9 Foraging in the Face of Danger Peter A. Bednekoff 9.1 Prologue A juvenile coho salmon holds its position in the flow of a brook. To conserve energy, it positions itself in the lee of a small rock. Distinc- tive blotches of color on its sides, called parr marks, provide effective camouflage. As long as it holds its position, it is virtually impossible to see. The simple strategy of keeping still hides it from the prying eyes of potential salmon-eaters. Kingfishers and herons threaten from above, and cutthroat trout, permanent residents of the stream, seldom reject a meal of young salmon. The threat posed by these and other predators is ever present. The clear water flowing past the salmon presents a stream of food items: midges struggle on the surface; mayfly nymphs drift in the cur- rent. But, here’s the rub: to capture a prey item, the salmon must dash out from its station, potentially telegraphing its position to unwelcome observers. When the salmon feels safe, it will travel quite a distance to intercept a food item, making a leisurely excursion to collect a drifting midge as far as a meter away from its location. Detecting a predator changes the salmon’s behavior. Depending on the level of the perceived threat, the salmon has several options. It may flush to deep water or another safe location. It may stop feeding alto- gether, but hold its position. It may continue feeding, but dramatically 306 Peter A. Bednekoff Figure 9.1. Patch residence time increases with travel time between patches (as predicted), but blue jays stay in patches much longer than the optimal residence time. Solid squares show observed residence times; open squares show the predicted optimal residence times. (After Kamil et al. 1993.) reduce the distance it will travel to intercept food. This series of graded re- sponses represents a sophisticated and often effective strategy to avoid preda- tors. Sophisticated or not, all of these responses reduce the salmon’s feeding efficiency. The salmon’s problem is far from unique; virtually all animals face a trade-off between acquiring resources and becoming a resource for another. 9.2 Overview and Road Map Resource acquisition is necessary for fitness, but it is not sufficient. Food is generally good for the forager, but not if the forager is dead. Danger affects animal decisions in many ways (see reviews in Lima and Dill 1990; Lima 1998). Animals often face a trade-off between food acquisition and danger: the alternative that yields the highest rates of food intake is also the most dangerous. A growing area of research focuses on this fundamental trade-off. This chapter examines how danger from predators affects foraging behavior. Early theory assumed that fitness was highest when the net rate of foraging gain (i.e., net amount of energy acquired per unit time) was highest. Early empirical tests consistently showed that foragers are sensitive to foraging gain (see Stephens and Krebs 1986). As predicted, many animals spend more time feeding in each patch when patches are farther apart (Stephens and Krebs 1986; Nonacs 2001). Animals often stay in patches longer, however, than the time that would maximize their overall rate of energy gain (Kamil et al. 1993; Nonacs 2001; fig. 9.1). Tests suggested that early rate-maximizing models were partly right: foragers are sensitive to their rate of energy gain, but often do not fully maximize it (see also Nonacs 2001). This observation Foraging in the Face of Danger 307 Distance from cover (m) % carried to cover 0 20 40 0 5 10 60 80 100 15 20 after hawk scare before hawk scare Figure 9.2. Black-capped chickadees are more likely to carry small food items to cover before eating them when cover is close and after a simulated hawk attack. (After Lima 1985a.) suggested that some non-energetic costs must be important. By pointing out the importance of such costs, early tests of rate-maximizing models provided the springboard for the study of foraging and danger. Black-capped chickadees often carry food items from a feeder to a bush before consuming them. They are more likely to carry larger items and are more likely to carry items if the feeder is closer to a bush (Lima 1985a; fig. 9.2). Carrying an item to a bush decreases a chickadee’s rate of intake, but intake rate is decreased less with large items and close distances. Steve Lima hypothesized that chickadees carried food to cover in order to reduce their exposure to predators. He tested this hypothesis by flying a hawk model in the area during some trials. After having seen the hawk model, chickadees were more likely to carry food to safety (Lima 1985a; fig. 9.2). Thus, animals are willing to reduce their intake rate in order to reduce danger. To begin this chapter, I examine why foraging gain and danger are gen- erally linked, and I build a life history framework for modeling foraging and danger. I discuss how danger may change with the internal state of the ani- mal, time, and group size. These topics lead to inquiries on how animals assess danger and whether they should overestimate danger. I close with my view of the prospects of the field. Within each section, I outline some principles, often with the help of simple models, and illustrate those principles with a sampling of examples. 