SEARCH

SEARCH BY CITATION

Keywords:

  • trail-mediated interactions;
  • non-lethal effects;
  • starvation-predation risk trade-off

Abstract

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgments
  7. REFERENCES

Redshanks Tringa totanus that are preyed upon by Sparrowhawks Accipiter nisus at the Tyninghame Estuary, Firth of Forth, Scotland, provide an example of how the starvation–predation risk trade-off results in mortality. In this trade-off, animals cannot always optimize anti-predation behaviour because anti-predation behaviours, such as avoiding predators, are usually incompatible with foraging behaviours that might maximize intake rates. Therefore, as animals compensate for starvation risk, predation risk increases. Sparrowhawks are the main direct cause of death in Redshanks at Tyninghame. Sparrowhawk attack rate is determined by Redshank vulnerability, and vulnerability decreases as group size and distance to cover increase, and probably as spacing decreases. But reduction of predation vulnerability reduces feeding rate because areas away from cover are less food-profitable and grouping results in increased interference competition. Increased starvation risk in midwinter means Redshanks are forced to feed on highly profitable prey, Orchestia amphipods, the behaviour of which means that Redshanks are forced to feed vulnerably, in widely spaced groups, close to predator-concealing cover. Therefore, it is the constraints that limit the ability of Redshanks to feed in large, dense flocks away from cover that ultimately lead to mortality. We investigate this hypothesis further by testing the prediction that mortality can be predicted directly by cold weather and population density. We demonstrate that the overall number of Redshanks and the proportion of Redshanks killed increase in cold months when controlling for population size. We also demonstrate that the proportion of Redshanks killed increases when there are fewer Redshanks present, because the success rate of hunting Sparrowhawks increases, probably because effective management of predation risk through flocking is constrained by a low population size. Redshanks therefore provide an example of how directly mortality caused by predation arises from starvation risk and other constraints that prevent animals from optimizing anti-predation behaviour.

Predation risk arises mainly because of starvation risk, where an inability to gain enough food via safe foraging options leads to increased predation risk (Lima & Dill 1990). Although the direct cause of mortality for an animal may be because a predator has killed it, the underlying cause may be because a starving animal is forced to take risks (Houston et al. 1993). Therefore, although a predator–prey system may be characterized by high mortality due to predators, the system may be driven by starvation risk (e.g. Watson et al. 2007). Previous studies suggest that Redshanks Tringa totanus that are preyed upon by Sparrowhawks Accipiter nisus at the Tyninghame Estuary, Firth of Forth, Scotland, provide a good example of how the starvation–predation risk trade-off results in mortality. In particular, they apparently provide an example of where the direct cause of death is almost entirely due to predation, yet the underlying cause of this mortality is starvation risk. This paper tests hypothesized predictions that high predation mortality is driven by constraints on safe foraging because of increased starvation risk.

Sparrowhawks are the main direct cause of death in Redshanks at Tyninghame (Cresswell & Whitfield 1994). The attack rate of Sparrowhawks on Redshanks is determined by their vulnerability not their availability, with Sparrowhawks attacking Redshanks only when their chance of success is high (Cresswell & Quinn 2004, Quinn & Cresswell 2004). Vulnerability is determined by group size, distance to cover and spacing. This is because success rate of attacking Sparrowhawks decreases with increasing group size (Cresswell 1994a) and distance to cover (Cresswell 1994b) and probably with decreasing spacing within groups (Whitfield 1988, 2003b, Hilton et al. 1999, Quinn & Cresswell 2006). Management of vulnerability, however, comes at a cost of reduced feeding rate because less vulnerable areas away from cover are less profitable (Cresswell 1994b). Increasing flock size also results in reduced intake rate per unit time spent feeding and increased interference competition (Cresswell 1994a, Minderman et al. 2006). There is therefore a clear starvation–predation risk trade-off in foraging Redshanks: foraging rate can only be maximized at the cost of increasing predation risk.

