert MacNulty 2002 i ACKNOWLEDGMENTS The privilege of writing this thesis would not have been possible without the commitment, encouragement, and patience of many individuals at many levels. In 1996 while I was working as a volunteer for the Yellowstone Wolf Project, Dr. David Mech suggested I summarize the early observations of wolves hunting elk in the park's Northern Range. Dr. Mech subsequently became my faculty advisor at the University of Minnesota and continued to encourage me to study the hunting behavior of wolves in YNP. Throughout my research Dr. Mech gave me the freedom to pursue my ideas, while providing important guidance at critical junctures in the research process. Dr. Douglas Smith, Yellowstone Wolf Project leader, was instrumental in conceiving and supporting the idea of a study of wolf hunting behavior. Dr. Smith handled administrative affairs to facilitate field research, and directed the Wolf Project’s biannual winter study, which included the study on wolf hunting behavior. Mike Phillips also encouraged the study and helped secure vital funding during his early tenure as Wolf Project leader. Deborah Guernsey, administrative assistant, Yellowstone Wolf Project, patiently endured my endless requests for information. Ms. Guernsey also tended to many small but critical details that allowed field research to occur. Wolf Project volunteers were the heart and soul of the data collection process. Since 1995 many dedicated volunteers passed through the Wolf Project field program and provided the bulk of the observations that comprise this thesis. Without their hard work and dedication this study would not have been possible. I owe a special thanks to Wolf Project volunteers Nathan Varley, Kevin Honness, Daniel Stahler, and Paul Frame for going the extra mile (literally) to watch wolves hunt bison in Pelican Valley. Also, without the interest and support of Lake District Rangers John Lounsbury and Lloyd Kortge our research in Pelican Valley would not have been possible. Robert Landis, Landis Wildlife Films, generously shared with me his film footage of wolves hunting in Yellowstone. Bob’s films played a key role in clarifying my understanding and interpretation of wolf hunting behavior. Dr. Thomas Drummer, Michigan Technological University, provided important statistical expertise early in the study, and generously provided statistical advice at various times throughout the study. Dr. Lynn Eberly, University of Minnesota, instructed me in statistical methods for analyzing correlated data, and thus opened my eyes to a new and indispensable area of statistics. Dr. Eberly patiently responded to every email and every question without exception. I am also indebted to Dr. James Halfpenny for initially introducing me to Yellowstone National Park, and the Yellowstone Wolf Project, shortly after wolves were restored to the park in 1995. Funding for this project was provided by the National Geographic Society, Yellowstone National Park Foundation, Dayton-Wilkie Natural History Fund, and the Department of Ecology, Evolution, and Behavior at the University of Minnesota. Housing and transportation in Yellowstone were provided by the Yellowstone Center for Natural Resources, Yellowstone National Park. Finally, Cory Counard provided valuable input and criticism throughout my research, and provided essential moral support during preparation of the thesis. ii ABSTRACT To study the hunting behavior of the wolf (Canis lupus) in Yellowstone National Park (YNP), I first define the wolf predatory sequence as consisting of six distinct behaviors: travel, approach, watch, attack, target, and capture. These behaviors are organized into three nested groups: predation attempt, prey encounter, and hunting bout. Using this framework, I first evaluate general patterns of wolf hunting behavior and estimate success rates for wolves hunting various prey in YNP. I then compare reported success rates for wolves hunting various prey species in North America to demonstrate the general relationship between hunting success and prey size. Next I show that prey are dangerous to wolves and that risk of prey-caused injury is related to prey size. Finally, I evaluate the influence of injury risk on patterns of wolf hunting behavior. From May 1995 to March 2000, 62 hunting bouts, 267 prey encounters, and 565 predation attempts were observed in their entirety. The typical hunting pattern involved a brief hunting bout (< 60 min.) including at least one prey encounter (< 15 min.) and at least one predation attempt (< 4 min.). Wolves encountered prey within 25 minutes of hunting, and approximately once every 20 minutes thereafter (3.00 ± 0.42 encounters/hour/bout N = 62). Multiple prey encounters during hunting bouts were neither a prominent nor important feature of hunting behavior patterns. Overall, the estimated rate of success was 0.21 ± 0.03 kills per encounter, and bison (Bison bison) were more difficult to kill (0.04 kills/encounter) than elk (Cervus elaphus) (0.24 kills/encounter). Comparisons with other studies indicate a broad association between hunting success and prey size. In general, prey that confronted wolves were more aggressive, and therefore less likely to be killed than prey that fled. In YNP, bison confronted wolves more frequently than elk (79% vs. 55% of encounters; χ2 = 8.60, d.f . = 1, P < 0.01) and charged wolves more frequently than did elk (62% vs. 26% of encounters; χ2 = 22.20, d.f. = 1, P < 0.001), suggesting that elk are less dangerous and more vulnerable to wolf predation than bison. Wolf hunting behavior differed between encounters with bison and elk. During bison encounters, wolves made fewer predation attempts (60% vs. 80%; χ2 = 8.50, d.f. = 1, P < 0.01) and shorter predation attempts (2.90 ± 0.51 min. vs. 4.00 ± 0.38 min.; t = 4.04, d.f. = 165, P<0.001) than during elk encounters. Wolf encounters with bison also included periods of watching from within 10 m. Difference in wolf hunting behavior between bison and elk encounters suggest that wolves assess their risk of injury and incorporate this information into their foraging decisions. The tendency for wolves to attack elk more often than bison suggests that wolf preference for vulnerable prey is an adaptive strategy to acquire food while minimizing the risk of prey-caused injury. iii TABLE OF CONTENTS ACKNOWLEDGMENTS…………………………………………………… i ABSTRACT…………………………………………………………………. ii TABLE OF CONTENTS……………………………………………………. iii LIST OF TABLES…………………………………………………………… iv LIST OF FIGURES………………………………………………………….. v INTRODUCTION…………………………………………………………… 1 Components of the Predatory Sequence………………………………….. 8 A Framework for the Predatory Sequence………………………………... 14 METHODS…………………………………………………………………... 16 Study Area………………………………………………………………... 16 Study Population………………………………………………………….. 17 Hunting Observations…………………………………………………….. 18 Statistical Methods………………………………………………………... 21 RESULTS……………………………………………………………………. 23 Hunting Bouts…………………………………………………………….. 23 Prey Encounters…………………………………………………………... 23 Predation Attempts……………………………………………………….. 25 Hunting States…………………………………………………………….. 26 Hunting Success…………………………………………………………... 27 Anti-Predator Response and the Risk of Injury…………………………... 29 Hunting Behavior in Bison and Elk Encounters………………………….. 30 Success Rates for Wolves Hunting Various North American Prey………. 31 DISCUSSION………………………………………………………………... 31 The Predatory Sequence………………………………………………….. 31 General Patterns of Hunting Behavior……………………………………. 33 General Patterns of Hunting Success……………………………………... 37 Hunting Success and Prey Size…………………………………………… 41 Risk of Injury and Prey Size……………………………………………… 41 Wolf Behavioral Response to the Risk of Injury…………………………. 42 LITERATURE CITED………………………………………………………. 47 iv LIST OF TABLES Table 1. Number of hunting bouts, prey encounters, and predation attempts observed in their entirety and partially observed (in parentheses) for various wolf packs in Yellowstone National Park, May 1995 – March 2000. 53 Table 2. Results of General Linear Mixed Model (GLMM) evaluating the effects of hunting-state type and prey species on huntingstate duration (min.) in wolf encounters with elk and bison herds in Yellowstone National Park, May 1995 – March 2000. Predicted mean hunting-state duration with 95% confidence intervals is shown in Figure 6. 54 Table 3. Success rates for wolves hunting various prey species in Yellowstone National Park, May 1995 - March 2000, based on known outcomes from completely observed prey encounters only, and on both complete and incompletely observed encounters (in parentheses). 55 Table 4. Age and sex of prey killed by wolves in Yellowstone National Park, May 1995 – March 2000. The proportion killed in each age/sex class for each species is shown in parentheses. 56 Table 5. Results of General Linear Mixed Model (GLMM) evaluating the effects of hunting-state outcome, hunting-state type, and prey species on hunting-state duration (min.) in wolf encounters with elk and bison herds in Yellowstone National Park, May 1995 – March 2000. Predicted mean hunting-state duration with 95% confidence intervals is shown in Figure 9. 57 Table 6. Results of General Linear Mixed Model (GLMM) evaluating the effects of hunting-state type on hunting-state duration (min.) in wolf encounters with bison herds in Yellowstone National Park, May 1995 – March 2000. Predicted mean hunting-state duration with 95% confidence intervals is shown in Figure 11. 58 Table 7. Reported success rates for wolves hunting various North American prey species. 59 v LIST OF FIGURES Figure 1. The predatory sequence for wolves hunting herds of prey. 60 Figure 2. Study area and general location of study wolf packs, Yellowstone National Park, May 1995 - March 2000. 61 Figure 3. Time of year wolf hunting bouts were observed in Yellowstone National Park, May 1995 - March 2000. 62 Figure 4. Time of day wolf hunting bouts were observed during intensive winter study periods (mid-November to mid-December & March) in Yellowstone National Park, 1995 - 2000. 63 Figure 5. Number of prey present during wolf encounters with various prey species in Yellowstone National Park, May 1995 - March 2000. 64 Figure 6. Predicted mean duration (min.) of hunting states with 95% confidence intervals in wolf encounters with elk and bison herds in Yellowstone National Park, May 1995 – March 2000. Fitted means and confidence intervals are derived from GLMM results (Table 2). 65 Figure 7. Results of completely observed wolf encounters with elk herds and solitary elk in Yellowstone National Park, May 1995 March 2000. 66 Figure 8. Results of completely observed wolf encounters with bison herds and solitary bison in Yellowstone National Park, May 1995 - March 2000. 67 Figure 9. Predicted mean duration (min.) of hunting states with 95% confidence intervals in failed and successful wolf encounters with elk and bison herds in Yellowstone National Park, May 1995 - March 2000. Fitted means and confidence intervals are derived from GLMM results (Table 5). 68 Figure 10. The association between mean wolf hunting success (kills/encounter) and season (early winter: Nov 1 - Dec 31, midwinter: Jan 1 - Feb 28, late winter: Mar 1 - Apr 30, spring: May 1 - Jun 30) in wolf encounters with elk and bison in Yellowstone National Park, May 1995 - March 2000. 69 vi Figure 11. Predicted mean duration (min.) of hunting states with 95% confidence intervals in wolf encounters with bison herds in Yellowstone National Park, May 1995 - March 2000. Fitted means and confidence intervals are derived from GLMM results (Table 6). 70 Figure 12. The association between hunting success (kills/encounter) and prey size for wolves hunting various North American prey. 71 1 INTRODUCTION Previous attempts to observe the behavior of wolves hunting have been frustrated by a number of factors including dense vegetation, rugged topography, and logistical constraints such as access to remote study sites and fuel limitations during aerial observation (Mech 1966a; Clark 1971; Haber 1977; Carbyn et al. 1993). As a result, studies of wolf predation have depended largely on the examination of remains from wolf-killed ungulates for insight into wolf hunting behavior. Most studies have found that wolves kill mainly vulnerable prey (e.g., prey easily captured due to their circumstantial, behavioral, or physical condition) (Murie 1944; Mech 1970; Carbyn 1974; Peterson 1977; Mech et al. 1998). In general, predators kill vulnerable prey when prey are difficult to capture, and very difficult prey are aggressive and can injure a predator (Temple 1987). Therefore, for predators that rely on dangerous prey for food, vulnerable prey are probably safer to kill. Wolves feed mainly on ungulates that can injure (Mech 1966a; Rausch 1967; Phillips 1984; Carbyn and Trottier 1988) or kill them (Ballard et al. 1987; PasitschniakArts et al. 1988; Mech and Nelson 1990; Weaver et al. 1992). Since failure to avoid injury or death greatly decreases fitness, the risk of prey-caused injury may be a strong selective force favoring wolf preference for vulnerable prey. If preference for vulnerable prey is an adaptive strategy to acquire food while avoiding injury wolves should be able to assess their risk of being injured or killed by prey and incorporate this information into their foraging decisions. 