FORAGING BEHAVIOUR AND DIET CHOICE

11.2 FORAGING BEHAVIOUR AND DIET CHOICE

where the net energy value of prey type i, E i =A i mb i -C h H i ,A i is the assimilable fraction of the energy

11.2.1 Feeding preference

content of prey type i, b i is the mass of prey type i, Most fish are selective foragers: they prefer to feed m is a conversion constant for prey mass to energy, on some prey types and not on others (Gerking C h is the energy costs of handling prey, H i is

1994). Preference can be defined as the difference handling time of prey type i, C s is the energy cost of between the proportion of a prey type in the diet searching, l i is the average per-capita encounter and the proportion of that prey type in the envi- rate with prey of type i, N i is the average density of ronment. Although feeding preference can be prey type i, P i (a) is the probability that an encoun- measured by a variety of mathematical indices tered prey of type i is attacked, and P i (s) is the (Confer and Moore 1987), the index developed by probability that an attacked prey of type i is Manly (1974) and further extended by Chesson successfully consumed. Feeding rate is equal to (1978, 1983) is the most widely accepted. The the total amount of assimilable food that can be Manly–Chesson index calculates the selectivity collected by a fish during one unit of search time, for prey type i as:

minus the energy cost of foraging, divided by the total foraging time that results from one unit

a i = ( dN i i ) Â ( dN j j )

of search time (Osenberg and Mittelbach 1989;

j = 1 Mittelbach and Osenberg 1994). Preference (positive or negative) for a given prey

Where i = 1, 2, . . . , k and k is the number of prey type may result from factors that influence the rate categories, d i is the number (or proportion) of prey at which a prey type is encountered (l i ) or that of type i in the diet, and N i is the density (or propor- influence the probability that an encountered tion) of prey of type i in the environment. The prey will be consumed (P i (a) and P i (s)). For example, index a i ranges from 0 to 1. Prey types that are larger prey are visible at further distances (Werner consumed in proportion to their abundance in the and Hall 1974; Confer and Blades 1975; Vinyard environment (i.e. no preference) have a i = 1/k;

and O’Brien 1976) and prey that are more active are

a i > 1/k indicates selection for a prey type and more easily seen by a searching fish (Ware 1973).

a i < 1/k indicates selection against a prey type. Any factor that affects the rate of encounter be-

Feeding preference may result from active prey tween predator and prey may result in a predator choice by the predator or from a variety of other showing a positive or a negative preference for a factors that may affect the rate of encounter prey type. It is important to note, however, that between predator and prey. We can best under- while differential prey encounter may lead to stand the role of behaviour in fish diet selection by selective foraging, meaning that there is a differ- breaking down the act of predation into its com- ence between the abundance of prey types in the

Fish Foraging and Habitat Choice

Number of prey type i in the environment

Prey size and reactive

distance, prey and

Probability of

predator activity, prey

prey encounter

crypsis, prey refuge use

(λ i N i )

Number of prey type i encountered

Probability that

Active choice by the

an encountered

predator

prey is attacked (P i (a))

Fig. 11.1 Stages in the predation

Number of prey

process. The right-hand column

type i attacked

lists the components of a multi- species Holling functional

Probability that response model (i.e. equation

an attacked prey 11.2) that determine the

Prey size, mobility and

is successfully probability that a given prey type

escape behaviour, prey

consumed (i) will be consumed. Listed to the

defence (e.g. shell

(P i (s)) left are attributes of predator and

hardness, spines)

prey that influence these

Number of prey

components. (Source: modified

type i consumed

from Sih 1993.)

environment, this mechanism does not involve 1989; Huckins 1997). The same problem is faced by active predator choice.

piscivorous fish. Christensen (1996) and Sih and

Predator behaviour may enter into diet selec- Christensen (2001) argue that piscivore diets are tion through a predator’s choice of which prey strongly influenced by the differential capture items to pursue (P i (a)) and the probability that a success of prey and that this is a major reason why pursuit will result in successful capture (P i (s)). piscivores generally prefer smaller prey than When prey are small relative to the predator, would be expected based on maximizing their capture success generally will be high. For ex- energy intake per handling time (see Juanes 1994; ample, adult planktivores have very high capture Juanes et al., Chapter 12, this volume). success when feeding on zooplankton. How- ever, even for adult planktivores, some types of

11.2.2 Diet selection and

zooplankton such as copepods are sufficiently evasive that fish must use alternative capture

optimal foraging

techniques (Vinyard 1980). Fish that feed on hard- Much of the theoretical development of foraging bodied prey like molluscs routinely encounter models has revolved around the question of prey that differ in capture success, brought about predator choice: when should a predator choose by a variable probability of successfully crushing a to pursue a prey item that it has encountered?