Increased starvation risk at Tyninghame means Redshanks (and these are almost entirely juvenile birds) are forced to feed in more profitable saltmarsh areas closer to predator-concealing cover (Yasuéet al. 2003). In cold midwinter periods, their energy budgets can only be met by feeding on high profitability prey, i.e. Orchestia amphipods. The anti-predation behaviour of Orchestia itself then means that the Redshanks on the saltmarsh are forced to feed for some part of the day in a very vulnerable fashion, in widely spaced groups, close to predator-concealing cover, because Orchestia burrow and become largely unavailable when Redshanks feed where they occur (Minderman et al. 2006). As a result, flocks end up feeding in areas close to cover that have not been previously disturbed and in widely spaced flocks as they constantly move out from the flock centre to find undisturbed prey areas.

A further potential constraint to safe foraging is group size. For any population size of Redshanks, a certain proportion feed on the risky saltmarsh for some part of the day, rather than safer but less profitable mudflat areas, so that they can maintain their daily energy budgets (Yasuéet al. 2003). When population size is low so that only a small number of Redshanks feed on the saltmarsh, then only relatively small flock sizes can form. Small flocks are attacked more frequently by Sparrowhawks (Cresswell & Quinn 2004, Quinn & Cresswell 2004) and attacks on small flocks are likely to be more successful (Cresswell 1994a). Therefore, it is possible that there is positive feedback at small population sizes so that there are higher levels of mortality when there are fewer Redshanks feeding on the saltmarsh, further increasing the strength of the population-size-dependent mortality.

Therefore, it is the constraints – cold weather and short days increasing the chance of negative energy budgets and the characteristics of foraging groups – that limit the ability of Redshanks to feed in large, dense flocks away from cover that is the real underlying cause of the mortality. Here, we test two predictions:

  • Prediction 1: If mortality depends on cold weather, then we predict that more Redshanks, controlling for population size, will be killed in cold weather because there will be more Redshanks adopting risky foraging behaviour in order to balance their energy budgets.

  • Prediction 2: If mortality depends on the density of conspecifics, because modifying vulnerability via grouping requires the presence of conspecifics, then we predict that when we control for the effects of temperature, the proportion of Redshanks killed will increase when population size is low. If this hypothesis is true, then we would also predict that attack rates will increase and success rates decrease when population size is high, because vulnerability to Sparrowhawk attack partly depends on group size.

METHODS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgments
  7. REFERENCES

Observations were made from September to April 1989–90 (Winter 1), August to April 1990–91 (Winter 2) and August to March 1991–92 (Winter 3), during 2557 h at the Tyninghame estuary, East Lothian, Scotland (for site details see Cresswell & Whitfield 1994). The study site was visited daily throughout the winter to count the Redshank population, to find the remains of Redshanks killed by Sparrowhawks, and to observe Sparrowhawk attacks and kills on Redshanks.

Redshank population was estimated by counting every 2 weeks on maximum-height spring tides, when all the Redshanks within the estuary were spread out linearly on a small area of saltmarsh. All could be counted easily from one point in about 5 min so that problems of detectability and recounting birds were avoided. The Redshanks were counted until counts became a consistent maximum (i.e. three or more consecutive counts gave a maximum figure ± 20 birds) and then the two highest counts for that day were averaged. During Winter 2, 47 Redshanks were cannon-netted and colour-ringed so that individual recognition was possible. At least 36 of these returned to the estuary for Winter 3, and so the percentage of Redshanks colour-ringed was recorded on each Redshank census for that winter to confirm the assumption that high tide counts assessed the entire Redshank population of the study area.