2 In addition, if safer prey provide less food than dangerous prey, wolves cannot simultaneously maximize food intake and minimize injury risk. Therefore, wolf preference for vulnerable prey may represent a trade-off between food and safety. For example, assuming prey size is related to injury risk (i.e. large prey are more dangerous than small prey) (Weaver et al. 1992), wolves may prefer small prey at the expense of food intake in order to minimize injury risk. Trade-offs between food and safety have only been examined for foragers that attempt to maximize food intake while minimizing the risk of predation (Krebs 1980; Newman and Caraco 1987; Lima and Dill 1990). Predators that elicit a conspicuous fear response from prey (e.g. trade food for safety) are considered “fierce” (Brown et al. 1999). In this model system, fierce predators freely pursue timid prey and contend only with the risk of starvation. However, predators are known to respond to the risk of preycaused injury, either by avoiding dangerous prey or modifying their behavior while foraging on dangerous prey. For instance, piscivorous birds attack fish with dangerous spines less often, and handle them longer than fish without spines (Forbes 1989). The dangerous-prey hypothesis proposes that predators respond to an increase in injury risk by handling dangerous prey more carefully, leading to longer handling times (Forbes 1989). In this case, predators manage the trade-off between food and safety through adjustments in handling time that lower injury risk at the expense of prey profitability, since handling time and prey profitability (e.g., net energetic return) are inversely related (Stephens and Krebs 1986). Although the tendency for wolves to kill vulnerable prey suggests that wolves trade between food and safety, the difficulty of 3 observing wolves hunt has hindered a close examination of the influence of injury risk on their hunting behavior. To examine wolf response to injury risk, one must first clearly characterize and define their hunting behavior. In most studies, observations of wolves hunting are limited in number, lacking in detail, and many are incomplete because the beginning or end of the hunt was not observed. As a result, no single study has produced a definitive account of wolf hunting behavior. Rather, current knowledge of wolf hunting behavior is the result of information accumulated over several generations of field studies. Initial studies provided the first general descriptions of wolf hunting behavior, indicating that wolves engage in several different types of behavior while hunting (Murie 1944; Banfield 1954; Tener 1954; Crisler 1956; Kelsall 1957, 1960). However, these early studies neither identified the individual behaviors explicitly, nor examined the relationships among them. A second generation of studies identified and described several types of wolf hunting behavior, and recognized that the behaviors occur in a logical sequence (Mech 1966a, 1970; Gray 1983; Carbyn and Trottier 1987). Mech (1970) decomposed hunting behavior into five “stages”: travel, stalk, encounter, rush, and chase. Gray (1983) described the behavior of wolves hunting muskoxen (Ovibos moschatus) as a sequence of six “events”: approach, circle herd, attack herd, cut off single individual, contact individual, and kill individual. For hunts of bison Carbyn and Trottier (1987) described five “categories”: watch, trail, trail and follow-up, harass, and rush. 4 The three predatory sequences differ in three main respects. First, the three sequences do not share the same set of behaviors. Second, different definitions are assigned to the same behavior. For example, Mech (1970) described the rush as an initial charge toward prey when prey are first encountered, while Carbyn and Trottier (1987) considered the rush to be the point during the encounter at which wolves grab prey. Third, similar definitions are assigned to different behaviors. For instance, the initial period of movement toward prey preceding attack has been described as ‘stalk’ (Mech 1970), ‘approach’ (Gray 1983), and ‘trail and follow-up’ (Carbyn and Trottier 1987). Lack of consensus on the type and definition of behaviors that constitute the wolf predatory sequence hinders further study of wolf hunting behavior, and highlights the need for further clarification and explanation. In addition, the exact relationship between behaviors in the predatory sequence and more general types of wolf hunting behavior is uncertain. Specifically, studies that describe sustained periods of hunting activity in which wolves travel from one prey to another in succession, and make one or more attempts to kill prey during each encounter (Murie 1944; Mech 1966a, b; Haber 1977; Carbyn and Trottier 1987; Carbyn et al. 1993; Huggard 1993; Mech et al.1998), imply that behaviors from the predatory sequence are components of more general types of hunting behavior (e.g., hunting bouts, prey encounters, and predation attempts). These general types of hunting behavior and their relationship with specific behaviors of the predatory sequence have not been explicitly defined. 5 General patterns of wolf hunting behavior have been considered mainly in terms of the behavior necessary to locate uncommon and widely dispersed vulnerable prey. For instance, to increase the probability of locating vulnerable prey, wolves are believed to encounter and attack several different sets of prey during a hunt (Murie 1944; Mech 1970; Mech et al. 1998), and to prefer prey herds to solitary prey (Huggard 1993; Hebblewhite 2000). As a result, long hunts should have more prey encounters than short hunts, and successful hunts should be marked by higher prey encounter rates than unsuccessful hunts. Also, where prey herds are available, wolves should encounter prey herds more frequently than solitary prey. Multiple prey encounters during a hunt might occur simultaneously or consecutively (Murie 1944; Mech 1966a, b; Clark 1971; Haber 1977; Gray 1983; Mech et al. 1998). It is unknown whether the outcome of an encounter has any influence on the occurrence, or outcome, of a consecutive encounter. Prey encounters can also involve prey individuals that wolves encountered previously in the hunt (Clark 1971), or during a different hunt (Fuller 1962; Mech 1966a; Carbyn et al. 1993). The various types of prey encounter have not been defined, nor has the frequency of their occurrence been measured. Multiple attempts to kill prey during a single prey encounter have also been observed. Multiple attempts can occur simultaneously (Murie 1944, Mech 1966a, b, 1988; Clark 1971; Carbyn 1974; Mech et al. 1998) or consecutively (Fuller 1957; Mech 1966a, b; Miller and Gunn 1977; Miller et al. 1985; Carbyn and Trottier 1987; Gray 1987; Carbyn et al. 1993; Mech et al. 1998). It is also unknown whether the outcome of a 6 predation attempt has any influence on the occurrence, or outcome, of a consecutive predation attempt. The frequency of multiple predation attempts during prey encounters is not known. Overall, detailed information requiring extended and uninterrupted observations of wolves hunting is scarce. For instance, no observational data are available describing prey search time or prey encounter rate. In the absence of a clearly defined predatory sequence, estimates of wolf hunting success have involved a variety of measures defined differently for different studies. As a result, comparisons of hunting success across studies can be confounded if the type of measure and its definition are not taken into account. Previous measures of wolf hunting success based on direct observation include: (1) number of kills per prey animal tested (Mech 1966a; Haber 1977; Peterson 1977; Mech et al. 1998), (2) number of kills per encounter (Carbyn et al. 1993; Mech et al. 1998), (3) number of kills per chase (Clark 1971; Carbyn et al. 1993; Nelson and Mech 1993), and (4) number of ‘successful’ behaviors per behavior of the same type (e.g., number of approaches resulting in an attack per approach) (Mech 1970; Peterson 1977; Carbyn and Trottier 1987). Confusion also arises over the term 'test'. A ‘test’ can be a prey encounter where wolves pursue or hold an individual prey at bay (Mech 1966a; Peterson 1977; Mech et al. 1998), or a prey encounter where wolves simply move towards an individual prey without necessarily pursuing or holding it at bay (Haber 1977). A 'test' can also apply to entire herds of prey where hunting success is expressed as the number of kills per herd 7 tested, such that if a herd split into several smaller groups the encounter is still treated as a single test (Haber 1977). Similarly, estimates of success measured as kills per encounter score the outcome of the overall encounter rather than the outcome of each predation attempt that might comprise an encounter. In this case an encounter is considered to occur when wolves watch prey (Carbyn et al. 1993), or at least approach prey (Mech et al. 1998). Where success is estimated as the proportion of all chases that resulted in a kill, it is uncertain whether ‘chases’ included predation attempts when prey were only held at bay (e.g., when pursuit did not occur). Here I first review and clarify the predatory sequence for wolves to establish a general framework within which to analyze their hunting behavior. I apply this framework to a highly observable population of gray wolves recently restored to YNP to (1) elucidate general patterns in wolf hunting behavior that have been difficult to quantify previously, and (2) estimate rates of hunting success for various prey in YNP. I compare reported rates of success among several different prey in North America to determine if wolf hunting success is generally related to prey size. Next I examine the influence of injury risk on patterns of wolf hunting behavior during encounters with bison (350-1000 kg) and elk (75-340 kg). First, I demonstrate that (1) prey in general are dangerous to wolves, and (2) bison are more dangerous than elk. Second, I evaluate whether wolves (1) make foraging decisions based on injury risk, and (2) trade food for safety. If elk are more vulnerable than bison, wolves should attempt to 8 kill elk more frequently than bison, and spend less time in elk encounters than bison encounters. Components of the Predatory Sequence A review of all available published accounts of wolf hunting behavior indicates that the predatory sequence in wolves is composed of six general hunting states which I describe as: travel, watch, approach, attack, target, and capture. In general, hunting states can be characterized according to the type of gait, or lack thereof, used by the wolves. All hunting states are defined in terms of the behavior of the wolves, independent of prey behavior. Although wolves may engage in each hunting state in a predatory sequence, they may also skip or repeat one or more hunting states. For instance, during encounters with bison in Wood Buffalo National Park, wolves sometimes go directly from the watch state to the attack state, or begin at the attack state, or even the capture state (Carbyn and Trottier 1987). As a result, I refer to the components of the predatory sequence as hunting states, where a state is considered to be any distinct behavior with a measurable duration (Martin and Bateson 1993). I use the term ‘state’ to avoid the implication that components necessarily follow a definite order as suggested by ‘stage’. Travel State The need to locate vulnerable prey requires that wolves travel widely (Murie 1944; Mech 1966a); therefore wolves are considered hunting anytime they are traveling 9 (Murie 1944; Kelsall 1960; Mech 1966a, 1970; Haber 1977; Mech et al. 1998). Traveling involves wolves using any type of gait to move across the landscape without an obvious intention to move toward a particular prey. While traveling, wolves locate prey either by sight, direct scent, chance encounter, or tracking (Mech 1970). Although there are no reports on the duration of travel necessary before encountering prey in general, Mech (1966a) reported one case in which wolves traveled 100 km or more before encountering vulnerable moose (Alces alces). Watch State Watching involves intent staring at prey (Clark 1971; Haber 1977; Nelson and Mech 1994; Mech et al.1998), and has been described by some as surveillance (Clark 1971; Carbyn and Trottier 1988; Carbyn et al. 1993). Watching is believed to be a means by which wolves assess the vulnerability of prey and thereby their chance for success (Murie 1944; Mech et al. 1998). Wolves may assess prey by identifying a strategic advantage (Mech et al. 1998). For example, wolves hunting bison have been observed to wait until bison flee before they attempt to capture a calf (Carbyn and Trottier 1988), and wolves hunting muskoxen are known to wait until a calf moves outside the protective ring of the adults before attacking (Tener 1954). When wolves are seeking a strategic advantage, they may appear to sleep while watching prey, but are quick to strike when an opportunity presents itself (Carbyn and Trottier 1988; Mech 1988). In addition, Murie's (1944) observation that wolves approach caribou (Rangifer tarandus) and then watch them flee suggests that wolves may assess the physical condition of prey by observing 10 their locomotion, similar to the function of watching behavior described for spotted hyena (Crocuta crocuta) (Kruuk 1972; Holekamp et. al 1997) and wild dog (Lycaon pictus) (Fitzgibbon and Fanshawe 1988) hunting Thompson's gazelle (Gazella thomsonii). Wolves may watch when they first encounter prey (Haber 1968; Carbyn 1974; Carbyn & Trottier 1987), after an initial approach (Murie 1944; Kelsall 1960; Haber 1968; Clark 1971; Carbyn 1974; Mech 1988; Nelson and Mech 1994), or following a failed attack (Murie 1944; Tener 1954; Mech 1966a; Gray 1983, 1987; Carbyn and Trottier 1988; Mech 1988). Except for an observation by Murie (1944) of a solitary wolf watching Dall sheep (Ovis dalli) after a failed attack, all reported watches following an attack have involved large, formidable prey, including moose (Mech 1966a), muskoxen (Tener 1954; Gray 1983, 1987; Mech 1988), and bison (Carbyn and Trottier 1987, 1988; Smith et al. 2000). In general, the distance at which wolves watch prey tends to be greatest for medium-size prey such as elk and caribou, and least for large prey such as