9.3 Why Does Increased Foraging Lead to Greater Danger? Animals often face alternatives that differ in both foraging gain and danger. Obviously, foragers should avoid options that combine poor feeding with 308 Peter A. Bednekoff great danger and choose options that offer good feeding with little danger. Most often, however, animals face difficult choices in which the options for better feeding also entail greater danger. Such difficult choices are ubiquitous for several reasons, and wherever one or more of these reasons applies, organ- isms face a trade-off between feeding and danger. After sketching out various routes to a trade-off, I return to a general conceptual approach because the many routes to a trade-off converge on the same basic consequences. Time Spent Exposed Guppies feed day and night when no predators are around, but only during the day if predators are around (Fraser et al. 2004). In response to indications of danger, many animals restrict their feeding time (Lima 1998, see especially table II). An animal that feeds part of the time can restrict its feeding to the safest period, but it must extend its feeding time into more dangerous periods in order to feed for longer. For example, small birds must extend their feeding time into the twilight periods around dusk and dawn, when they are less able to detect attacks in the low light and deep shadows (see Lima 1988a, 1988c; Krams 2000). For bats that feed on insects, darkness is safer, but emerging before nightfall may allow greater feeding ( Jones and Rydell 1994). Feeding at nightisalso saferfor minnows (Greenwoodand Metcalfe 1998)and juvenile salmon (Metcalfe et al. 1999). In order to increase feeding, however, these fish have to feed during the more dangerous daylight period. Habitat Choice While actively foraging, animals often choose between habitats that differ in danger and productivity. For example, aquatic snails feeding on algae face a trade-off because more algae grows on the sunny side of rocks, but the tops of rocksarealso moreexposed tofish predators (Levri1998). Thebasic ecology of exposure leads to the trade-off: exposure to sunlight allows more photo- synthesis, but exposure often leaves foragers more vulnerable to predators. Similarly, sunfish can find more zooplankton to eat in the open-water por- tions oflakes because theseareas produce morephytoplankton, which support the zooplankton. The open areas, however, provide no refuge from attack, whereas the weedy littoral zone provides refuge, but less food (Werner and Hall 1988). Animals switch between these two kinds of areas during growth because both foraging gain and danger change as they grow (Werner and Gil- liam 1984). In other cases, the attack strategy of the predator and the escape strategy of the prey combine to create the trade-off. In boreal forests, the swooping Foraging in the Face of Danger 309 attacks of pygmy owls make the outer, lower branches of trees particularly dangerous (Kullberg 1995), and small birds avoid these branches unless com- petition or hunger forces them there (e.g., Krams 1996; Kullberg 1998b). Within a foraging group, individuals on the leading edge will first encounter new sources of both food and danger (Bumann et al. 1997). Animals often move to edge positions when hungry (Romey 1995) and to central positions when alarmed (Krause 1993). Habitat choice may involve another layer of compromise when foragers face conflicting pressures from different kinds of predators. For example, grasshoppers can reduce bird predation by staying low on a blade of grass, but they can minimize predation by lizards and small mammals by positioning themselves high on grass stems. When both kinds of predators are around, grasshoppers choose intermediate positions (Pitt 1999). As theseexamples emphasize,animals choosebetween habitatson small aswell as large spatial scales, and both kinds of choices have ecological consequences. Movement Creatures great and small move less when predators are around (Lima 1998, table II). A forager actively searching for food can cover