Tyninghame was searched regularly for Sparrowhawk kill remains, in order to assess the number of Redshanks being killed by Sparrowhawks each month. Three days every month were spent searching a fixed area covering the entire north shore of the estuary and the woodland or cover adjacent to it for about 100–200 m inland. The south shore was found to contain few kills and was searched once each winter. The saltmarsh was searched before each set of spring tides, and remains were also found opportunistically. The efficiency of recovery of the wader remains was tested by one of us setting out simulated Sparrowhawk kills (piles of Redshank feathers) in typical plucking areas and the other then seeing how many were found during normal monthly searches. At least 97% (n = 38) of remains were found. Recovery of kills was also estimated by direct observation of the frequency of Sparrowhawks leaving prey remains within the whole search area: 98.8% of Sparrowhawks (n = 81) seen carrying wader prey caught within the previous minute left remains in the search area. The remains were searched for immediately after the raptor had finished feeding, to confirm the location of visible remains within the normal monthly search area (for further details see Cresswell & Whitfield 1994). Prey remains were classified as being due to Sparrowhawk on the basis of location and conditions of the remains: feathers in cover (under low-canopy woodland or bushes, in tall grass or bracken); all neatly plucked, usually on the ground or from a low (< 1 m) perch; bones rarely present, if present only front part of the keel and wing bones uneaten and wing bones entirely or almost entirely clean of feathers; occasional presence of distinctive Sparrowhawk faeces (for further details see Cresswell & Whitfield 1994). Sparrowhawks were also frequently (about 25% of observed kills) kleptoparasitized by Carrion Crows Corvus corone (Cresswell 2004) and so Redshank kills were also ascribed to Sparrowhawks when kills were found with: remains not in cover; whole body intact or intact keel with wings attached, and legs still attached to pelvis; wings raggedly plucked, rarely past humerus; large blunt notches in the sternum occasionally present (Cresswell & Whitfield 1994). As a check of the above classification, all observations of identified Sparrowhawks eating prey were followed up by a description and classification of the remains. For Sparrowhawk, 71% (n = 87) of remains, all after a Sparrowhawk was seen carrying prey in the area into cover, were classified as Sparrowhawk, and the remaining 29% were classified as Sparrowhawk/kleptoparasitism, all after the Sparrowhawk was seen being kleptoparasitized by Crows.

Redshank remains were aged as juvenile (birds in their first winter) or adult (birds in at least their second winter) on the basis of distinctive juvenile feathers (Cresswell & Whitfield 1994) and age-specific differences in outer primary tip shape and wear (Whitfield et al. 1999). Redshank mortality is strongly age-dependent due to Sparrowhawk predation (Cresswell & Whitfield 1994), because juveniles are excluded by adults from safer rocky shore areas far from cover and feed on the mudflats and saltmarsh closer to cover where most of the Sparrowhawk predation risk occurs (Cresswell 1994b). Analyses were therefore predominantly restricted to the population of juvenile Redshanks only (Cresswell & Whitfield 1994), where juvenile population size refers to those birds that feed on the mudflats and saltmarsh system (Yasuéet al. 2003). In this paper, overall population size refers to the juvenile population plus the almost entirely adult population that feeds on the rocky shore and that only very rarely feeds on the saltmarsh habitat (Cresswell 1994b). We observed hunting Sparrowhawks opportunistically by using 10× binoculars or a 15–60× zoom telescope and the majority of raptor observations were made at less than 300 m. A Sparrowhawk attack was defined as a rapid directed flight at a clearly identifiable Redshank or group of Redshanks. A successful attack by a raptor (a kill) was one that resulted in the raptor catching hold of the prey, which in almost all cases resulted in the death of the Redshank (Cresswell 1996).

Prediction 1: mortality is determined by cold weather

A general linear model (Table 1) to test the relationship between number killed and temperature was constructed with total number of juvenile Redshanks killed as the dependent variable and mean monthly temperature and population of juvenile Redshanks as independent variables, controlling for month and winter season. The temperature*population interaction was then added because it is a direct prediction that mortality in Redshank via Sparrowhawk predation should occur particularly at low temperatures. Whether the effects of temperature on the number killed was the same across the three winter seasons was tested by the addition of the interaction of winter period*temperature to the model. Whether the dependence of the relationship between mortality and population size on temperature was the same across the three winter seasons was tested by adding temperature*population*winter, and all sub-component two-way interactions into the model.