moose, bison, and muskoxen. Wolves have been reported to watch medium-size prey from 23 to 410 m (Murie 1944; Kelsall 1960; Haber 1968; Clark 1971; Carbyn 1974) and large prey from 3 to 200 m (Tener 1954; Mech 1966a; Haber 1977; Carbyn and Trottier 1988; Mech 1988). However, in one exceptional case three wolves watched an adult white-tailed deer from 25 m prior to an approach, and from 5 m following an approach (Nelson and Mech 1994). The distance at which wolves watch prey also tends to be greatest prior to initial attack. Prior to initial attack, wolves have been reported to watch prey from 3 to 410 m 11 (Murie 1944; Kelsall 1960; Haber 1968; Clark 1971; Carbyn 1974; Haber 1977; Mech 1988), and following an initial attack from 23 to 200 m (Tener 1954; Mech 1966a; Carbyn and Trottier 1988). Approach State An approach involves wolves walking (Murie 1944; Banfield 1954; Crisler 1956; Kelsall 1957, 1960; Haber 1968; Mech 1970, 1988; Clark 1971; Carbyn 1974; Gray 1983, 1987; Miller et al. 1985; Carbyn and Trottier 1988) or trotting (Murie 1944; Banfield 1954; Haber 1968, 1977) toward prey. In one observation, Gray (1970) characterized the approach of one wolf toward an adult male muskox as a gallop. When wolves walk toward prey, they may do so casually with no attempt at concealment (Murie 1944; Kelsall 1960; Carbyn 1974; Gray 1983, 1987; Miller et al. 1985; Carbyn & Trottier 1988; Mech 1988; Nelson and Mech 1994) or they may stalk prey by walking either upright, slowly and deliberately (Banfield 1954; Crisler 1956; Kelsall 1957, 1960; Haber 1968; Mech 1970, 1988) or in a crouch, using topography (Clark 1971; Mech, USGS, unpublished) or vegetation to conceal themselves (Haber 1977). Reports of wolves stalking prey tend to be more common for encounters with medium-size prey such as caribou (Banfield 1954; Crisler 1956; Kelsall 1957, 1960; Haber 1968, 1977; Clark 1971) and Dall sheep (Murie 1944; cf. Haber 1968), than for encounters with large prey such as moose (Mech 1966a) and muskoxen (Mech 1988), although they do stalk the latter (Mech, USGS, unpublished). Wolves might stalk medium-size prey to reduce the chance that they flee (Mech 1970). However, Murie 12 (1944) found that without stalking, wolves could approach to within a few hundred meters of caribou. If wolves do not leave prey, an approach can be followed by either a period of watching (Murie 1944; Banfield 1954; Mech 1966b, 1988; Haber 1977; Carbyn and Trottier 1988), or by some form of attack (Murie 1944; Banfield 1954; Crisler 1956; Kelsall 1957; Mech 1966a, b, 1988; Haber 1977; Peterson 1977; Gray 1983, 1987; Miller et al. 1985; Carbyn and Trottier 1988; Mech et al. 1998; Mech and Adams 1999). Attack State An attack involves wolves pursuing (Murie 1944; Banfield 1954; Crisler 1956; Kelsall 1957, 1960; Mech 1966a, b, 1988; Haber 1977; Peterson 1977; Gray 1983, 1987; Miller et al. 1985; Carbyn and Trottier 1987, 1988; Carbyn et al. 1993; Nelson and Mech 1993; Mech et al. 1998; Mech and Adams 1999) and/or holding prey at bay (Tener 1954; Mech 1966a, 1988; Gray 1970; Miller and Gunn 1977; Peterson 1977; Gray 1983, 1987; Carbyn and Trottier 1987, 1988; Carbyn et al. 1993; Mech et al. 1998; Mech and Adams 1999). Elsewhere, the term ‘attack’ has also been used to describe wolves and coyotes biting and grabbing prey (Mech 1966a; Gese and Grothe 1995) or pursuing an individual prey (Lingle 2002). The gait of wolves pursuing prey is usually a gallop (Murie 1944; Crisler 1956; Kelsall 1960; Miller et al. 1985; Carbyn and Trottier 1988) or trot (Murie 1944). If a prey herd fragments into several smaller groups during a pursuit, wolves may move from one group to another in succession (Murie 1944; Gray 1987), presumably to locate a 13 vulnerable individual (Murie 1944; Mech et al. 1998). Otherwise, if the herd continues to flee as a single group, wolves may keep to the rear of the herd and wait for a vulnerable individual to fall behind (Murie 1944; Crisler 1956; Kelsall 1960; Haber 1977; Carbyn and Trottier 1988). All reports of wolves holding prey at bay, except 2 involving white-tailed deer (Odocoileus virginianus) (Mech 1984; Nelson and Mech 1994), involve large prey including moose (Mech 1966a; Peterson 1977; Mech et al. 1998), muskoxen (Tener 1954; Gray 1970, 1983, 1987; Miller and Gunn 1977; Mech 1988; Mech and Adams 1999), and bison (Carbyn and Trottier 1987, 1988; Carbyn et al. 1993). If wolves attack more than one prey and single out an individual, the next hunting state in the predatory sequence would be target. If wolves attack a solitary prey and grab it, the next hunting state would be capture. Target State The target state involves wolves pursuing and/or holding at bay a specific individual from a prey group (i.e. 2 or more individuals) (Murie 1944; Banfield 1954; Crisler 1956; Fuller 1957; Kelsall 1960, 1968; Mech 1966a, b, 1988; Haber 1977; Miller and Gunn 1977; Gray 1983, 1987; Carbyn and Trottier 1987, 1988; Carbyn et al. 1993; Mech et al. 1998; Mech and Adams 1999). If wolves are in pursuit, their running speed typically increases from the attack state to the target state (Murie 1944; Crisler 1956; Kelsall 1960; Carbyn and Trottier 1987). Prey targeted by wolves are often young (e.g., < 12 months) (Murie 1944; Kelsall 1968; Haber 1977; Miller and Gunn 1977; 14 Carbyn and Trottier 1987, 1988; Mech 1988; Mech et al. 1998) or crippled (Mech and Frenzel 1971; Peterson 1977; Carbyn and Trottier 1987). Sometimes wolves target a prey that falls behind the herd (Crisler 1956; Haber 1977; Mech 1988) or stumbles (Kelsall 1960). The target prey can be pursued or held at bay repeatedly if wolves fail to initially subdue it (Fuller 1957; Mech 1966a; Carbyn et al. 1993). If wolves grab and restrain the prey, the next hunting state in the predatory sequence would be capture. Capture State The capture state involves wolves grabbing and restraining an individual prey (Murie 1944; Kesall 1960; Mech 1966a; Peterson 1977; Carbyn and Trottier 1987; Gray 1983, 1987; Carbyn and Trottier 1988; Carbyn et al. 1993; Nelson and Mech 1993; Mech and Adams 1999). The capture state results in wolves killing prey (Murie 1944; Kelsall 1960; Mech 1966a; Peterson 1977; Smith 1980; Carbyn et al. 1993; Mech and Adams 1999) or prey escaping wolves (Mech 1966a; Peterson 1977; Gray 1983; Carbyn and Trottier 1988; Nelson and Mech 1993; Mech and Adams 1999). A Framework for the Predatory Sequence To examine general patterns of wolf hunting behavior that involved one or more hunting states, I developed a framework that defined the relationship between the individual hunting states and three nested groups of hunting states: predation attempt, prey encounter, and hunting bout (Fig. 1). A hunting bout was a discrete period beginning when wolves began traveling or encountering live prey (whichever came first) and ending 15 when wolves stopped encountering live prey and rested (e.g., sat or laid down) or stopped traveling (whichever came last). A hunting bout could contain at least one hunting state (except watch, see below) or a continuous sequence of hunting states. Wolves often howl or socialize while hunting (Banfield 1954; Mech 1966a; Haber 1977), and I considered this to be part of the hunting bout, rather than marking its conclusion. One or more hunting bouts could occur per day. A prey encounter was a period during a hunting bout involving prey and containing one or more of the following hunting states: approach, watch, attack, target, capture. The unit of encounter during a hunting bout was either a solitary prey or herd of prey. An individual prey was considered to be in a herd if its nearest neighbor was ≤ 20 m away. I considered a prey encounter to begin when wolves sighted prey and walked or ran toward them. A prey encounter ended when wolves stopped staring at prey or otherwise ceased to pay attention to prey (e.g., traveled away from prey). Situations that involved wolves watching prey without initially walking or running toward them were not considered encounters by my definition. One or more prey encounters could occur per hunting bout. Prey encounters were categorized as new, consecutive, repeat, return, or simultaneous. A new encounter was the first encounter to occur in a hunting bout and involved prey not previously encountered earlier in the day or in preceding days. A consecutive encounter denoted the second, third, fourth, etc. encounter during a hunting bout, and involved prey not previously encountered in the hunting bout or during a different hunting bout earlier in the day or in preceding days. A repeat encounter 16 involved prey encountered earlier in the same hunting bout, while a return encounter involved prey encountered in a different hunting bout earlier in the same day or in preceding days. An encounter was simultaneous if it occurred at the same time as another encounter in the same hunting bout. A predation attempt occurred when wolves pursued, held at bay, or grabbed prey. It was a period during a prey encounter that involved the sequential occurrence of one or more of the following hunting states: attack, target (prey groups only), or capture. A predation attempt failed if (1) the sequence leading from attack to capture was interrupted or (2) a capture did not result in a kill. A subsequent attempt began when the sequence restarted at one of the three hunting states. Multiple predation attempts could be consecutive and/or simultaneous. A consecutive predation attempt was one that occurred following a preceding attempt, and a simultaneous attempt was one that occurred at the same time as another attempt in the same prey encounter. METHODS Study Area Yellowstone National Park extends across 891,000 ha of a primarily forested plateau in northwestern Wyoming (Fig. 2). Elevations range from 1,500 m to 3,300 m. Several large montane grasslands punctuate the Yellowstone plateau and provide unobstructed views of wildlife. However, continuous viewing can be inhibited by forests on the periphery of grasslands, by occasional trees within grasslands and by varied 17 topography. Approximately 35,000 elk, 4,000 mule deer (Odocoileus hemionus), 3,000 bison, 700 moose, 200 pronghorn (Antilocapra americana), 200 bighorn (Ovis canadensis) and scattered mountain goats (Oreamnos americanus) reside in YNP during all or part of the year (D.W. Smith, National Park Service, unpublished data). Observations of wolves hunting were made primarily in a 100,000 ha complex of montane grasslands situated in the northern quarter of YNP referred to as the Northern Range (Fig. 2). The Northern Range is a series of open valleys, ridges, and minor plateaus linked by the Lamar and Yellowstone Rivers. Low elevations (1,500 m to 2,400 m) on the Northern Range create the warmest and driest conditions in YNP during winter. As a result, the Northern Range serves as the principal winter range for nearly 12,000 elk and 700 bison (D.W. Smith, National Park Service, unpublished data). Elk and bison occurred in singles or in herds of up to 800 and 75 animals, respectively. A single paved road runs the length of the Northern Range and provides year-round access. Wolves were routinely visible from observation points on or near the road. In winter, wolves were also observed from a hilltop observation point in Pelican Valley in the interior of YNP at approximately 2,500 m. Pelican Valley was accessed in winter by ski. Study Population A combined total of 31 radio-collared wolves were reintroduced to YNP in 1995 and 1996 (Bangs and Fritts 1996; Phillips and Smith 1996). Each subsequent year YNP personnel radio-collared 30-50% of the pups born (Smith et al. 2000). Wolves observed in the study were either members or descendents of the original reintroduced population. 18 From 1995 to 2000, 14-110 wolves comprised 2-7 packs of 2 to 27 wolves per pack (9.9 ± 1.0 wolves/pack, N = 37 pack-years). Approximately 20-60 of the wolves studied were radio-collared and 20-25 were individually recognizable by combination of color pattern, radio-frequency, and body conformation. Observations of wolves hunting were recorded from May 1995 to March 2000. During the study the number and location of packs changed. In 1995, the study population was limited to the three initial packs reintroduced to the Northern Range: Crystal Creek, Rose Creek and Soda Butte. The Soda Butte pack eventually moved outside YNP onto private lands, was returned to YNP, and was released in a remote southern region of the park. In early 1996, a female and male dispersed from the Rose Creek and Crystal Creek packs, respectively, and formed a third pack on the Northern Range called Leopold. Shortly after their release in 1996, the Druid Peak pack replaced the Crystal Creek pack on the Northern Range, and the Crystal Creek pack relocated to the more remote Pelican Valley. For the remainder of the study, the Rose Creek, Leopold, and Druid Peak packs were the focus of study because they inhabited the easily accessed, and sparsely forested Northern Range (Fig. 2). In 1996, the Nez Perce and Chief Joseph packs were also released, but they inhabited areas too forested and/or inaccessible to allow observation from the ground. Hunting Observations Wolves were mainly observed hunting during two annual 30-day intensive study periods in March and mid-November to mid-December (Fig. 3). In general, wolves on the 19 Northern Range were more easily observed during winter because they were attracted to ungulates concentrating on low elevation winter range that was easily accessed by observers. Observations were also made in April, May, and June during annual wolf-denmonitoring studies. Observations during other months were recorded opportunistically. During each study period, teams of two observers were assigned to daily monitor each Northern Range wolf pack from the ground from dawn to dusk. Observation effort