Fear for the family has negative consequences: indirect effects of nest predators on chick growth in a farmland bird

Authors


Correspondence author. RSPB, The Lodge, Potton Road, Sandy, Bedfordshire SG19 2DL, UK. E-mail: jenny.dunn@rspb.org.uk

Summary

1. The management of habitat structure can limit access to food and can also alter perceived predation risk. Minimising the risk of predation, by changing behaviour, may have negative impacts similar to predation itself across a wide range of species. Predation risk influences the behaviour of adults foraging for altricial young, so that they avoid disclosing the location of their offspring to predators. The consequences of these behavioural changes for offspring are unknown.

2. We investigate whether predator-induced changes in provisioning rates can have impacts upon avian nestlings through reductions in growth and condition, and whether this is influenced by resource availability, using the declining yellowhammer Emberiza citrinella as a model species.

3. We show a sizeable negative impact of nest predator activity upon brood provisioning rate, indicating that parents can assess nest predation risk and adjust their behaviour accordingly.

4. Chick condition and growth were both negatively influenced by corvid nest predator abundance and positively influenced by food abundance in large broods, suggesting that parents raising large broods in unfavourable conditions were unable to compensate fully for the effect of corvid activity on provisioning rate.

5. In areas of low food availability, corvid abundance was associated with reduced chick growth and condition; in areas of higher food availability no association was found indicating that where food resources are abundant, parents can compensate for reductions in provisioning rate when corvids are active, with no long-term implications for chicks.

6.Synthesis and applications. We propose a mechanism by which two distinct trends linked with the intensification of agriculture, namely increasing corvid abundance combined with a decreasing food supply, may have indirectly precipitated population declines in farmland passerines through delayed life-history effects across generations. As the impacts of corvids are reduced where invertebrates are abundant, we suggest that management should concentrate on improving the quality of foraging habitat by creating mosaics of long and short vegetation, rather than on the control of corvids. This will allow adult birds to compensate for the indirect effects of high corvid abundance by increasing their provisioning effort when nest predation risk is low and thus buffer any long-term consequences for nestlings.

Introduction

The modification of behaviour to minimise predation risk (‘fear’), has profound consequences for individuals, populations and ecosystems (Krebs et al. 1995; Beckerman, Uriarte & Schmitz 1997; Clinchy et al. 2004; Ripple & Beschta 2004; Cresswell 2008). Increased predation risk can alter habitat use, foraging and food provisioning behaviour, with consequences for reproductive success and long-term population dynamics (Krebs et al. 1995; Boonstra et al. 1998; Clinchy et al. 2004). The consequences of these sub-lethal or non-consumptive effects on individuals and populations may be more important than those of predation itself (reviewed by Cresswell 2008).

Resource availability is also crucial in determining foraging behaviour, and predation risk and food availability can interact to determine where animals forage (Searle, Stokes & Gordon 2008). Environments with unpredictable resources can emphasise the effects of sub-lethal predation risk on prey demography (Preisser, Bolnick & Grabowski 2009), possibly through mass-dependent predation risk (MacLeod et al. 2006). Food availability and predation risk can act synergistically to induce chronic stress in individuals (Clinchy et al. 2004) with detrimental effects upon reproductive success (Zanette et al. 2003).

In altricial systems (where offspring need considerable parental care and are typically unable to escape predators), parental activity around the nest or burrow increases the risk of predation by visually oriented predators; as a result parents may minimise predation risk by reducing their activity and consequently their ability to provide for the brood (e.g. Eggers, Griesser & Ekman 2005). The magnitude of behavioural compensation by parents is, similarly to that of predation risk on foraging adults, influenced both by habitat, with reduced provisioning to poorly concealed nests (Eggers, Griesser & Ekman 2008), and by the activity of predators, with reduced provisioning at times of high predator activity (Eggers, Griesser & Ekman 2005).

High provisioning rates are associated with high chick growth rates and good nestling condition (e.g. Takahashi et al. 2003) increasing the likelihood of fledging success (e.g. Schwagmeyer & Mock 2008). However, if parents reduce foraging behaviour in response to predation risk it may directly reduce their growth rate, but indirectly it may elongate their growth period and therefore expose them to risk for longer (Bize et al. 2003). Brood size also influences provisioning behaviour: larger broods have a greater demand for food, which may result in parents providing lower quality food (Wright et al. 1998), thus constraining nestlings’ growth. Whilst mobile animals may be able to compensate for reductions in energy intake caused by predator-induced changes in foraging behaviour (Preisser & Bolnick 2008), it remains unclear whether, in altricial systems, parents can fully compensate for predator-induced reductions in provisioning rates to offspring, especially where food resources are limiting.

Here we investigate whether predator induced changes in parental provisioning rates can have impacts on nestlings through reductions in growth and condition, and whether this is compounded by reduced food resource availability. We study a model species: the yellowhammer Emberiza citrinella (Linnaeus), a farmland passerine whose population remains in decline in the UK (Eaton et al. 2008). The majority of nest failures in this species are due to predation (Crick et al. 1994) and corvids – such as magpies Pica pica, jackdaws Corvus monedula, rooks Corvus frugilegus– constitute the main nest predator of yellowhammers in the UK (Bradbury et al. 2000). The species is also influenced by invertebrate abundance, with chick growth, body condition and fledging success all associated with the abundance of invertebrates around the nest (e.g. Hart et al. 2006).