a greater area by movingfaster. Bycovering a greaterarea, itislikely toencountermore feeding opportunities, and may also encounter more predators (Werner and Anholt 1993). Besides simply crossing paths with more predators, moving foragers increase the likelihood of an attack. Anaesthetized tadpoles are less likely to be killed by aquatic invertebrate predators (Skelly 1994), and tadpoles generally move less when danger is greater (Anholt et al. 2000). When movement is in short bursts, as in degus, greater movement may involve both faster speeds while moving and shorter pauses between bursts (Vasquez et al. 2002). I will consider movement in further detail below after developing a general model of foraging in the face of danger. Detection Behavior Most animals perform behaviors that increase their chances of detecting and escaping from predators. The best studied of these behaviors are pauses dur- ing foraging to scan the environment for potential danger (see Bednekoff and Lima 1998a; Treves 2000). Animals can raise their rate of food consumption by scanning less frequently, but at the cost of detecting attacks less effectively (e.g., Wahungu et al. 2001). Investigators have often operationally defined vigilance as raising the head above horizontal. While this operational defini- tion works well for birds and mammals, animals with different body forms and lifestyles may require other operational definitions. For example, lizards 310 Peter A. Bednekoff basking with their eyes shut and one or more limbs raised off the substratum seem to be showing little antipredator behavior (Downes and Hoefer 2004). Overall, the varied postures and attention required for foraging probably affect predator detection in many organisms. For example, guppies react less quickly to predators when foraging than when not foraging, and even less quickly when foraging nose down (Krause and Godin 1996). Depletion and Density Dependence For a burrowing animal such as a marmot, safety comes from fleeing back to the burrow (Holmes 1984; Blumstein 1998). Marmots feed near their bur- rows, and so deplete food in the area (Del Moral 1984). Due to this depletion, a marmot can feed at a higher rate, but at greater danger, by venturing farther from the burrow. Thus, reactions to initial differences in danger produce dif- ferences in foraging. Many lizards also feed from a safe central place (Cooper 2000). Such lizards can find more prey farther out, but at a cost. The actions of a lizard also produce a gradient of food and danger for its potential prey. The grasshoppers the lizard preys on can find a richer, less depleted food supply near the lizard’s perch, but obviously, feeding near the lizard increases the possibility of attack (Chase 1998). Thus, a spatial trade-off at one trophic level may have cascading effects on other trophic levels. In a manner similar to food depletion, density dependence can produce a trade-off whenpotential preycongregate. By congregating,prey decreaseone another’s feeding rates through competition and also decrease one another’s danger through safety-in-numbers advantages. When avoiding predatory perch, 92% of small crucian carp concentrate in the safer shallows, compoun- ding the differences in food availability between shallows and open waters (Paszkowski et al. 1996). In theory, the outcome depends on the balance of competitive and safety-in-numbers effects and on how free predators are to choose habitats, but we may often expect habitats to be made either safe but poor orrich butdangerous bythese mechanisms(Hugie and Dill 1994; Moody et al. 1996; Sih 1998). 9.4 Modeling Foraging under Danger of Predation Foraging for a Fixed Time Bluehead chubs alter their foraging in response to changes in energetic returns and danger from green sunfish. The best explanation for their behavior com- bines food and danger in a life history context (Skalski and Gilliam 2002). To build models of foraging under danger of predation, we start from the first Foraging in the Face of Danger 311 principle of foraging theory—that food is good. We assume that higher for- aging success leads to greater reproductive success in the future. To include danger, we need a second principle—that death is bad for fitness. Early re- search was uncertain on how to incorporate danger into foraging models (see box 1.1), perhaps because it is not obvious how to combine the benefits of foraging and the costs of predation. Because the costs and benefits are in different units, we need to translate both foraging gain and danger into some measure of fitness. A life history perspective isessential, and it leads to a simple solution that exists precisely