Introduction

Speed while foraging is a key factor determining foraging success (Pyke 1981). The area searched per unit time by a forager depends critically on foraging speed. Moving more rapidly during foraging will increase the area searched, and prey intake rate will increase if the probability of detecting prey is not negatively affected by foraging speed. Foraging speed thus enters directly into equations relating prey intake rate to food density, which are the basis for functional responses in ecology (Holling 1959).

Models of functional responses typically consider only one predator searching for prey. When multiple predators are present, competition may take place through exploitation of the resources or through interference among foragers (Sutherland 1996). Scramble competition is concerned with competition via exploitation of resources (Parker 2000). To compete in a scramble, individuals are expected to adopt behavioral patterns that increase their share of resources at the expense of others. Models that predict behavioral adjustments in response to scramble competition, such as changes in foraging speed, are based on evolutionary stable strategies (ESS), which prescribe best responses by foragers in the context of responses exhibited by the rest of the population (Maynard Smith 1974). Such modeling in the context of foraging speed is quite recent, and few empirical studies have examined their predictions. The obvious prediction that animals should feed faster when facing competition for limited resources is not always the only stable outcome (Shaw et al. 1995).

A recent comprehensive model considers behavioral adjustments made by foragers competing for a limited food supply in an ESS context (Bednekoff and Lima 2004). An individual in this model can increase its share of the resources, and thus experience a higher intake rate, by increasing its feeding effort relative to others in the group. Feeding effort can be empirically related to any foraging behavior that increases rate of encounter with prey, such as foraging speed. At the ESS, feeding effort is expected to increase with the number of competitors and also to decrease when attacks by predators are more likely. The latter prediction arises when maintaining a higher speed is less compatible with detecting predators. The effect of resource density was not considered explicitly in this model. It seems reasonable to assume that, as resources become more abundant, increasing foraging speed relative to others is less and less likely to produce a large increase in relative shares of the resources, and the need to forage faster should thus decrease with resource density. Any increase in locomotion costs with foraging speed should also lead to a decrease in foraging speed in richer food patches (Speakman and Bryant 1993).

While foraging speed has been measured directly or indirectly in many species, few empirical studies investigated adjustments in foraging speed in feeding scrambles taking into account both food density and competitor density. For instance, foraging speed was related to competitor density in shorebird species but food density was not measured directly (Minderman et al. 2006; Speakman and Bryant 1993). Another study related indirect measures of speed, such as handling time, to competitor density, but in a context where individuals did not have to compete for resources (Lima et al. 1999).

Here, I examined changes in foraging speed in relation to competitor density and prey density along a gradient of predation risk in flocking semipalmated sandpipers (Calidris pusilla). During fall staging, sandpipers gather in large groups foraging predominantly for a burrowing prey, Corophium volutator (Hamilton et al. 2003; Hicklin and Smith 1984; Peer et al. 1986). While density of this amphipod is high on the mudflats where sandpipers forage, Corophium retreats to its burrow following close encounters with birds (Selman and Goss-Custard 1988), as is the case in other burrowing invertebrate species (Minderman et al. 2006). The presence of many sandpipers is thus expected to decrease food availability for those companions immediately behind. Given the small size of each prey and the large size of sandpiper groups, interference in the form of prey stealing and aggressive food defense is not expected and rarely documented (Boates 1980; Wilson and Vogel 1997). Instead, sandpipers should compete with one another by adjusting their search paths and foraging behavior (Beauchamp 2007; Stillman et al. 2000).

I predicted that, under scramble competition, individuals would forage faster in denser groups and slower in areas with higher food density. Assuming that increases in foraging speed reduce the ability to detect predators (Bednekoff and Lima 2004), foraging speed should increase as the distance to obstructive cover increases. Further away from cover, sandpipers have more time to detect predators, such as falcons (Falco spp.), that launch surprise attacks from wooded cover (Beauchamp 2008; Dekker et al. 2011).