Table 1.  Results of a GLM to test Prediction 1 – the number of Redshanks killed depends on population, but only at low temperatures controlling for seasonal effects.
 Type III sumd.f.FPB
  1. Monthly sums or means are used (n = 22 months over three consecutive winters). B is the model parameter estimate; e.g. y = Bx+Bz ... + intercept. Dependent variable: total number of juvenile Redshanks killed by Sparrowhawks each month. Addition of the interaction winter period*mean monthly temperature was not significant (F2,13 = 0.9, P = 0.42), showing that the effect of temperature on proportion killed was the same for all three winters.

  2. Addition of the interaction winter period*mean monthly temperature*Redshank population (along with all two-way component interactions) was not significant (F2,9 = 3.6, P = 0.07), showing that the relationship between temperature and population size was similar in all three winters.

Corrected model3020.0 614.7< 0.001 
Intercept  32.5 1 1.0  0.34 
Redshank population 471.3 113.7  0.002 0.29 ± 0.08
Temperature   0.5 1 0.01  0.91–0.15 ± 1.3
Winter 229.7 2 3.4  0.062–8.5 W1; –4.7 W2
Month  28.9 1 0.8  0.37–1.5 ± 1.7
Population*temperature 269.9 1 7.9  0.013–0.027 ± 0.01
Error 513.315   
Total929422   
Corrected total   1.0421   
Corrected R2 = 0.79

Prediction 2: mortality is determined by population density

A general linear model (Table 2) to test the relationship between proportion killed and population was constructed with proportion of juvenile Redshanks killed as the dependent variable and the population of juvenile Redshanks as an independent variable, controlling for mean monthly temperature, month and winter season. The interaction of temperature*population was then tested with the prediction that this would not be significant. Although it is a direct prediction that mortality in Redshank via Sparrowhawk predation should occur particularly at low temperatures, the relationship between population size affecting mortality through group size constraints should not be affected by temperature: Redshanks only feed in groups on the saltmarsh so any variation in vulnerability due to temperature has already been removed. Whether the effects of population size and temperature on the number killed were the same across the three winter seasons were tested by the addition of the interaction of winter period*population and winter period*temperature to the model, respectively.

Table 2.  Results of a GLM to test Prediction 2 – the proportion of Redshanks killed depends on population, controlling for temperature and seasonal effects.
 Type III sumd.f.FPB
  1. Monthly sums or means are used (n = 22 months over three consecutive winters). B is the model parameter estimate. Dependent variable: proportion of total population of juvenile Redshanks killed each month. The relationship between the proportion killed and overall population was not significantly affected by temperature (population*temperature added to the model, F1,15 = 1.9, P = 0.18).

  2. The relationship between proportion killed and overall population was the same for all three winters (winter*population interaction added to the model, F2,14 = 1.7, P = 0.22).

  3. The relationship between proportion killed and temperature was the same for all three winters (winter*temperature interaction added to the model, F2,14 = 0.9, P = 0.42).

Corrected model0.36 512.8< 0.001 
Intercept0.10 117.2  0.0011.6 + 0.4
Redshank population0.11 119.4< 0.001–0.003 ± 0.001
Temperature0.08 114.2  0.002–0.026 ± 0.007
Winter0.16 214.5< 0.001–0.22 W1; –0.18 W2
Month0.05 1 8.3  0.011–0.059 ± 0.02
Error0.0916   
Total1.1322   
Corrected total0.4521   
Corrected R2 = 0.74

A generalized linear mixed model (Table 3) to test the relationship between attack rate and population was constructed with number of attacks per observation day by Sparrowhawks as dependent variable (data approximated a Poisson distribution) and population of juvenile Redshanks as independent variable, controlling for time spent at the study site (to control for variation in daily observer effort), mean monthly temperature, and month and winter season (the last two as random variables to control for uneven numbers of sampling days in each month). The interaction of population*temperature was included in the model because it is a direct prediction that attack rate should decrease particularly at lower temperatures because vulnerability, and so success rate of attacks, should increase.