because the costs and benefits of foraging under danger of predation are linked. Decisions made under danger of predation are life history problems be- cause, if predation occurs, the forager’s life is history. In a life history, the basic currency to maximize is expected reproductive value, b + SV, where b is current reproduction, S is survival to the following breeding season, and V is the expected reproductive value for an animal that does survive to the next breeding season (see Stearns 1992). I concentrate here on foraging and fitness during a period without current reproduction (b = 0), so the measure of fitness is SV, the future benefits multiplied by the odds of surviving to realize them. I expect increased foraging to decrease survival to the time of reproduction, but to increase future reproduction if the animal does survive. Death lowers expected future fitness to zero. Therefore, the cost of being killed is the reproductive success a forager could have had if it had survived. This linkage means that when we ask how much risk a forager should accept to produce one additional offspring, we need to know how many offspring it would produce otherwise. For example, a forager that would otherwise ex- pect to produce one offspring might risk a lot to produce a second, while a forager that would otherwise expect to produce three offspring should risk less to produce a fourth,and a forager that would otherwise expect to produce a dozen should risk little to produce a thirteenth. This linkage of costs and benefits sets up an automatic state dependence: the potential losses from being killed increase with previous foraging success, so the relative value of further foraging is likely to be lower (see Clark 1994). Even if the fitness gains of foraging are constant, the costs should increase, since the expected reproduc- tive value increases, and that entire value would be lost in death. In line with this logic, juvenile coho salmon are more cautious when they are larger, be- cause larger individuals expect greater reproduction if they survive to breed (Reinhardt and Healey 1999). Now I will repeat these arguments mathematically. For a nonreproducing animal, fitness equals the future value of foraging discounted by the probabil- ity of surviving from now until then, W(u) =S(u)V(u), where uis a measure of foraging effort, W is fitness, S is survival, and V is future reproductive value. 312 Peter A. Bednekoff Fitness, survival, and future reproductive value are all functions of foraging effort u. In general, we expect survival to decrease and future reproductive value toincrease withforagingeffort. Morespecifically,we expectsurvivalto decre- ase exponentially with mortality, S(u) =exp[−M(u)], where M(u) is mortality. Mortality rate, M(u), and future reproductive value, V(u), could take var- ious mathematical forms. For simplicity, I define foraging effort as a fraction of the maximum possible effort, so that u varies from zero to one and does not have units. This allows mortality, M(u), and future reproduction, V(u), to be given as simple functions of foraging effort. Mortality is a function of the amount of time spent exposed to attack, the attack rate per unit time, and the probability of dying when attacked (see Lima and Dill 1990). Greater overall foraging effort could affect any of these components. For now I use a descriptive equation for mortality, M(u) =ku z , where k is a constant and the exponent z gives the overall shape of the trade-off. Later we will examine two specific cases to see what k and z might mean biologically, but for now I will simply label k as the mortality constant and z as themortality exponent. The general principle is that mortality should increase with foraging effort at a linear or accelerating rate; that is, M(u) = ku z with z ≥ 1. If foragers exercise their safest options first, we expect an accelerating function because additional food comes from increasingly dangerous options. A mathematically convenient value for the exponent, z = 2, matches observed changes in behavior well enough (Werner and Anholt 1993), but othervalues are notruled out, so Ialso examine alinear relationship (z = 1) as well as more sharply accelerating ones (z = 3andz = 4). For all values, survival declines as foraging effort increases, but the contours of the decline depend on the exponent of the mortality function, z (fig. 9.3). As we Figure 9.3. Survival declines with foraging effort. The swiftness of the decline varies with z, the exponent of the curve relating foraging effort to mortality. Foraging in the Face of Danger 313 shall see near the end of this chapter, the value of this exponent determines