per hour was generally constant throughout the day. Nighttime observation was attempted with night vision goggles but was ineffective due to long distances between wolves and observers. In the eight study periods from May 1995 to March 2000, observers on the Northern Range watched wolves for a total 1,901 hours. At least two observers monitored the Crystal Creek pack in Pelican Valley during March 13-19, 1999 and March 23-31, 2000. Observers in Pelican Valley watched wolves for a total of 80 hours. All packs were located daily from fixed-wing aircraft, weather permitting, during each study period. Outside study periods, wolves were located from the air weekly. Hunting behavior was observed from the aircraft as well as from the ground. Aerial observers recorded 24 wolfprey encounters, and ground observers recorded 560 wolf-prey encounters. To standardize data collection, each observer was trained to record wolf hunting behavior prior to each study period. Observers on the ground first located wolf packs with radio-telemetry to obtain a directional fix, and visually located and observed wolves with binoculars and spotting scopes. Observers watched wolves at distances of 0.1 - 6.0 km for as long as they remained in view and recorded hunting behavior using hand-held voice recorders and 20 digital stopwatches. Recorded observations were subsequently transcribed onto data forms. Some observations were also recorded on video. During each hunting state, observers recorded the following: duration of the hunting state, number, age, and gender of wolves, number and age/sex class of prey, and prey behavioral response. Wolf age and gender were determined during the annual effort to capture and radio-collar wolves (Smith et al. 2000). Sex of wolves not captured was determined by noting wolf body position while urinating (Mech 1970). Wolf age was also determined during den monitoring by noting when individuals were born. Hunting states that started or ended out of view were excluded from duration analyses. For this study only watch states that occurred at close range (< 10 m) were noted. Unless otherwise noted, all estimates of hunting success were based on prey encounters, predation attempts, and hunting states that were observed in their entirety. Estimates of hunting success included encounters involving both solitary and group hunts. Hunting success was measured at the level of prey encounter (kills/encounter), predation attempt (kills/attempt), and hunting state. For prey encounters and predation attempts, a success was considered to be the occurrence of a single ungulate kill. At the level of hunting state, success was measured according to whether the subsequent hunting state in the predatory sequence occurred. For example, the success of wolves approaching was measured as the proportion of approaches that resulted in an attack (i.e., number of attacks per approach). Data on the association between hunting group size and hunting success will be presented elsewhere. 21 Statistical Methods Means are reported with standard errors throughout, and for all analyses, results were considered significant at P < 0.05. P-values shown are for two-tailed tests. Frequency data, such as kills/encounter and kills/attempt, were evaluated with Pearson's chi-square test, or if more than one-fifth of fitted cells were sparse (e.g., frequency < 5), with Fisher's exact test. All continuous data were checked for normality prior to analysis. To satisfy normality assumptions, duration and count data (e.g., number of encounters and attempts) were log and square-root transformed, respectively. However, results were plotted in the original scale to aid interpretation. Prey encounter and predation attempt duration were evaluated with Student's t-test. Continuous data were analyzed with a Mann-Whitney U-test and Spearman’s rank correlation coefficient if sample sizes were small (< 30), or transformations were not adequate. The relationship between prey mass and anatomical point of capture was evaluated using ANOVA. Assumed weights (kg) for elk were: cow, 226; yearling, 165; calf, 103 and bull, 266 (K. E. Murphy, National Park Service, unpublished data). Assumed weights (kg) for bison were: calf, 271; cow, 430; and bull, 679 (Meagher 1973). All the above tests assume independence of observations. Analyses of hunting-state duration were performed with general linear mixed models (GLMMs) (Verbeke and Molenberghs 2000) using the SAS 8.0 analysis package (SAS Inc. 2000). A mixed linear model is a generalization of the standard linear model (i.e. simple linear regression) which accounts for correlation and non-constant variance in the data. Hunting states were correlated if they occurred during the same hunting bout or prey encounter. In these models, an unstructured correlation matrix was used, which 22 allows for any level of correlation among hunting states within the same prey encounter and within the same hunting bout. Model parameters were estimated using maximumlikelihood estimation, and significance of effects was determined by an approximate ttest. Model reduction was performed using the likelihood-ratio test. Results were robust to other choices of correlation matrices. Predicted mean hunting-state duration was derived from the GLMM analysis and plotted with confidence limits to illustrate the significance of comparisons between different types of hunting state. Sample sizes varied considerably among tests because not all observations contained the same quality of information. For example, 267 prey encounters were observed in their entirety, but accurate duration data were only recorded in 175 of those encounters. Thus, analyses of encounter duration were restricted to those 175 prey encounters. The association between prey size and hunting success was examined by first summarizing reported rates of success for various North American prey species. Where more than one estimate was available for a particular prey species, estimates were compared within prey species, using Pearson’s chi-square or Fisher’s exact test, to identify if differences existed among studies. To quantify prey size, mean weights (kg) were estimated for each prey species based on Nowak (1999). Spearman's rank correlation coefficient (rs) was used to test the association between success rate and mean prey weight. 23 RESULTS Hunting Bouts Observers watching the Rose Creek, Leopold, Druid Peak, and Crystal Creek packs recorded 62 hunting bouts in their entirety and portions of an additional 400 (Table 1). I personally observed and recorded 78 (17%) of those 462 hunting bouts. During intensive study periods, packs made on average 1.08 + 0.22 hunting bouts/observation hour (N = 181 pack-observation days), and hunting bouts were observed mainly in the morning and evening (Fig. 4). Wolf behavior immediately preceding a hunt was noted in 91 hunting bouts and included 36 (40%) resting, 21 (23%) sleeping, 12 (13%) feeding, 17 (19%) rallying (e.g., excited greeting), and 5 (5%) howling. The initial hunting state in 144 hunting bouts observed at the start included 101 (70%) travel, 36 (25%) approach, and 7 (5%) attack states. In 65 hunting bouts that began with travel, wolves encountered prey within 1 to 118 minutes (22.30 + 3.10 min., N = 65). Duration of hunting bouts was 3-594 minutes (48.10 ± 9.80 min., N = 62), and included 0-3 prey encounters (1.20 ± 0.09 encounters/bout, N = 62). Number of prey encounters/bout was not significantly associated with duration of hunting bout (Spearman rank correlation coefficient, rs = 0.22, N = 62, P=0.24). Prey Encounters Observers recorded 267 prey encounters in their entirety and portions of an additional 317. The initial hunting state among 336 prey encounters observed at the start included 291 (87%) approach, 44 (13%) attack, and 1 (0%) target states. Of the 584 total 24 prey encounters observed, 486 (83%) involved elk, 75 (13%) bison, 12 (2%) pronghorn, 6 (1%) bighorn sheep, 3 (0.5%) mule deer, and 2 (0.5%) moose encounters. Overall, wolves encountered prey at a rate of 3.10 ± 0.42 encounters/hour/bout (N = 62). Among 55 completely observed hunting bouts in the Northern Range where wolves encountered elk or bison, wolves encountered elk at a slightly higher rate (3.28 + 0.48 encounters/hour/bout, N = 49) than they encountered bison (3.18 + 0.80 encounters/hour/bout, N = 6) but the difference was not significant (Mann-Whitney Utest, z = 0.16, P = 0.68). Among 529 encounters involving elk or bison on the Northern Range, wolves did encounter elk more frequently 472 (89%) than bison 57 (11%). Data were unavailable to adequately determine whether wolves were encountering prey species proportionate to their occurrence in the study area. Overall, wolves encountered herds of prey more often (85% of N = 584 prey encounters) than solitary prey (15%; χ2 = 267, d.f. = 1, P < 0.001; Fig. 5). Again, data were not available to determine whether wolves were encountering prey herds in proportion to their occurrence in the study area. Among 175 prey encounters for which duration data were available, duration of prey encounters ranged from < 1 to 553 minutes (12.40 ± 3.30 min., N = 175). Of 134 multiple encounters that occurred during hunting bouts, 118 (88%) were consecutive, 8 (6%) simultaneous, 4 (3%) repeat, and 4 (3%) unknown. Among 178 prey encounters observed at their finish, a consecutive encounter was more likely to occur after an unsuccessful encounter (34.2%, N = 105) than after a successful encounter (8.2%, N = 73; χ2 = 16.23, d.f. = 1, P < 0.001). Among 123 consecutive prey encounters 25 that followed a prey encounter with known outcome, a consecutive encounter was less likely to be successful if it followed an unsuccessful encounter (15%, N = 115) than if it followed a successful encounter (50%, N = 8; Fisher exact test, P = 0.03). For 40 known intervals between consecutive encounters, time between consecutive encounters ranged from 0 to 41 minutes (7.50 ± 1.50 min., N = 40). Only 12 return encounters were observed, and 5 (42%) involved prey that were previously wounded. In 267 prey encounters observed in their entirety, the number of predation attempts per prey encounter was 0-45 (1.30 ± 0.18 attempts/encounter, N = 267). In the 175 prey encounters with known duration, the number of predation attempts was positively related to duration of prey encounter (Spearman rank correlation coefficient, rs = 0.39, N = 175, P < 0.001). During encounters, wolves attempted to kill prey at an overall rate of 0.30 ± .04 attempts/min./encounter (N = 175). Predation Attempts Observers recorded 565 predation attempts in their entirety and portions of an additional 171. In 320 predation attempts with known duration, predation attempt duration was 0.07-41 minutes (3.70 ± 0.31 min., N = 320). The initial hunting state among 564 predation attempts involving prey herds and observed from the start included 468 (83%) attack, 90 (16%) target, and 6 (1%) capture states. All 6 attempts beginning with a capture involved newly born elk calves. Among 260 multiple predation attempts that occurred during prey encounters, 196 (75%) were consecutive, 55 (21%) simultaneous, and 9 (4%) unknown. Among 443 26 predation attempts observed at their finish, a consecutive predation attempt tended to occur more often after an unsuccessful predation attempt (29%, N = 377) than after a successful predation attempt (17%, N = 66; χ2 = 3.60, d.f. = 1, P = 0.057). Among 92 consecutive predation attempts that followed a predation attempt with a known outcome, a consecutive attempt was less likely to be successful if it followed an unsuccessful predation attempt (5%, N = 83) than a successful predation attempt (22%, N = 9), but the difference was not significant (Fisher exact test, P = 0.10). For 75 known intervals between consecutive attempts, time between consecutive predation attempts was 0 - 88.5 minutes (6.60 ± 1.70 min., N = 75). Hunting States Observers recorded 1,472 complete hunting states and portions of an additional 391. Among the hunting states observed completely, 188 (13%) were travel, 83 (6%) watch, 369 (25%) approach, 494 (33%) attack, 225 (15%) target, and 113 (8%) capture. During elk and bison encounters, hunting-state duration remained the same between approaches and attacks, decreased during target, and increased during capture (Fig. 6). For both elk and bison encounters, target duration was significantly shorter than other hunting states (Table 2.) Wolves captured prey by grabbing the hindquarters, the neck, or both. Overall, wolves grabbed prey heavier than 270 kg by the hind end (N = 32), prey between 200 and 270 kg by the hind end and neck (N = 35), and prey < 200 kg by the neck only (N = 31) (ANOVA, F = 4.472, 95, P = 0.014). 