Table 3.  The results of a Generalized Linear Model testing Prediction 2 – attack rate depends on population size and temperature individually, but the strength of the relationship between attack rate and population size depends on temperature, controlling for seasonal effects.
Independent variablesd.f.FPB
  1. Daily sums (n = 299 days) of attacks (overall n = 546) are used with daily temperature at 09:00 h, daily total of hours spent observing for attacks, and monthly population size (n = 22 months over three consecutive winters). B is the model parameter estimate. Dependent variable: total number of attacks per day. The relationship between attack rate and population size was dependent on temperature (population size*temperature interaction added to the model; F1,289 = 21.5, P < 0.001).

Daily temperature1,29047.9< 0.001–0.21 ± 0.03
Juvenile population size1,29010.5  0.001 0.0086 ± 0.003
Effort (time in field)1,29012.9< 0.001 0.22 ± 0.06
Month1,290 0.5  0.49–0.01 ± 0.02
Winter2,290 1.0  0.37–0.13 W1; –0.26 W2
Intercept   –0.75 ± 0.61

The model in Table 3 was then used to calculate how attack rate would vary from a population size of 20 Redshanks to 100 Redshanks. This was compared with the predicted number of kills over the same population range from the model in Table 1 (all other variables being set at the same average values in both models) to test whether the increase in kill rate with population size was proportional to the increase in attack rate. If success rate was independent of population size then the number of kills predicted from the attack rate should increase with population size, i.e. [(attack rate at population size 100/attack rate at population size 20) × number of kills at population size 20] should be the same as the increase in the rate of kills, i.e. number of kills at population size 100 – number of kills at population size 20. If, however, success rate is lower for large populations then the predicted number of kills from the attack rate should be greater than that simply predicted from change in population.

Finally, a general linear model (Table 4) to test the relationship between success rate and population was constructed with proportion of successful attacks per month (this was normally distributed so was not transformed) by Sparrowhawks as the dependent variable and the population of juvenile Redshanks as an independent variable, controlling for mean monthly temperature, and month and winter season.

Table 4.  Results of a GLM to test Prediction 2 – success rate depends on population size controlling for temperature and seasonal effects.
 Type III sumd.f.FPB
  1. Monthly sums or means are used (n = 22 months over three consecutive winters). B is the model parameter estimate. Dependent variable: success rate of attacks per month. The relationship between success rate and population size was not dependent on temperature (population size*temperature interaction added to the model; F1,15 = 3.6, P = 0.079).

Corrected model0.43 52.20.10 
Intercept0.13 13.20.09 1.8 ± 0.9
Redshank population0.23 160.026–0.005 ± 0.002
Temperature0.00 10.010.92 0.002 ± 0.02
Winter0.26 23.30.062–0.29 W1; –0.21 W2
Month0.09 12.20.15–0.08 ± 0.05
Error0.6216   
Total1.5722   
Corrected total1.0421   
Corrected R2 = 0.23

Statistical considerations

The proportion of Redshanks killed was calculated as the total number of juvenile Redshanks found dead at Tyninghame at the end of a month divided by the total count of juvenile Redshanks present on the estuary during the high tide count closest to the end of that month. Mean daily temperature for a month was calculated as the mean of each day's temperature at 09:00 h at a weather station at Dunbar, 4 km from the study site (and the actual daily value at 09:00 h was used for the attack rate analysis in Table 3). Data to calculate success rate of attacks from directly observed kills were limited, with only 85 kills being observed, and several months having success rate estimates based on only one or two kills.