whether foragers should over- or underestimate danger. For therelationship betweenforaging effortand future reproductive value, we will useV(u) =κu. Inthisequation, theconstant κ translates foragingeffort into future offspring, and u is foraging effort. We expect future reproductive value to increase with total foraging effort. Studies have shown that greater foraging success leads to greater fitness in adult crab spiders (Morse and Stephens 1996), water striders (Blanckenhorn 1991), and water pipits (Frey- Roos etal. 1995).Particularly forany organismsthat areable togrow, reduced foraging in the presence of predators can lead to considerable long-term losses of potential reproduction (Martin and Lopez 1999; see also Lima 1998, table III). Alinear relationshipbetweenforaging andfuturereproductive valueisuse- ful forits simplicity.Other relationshipsmay occurin nature,and therelation- ship may differ between the sexes even within a species (Merilaita and Jor- malainen 2000). I use alinear relationship here because it allows simplemodels with clear conclusions, even though these models may somewhat understate the effects of danger. The results of more complex models, in which future fitness is a decelerating function of foraging gain, strongly support the con- clusions I reach using this simpler linear relationship. To complete the modeling framework, assume that foraging effort must be greater than some required effort, R. This requirement, R,istherequired rate of feeding divided by the maximum rate of feeding and so is a proportion without units. A forager starves if its foraging effort is less than the require- ment, and avoids starvation as long as its foraging effort is greater than the requirement. A forager gains some amount of fitness, V(R), by just meeting the requirement, but increases its future reproductive value by foraging at a rate higher than the requirement. Assembling the pieces described above, we get the overall equation for fitness: W(u) =S(u)V(u) =[exp(−ku z )][κu]. We can find the optimal foraging effort, u ∗ , if we differentiate W(u), set the derivative to zero, and solve for u. We find that u ∗ = 1 z √ kz , (9.1) so long as u ∗ ≥ R. The foraging effort that maximizes fitness (i.e., is optimal) decreases as the danger constant k increases. As the mortality exponent z increases, optimal foraging effort decreases less sharply with increases in k (fig. 9.4). Modelers sometimes assume that animals maximize survival during the nonbreeding season (see McNamara and Houston 1982, 1986; Houston and McNamara 314 Peter A. Bednekoff Figure 9.4. Optimal foraging effort declines with the expected number of attacks by predators. The swift- ness of the decline varies with z, the exponent of the curve relating foraging effort to mortality. 1999). This assumption is justified whenever the requirement is greater than the feeding rate that would otherwise be optimal, R > 1/( z √ kz) . Thus, a life history approach can converge on models that assume survival maximization even when future reproductive value increases linearly with foraging effort. In order to examine our model further, we need to look more closely at the relationship between mortality and foraging effort, m(u) = ku z .Mor- tality depends on the encounter rate with predators, time spent exposed to predators, and the probability of being killed per encounter. The relationship between mortality and foraging effort includes effects on any of these three components. I consider two situations here. First, considertadpoles encountering predatorydragonfly larvae.Tadpoles move while foraging, while dragonfly larvae sit and wait for prey. If a tadpole moves faster, it encounters both more food and more predators. In this case, the exponent, z, reflects changes in metabolic cost per distance moved and the constant, k, combines the relative encounter rate with predators and the probability of being killed in an encounter. This logic applies to any forager moving at different speeds with relatively immobile food and predators. Second, consider birds hunted by Accipiter hawks, which move a great deal while hunting. Because the hawks seek them out, greater foraging effort will not cause foragers to encounter more predators, but it may make them more likely to be killed when they do encounter a predator. In this case, the constant k includes a constant attack rate, α, and the exponent, z, reflects how foraging effort increases the probability of being killed in an attack. This logic applies when predators move rapidly and foragers are relatively im- mobile. [...]