27 Hunting Success Overall, the estimated rate of success was 0.21 ± 0.03 kills per encounter and 0.16 ± 0.02 kills per predation attempt (Table 3). Among prey encounters observed in their entirety, wolves were more successful hunting elk (0.24 ± 0.03 kills/encounter, N = 211) than bison (0.04 ± 0.03 kills/encounter, N =47; Mann-Whitney U-test, z = 8.11, P = 0.004), despite nearly twice as many predation attempts per bison encounter. In encounters with prey herds, the sequence of hunting states leading to a kill generally progressed from approach to attack, target and capture (Fig. 7a, 8a). Wolves that approached elk herds attacked more frequently (63%, N = 177) than wolves that approached bison herds (47%, N = 64; χ2 = 4.86, d.f. = 1, P = 0.03). During an attack on a prey herd, wolves were more likely to target an individual from an elk herd (42%, N = 180) than an individual from a bison herd (17%, N = 53; χ2 = 10.82, d.f. = 1, P = 0.001). Wolves were also more likely to capture a targeted elk (43%, N = 108) than a targeted bison (23%, N = 35; χ2 = 4.38, d.f. = 1, P = 0.04), and more likely to kill a captured elk (85%, N = 53) than a captured bison (20%, N = 10; Fisher exact test, P < 0.001). Wolves were just as likely to target an individual bison following a watch (19%, N = 42) or attack (17%, N = 53; χ2 = 0.07, d.f. = 1, P = 0.79). In encounters with solitary prey, the sequence of hunting states generally progressed from approach to attack, and to capture (Fig. 7b, 8b). Wolves that approached solitary elk tended to attack more often (76%, N = 25) than wolves that approached solitary bison (45%, N = 11), but the difference was not significant (Fisher exact test, P = 28 0.12). Wolves always killed captured solitary elk (N = 5), but never killed captured solitary bison (N = 10). We did not observe wolves watching solitary bison. Among elk, wolves killed adult females (48%), young-of-the-year (25%), adult males (19%), yearlings (2%), and elk for which age and sex were unknown (6%) (Table 4). Among bison, wolves killed young-of-the-year (80%) and only 1 adult (20%). Among 62 hunting bouts observed in their entirety, successful hunting bouts tended to be longer (64.80 ± 23.50 min., N = 25) than unsuccessful hunting bouts (36.70 ± 3.70 min., N = 37), but the difference was not significant (t = -0.77, d.f. = 45, P = 0.44). The rate of encounter did not differ between successful (3.10 ± 0.47 encounters/hour/bout, N = 25) and unsuccessful hunting bouts (3.10 ± 0.63 encounters/hour/bout, N = 37). Among the 62 hunting bouts observed in their entirety, only 2 (3%) contained multiple kills, all involving elk. Among 175 prey encounters with known duration, successful prey encounters were longer (27.68 ± 13.73 min., N = 40) than unsuccessful prey encounters (7.90 ± 1.24 min., N = 135; t = -4.39, d.f. = 66.8, P < 0.001). Among 134 prey encounters with at least one predation attempt, successful prey encounters involved a lower attempt rate (0.20 ± 0.02 attempts/min./encounter, N = 40) than unsuccessful encounters (0.47 ± 0.07 attempts/min./encounter, N = 94; t = 3.17, d.f. = 94, P < 0.01). Likewise, among 320 predation attempts with known duration, successful predation attempts were longer (6.60 ± 1.20 min., N = 46) than unsuccessful predation 29 attempts (3.20 ± 0.29 min., N = 274; t = -6.32, d.f. = 74.1, P < 0.001). Also, successful captures of elk and bison were significantly longer than failures (Fig. 9 and Table 5). In wolf encounters with elk, hunting success varied significantly over season and was highest during spring (Kruskal-Wallis one-way ANOVA, H = 9.24, P = 0.03; Fig. 10). In wolf encounters with bison, hunting success did not vary significantly over season (Kruskal-Wallis one-way ANOVA, H = 2.99, P = 0.22; Fig. 10) despite high hunting success during late winter. Anti-predator Response and the Risk of Injury During 258 encounters (211 with elk and 47 with bison) in which the behavioral response of prey to wolves was noted, prey stood and confronted (60%, N = 154) wolves more often than they fled (40%, N = 104; χ2 = 9.69, d.f. = 1, P = 0.02). Prey that confronted wolves were more aggressive than prey that fled from wolves, being more likely to charge wolves (48%, N = 154) than prey that fled (10%, N = 104; χ2 = 41.77, d.f. = 1, P < 0.001), and more likely to kick at wolves (11%, N = 154) than prey that fled (4%, N = 104; χ2 = 4.29, d.f. = 1, P = 0.04). Overall, prey that confronted wolves were less likely to be killed (14%, N = 154) than prey that fled from wolves (26%, N = 104; χ2 = 4.29, d.f. = 1, P = 0.02). Bison were more aggressive than elk during encounters with wolves. Bison stood and confronted wolves more frequently (79%, N = 47) than elk (55%, N = 211; χ2 = 8.6, d.f. = 1, P < 0.01), and charged wolves more frequently (62%, N = 47) than elk did (26%, N = 211; χ2 = 22.2, d.f. = 1, P < 0.001). 30 Hunting Behavior in Bison and Elk Encounters In general, wolf encounters with bison were longer (29.50 ± 17.50 min., N = 32) than with elk (8.80 ± 1.00 min., N = 139), but not significantly (t = -0.65, d.f. = 39.5, P = 0.52). One 33.5-hour bison encounter, monitored by radio-telemetry during darkness, was excluded from duration estimates, because it included the period from dusk to dawn when direct observation was not possible. Overall, a predation attempt was more likely during an elk encounter (80%, N = 211) than a bison encounter (60%, N = 47; χ2 = 8.5, d.f.=1, P < 0.01). The number of predation attempts tended to be greater during encounters with bison (1.90 ± 0.95 attempts/encounter, N = 47) than with elk (1.20 ± 0.09 attempts/encounter, N = 211; Mann-Whitney U-test, z = 3.12, P = 0.08). However, based on encounters with at least one predation attempt, rate of predation attempt did not differ between bison (0.27 ± 0.06 attempts/min./encounter, N = 32) and elk (0.30 ± 0.05 attempts/min./encounter, N = 139; t = 0.032, d.f. = 32, P = 0.98). Predation attempts were shorter in encounters with bison (2.90 ± 0.51 min., N = 101) than with elk (4.00 ± 0.38 min., N = 215; t = 4.04, d.f. = 165, P < 0.001). Time between consecutive predation attempts also tended to be longer for bison (8.80 ± 3.00 min., N = 41) than for elk (3.90 ± 0.88 min., N = 34), but not significantly (t = -0.493, d.f. = 56.3, P = 0.62). Among hunting states, an approach to a bison herd was more likely to result in a target (17%, N = 64) than was an approach to an elk herd (8%, N = 177; χ2 = 4.32, d.f. = 31 1, P = 0.04). Successful bison captures tended to take longer than successful elk captures, though the sample was small (Fig. 9). Wolves did not watch elk at close range (< 10 m). In encounters with bison herds, watch duration was significantly greater than the duration of other hunting states, except for the capture state (Fig. 11 and Table 6). Success Rates for Wolves Hunting Various North American Prey Wolves hunting bison tended to experience a lower rate of success in YNP than in Wood Buffalo National Park (Table 7), but the difference was not significant for kills/encounter (Fisher exact test, P = 0.69) or kills/attempt (Fisher exact test, P = 0.39). Reported success rates for moose were not significantly different among studies (Table 7) for kills/encounter (Fisher exact test, P = 0.13) or kills per animal (χ2 = 5.95, d.f. = 1, P = 0.11). Estimates of hunting success among studies differed for caribou (χ2 = 17.90, d.f. = 1, P < 0.001), but not for Dall sheep (Fisher exact test, P = 0.39). Wolves were more successful killing small prey than large prey. Overall, the rate of success decreased as prey size increased (Spearman rank correlation coefficient, rs = 0.60, N = 11, P < 0.05; Fig. 12). DISCUSSION The Predatory Sequence The predatory sequence presented for wolves is very similar to one reported for spotted hyenas. Kruuk (1972) defined the predatory sequence for spotted hyena as (1) 32 search, (2) random dash, (3) chase, and (4) kill. Similar to the attack state in wolves, hyenas gallop leisurely during the random dash and never at full speed. Kruuk (1972) speculated that the main function of the random dash is to make prey run, which might allow hyenas to identify any physically inferior individuals. Like the target state in wolves, hyenas increase their running speed and focus on a single prey during the chase (Kruuk 1972). The similarities in the predatory sequence between wolves and spotted hyenas are further evidence of the evolutionary convergence in foraging behavior between Canidae and Hyaenidae first noted by Kruuk (1972). In wolves, the sequence of hunting states that led to a kill was determined by several factors. First, whether wolves encountered herd prey or solitary prey determined if a target state was, by definition, included in a sequence. Second, the outcome of a hunting state also affected the actual sequence of hunting states comprising a hunting bout since hunting states that failed to lead to the next state (e.g., approach to attack, attack to target, target to capture) were often repeated. Third, some states were occasionally skipped such as when wolves approached a prey herd and immediately targeted an individual. A hunting bout usually began with travel. It only started with an approach or attack if wolves detected prey while resting. A prey encounter commonly started with an approach, and sometimes with an attack. If an encounter involved a prey herd, a predation attempt usually began with an attack, and rarely with a target. A predation attempt might start with a capture if newborn prey were encountered, in which case it was only necessary for wolves to approach and capture. 33 Brief target duration likely reflected constraints imposed by limited energetic reserves or the risk of prey-caused injury. Several observers have noted that the speed at which wolves pursue prey tends to increase when they transition from 'casual' pursuit of an entire herd to focused pursuit (i.e. target) of an individual (Murie 1944; Crisler 1956; Kelsall 1960; Carbyn and Trottier 1987). In a chase filmed in YNP (Landis, unpublished), wolves increased their speed from approximately 14 km/hr during the attack to 45 km/hr during the target state (calculated from the film [MacNulty, unpublished]). Since speed is positively associated with energetic cost (Alexander et al. 1980; Heglund and Taylor 1988), wolves might minimize target duration to minimize energy expenditures. When prey stand and confront wolves during a target, wolves may limit target duration to reduce the risk of injury from aggressive prey. Extended capture duration was likely due to the wolf’s small size and limited ability to subdue prey. Unlike big cats (Hornocker 1970; Schaller 1967, 1972), or grizzly bears (Ursus arctos) (MacNulty et al. 2002), wolves lack mass, muscular forelimbs, and longer claws that enable other carnivores to quickly grab and overcome prey. Wolves also cannot generally deliver a quick killing bite commonly used by the big cats. Instead, wolves rely solely on their teeth to grab prey and tear into a vital area until the prey weakens and falls. General Patterns of Hunting Behavior Previous studies indicate that wolves tend to socialize by howling, excited greeting (e.g., rally), or both, prior to initiating a hunting bout (Murie 1944; Mech 1966a; 34 Haber 1977). However, the function of group socializing in the hunting behavior of wolves is not clear. Group howling might serve to assemble the pack before setting out on a hunt (Mech 1970). A rally could be a food-begging ritual initiated by young wolves toward adults in anticipation of a kill, or a means for adults to motivate a pack leader to initiate a hunt (Mech 1970). Wild dogs also rally prior to hunting (Estes and Goddard 1967; Schaller 1972; Creel and Creel 1995), and it may focus the attention of all pack members on the hunt (Estes and Goddard 1967). In YNP, wolves were usually sleeping or resting, and occasionally feeding, prior to a hunting bout, and group socializing then was infrequent. Though few comparable data for wolves elsewhere are available, the tendency for YNP wolves to hunt more frequently during morning and evening than during mid-day (Merrill 2002; Theuerkauf et al. in press) is consistent with the observed pattern of hunting behavior for other carnivores including the African wild dog (Estes and Goddard 1967; Fuller and Kat 1990; Creel 2001), spotted hyena (Kruuk 1972), and lion (Schaller 1972). However, while the occurrence of a mid-day lull in hunting activity in YNP was clear, the exact time of maximal hunting activity, and the magnitude of the observed morning and evening peaks in hunting activity were likely affected by our inability to observe wolves before dawn and after dusk. The crepuscular hunting activity is in accord with the basic daily activity patterns of wolves (Mech and Merrill 1998; Merrill 2002). In general, the typical hunting pattern for wolves in YNP involved a hunting bout < 60 minutes long, containing at least one prey encounter (< 15 min.) and at least one predation attempt (< 4 min.). I found no evidence to support the hypothesis that wolves 35 locate vulnerable prey by making multiple encounters during a hunt (Murie 1944; Mech 1970; Mech et al. 1998). Contrary to expectation, number of prey encounters in a hunting bout did not differ between long and short hunting bouts, and encounter rates did not differ between successful and unsuccessful hunting bouts. Multiple prey encounters during hunting bouts were neither a prominent, nor important, feature of the overall pattern of hunting behavior for wolves in YNP. However, if locating vulnerable prey depends on making multiple prey encounters, wolves in YNP might make several prey encounters during a succession of short hunting bouts over the course of day rather than during a single hunting bout. Unfortunately, data were not sufficient to test this hypothesis. Nonetheless, generally short hunting bouts containing an average of 1 encounter per bout in YNP, and previous reports that wolves rest following a prey encounter involving an attack (Mech 1970), support the hypothesis that wolves make multiple prey encounters over several hunting bouts. If a hunting bout did include multiple encounters, they were more often consecutive than simultaneous or repeat. A consecutive encounter was more likely to occur, but less likely to succeed, if the previous encounter failed. Wolves that succeeded during an encounter were probably preoccupied with feeding and less motivated to make another encounter. My estimate of an approximate 8-minute interval between consecutive encounters is likely biased low due to a tendency for wolves to go out of sight during extended intervals between consecutive encounters. 