For the models in Tables 1, 2 and 4, monthly summary data for three winters were used from 1989–90 (n = 8, September to April), 1990–91 (n = 7, September to March) and 1991–92 (n = 7, September to March). Both month since September (covariate with September = 9 and April = 16) and winter period (three-way factor) are included in all models to account for any seasonal order effects. Analysis was carried out using General Linear Models in SPSS v.15: dependent variables in the models all had distributions that did not differ significantly from a normal distribution. For the model in Table 3, data from individual days were used (n = 299 days). Total number of attacks (n = 546) and total amount of time spent observing in the field [7.3 ± 0.1 h daily (mean ± se)] were summed for each day, and monthly estimates were used for population. Analysis was carried out using Generalised Linear Mixed Models: procedure GLIMMIX in GLIM for SAS v.9, with a Poisson distribution.

Where significant interactions between covariates were identified in the models, these were further investigated by splitting one of the covariates into two classes so that 50% of observations were in each class. The effects within the two classes were then analysed separately. For example, where the relationship between kill rate and population size depended on temperature, temperature was split into two class intervals (i.e. greater than and less than 5 °C) so that an equal number of observations were in each temperature class. The relationship between kill rate and population size was then analysed separately for each class. Note that temperatures for each class vary slightly between analyses because of the different resolutions of the data (monthly means and daily readings, respectively), resulting in different numbers of observations for each temperature and so a different temperature value for the 50th percentile of observations. Note that where such analyses were carried out, and also for analyses presented in the figures, these are for illustrative purposes only, to show clearly the direction and magnitude of effects: the significance of the effects should only be assessed from the models presented in Tables 1–4. All tests are two-tailed and means are given ± 1 standard error.

RESULTS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgments
  7. REFERENCES

Prediction 1: mortality is determined by cold weather

The number of juvenile Redshanks killed decreased with decreasing temperature in all winters (F1,16 = 28.3, P < 0.001) controlling for overall size of the juvenile Redshank population (F1,16 = 4.1, P = 0.06, B = 0.14 ± 0.07), month and winter period (Fig. 1): for any given population size, 3.4 more Redshanks died each month on average for every one degree decrease in temperature. The number of juvenile Redshanks killed increased with juvenile population size, but only at relatively low temperatures (for months with temperatures < 6.4 °C F1,9 = 20.5, P < 0.001; for months with temperatures > 6.4 °C F1,9 = 2.5, P < 0.15; Table 1, Fig. 2): 2.2 more Redshanks died each month for every 10 birds increase in the population at low temperatures.

image

Figure 1. The relationship between the number of juvenile Redshanks killed per month (controlling for overall population size of juvenile Redshanks) and mean monthly temperature.

Download figure to PowerPoint

image

Figure 2. The relationship between the overall number of juvenile Redshanks killed per month and the population size of juvenile Redshanks. The relationship is temperature-dependent, with the filled squares and solid line plotted being the significant relationship for months in which temperature was less than 6.4 °C, and the open squares and the dotted line being the non-significant relationship for months in which the temperature was greater than 6.4 °C.

Download figure to PowerPoint

Prediction 2: mortality is determined by population density

The proportion of juveniles killed decreased with increasing population size, when controlling for mean monthly temperature, month and winter period (Fig. 3, Table 2): 3% less of the Redshank population died for every extra 10 birds added to the population. The proportion of juveniles killed increased with decreasing temperature, controlling for population size, month and winter period (Table 2): 2.6% fewer Redshanks died for each one degree increase in temperature. The negative relationship between proportion killed and population size was not significantly dependent on temperature (Table 2).

image

Figure 3. The relationship between the proportion of the total juvenile Redshank population killed per month, controlling for monthly mean temperature and overall population size of juvenile Redshanks.