Materials and methods

Study site

The study was conducted from late July to early August during the peak of fall migration at Mary’s Point in 2005 and 2006, and at Daniel’s flat from 2008 to 2011. The two sites are within kilometers of each other and are located in the upper Bay of Fundy, New Brunswick, Canada (45.73°N, 64.65°W). Each site is bordered by wooded cover 30–50 m from shore. Wooded cover consists of a mixture of tall deciduous and conifer trees from where falcons launch surprise attacks. Tides averaging 11.5 m in height expose mudflats in the area twice daily. I monitored birds from shore with a 10–45× or a 15–60× spotting scope.

Setting plots and Corophium sampling

Bamboo canes were used to stake 6 × 6 m plots each year set at distances ranging from 55 to 285 m from wooded cover. Depending on the year, I set five or six plots at fixed distances along two transect lines perpendicular to the shore. I took two to three core samples from each plot at different times during the study period using a circular corer (79 cm2) pressed into the sediment to the top of the anaerobic layer (approximately 5–10 cm). Contents were sieved through a 0.85-mm sieve known to retain the large Corophium individuals (Crewe et al. 2001) that are preferentially selected by sandpipers at this time of year (Peer et al. 1986).

Sampling sandpiper behavior

Observations took place during daylight hours and started as the birds left high tide roosts to start foraging 2–3 h following high tide. I made observations as birds foraged near the receding water line and passed through a plot. All plots could be visited within an hour after the start of foraging.

Estimating foraging speed directly is difficult in the field since individual foraging path is not easy to control. Therefore, I used the time needed by sandpipers to cross one plot as a proxy for foraging speed. At the tideline, individuals cannot forage in the water and are thus forced to move primarily along the edges of the water. In addition, food availability is greatest in recently emerged areas (Morgan 1965), again restricting search to a limited area along the tideline. Near the tideline, individuals thus spread rapidly along the edges of the water and this causes movements to be relatively linear.

Upon the first visit by a flock at a given plot, I monitored the behavior of one to several individuals selected haphazardly from those present until all birds left the plot. Repeated sampling of the same subjects appears quite unlikely given the large number of birds using the site every day (from 1,000 to 75,000). A focal observation started when the focal bird crossed one boundary of the plot and ended when the bird reached the opposite boundary. Focal observations were not retained if the focal bird disappeared from sight or flew away before crossing the boundary. I also dropped focal observations for which the density of birds changed in the plot due to the sudden arrival or departure of other birds. The number of birds present in the plot was counted at the beginning of each focal observation and served as an estimate of sandpiper density for this focal observation.

In addition to the time needed to cross a plot by each focal subject, I collected data on foraging behavior. For each focal bird, I dictated the occurrence of captures (pecks followed by visible swallowing or accompanied by handling movements) on a portable cassette recorder as events unfolded. Captures can be difficult to detect especially if birds swallow after capturing more than one prey or if swallowing movements are missed (Boates 1980). Errors in the estimation of capture rate should be fairly constant from plot to plot as there was only one observer. Capture rate could not be evaluated for each focal observation due to poor visibility.

Statistical analysis

Prior to analysis, sandpiper density and food density were log-transformed to normalize distributions. I calculated the average density of Corophium for each plot using all available estimates obtained during the study period. I used a mixed linear model (SAS v.9.2; Cary, NC) with duration of a focal observation as the dependent variable and sandpiper density, Corophium density and distance from cover as independent variables. Site, year within site, and day within a given year and site were treated as random variables. The latter random variable controls for variation in environmental factors, such as temperature, that are much greater between days rather than within observation periods. Such environmental factors are known to influence foraging behavior in staging sandpipers (Beauchamp 2006). As the distribution of captures was right-skewed, I used a negative binomial regression model with the natural logarithm of focal observation duration as an offset to control for differences in duration among focal observations. Only one random effect could be used in the negative binomial regression model. The model thus included day nested within year and site as a random factor and the aforementioned fixed factors including prey density squared to model a non-linear relationship.

Results

A total of 851 focal observations met the inclusion criteria. Focal observations lasted on average 41.5 s, ranging from 10.2 to 105.7 s. Food density ranged from 191 to 4,201 prey/m2, while sandpiper density ranged from 0.028 to 55.6 birds/m2.

Capture rate ranged from 0 to 14.2 per min with a median of 1.97. The expected number of captures increased with food density (χ 2 = 6.7, p = 0.01) at a decreasing rate (food density squared, χ 2 = 4.5, p = 0.03). The expected number of captures decreased by 16% per unit of sandpiper density in logarithm [β (SE) −0.17 (0.058), χ 2 = 5.9, p = 0.02]. Distance from cover was not statistically significant when added to the above model (χ 2 = 0.11, p = 0.74).