36 Repeat encounters during a hunting bout may be infrequent due to increased prey wariness following the initial encounter, which may make it more difficult for wolves to relocate prey and/or kill prey once relocated. Also, wolves may not return to prey if they have already determined that a vulnerable prey is not present. Return encounters involving prey encountered in a previous hunting bout were also infrequent, perhaps for the same reason. However, because not all prey encountered and abandoned were wounded or otherwise identifiable, it was difficult to determine whether a prey had been encountered in a previous hunting bout. Therefore, the frequency of return encounters may have been underestimated. Consistent with earlier expectations that wolves prefer prey herds to solitary prey (Huggard 1993; Hebblewhite 2000), wolves in YNP encountered herds of prey more frequently than solitary prey. However, without our having information on the abundance of prey herds and solitary prey in the study area, we had no way of knowing whether wolves simply encountered herds in proportion to their occurrence. Visibility bias could also have influenced the tendency for wolves to encounter herds of prey more often than solitary prey if solitary prey inhabit forests more frequently than grasslands. However, evidence based on snow tracking in Banff National Park indicates that wolves encounter prey herds more often than solitary prey in both open and forested habitats (Huggard 1993; Hebblewhite 2000). Yellowstone wolves tended to encounter elk at a higher rate (encounters/hour/bout) than bison but not significantly so, likely due to the small sample of completely observed hunting bouts involving bison. Nonetheless, wolves did 37 encounter elk more frequently (proportion of the total number of prey encounters) than bison, which was consistent with greater elk densities in YNP (Smith et al. 2000). For other prey species found in or near forest such as moose and mule deer, visibility bias may have underestimated prey encounter frequencies. General Patterns of Hunting Success Overall, the duration of wolf hunting behavior was more important to hunting success in YNP than the frequency of hunting behavior. For example, while long prey encounters and predation attempts were significantly more successful than short ones, neither the rate at which wolves encountered prey, nor the rate at which wolves attempted to kill prey had an affect on hunting success. Successful encounters were characterized by significantly lower predation-attempt rates. Since long encounters were more successful than short ones it is not surprising that long hunting bouts contained no more prey encounters than short hunting bouts. Rather than make several encounters during a hunting bout, it may be more advantageous for wolves to make a few extended encounters during the same period. A similar association between hunting success and encounter duration, found for wolves hunting bison in Wood Buffalo National Park, Canada, was considered the result of prey distribution patterns. Wolves there were thought to retain contact with bison herds for long periods due to the difficulty in locating widely dispersed prey (Carbyn et al. 1993). Considering the high prey encounter rate in YNP, other factors may also be involved. 38 Clearly, one would expect a hunting success to be longer than a failure for the simple reason that it takes longer to kill than to give up. However, the exact factors that cause successful encounters to be prolonged are unknown. The relationship between hunting success and encounter duration may result from time expenditures required (1) to kill an individual prey once it has been captured, (2) to identify vulnerable prey upon encounter, and (3) to wait to capture a vulnerable prey once it has been identified (e.g., to identify a strategic advantage). The time required to subdue prey once it has been captured necessarily results in increased encounter duration, and in part explains the difference in duration between successful and unsuccessful encounters. However, since a successful capture only comprises an average 2.5 (elk) to 7 (bison) minutes of an encounter that is on average nearly 28 minutes long, capture duration by itself does not explain the relationship between hunting success and encounter duration. Identifying a prey as vulnerable requires a period of information gathering (Mech et al. 1998) that is likely not instantaneous. Failure to obtain adequate information may lead wolves to attack when they should not or to not attack when they should, resulting in unnecessary energetic expenditures or missed feeding opportunities, respectively (Hasson 1991). Wolves may also face unnecessary injury risk if they attack when they should not. Therefore, accurate prey assessment may require extended contact with prey and may be especially prolonged if prey mask their vulnerability (e.g., bluffing), and/or live in large herds. Prolonged encounter duration may also be related to an inability to identify vulnerability in prey. Since hunting experience likely influences accurate prey 39 assessment, inexperienced wolves may retain contact with prey for longer periods than experienced wolves. Haber (1977) believed that adult wolves (≥24 months) generally identified a potential victim as not vulnerable in less than 2-3 minutes, while young wolves (< 24 months) lingered near prey for hours or even a day longer making repeated attempts to capture a prey, although his method for determining age of wolves was not clear. Once wolves identify a prey as vulnerable, additional time is often necessary to grab the individual if other non-target prey are protecting it. For instance, female moose will aggressively protect their offspring from attacking wolves (Mech 1966a; Peterson 1977). In herd prey, such as bison (Carbyn et al. 1993; MacNulty et al. 2002) and muskoxen (Tener 1954; Gray 1987; Mech and Adams 1999), non-target individuals will shield the target individual, and attempt to drive off attacking wolves. To minimize their risk of injury from non-target prey, wolves must often wait and watch for an opportunity to strike at the target. Extended duration in successful encounters might also result from wolves capturing and wounding prey, followed by a period in which wolves lie and wait for the prey to weaken (Mech 1966a; Miller and Gunn 1977; Mech et al. 1998). The negative relationship between predation-attempt rate and hunting success may be explained in terms of wolf energetics, and the behavioral response of prey. Because predation attempts are likely energetically expensive causing wolves to tire, each additional attempt wolves make during an encounter may be less effective than the previous attempt. Also, the effectiveness of high predation-attempt rate during an 40 encounter may be diminished by variation in the intensity of anti-predator behavior. For instance, after an initial predation attempt, prey may become more alert or, if in a herd, they could consolidate into a tight group, thereby decreasing the chance of targeting, and thus that a subsequent predation attempt will succeed. Also, if the intensity of antipredator behavior decreases over a long period during an encounter, prey may be more susceptible to a few predation attempts over an extended encounter rather than a rapid series of predation attempts in a short encounter. Hunting success in elk encounters was highest during spring, a period during which elk in YNP experience nutritional stress due to low forage quality and abundance (Houston 1982). Consequently, wolves probably enjoyed greater hunting success during spring as a result of reduced elk condition (Mech 1977; Carbyn 1983). Hunting success in bison encounters probably did not vary significantly over season due to the small sample of bison encounters. Overall, bison were more difficult for wolves to kill in YNP, confirming an earlier result (Smith et al. 2000). In this study wolves infrequently killed bison (1 kill every 25 encounters) and did not make multiple kills. In contrast, wolves more frequently killed elk (1 kill every 4 encounters) and sometimes more than one at a time. Differences in success rate between wolves preying on bison and elk in YNP suggest that bison are more difficult to capture and kill than elk, perhaps because of size differences. 41 Hunting Success and Prey Size Analysis across several studies and prey species suggests that hunting success and prey size are broadly related. Success was lowest among the large prey, including bison, moose, and muskoxen, and higher among smaller prey including elk, caribou, and Dall sheep. The negative relationship between prey size and hunting success is most likely due to differences in injury risk among different prey sizes. Rates of success for wolves hunting caribou likely differ due to different encounter definitions that were not explicit for each study. Lack of significant difference in success rate of wolves hunting bison in WBNP and YNP is interesting considering that wolves in YNP have only recently learned to kill bison (Smith et al. 2000), while wolves in WBNP have always killed bison (Carbyn et al. 1993). Risk of Injury and Prey Size Overall, prey encountered by wolves were aggressive and dangerous. Over 50% of prey encounters involved prey that stood and confronted wolves. Prey that confronted wolves were more likely to charge and kick at wolves than prey that fled. As a result, prey that confronted wolves were less likely to be killed (Mech 1966a; Peterson 1977). Prey that confront wolves are subject to intense scrutiny, including brief probing attacks. The extent to which a prey represents a genuine injury risk will determine whether wolves initiate an attack or escalate a probing attack. As a result, prey that 42 confront wolves are probably limited to those relatively few individuals that are sufficiently fit and aggressive to intimidate and deter wolves. Bison, which are generally larger than elk, were more dangerous to wolves than elk. Bison stood and confronted wolves in nearly 80% of encounters, and charged wolves in over 60% of encounters. In contrast, elk stood and confronted wolves in just over 50% of encounters, and charged wolves in slightly more than 25% of encounters. These differences in anti-predator behavior indicate that elk are less risky for wolves to kill than are bison. As a result, elk are more vulnerable to wolf predation than are bison. Observed differences in aggression toward wolves between elk and bison were consistent with previous evidence that prey size and injury risk are positively related (Weaver et al. 1992). In general, small prey are probably less risky than large prey because they are less able to use their size to intimidate and physically threaten wolves. Wolf Behavioral Response to the Risk of Injury During this study, the mortality of 6 wolves was attributed to prey-caused injury. Necropsy indicated that 4 wolves were definitely killed by elk, and 2 were probably killed by elk and moose, respectively (D.W. Smith, National Park Service, unpublished data). These wolf mortalities demonstrate that the risk of injury is significant during wolf encounters with prey in YNP. Comparisons of wolf hunting behavior between encounters with bison and elk provide strong evidence that wolves assess their risk of injury and make foraging decisions based on injury risk. Evidence that wolves modify their behavior to reduce 43 injury risk during bison encounters included (1) longer encounters, (2) infrequent predation attempts, (3) brief predation attempts, (4) longer time between consecutive attempts, and (5) frequent and extended watching behavior following an approach. Longer bison encounters and infrequent predation attempts on bison also reflect apparent trade-offs between food and safety. Although the difference was not statistically significant, the tendency for bison encounters to be longer than elk encounters is consistent with the hypothesis that predators handle dangerous prey more carefully. However, prolonged encounters may also result from the reluctance of bison to flee, thus allowing wolves to loiter near bison for extended periods. Long bison encounters support the hypothesis that wolves accept reduced prey profitability in exchange for safety. However, if wolves kill an adult bison it is possible that large energetic returns compensate for large time expenditures during encounters. Wolf encounters with large prey elsewhere also tend to be prolonged. Encounter duration has lasted more than 36 hours with moose (Mech et al. 1998), 11 hours with bison (Carbyn and Trottier 1988), and 2.5 hours with muskoxen (Gray 1983). In the Northwest Territories, Canada, wolves spent more time in encounters with muskoxen than with Peary caribou, a much smaller prey (Gray 1983). More frequent predation attempts during encounters with elk suggests that wolves prefer attacking elk and avoid attacking bison. Since elk are less dangerous than bison, preference for attacking elk suggests that wolves possess the ability to assess their risk of injury and incorporate this information into their predation-attempt decisions. Wolves likely minimized predation attempts on bison because (1) the probability of injury was 44 too high, and/or (2) long encounter duration (e.g., handling time) made bison energetically unprofitable. Given that more biomass is generally available on a bison than an elk, preference for elk is another indication that wolves trade food for safety. Contrary to an earlier suggestion that prey encounter rates are the most important influence on wolf diet (Huggard 1993), apparent differences in patterns of wolf hunting behavior between bison and elk encounters in YNP suggests that risk of injury is an equally important influence on the species of prey included in the wolf diet. Wolves that pursued or held bison at bay also reduced their risk of injury by minimizing the duration of the predation attempt. By minimizing the time they pursued or held bison at bay, wolves minimized the amount of time they were exposed to a bison counterattack. The tendency for extended periods between consecutive attempts also likely reflected an effort to minimize injury risk. In such cases, wolves may have spent extra time resting and/or waiting for a ‘safe’ opportunity to attack. The tendency for YNP wolves to approach bison within close range and watch, rather than immediately attack, is another indication that wolves use caution during bison encounters. Periods of watching were prolonged, lasting an average of 13 minutes. While the tendency of wolves to watch bison may be the result of fear, it also reflects a lack of fear among bison towards wolves. For instance, targeting usually follows an approach because generally bison do not flee from wolves (Smith et al. 2000), and attacks on individuals located at the periphery of the herd often occur while non-target individuals elsewhere in the herd continue to graze uninterrupted (MacNulty et al. 2002). 