Download figure to PowerPoint

Attack rate increased with increasing population size overall (Table 3): there were approximately 8.3 attacks per month more (assuming an 8-h day and a 31-day month) for every extra 10 birds added to the population. Attack rate increased significantly only at low temperatures (for months with temperatures greater than 5.6 °C F1,148 = 1.2, P = 0.27; for months with temperatures less than 5.6 °C F1,135 = 3.5, P < 0.001; Table 3, Fig. 4). The predicted increase in attack rate (from Table 3) from a population of 20 to 100 was 6.7 attacks, which would, assuming a constant success rate, result in an increase from 19 to 125 kills; the predicted increase in number of kills (from Table 1) over the same population range and assuming a constant attack rate was from 19 to 29 kills. Therefore, the increase in attack rate exceeded the kill rate by 10 times, suggesting that success rate decreased substantially as population size increased. The success rate of attacks of Sparrowhawks was dependent on overall population size, being greater when populations were smaller, confirming this result (Table 4). This last analysis is weak, however, because significance is dependent on a single outlier and there were several months with only a small sample size of kills.

image

Figure 4. The relationship between attack rate of Sparrowhawks per day and the overall population size of juvenile Redshanks. The relationship is temperature-dependent, with the filled squares and solid line plotted being the significant relationship for months in which temperature was less than 5.6 °C, and the open squares and the dotted line being the non-significant relationship for months in which the temperature was greater than 5.6 °C.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgments
  7. REFERENCES

We demonstrated that the overall number of Redshanks and the proportion of Redshanks killed increased in cold months. This was consistent with the hypothesis that despite the direct cause of mortality being raptors at Tyninghame (for example, over the period of the study there were only 1.0% or n = 6/587 Redshank deaths that were attributed to direct starvation; Cresswell & Whitfield 1994), the underlying cause of mortality was starvation.

Cold weather has often been linked to mortality in waders directly through starvation (Hale 1980, Davidson 1981, Davidson & Evans 1982). Deaths through starvation arise because of sudden extreme cold weather events (Beecroft & Clarke 1986) or because predators have been extirpated from the area so that predation cannot be a direct cause of mortality (Cresswell & Whitfield 1994). High mortality of waders due to raptor predation in communities where raptors are common has been described (Page & Whitacre 1975, Whitfield 1985, Buchanan et al. 1988). However, only recently has the relationship between this potential mortality and the starvation–predation risk trade-off been identified as dominating the foraging decisions of shorebirds (Pomeroy et al. 2006). Other recent studies have also shown that raptors select their prey according to vulnerability and that vulnerability arises through changes in prey behaviour linked to foraging requirements (Pomeroy 2006, van den Hout et al. 2008). The relationship between cold weather and actual mortality through increased vulnerability to predation has, to our knowledge, not been described before, because even though many studies measure behaviours that reflect variation in predation risk (Lima & Dill 1990, Lima 1998b, Caro 2005), this may not reflect mortality because of compensation via other behaviours (Lind & Cresswell 2005) and because few studies are able to measure actual mortality rates due to predation and starvation and to relate them to behaviours. However, it seems likely that starvation risk is the main impetus behind the trade-off in many, if not most, predator–prey systems and it is only the difficulty in proving this that prevents empirical data of mortality due to predators being discussed in terms of the starvation–predation risk trade-off.

Our results also show how without the constraints of cold weather, predation risk and resultant mortality may not arise. Figure 2 shows clearly how mortality dependent on population size only occurs at low temperatures. The system at Tyninghame is only obvious as a system where predation risk is important when temperatures are absolutely low: in mild winters, as with many of the more recent winters in the UK, Redshanks are rarely forced onto the saltmarsh and few die during a winter. Our results also show mortality in Redshanks to be density dependent at Tyninghame, as shown for Redshanks on a study site immediately adjacent (Whitfield 2003a), but with the caveat that this may arise only in cold winters.