The time needed to cross a plot decreased with sandpiper density [β (SE) −6.91 (1.15), F 1,770 = 35.9, p < 0.0001; Fig. 1a], increased with food density [β (SE) 6.10 (2.51), F 1,770 = 5.91, p = 0.01; Fig. 1b] and increased with distance from cover [β (SE) 0.043 (0.013), F 1,770 = 11.0, p = 0.001; Fig. 1c]. Variation between sites was not statistically significant (0% of total variation) and nor was variation among years within each site (11% of the total variation Z = 1.32, p = 0.09). Statistically significant variation occurred among days (16% Z = 3.41, p = 0.0003).

Fig. 1
figure 1

Changes in foraging speed [duration of movement (s) across a standard plot] in fall staging semipalmated sandpipers (Calidris pusilla) as a function of a log10 forager density, b log10 prey density, and c distance to cover (m) in the upper Bay of Fundy (New Brunswick, Canada), 2005–2011. A linear regression line is shown to illustrate the linear trend

Full size image

Discussion

Prey intake rate decreased with sandpiper density indicating competition for resources. Given that overt aggression is relatively rare in this species, competition can be characterized as a scramble. The ever decreasing positive effect of food density on prey intake rate is typical of type II functional responses, which has been documented in this species (Beauchamp 2009a) and other shorebirds (Goss-Custard et al. 2006).

Foraging speed, as estimated from the time taken to cross opposite boundaries of standard plots with known food and sandpiper densities, did not vary to a large extent between the two sites and across several years within each site, suggesting that the results are robust across space and time. Foraging speed did vary among days during the study period, reflecting the role of environmental factors, such as temperature and migration phenology, which vary from day to day and influence foraging behavior in sandpipers (Beauchamp 2006). The largest source of variation was within plot, and three statistically significant sources of variation in foraging speed emerged: foraging speed decreased with distance to cover and food density, but increased with sandpiper density.

In the context of scramble competition, the negative effect of prey density on foraging speed probably reflects the reduction in competition intensity as the availability of prey increases. Potential gains from unilateral increases in foraging speed, in terms of relative share of resources, become more and more negligible as prey availability increases, and may not be sufficient to compensate for increased energy costs associated with more rapid movements. In an earlier study of sandpipers foraging on Corophium, step rate decreased with prey density in a limited sample of three plots, but sandpiper density was not measured (Wilson and Vogel 1997). A similar finding was documented in redshanks (Tringa totanus), but food density could not be measured independently of the birds’ behavior (Minderman et al. 2006). In another shorebird, step rate was also reduced in enriched food patches (Santos et al. 2009). In a rare instance where foraging speed was measured directly, foraging speed decreased with food density in redshanks foraging on Corophium and in oystercatchers (Haematopus ostralegus) searching for larger prey (Speakman and Bryant 1993), but here again food density could not be estimated directly. As an aside, it is not clear in general whether step rate is a good proxy for foraging speed given that step size may not be constant (Hayes and McNeil Alexander 1983; Speakman and Bryant 1993).

The positive effect of sandpiper density on foraging speed is compatible with the scramble hypothesis, and indicates that, to obtain a sizeable share of resources, individuals need to increase speed of foraging when competing with more companions. Results from the literature confirm this expectation. Foraging speed in fish increased with competitor density when the ratio of food density to the number of foragers remained constant (Shaw et al. 1995). Foraging activity, a more indirect measure of foraging speed, was found to increase with group size independently of predation risk level in another fish species, again suggesting a role for scramble competition (Johnsson 2003). In a recent experiment manipulating the perceived level of competition, individual birds interrupted feeding less often when companions were feeding at a higher rate (Rieucau and Giraldeau 2009), suggesting a role for resource competition. In redshanks foraging on burrowing prey, step rate did not vary with flock size, but neither competitor nor prey density was measured directly (Minderman et al. 2006). In dunlins (Calidris alpina), step rate decreased with flock size, but again both densities were not measured (Barbosa 1996).