254

Chapter 11

optimization criteria to address this question of predator choice, arguing that natural selection should result in predator behaviours that maxi- mize their rate of energy gain, which is a com- ponent of fitness. Charnov (1976) developed one of the first optimal diets models, using an equation similar to equation 11.2 to predict a predator’s diet. In Charnov’s model, predators adjust their attack probabilities (P i (a)) to maximize their total energy gain (E/T). Two basic predictions that arise from Charnov’s model and from similar models devel- oped by Schoener (1971), Pulliam (1974), Werner and Hall (1974) and others are:

1 predators should prefer prey that yield more energy per unit handling time;

2 as the abundance of higher value prey increases in the environment, lower value prey should

be dropped from the diet and predators should become more selective (Stephens and Krebs 1986; Sih and Christensen 2001).

Fish were used in some of the first tests of opti- mal foraging theory (OFT) (Werner and Hall 1974; Kislalioglu and Gibson 1976; Mittelbach 1981), and a number of articles have reviewed the applica- tion of OFT to understanding fish diet selection and feeding behaviour (Werner and Mittelbach 1981; Townsend and Winfield 1985; Hart 1986, 1989). In addition, two general reviews of OFT by Stephen and Krebs (1986) and Sih and Christensen (2001) have evaluated the overall success of OFT in predicting predator diets, including those of fish. What can we conclude from these and other stud- ies about the applicability of OFT to understand- ing diet choice in fish? As Sih and Christensen (2001) note, this is a bit like asking whether the glass is half full or half empty. In many ways, OFT has been successful. Fish do prefer to feed on prey that yield higher energetic returns. However, we now recognize that other factors besides handling time need to be considered in measuring the cost of feeding on a prey item. In particular, for evasive prey, we need to also include the probability of cap- ture (P i (s)) in calculating the expected energetic re- turn from prey type i (Christensen 1996). Also, for some prey types, digestive time may need to be in- cluded in the total ‘handling’ time for a prey item, although current evidence suggests that including

digestive time may not change the relative rank- ings of prey types (Kaiser et al. 1992; Nilsson 2000). Some studies have shown that prey selection by

a fish changes as the stomach of the fish fills.

A review of this area can be found in Clark and Mangel (2000).

Fish have also been shown to follow the OFT prediction that diets should become more selec- tive as the abundance of more profitable prey increases in the environment (Werner and Hall 1974), and in a few cases predicted diets based on OFT closely matched observed diets (Mittelbach 1981; Galis and de Jong 1988; Persson and Greenberg 1990). However, in most cases, fish consumed some prey outside the optimal diet. The reviews by Stephen and Krebs (1986) and Sih and Christensen (2001) list 18 studies that tested the predictions of OFT with fish. In only a few of these studies (<20%) was there a good quantitative fit between predicted and observed diets. Given the simplicity of most optimal diet models, and the fact that these models do not incorporate many important components of feeding behaviour such as prey detection and recognition, prey capture success, predator learning and memory, it is not surprising that a quantitative fit between theory and observation is missing. However, most (>80%) of the fish studies reviewed by Stephen and Krebs (1986) and Sih and Christensen (2001) found at least qualitative agreement between predicted and observed patterns of diet selection. Further, the usefulness of OFT may extend beyond the study of feeding behaviour, as OFT provides one of the few means of predicting how the attack probabil- ities for a prey type (P i (a) in equation 11.2) should change in different environments. These attack probabilities are needed if we are to use a foraging model such as equation 11.2 to predict fish feeding rates under different environmental conditions.