45 Sensitivity to the risk of prey-caused injury in YNP wolves was also apparent in the wolves’ selection of capture point on prey. In YNP, wolves tended to grab the largest prey by the hind end, medium-size prey by the hind end and neck, and small prey by the neck. Grabbing the hind end of prey may lower the risk of injury in two ways. First, a prey may have more difficulty delivering lethal blows to wolves it cannot keep in direct sight (Mech 1966a). Second, wolves grabbing the hind end might be better positioned to quickly retreat from an aggressive prey than wolves grabbing the neck. Other studies confirm the association between capture point and prey size. Wolves tend to grab deer almost anywhere (Mech 1970; Mech and Frenzel 1971) and caribou by the front end (Murie 1944; Kelsall 1960; Smith 1980), and larger prey including bison (Carbyn et al. 1993), moose (Burkholder 1959; Mech 1966a; Haber 1977), and muskoxen (Gray 1983; Mech and Adams 1999) by the hind end. Occasionally wolves will grab the nose of large prey (Mech 1966a; Gray 1970; Nelson and Mech 1993), but this often appears to be an effort to distract the prey while others are attacking the hind end (Mech 1966a). Although many predators appear to make foraging decisions based on their risk of injury from prey (Forbes 1989), predators are usually considered "fierce" (Brown et al. 1999) because they elicit a fear response from prey that involves trade-offs between food and safety (Krebs 1980; Newman and Caraco 1987). Foragers must trade between food and safety to survive and reproduce, and the fitness consequences related to this trade-off result in the evolution of adaptive foraging strategies. Most foraging studies focus on the 46 behavior of a timid prey pursued by a fierce predator and assume that behaviorally unresponsive predators hunt prey with no risk of prey-caused injury (Sih 1980; Caraco 1981; Dill 1987). Comparisons of wolf hunting behavior between bison and elk encounters demonstrate that predators hunting dangerous prey respond to the risk of injury and make trade-offs between food and safety. In this case, safety represents the critical need to avoid being killed or injured by a belligerent prey. However, since predators cannot simultaneously maximize food intake and minimize injury risk, conflict arises in deciding whether to attack or avoid dangerous prey. Wolves appear to have resolved this conflict by evolving a preference for vulnerable prey. As a result, wolves are able to acquire food while avoiding or minimizing the risk of injury (Mech 1970). Since the fitness consequences of prey-caused injury are severe, the risk of injury is likely a strong selective force over evolutionary time on the foraging behavior of predators that rely on dangerous prey for food. Further insight to foraging behavior is possible by moving beyond the traditional view of a timid prey pursued by a fierce predator (Lima and Dill 1990; Brown et al. 1999) and considering more behaviorally sophisticated systems where prey are not invariably timid, and predators are not consistently fierce. 47 LITERATURE CITED Alexander, R.M., A.S. Jayes, and R.F. Ker. 1980. Estimates of energy cost of quadrupedal running gaits. Journal of Zoology 190:155-192. Ballard, W. B., J. S. Whitman, and C. L. Gardner. 1987. Ecology of an exploited wolf population in south-central Alaska. Wildlife Monograph, Number 98. Banfield, A.W.F. 1954. Preliminary investigation of the barren-ground caribou. Canadian Wildlife Service, Wildlife Management Bulletin Series 1, Number 10B. Bangs, E.E., and S.H. Fritts. 1996. Reintroducing the gray wolf to central Idaho and Yellowstone National Park. Wildlife Society Bulletin 24:402-413. Brown, J.S., J.W. Laundre, and M. Gurung. 1999. The ecology of fear: optimal foraging, game theory, and trophic interactions. Journal of Mammalogy 80:385-399. Burkholder, B. L. 1959. Movements and behavior of a wolf pack in Alaska. Journal of Wildlife Management 23:1-11. Caraco, T. 1981. Energy budgets, risk and foraging preferences in dark-eyed juncos (Junco hymelais). Behavioral Ecology and Sociobiology 8:820-830. Carbyn, L. N. 1974. Wolf predation and behavioral interactions with elk and other ungulates in an area of high prey diversity. Ph.D. Thesis, University of Toronto, Ontario Carbyn, L.N. 1983. Wolf predation on elk in Riding Mountain National Park, Manitoba. Journal of Wildlife Management 47:963-976. Carbyn, L.N., and T.Trottier. 1987. Responses of bison on their calving grounds to predation by wolves in Wood Buffalo National Park. Canadian Journal of Zoology 65:2072-2078. Carbyn, L.N., and T. Trottier. 1988. Descriptions of wolf attacks on bison calves in Wood Buffalo National Park, Canada. Arctic 41:297-302. Carbyn, L.N., S.M. Oosenbrug, and D.W. Anions. 1993. Wolves, bison and the dynamics related to the Peace-Athabasca Delta in Canada's Wood Buffalo National Park. Art Design Printing Inc., Edmonton, Alberta. Clark, K.R.F. 1971. Food habits and behavior of the tundra wolf on central Baffin island. Ph.D. Thesis, University of Toronto, Ontario 48 Creel, S., and N.M. Creel. 1995. Communal hunting and pack size in African wild dogs, Lycaon pictus. Animal Behaviour 50:1325-1339. Creel, S. 2001. Cooperative hunting and sociality in African wild dogs, Lycaon pictus. Pp. 466-490 in Model systems in behavioral ecology: integrating conceptual, theoretical, and empirical approaches (L. A. Dugatkin, ed.) Princeton University Press, Princeton, New Jersey. Crisler, L. 1956. Observations of wolves hunting caribou. Journal of Mammalogy 37:337-346. Dill, L.M. 1987. Animal decision making and its ecological consequences: the future of aquatic ecology and behavior. Canadian Journal of Zoology 65:803-811. Estes, R. D., and J. Goddard. 1967. Prey selection and hunting behavior of the African wild dog. Journal of Wildlife Management 31:52-70. Fitzgibbon, C.D., and J.H. Fanshawe. 1988. Stotting in Thomson's gazelle: an honest signal of condition. Behavioral Ecology and Sociobiology 23:69-74. Forbes, L. S. 1989. Prey defenses and predator handling behaviour: the dangerous prey hypothesis. Oikos 55:155-158. Fuller, W.A. 1957. The biology and management of the bison of Wood Buffalo National Park. Ph.D. Thesis, University of Wisconsin, Madison. Fuller, W.A. 1960. Behavior and social organization of the wild bison of Wood Buffalo National Park, Canada. Arctic 13:3-19. Fuller, W.A. 1962. The biology and management of the bison of Wood Buffalo National Park. Canadian Wildlife Service, Wildlife Management Bulletin Series 1, Number 16. Gray, D.R. 1970. The killing of a bull muskox by a single wolf. Arctic 23:197-199. Gray, D.R. 1983. Interactions between wolves and muskoxen on Bathurst Island, Northwest Territories, Canada. Acta Zoologica Fennica 174:255-257. Gray, D.R. 1987. The muskoxen of Polar Bear Pass. National Museum of Natural Sciences. Markham, Ontario Haber, G.C. 1968. The social structure and behavior of an Alaskan wolf population. M.A. Thesis, Northern Michigan University, Marquette. 49 Haber, G.C. 1977. Socio-ecological dynamics of wolves and prey in a subarctic ecosystem. Ph.D. Thesis, University of British Columbia, Vancouver. Hasson, O. 1991. Pursuit deterrent signals: the communication between prey and predator. Trends in Ecology and Evolution 6:325-329. Hebblewhite, M. 2000. Wolf and elk predator-prey dynamics in Banff National Park. M.S. Thesis, University of Montana, Missoula. Heglund N. C., and C. R. Taylor. 1988. Speed stride frequency and energy cost per stride how do they change with body size and gait? Journal of Experimental Biology 138:301-318. Holekamp K.E., L. Smale, and S.M. Cooper. 1997. Hunting rates and hunting success in the spotted hyena (Crocuta crocuta). Journal of Zoology 242:1-15. Houston, D.B. 1982. The northern Yellowstone elk: ecology and management. MacMillan, New York, New York. Hornocker, M.G. 1970. An analysis of mountain lion predation upon mule deer and elk in the Idaho Primitive Area. Wildlife Monograph, Number 21. Huggard, D.J. 1993. Prey selectivity of wolves in Banff National Park. I. Prey species. Canadian Journal of Zoology 71:130-139. Kelsall, J.P. 1957. Continued barren-ground caribou studies. Canadian Wildlife Service, Wildlife Management Bulletin Series 1, Number 12. Kelsall, J.P. 1960. Co-operative studies of barren-ground caribou 1957-58. Canadian Wildlife Service, Wildlife Management Bulletin Series 1, Number 15. Kelsall, J.P. 1968. The migratory barren ground caribou of Canada. Canadian Wildlife Service, Queen’s Printer, Ottawa. Kolensky, G.B. 1972. Wolf predation on wintering deer in east-central Ontario. Journal of Wildlife Management 44:583-602. Krebs, J.R. 1980. Optimal foraging, predation risk, and territory defense. Ardea, 68:8390. Kruuk, H. 1972. The spotted hyena. University of Chicago Press, Chicago, Illinois. Lima, S.L., and L. M. Dill. 1990. Behavioral decisions made under the risk of predation: a review and prospectus. Canadian Journal of Zoology 68:619-640. 50 Lingle, S., and S.M. Pellis. 2002. Fight or flight? Antipredator behavior and the escalation of coyote encounters with deer. Oecologia 131:154-164. MacNulty, D.R., N. Varley, and D.W. Smith. 2002. Grizzly bear, Ursus arctos, usurps bison calf, Bison bison, captured by wolves, Canis lupus, in Yellowstone National Park, Wyoming. Canadian Field-Naturalist 115:495-498. Martin P., and P. Bateson. 1993. Measuring behavior: an introductory guide. Cambridge University Press, Cambridge, United Kingdom. Meagher, M. 1973. The bison of Yellowstone National Park. Scientific Monograph No. 1 Department of the Interior, National Park Service. Mech, L.D. 1966a. The wolves of Isle Royale. U. S. National Park Service., Washington, D. C. Mech, L.D. 1966b. Hunting behavior of timber wolves in Minnesota. Journal of Mammalogy 47:347-348. Mech, L.D. 1970. The wolf: the ecology and behavior of an endangered species. Doubleday/Natural History Press, Garden City, New York. Mech, L.D. 1977. Population trend and winter deer consumption in a Minnesota wolf pack. Pages 55-83, in R.L. Phillips and C. Jonkel, eds., Proceedings of the 1975 Predator Symposium. Mech, L.D. 1984. Predators and predation. Pp. 189-200 in White-tailed deer: ecology and management (L. K. Halls, ed.). Stackpole Books, Harrisburg, Pennsylvania. Mech, L.D. 1988. The arctic wolf: Living with the pack. Voyageur Press. Stillwater, Minnesota. Mech, L. D., and M. Frenzel. 1971. Ecological studies of the timber wolf in northeastern Minnesota. North Central Forest Experiment Station, United States Department of Agriculture, Forest Service Research Paper NC-52., St. Paul, Minnesota. Mech, L.D., and M.E. Nelson. 1990. Evidence of prey-caused mortality in three wolves. American Midland Naturalist 123:207-208. Mech, L. D., L. G. Adams, T. J. Meier, J. W. Burch, and B. W. Dale. 1998. The wolves of Denali. University of Minnesota Press, Minneapolis, Minnesota. 51 Mech, L.D., and S.B. Merrill. 1998. Daily departure and return patterns of wolves, Canis lupus, from a den at 80 degrees N latitude. Canadian Field-Naturalist 112:515-517. Mech, L.D., and L.G. Adams. 1999. Killing of muskox, Ovibos moschatus, by two Wolves, Canis lupus, and subsequent caching. Canadian Field-Naturalist 113:673675. Merrill, S. 2002. An evaluation of the use of global position system telemetry in studying wolf biology. Ph.D. Thesis, University of Minnesota, St. Paul. Miller, F.L., and A. Gunn. 1977. Group of muskoxen attacked by a solitary arctic wolf, Prince of Wales Island, Northwest Territories. Musk-Ox 20:87-88. Miller, F. L., A. Gunn, and E. Broughton. 1985. Surplus killing as exemplified by wolf predation on newborn caribou. Canadian Journal of Zoology 63:295-300. Murie, A. 1944. The wolves of Mount McKinley. United States Government Printing Office, Washington, D. C. Nelson, M.E., and L.D. Mech. 1993. Prey escaping wolves, Canis lupus, despite close proximity. Canadian Field-Naturalist 107:245-246. Nelson, M.E., and L.D. Mech. 1994. A single deer stands-off three wolves. American Midland Naturalist 131:207-208. Newman J.A., and T. Caraco. 1987. Foraging, predation hazard and patch use in gray squirrels. Animal Behaviour, 35:1804-1813. Nowak, R.M. 1999. Walker's mammals of the world. 6th edition. John Hopkins University Press, Baltimore, Maryland. Pasitchniak-Arts, M., M. E. Taylor, and L. D. Mech. 1988. Skeletal injuries in an adult arctic wolf. Arctic and Alpine Research 20:360-365. Peterson, R.O. 1977. Wolf ecology and prey relationships on Isle Royale. National Park Service Scientific Monograph Series, Number 11. Phillips, M.K. 1984. The cost to wolves of preying on ungulates. Australian Mammalogy 8:99. Phillips, M.K., and D.W. Smith. 1996. The wolves of Yellowstone. Voyageur Press, Stillwater, Minnesota. 