Distance to cover was introduced as a covariate in the analysis. In shorebirds, predation risk is considered higher closer to obstructive cover because foragers have less time to react to a bird of prey launching surprise attacks (Cresswell and Whitfield 2008; Dekker and Ydenberg 2004; Leger and Nelson 1982). Foragers should be the wariest closer to obstructive cover, and therefore the increase in foraging speed closer to cover was not expected. Earlier work with sandpipers indicated that false alarm flights were less frequent further away from cover, suggesting that predation risk is indeed higher closer to cover (Beauchamp 2010). If vigilance against predators is not totally compatible with foraging, foraging speed should decrease closer to cover to allow more time to vigilance. Indeed, the scramble model by Bednekoff and Lima (2004) predicted a decrease in feeding effort when predation risk was higher. Nevertheless, foraging in sandpipers is often curbed closer to cover by numerous false alarms (Beauchamp 2010). Frequent interruptions in foraging can lead to an increase in foraging effort, and thus to an increase in foraging speed, to compensate for reduced foraging time (Beauchamp and Ruxton 2007; McNamara and Houston 1992). The present findings are compatible with this interpretation. A similar negative effect of distance to cover on step rate was documented in redshanks (Minderman et al. 2006).

The time to cross opposite boundaries in a standard plot was used as a proxy for foraging speed. The two measures would be equivalent if all sandpipers travelled the same distances in each plot. However, since I could not control entry and exit points nor the sinuosity of foraging paths within a plot, duration of movement is therefore only an approximation of foraging speed. It is conceivable that a longer duration of movement through a plot may reflect a more sinuous foraging path rather than a slower speed. Nevertheless, I would argue that even a sinuous path is compatible with a decrease in competition intensity. To increase their share of vanishing resources, sandpipers do best by avoiding nearby companions and by trying to reach unexploited areas by running straight ahead. Geese at the trailing edge of a flock, for instance, move much more rapidly to secure unexploited resources than those behind (Carbone et al. 2003). A sinuous path would then simply indicate that sandpipers are freer to turn left and right instead of going straight ahead to avoid neighbors. In other species, sinuous foraging paths are often considered an indication of low foraging speeds (Nolet and Mooij 2002). I thus consider duration of movement a good indicator of competition intensity under the conditions faced by the sandpipers. Direct measurements of speed would be welcome in future studies to establish the limits of this proxy.

In most species, foraging activities are performed at the expense of vigilance against predators. It is therefore reasonable to ask what is the relative contribution of perceived safety and competition intensity on changes in foraging speed (Bohlin and Johnsson 2004; Grand and Dill 1999). To illustrate the difficulty in disentangling competition from safety effects, consider the potential effect of sandpiper density and food density on vigilance. Foraging in denser groups is expected to reduce vigilance by facilitating intra-group communication about predators (Proctor et al. 2003). If vigilance interferes with locomotion, foraging speed may be expected to increase when individuals forage in denser groups, as documented here. With respect to food density, models predict that vigilance should decrease with food density when animals face time constraints (Ale and Brown 2007; Beauchamp 2009b; McNamara and Houston 1992). From the perspective of safety, foraging speed should thus increase with food density if vigilance interferes with locomotion. I documented the opposite pattern, but as noted earlier such models did not take into account the likelihood of foraging interruptions, which changes as a function of distance to cover in this species. It is important to stress that safety concerns, if present, must act together with scramble competition to explain the negative effect of sandpiper density on capture rate. If safety concerns acted alone, capture rate would increase in denser, safer groups of sandpipers.

Vigilance is not straightforward to document in rapidly moving shorebirds with short handling times and few pauses while foraging (Barbosa 2002). How vigilance influences foraging speed is not known in sandpipers, and as such it is difficult to make unambiguous predictions for the safety hypothesis. One possible way to disentangle the two mechanisms is to determine whether foraging speed varies with competitor density when competition is relaxed, for instance in very rich food patches. An increase in foraging speed with competitor density would only be compatible with the safety hypothesis. Pending further research, variation in foraging speed in staging flocks of semipalmated sandpipers is at the very least compatible with scramble competition. It is certainly possible that safety and competition act together in the same direction to affect foraging speed (Bednekoff and Lima 2004).