... (Lima 199 4a, 199 5a, 199 5b; Lima and Zollner 199 6) In other situations, one detector can put the flock to flight, but not as quickly as multiple detectors (Cresswell et al 2000; Hilton et al 199 9) Individual detectors flee a considerable fraction of a second ahead of the rest (Lima 199 4b; see also Elgar et al 198 6), and fractions of second could mean the difference between life and death (Bednekoff 199 6)... occasionally (Carrascal and Polo 199 9) or shown a model predator (Lilliendahl 199 7, 2000; van der Veen 199 9; Gentle and Gosler 2001), but in other tests they gained mass (Lilliendahl 199 8; Pravosudov and Grubb 199 8b) Warblers preparing for migration accumulated fat reserves faster, but attempted to leave at a lower mass, when in the presence of a simulated predator (Fransson and Weber 199 7) This seems a sensible... Current issues of journals in animal behavior and ecology provide many examples of the effects of danger on foraging Two wide-ranging reviews summarize older examples (Lima and Dill 199 0; Lima 199 8) Two excellent books give details about the interplay between foraging and danger in groups (Giraldeau and Caraco 2000; Krause and Ruxton 2002) Cuthill and Houston ( 199 7) investigate issues related to state... value by foraging too much, but can increase it only by some fraction, it might seem that the costs of foraging too much are generally higher than the costs of foraging too little Previous models have suggested that this intuition is sometimes correct, but not always (Bouskila and Blumstein 199 2; Abrams 199 4, 199 5; Bouskila et al 199 5; Koops and Abrahams 199 8) For the models used in this chapter, the... an impressive body of theory (Lima 198 6; McNamara and Houston 199 0; Bednekoff and Houston 199 4a, 199 4b; Brodin 2000; Pravosudov and Lucas 2000, 2001b) as well as a large collection of novel results that generally support the theory (e.g., Gosler et al 199 5; Bautista and Lane 2001; Thomas 2000; Olsson et al 2000; Cuthill et al 2000; see also Cuthill and Houston 199 7) Whether birds pay extra costs when... still learning details about individual (Lima and Bednekoff 199 9a) and collective detection, but these seem unlikely to negate any basic advantage in predator detection for larger groups An enormous literature documents that vigilance rates decrease and feeding rates increase with group size (Elgar 198 9; Roberts 199 6; Beauchamp 199 8; Blumstein et al 199 9) Given any safety advantage for groups, if group... of time and state on fitness (Ludwig and Rowe 199 0; Skalski and Gilliam 2002), but we should not be surprised when models produce solutions that relate to this surprisingly general rule (see also Houston et al 199 3; Werner and Anholt 199 3; Lima 199 8) 9. 7 Danger Often Depends on Group Size In one study, solitary male grey-cheeked mangabeys died at twelve times the rate of males in groups (Olupot and Waser... birds are taking off spontaneously (Metcalfe and Ure 199 5), but little effect when birds are startled into flight (Veasey et al 199 8) Large fat reserves slow escape flights by startled birds (Kullberg et al 199 6, 2000; Lind et al 199 9) and also lower takeoff angle and maneuverability during flight (Witter et al 199 4) The diurnal changes in body mass between dawn and dusk, however, have little effect on escape... literature on mixed-species groups (e.g., Bshary and Noe 199 7; Dolby and Grubb 2000), in which advantages may come about because individuals differ in complementary ways In addition, different positions with a group will generally bring different costs and benefits (Krause 199 3; Romey 199 5; Bumann et al 199 7) We expect different individuals to respond to their own costs and benefits, and therefore members... brings us to our next topic 9. 8 How Do Foragers Assess Danger? No living forager can really know its odds of being killed by a predator Therefore, we should not expect foragers to make accurate estimates of danger, but the details of their estimates may have profound effects on their behavior and ecology (Sih 199 2; Houtman and Dill 199 8; Brown et al 199 9) What they might know and how they might know it . (Carrascal and Polo 199 9) or shown a model predator (Lilliendahl 199 7, 2000; van der Veen 199 9; Gentle and Gosler 2001),but in other tests they gained mass (Lilliendahl 199 8; Pravosudov and Grubb 199 8b) area has expanded rapidly (see chap. 7) and now possesses an impressive body of theory (Lima 198 6; McNamara and Houston 199 0; Bednekoff and Houston 199 4a, 199 4b; Brodin 2000; Pravosudov and Lucas. their estimates may have profound effects on their behavior and ecology (Sih 199 2; Houtman and Dill 199 8; Brown et al. 199 9). What they might know and how they might know it are largely open questions.