52 Rausch, R. A. 1967. Some aspects of the population ecology of wolves, Alaska. American Zoologist 7:253-265. SAS Institute Inc., 2000. SAS 8.0. Cary, North Carolina. Schaller, G.B. 1967. The deer and the tiger. The University of Chicago Press, Chicago, Illinois. Schaller, G. B. 1972. The Serengeti Lion: A study of predator-prey relations. The University of Chicago Press, Chicago, Illinois. Sih, A.S. 1980. Optimal behavior: can foragers balance two conflicting demands? Science 280:1041-1043. Smith, T. G. 1980. Hunting, kill and utilization of a caribou by a single gray wolf. Canadian Field-Naturalist 94:175-177. Smith, D.W., L.D. Mech, M. Meagher, W.E. Clark, R. Jaffe, M.K. Phillips, and J.A. Mack. 2000. Wolf-bison interactions in Yellowstone National Park. Journal of Mammalogy 81:1128-1135. Stephens, D.W., and J.R. Krebs. 1986. Foraging theory. Princeton University Press, Princeton, New Jersey. Temple, S.A. 1987. Do predators always catch substandard individuals disproportionately from prey populations? Ecology, 68:669-674. Tener, J. S. 1954. A preliminary study of the musk-oxen of Fosheim Peninsula, Ellesmere Island, NWT. Canadian Wildlife Service, Wildlife Management Bulletin Series 1, No. 9. Theuerkauf, J., W. Jedrzejewski, K. Schmidt, H. Okarma, I. Ruczynski, S. Sniezko, and R. Gula. In press. Daily patterns and duration of wolf activity in the Bialowieza forest, Poland. Journal of Wildlife Management. Verbeke, G., and G., Molenberghs. 2000. Linear mixed models for longitudinal data. Springer-Verlag. New York, New York. Weaver, J. L., C. Arvidson, and P. Wood. 1992. Two wolves, Canis lupus, killed by a moose, Alces alces, in Jasper National Park, Alberta. Canadian Field-Naturalist 106:126-127. 53 Table 1. Number of hunting bouts, prey encounters, and predation attempts observed in their entirety and partially observed (in parentheses) for various wolf packs in Yellowstone National Park, May 1995 – March 2000. Pack No. wolves No. hunting bouts No. prey encounters No. predation attempts Rose Creek 10-22 23 (146) 90 (175) 160 (217) Leopold 2-13 17 (125) 68 (156) 113 (165) Druid Peak 5-9 18 (135) 75 (172) 154 (193) Crystal Creek 2-16 4 (34) 26 (57) 125 (139) Chief Joseph 2-11 0 (7) 1 (7) 3 (6) Nez Perce 2-13 0 (4) 0 (4) 0 (3) Soda Butte 3-8 0 (2) 1 (2) 1 (2) 0 (9) 6 (11) 9 (11) 62 (400) 267 (584) 565 (736) Lone wolves Total 54 Table 2. Results of General Linear Mixed Model (GLMM) evaluating the effects of hunting-state type and prey species on hunting-state duration (min.) in wolf encounters with elk and bison herds in Yellowstone National Park, May 1995 – March 2000. Predicted mean hunting-state duration with 95% confidence intervals is shown in Figure 6. Parameter Regression coefficient1 S.E. d.f. Approximate t P -0.04 0.19 220 -0.24 0.8126 Approach 0.61 0.14 307 4.51 < 0.0001 Attack 0.63 0.13 307 4.77 < 0.0001 Capture 0.93 0.20 307 4.64 < 0.0001 0.04 0.17 307 0.22 0.8295 Constant Hunting state 2 Prey species Elk 1 A coefficient is interpreted as the typical difference in mean hunting-state duration relative to the reference group for each covariate (i.e. target for hunting state, and bison for prey species) when all other covariates are held constant. For example, the coefficient for the approach state indicates that the mean hunting-state duration was on average 0.61 minutes more (because the coefficient is positive) during approach states than during target states. 2 The watch state was excluded from this analysis because it only occurred during wolf encounters with bison. 55 Table 3. Success rates for wolves hunting various prey species in Yellowstone National Park, May 1995 - March 2000, based on known outcomes from completely observed prey encounters only, and on both complete and incompletely observed encounters (in parentheses) 1. Prey species No. encounters No. attempts No. kills Kills per encounter Kills per attempt 211 (463) 246 (543) 50 (104) 0.24 (0.23) 0.20 (0.19) Bison 47 (74) 91 (154) 2 (5) 0.04 (0.07) 0.02 (0.03) Pronghorn 6 (10) 4 (7) 2 (2) 0.33 (0.20) 0.50 (0.29) Bighorn 2 (6) 1 (4) 0 (0) 0 (0) 0 (0) Mule Deer 1 (3) 1 (3) 1 (2) 1.00 (0.67) 1.00 (0.67) Moose 0 (2) 0 (2) 0 (0) 0 (0) 0 (0) Total 267 (558) 343 (713) 55 (113) 0.21 (0.21) 0.16 (0.16) Elk Weighted mean2 1 2 Encounters include both solitary and herd prey. Mean weighted using number of prey encounters or predation attempts for each species. 56 Table 4. Age1 and sex of prey killed by wolves in Yellowstone National Park, May 1995 - March 2000. The proportion killed in each age/sex class for each prey species is shown in parentheses. Prey species Adult male Adult female Yearling Young-ofthe-year Unknown N 20 (0.19) 50 (0.48) 2 (0.02) 26 (0.25) 6 (0.06) 104 Bison2 0 1 (0.20) 0 4 (0.80) 0 5 Pronghorn3 0 0 0 2 (0.67) 1 (0.33) 3 Mule deer 0 1 (0.50) 0 1 (0.50) 0 2 20 (0.17) 52 (0.46) 2 (0.02) 33 (0.29) 7 (0.06) 114 Elk Total 1 Age was classed as adult (>23 months), yearling (12-23 months), or young-of-the-year (<12 months). Kills of the adult female and two young-of-the-year bison were reported earlier (Smith et al. 2001). 3 Includes 1 fawn killed in an encounter that was partially observed. 2 57 Table 5. Results of General Linear Mixed Model (GLMM) evaluating the effects of hunting-state outcome, hunting-state type, and prey species on hunting-state duration (min.) in wolf encounters with elk and bison herds in Yellowstone National Park, May 1995 – March 2000. Predicted mean hunting-state duration with 95% confidence intervals is shown in Figure 9. Parameter Regression coefficient1 S.E. d.f. Approximate t P 1.93 0.48 217 4.00 < 0.0001 Approach -1.31 0.49 275 -2.70 0.0074 Attack -1.37 0.50 275 -2.75 0.0063 Target -2.27 0.51 275 -4.45 < 0.0001 -1.08 0.49 275 -2.18 0.0299 -0.97 0.46 275 -2.10 0.0371 Approach × Elk 1.08 0.50 275 2.16 0.0314 Attack × Elk 1.08 0.49 275 2.19 0.0294 Target × Elk 1.23 0.50 275 2.44 0.0155 Constant Hunting state 2 Prey species Elk Hunting-state outcome Failure Hunting state × Prey species Hunting state × Hunting-state outcome 1 Approach × Failure 1.02 0.50 275 2.05 0.0408 Attack × Failure 1.03 0.50 275 2.08 0.0382 Target × Failure 1.27 0.51 275 2.46 0.0145 A coefficient is interpreted as the typical difference in mean hunting-state duration relative to the reference group for each covariate (i.e. capture for hunting state, bison for prey species, and success for outcome) when all other covariates are held constant. 2 The watch state was excluded from this analysis because it only occurred during wolf encounters with bison. 58 Table 6. Results of General Linear Mixed Model (GLMM) evaluating the effects of hunting-state type on hunting-state duration (min.) in wolf encounters with bison herds in Yellowstone National Park, May 1995 – March 2000. Predicted mean hunting-state duration with 95% confidence intervals is shown in Figure 11. Parameter Regression coefficient1 S.E. d.f. Approximate t P 1.86 0.22 31 8.49 < 0.0001 Approach -1.31 0.21 183 -6.31 < 0.0001 Attack -1.19 0.21 183 -5.75 < 0.0001 Target -1.85 0.23 183 -7.89 < 0.0001 Capture -0.53 0.38 183 -1.39 0.1672 Constant Hunting state 1 A coefficient is interpreted as the typical difference in mean hunting-state duration relative to the reference group (i.e. watch). For example, the coefficient for the approach state indicates that the mean hunting-state duration was on average 1.31 minutes less (because the coefficient is negative) during approach states than during watch states. 59 Table 7. Reported success rates for wolves hunting various North American prey species. Prey species No. encounters No. attempts No. animals No. kills Kills per encounter Kills per attempt Kills per animal Moose 71 66 77 6 0.08 0.09 0.08 Mech 1966a 49 1 0.02 Peterson 1977 53 7 0.13 Mech et al. 1998 389 23 0.06 Haber 1977 Moose Moose 37 Moose 0.19 Reference Bison 31 46 3 0.10 0.07 Carbyn et al. 1993 Bison 74 154 5 0.07 0.03 This study Muskoxen 21 3 0.141 Caribou 34 2 0.06 Caribou 16 9 0.56 Caribou 26 4 0.15 24 0.24 Dall Sheep Whitetailed Deer Dall Sheep Elk 1 44 303 100 60 18 463 12 186 543 0.05 Clark 1971 Haber 1977 0.01 0.33 104 0.23 Mech et al. 1998 Haber 1977 Nelson and Mech 1993 0.20 6 Rate calculated from data reported in Grey (1983). Grey 1983 0.03 0.20 Mech et al. 1998 This study 60 Figure 1. The predatory sequence for wolves hunting herds of prey. Hunting Bout Prey Encounter Predation Attempt Hunting State 1 Travel Approach Watch Attack The target state would not be included in wolf encounters with solitary prey. Target1 Capture 61 Figure 2. Study area and general location of study wolf packs, Yellowstone National Park, 19951-2000. Chief Joseph Rose Creek Leopold Druid Peak Crystal Creek Nez Perce Soda Butte Wolf Pack Territory Park Boundary N Roads MT ID 1 2 WY 0 10 20 KM Only Rose Creek, Crystal Creek, and Soda Butte packs were present in 1995. Figures do not depict actual size or shape of wolf pack territories. Northern Range Lakes 2 62 Figure 3. Time of year wolf hunting bouts were observed in Yellowstone National Park, May 1995 - March 2000. 160 No. hunting bouts observed a 140 120 100 80 60 40 20 0 Jan. Feb. Mar. Apr. May June July Aug. Sept. Oct. Nov. Dec. Month 63 Figure 4. Time of day wolf hunting bouts were observed during intensive winter study periods1 (mid-November to mid-December & March) in Yellowstone National Park, 1995 - 2000. 40 No. hunting bouts observed a 35 30 25 20 15 10 5 00 19 00 18 00 17 00 16 00 15 00 14 00 13 00 12 00 11 00 0 10 90 0 80 0 70 0 60 50 0 0 Time of day (hrs.) 1 Daily monitoring was continuous during winter study and observation effort per hour was generally constant. 64 No. prey encounters observed a Figure 5. Number of prey present during wolf encounters with various prey species in Yellowstone National Park, May 1995 - March 2000. 50 Elk 45 Bison Mule deer 40 Pronghorn 35 Moose 30 Bighorn sheep 25 20 15 10 5 0 1 2-5 6-10 11-20 21-50 No. prey present 51-100 >100 65 Figure 6. Predicted mean duration (min.) of hunting states with 95% confidence intervals in wolf encounters with elk (S) and bison (z) herds in Yellowstone National Park, May 1995 - March 2000. Fitted means and confidence intervals are derived from GLMM results (Table 2). The number of hunting states is shown above each confidence interval. 4.5 10 4.0 29 3.5 Time (min.) 3.0 2.5 59 59 136 145 2.0 1.5 57 37 1.0 0.5 0.0 0.00 Approach 1.00 Attack 2.00 Hunting state Target 3.00 Capture 4.00 66 Figure 7. Results of completely observed wolf encounters with (a) elk herds and (b) solitary elk in Yellowstone National Park, May 1995 - March 2000. A single type of state can occur multiple times during an encounter. Therefore, percents indicate the proportion of the total number of hunting states (in parentheses) observed during an encounter that resulted in a specific outcome (indicated by arrow). Percents in bold indicate the first hunting state to occur in the encounter. (a) 179 encounters with elk herds 85% 14% 1% Approach (177) 63% Attack (180) 42% 8% Target (108) 43% 1%1 Capture (53) 85% 1 Kill (45) Newborn elk calves (b) 32 encounters with solitary elk 75% 25% Approach (25) 76% Attack (31) 16% Capture (5) 100% Kill (5) 67 Figure 8. Results of completely observed wolf encounters with (a) bison herds and (b) solitary bison in Yellowstone National Park, May 1995 - March 2000. A single type of state can occur multiple times during an encounter. Therefore, percents indicate the proportion of the total number of hunting states (in parentheses) observed during an encounter that resulted in a specific outcome (indicated by arrow). Percents in bold indicate the first hunting state to occur in the encounter. (a) 36 encounters with bison herds 97% 3% Approach (64) 17% 43% Watch (42) 19% 47% Attack (53) 17% 19% Target (35) 17% 23% Capture (10) 20% Kill (2) (b) 11 encounters with solitary bison 91% 9% Approach (11) 45% Attack (10) 0% Capture (0) 68 Figure 9. Predicted mean duration (min.) of hunting states with 95% confidence intervals in failed (S) and successful (z) wolf encounters with (a) elk and (b) bison herds in Yellowstone National Park, May 1995 - March 2000. Fitted means and confidence intervals are derived from GLMM results (Table 5). The number of hunting states is shown above each confidence interval. 4.0 (a) elk 26 3.5 Time (min.) 3.0 34 93 52 90 2.5 3 2.0 37 1.5 20 1.0 0.5 0.0 0 Approach 1 Attack 2 Target 3 Capture 4 Hunting state 20 (b) bison 18 3 16 Time (min.) 14 12 10 8 7 6 18 25 4 53 6 31 6 2 0 0 Approach 1 Attack 2 Target Hunting state 3 Capture 4 69 Figure 10. The association between mean wolf hunting success (kills/encounter)1 and season (early winter: Nov 1 - Dec 31, mid-winter: Jan 1 - Feb 28, late winter: Mar 1 - Apr 30, spring: May 1 - Jun 30) in wolf encounters with elk (S) and bison2 (z) in Yellowstone National Park, May 1995 - March 2000. The number of encounters is shown above each confidence interval. 0.5 Kills/encounter 0.4 84 0.3 209 105 51 0.2 46 0.1 12 14 Spring Season 1 Includes known outcomes from both complete and incompletely observed prey encounters. No bison encounters were observed during mid-winter. 2 4 Late Winter 3 Mid-winter 2 Early Winter 1 0 0 70 Figure 11. Predicted mean duration (min.) of hunting states with 95% confidence intervals in wolf encounters with bison herds in Yellowstone National Park, May 1995 March 2000. Fitted means and confidence intervals are derived from GLMM results (Table 6). The number of hunting states is shown above each confidence interval. 11 54 10 9 10 Time (min.) s 8 7 6 5 4 3 59 59 2 37 1 0 0 Approach 1 Watch 2 Attack Hunting state 3 Target 4 Capture 5 71 Figure 12. The association between hunting success (kills/encounter)1 and prey size2 for wolves hunting various North American prey (Spearman rank correlation coefficient, rs = -0.60, N = 11, P < 0.05). 0.70 Dall Sheep Caribou 0.60 Elk Muskoxen Kills/encounter a 0.50 Moose Bison 0.40 0.30 0.20 0.10 0.00 0 100 200 300 400 500 Mean prey weight (kg) 1 Hunting success data are from Table 7. Mean prey weights are estimated from Nowak (1999). 2 600 700
© Copyright 2026 Paperzz