We also demonstrated that the proportion of Redshanks killed increased when there are fewer Redshanks present, because success rate of hunting Sparrowhawks increased. Vulnerability of Redshanks to capture when attacked by Sparrowhawks is a direct consequence of the early warning benefits, the dilution effects and the confusion effects that accrue from being in a group (Whitfield 1988, Cresswell 1993, 1994a, 1996, Whitfield 2003b). When there is a small population of Redshanks, large flocks cannot be formed: flock sizes of Redshanks feeding on the saltmarsh can vary from one to about 100 birds, although more usually there are one or two flocks containing between 20 and 40 birds on average (Quinn & Cresswell 2004). At the end of some winters there may only be fewer than 20 individuals available to form flocks (Cresswell 1994b, Quinn & Cresswell 2004), declining from a potential saltmarsh feeding population of 200 or more birds (Cresswell & Whitfield 1994). Juvenile Redshanks, even at low population sizes, must remain in the mudflat–saltmarsh system because adult population size is not greatly changed, and so the exclusion from the safest rocky shore areas remains (Cresswell 1994b). Therefore, Redshanks are likely to be constrained in managing predation risk by forming groups. Although predation by Sparrowhawks at the spatial level of a local population is positively density-dependent (Whitfield 2003a), at the local level of the population of juvenile Redshanks forced to feed on the saltmarsh at Tyninghame, it is negatively density-dependent. In particular, large group sizes allow foraging rates to be maintained without compromising probability of detecting predators (Pulliam 1973). Small groups therefore must spend more time being vigilant, further increasing starvation risk (see also Watson et al. 2007). Larger groups may also suffer interference competition, however, reducing the benefits of flocking (Goss-Custard 1980, Vahl et al. 2005, Minderman et al. 2006). Our results suggest that the overall benefits of flocking in terms of reducing mortality outweigh the costs in terms of foraging competition: this is perhaps not surprising because Redshanks almost always seek out other Redshanks to join when foraging in areas close to cover at Tyninghame.

The result that the negative relationship between proportion of Redshanks killed and population size was not dependent on temperature was probably because temperature and population size effects on mortality act at different spatial levels. Low temperature forces Redshanks onto the saltmarsh from the rest of the estuary, and it is only when this more important effect is operating that selection with respect to group size limitations can occur within the saltmarsh. Because predation only really arises at Tyninghame when Redshanks are forced onto the saltmarsh and this only occurs in cold weather, predation with respect to group size will only occur when there is relatively little variation in temperature, and when the main effect of temperature on vulnerability is acting in a uniform way. This result helps to confirm the hypothesis that when population size is low, animals may be unable to manage predation risk effectively through group size maximization (Watson et al. 2007) and implies an Allee effect (e.g. Keitt et al. 2001).

Redshanks provide an example of how mortality through predation risk arises mainly from starvation risk and other constraints that prevent the starvation–predation risk trade-off from being resolved in favour of anti-predation behaviour. That there are not many such examples probably reflects the difficulty in studying predation and measuring mortality accurately, rather than their rarity. The study also provides an example of how the non-lethal effects of predation (see Lima 1998a, Sergio & Hiraldo 2008) can structure communities, and how an appreciation of the starvation–predation risk trade-off may often be the only way to measure the consequences of predation (Abrams 1993). Although predation does result in direct mortality at Tyninghame, this is only the case in cold weather. In warm winters there will be little direct mortality evidence of predation risk affecting Redshanks, and so the effect of predation can only be seen by determining whether Redshanks are maximizing their foraging rates in both time and space.

Acknowledgments

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgments
  7. REFERENCES

We thank Bobby Anderson, Pat Whitfield, Sue Holt, East Lothian District Council and the Tyninghame Estate for logistical help. We thank Richard Bradbury, John Quinn, Jim Reynolds, Tim Caro and a very helpful anonymous reviewer for comments on an earlier draft of this paper. W.C. is a Royal Society University Research fellow and the fieldwork here was funded by a NERC studentship.

REFERENCES

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgments
  7. REFERENCES