First, we test the hypothesis that corvid activity and invertebrate abundance influence the rate at which adult yellowhammers provision their young. We assess corvid activity and also invertebrate abundance in the vicinity of nests being provisioned by adult birds. We predict that high levels of corvid activity in the immediate vicinity of the nest will be associated with low provisioning rates because adults will reduce the risk of nest predation by reducing activity around the nest whilst corvids are active. We also predict that high invertebrate abundances will be associated with high provisioning rates as adult birds maximise the food intake of their chicks.

Secondly, we test whether any changes in provisioning rate induced by corvid or invertebrate abundance have longer term impacts upon chick condition or growth. For this, we use farm-scale corvid abundance data measured across the entire breeding season as a measure of long-term activity levels (i.e. the greater the abundance of corvids in the area, the higher corvid activity should be for a greater length of time) and invertebrate abundance. In areas of high corvid abundance and low invertebrate abundance we predict that foraging adults will be unable to compensate for reductions in provisioning rate during times of high corvid activity, resulting in a decrease in both chick condition and chick growth.

Materials and methods

Sites

Fieldwork was carried out during May–July 2007 and 2008 on eighteen farms in Wiltshire, Hampshire, Gloucestershire and West Sussex, UK (ranging from 50·84°N, −0·78°W to 52·75°N, −2·75°W). Farms consisted of pairs of organic and conventional farms (herein referred to as under differing managements), grouped into clusters of four and matched according to soil type, climate, topography, land use, farm size, crop type and ratio of arable to livestock (see Gabriel et al. 2010 for full details of farm selection methods). Sixteen farms were surveyed during each year, with two farms being replaced in 2008 due to changing management practices.

Chick growth

Nests were located and monitored using standard methodology (Bradbury et al. 2000). All nests monitored had first egg dates between 6 May and 7 July and were thus considered to be first broods (Cornulier et al. 2009). Chicks were measured on two occasions when between two and seven days old, the period of linear growth for this species (Bradbury et al. 2003) and were individually marked on the leg with a non-toxic marker pen to allow the identification of each nestling. Measurements were taken of tarsus length (from the foot to the inside of the knee), mass and wing length to allow calculation of growth rate. Measurements of tarsus and wing were taken using digital callipers (±0·1 mm) and mass was measured using a pocket scale (±0·1 g; Satrue, Taichung City, Taiwan).

Food provisioning and food availability

Observations of adult foraging behaviour were carried out on between one and four occasions when chicks were between two and nine days old, depending on the stage at which nests were located, and weather conditions (2·71 ± 0·24 foraging watches per nest). The observer was positioned between 50 and 100 m from the nest to ensure the birds’ behaviour was unaffected by their presence (Stoate, Moreby & Szczur 1998). Adults were watched for an hour between 6:00 and 21:00 h and food provisioning rate was calculated as the number of complete foraging trips per hour, where a complete foraging trip is defined as a bird observed leaving the nest and subsequently returning with food. At least one morning and one afternoon watch were carried out for every nest where possible, to account for any diurnal variation in foraging patterns. Nests were not observed during heavy rain.

Invertebrate samples were collected from foraging and control sites using a leaf-vacuum (Ryobi RGBV-3100; Marlow, UK) modified by the use of a fine mesh to trap invertebrates and a 1 cm wire mesh to keep vegetation out of the sample. Sampling followed the protocol of Douglas, Vickery & Benton (2009), whereby each sampling site consisted of a 1 m square and 5 × 5 s sucks were taken from each corner and from the centre of the square. Field margins are a favoured foraging habitat of yellowhammers and the majority of foraging trips are within 200 m of their nest with 60% within 100 m (Douglas, Vickery & Benton 2009). Thus, sampling was designed to obtain a comparable measure of invertebrate abundance within potential foraging habitat. During 2007, samples were taken from ten patches within the field margin on either side of the hedgerow between 0 and 50 m from the nest at 25 m intervals. Analysis showed no significant difference in invertebrate abundance between ten samples and a subset of four samples (GLMM, χ21,17 = 0·014, P = 0·906), so in 2008 four samples were taken within the margin habitat on both sides of the hedgerow 50 m from the nest. Samples were collected immediately following each foraging watch but not if conditions were wet as this impeded the efficiency of the vacuum sampler (Hart et al. 2006). Samples were stored in 70% methanol prior to being identified to order, and the abundance of invertebrates in orders important in yellowhammer chick diet was calculated as per Hart et al. (2006), based on the previous identification of taxonomic groups known to be important in the diet of yellowhammer nestlings (Stoate, Moreby & Szczur 1998; Moreby & Stoate 2001). Invertebrates smaller than 2 mm in length were excluded from the total count as these are unlikely to be important in nestling diet (Morris & Bradbury 2002).

Predator abundance

Corvids [magpies Pica pica (Linnaeus), carrion crows Corvus corone (Linnaeus), rooks Corvus frugilegus (Linnaeus) and jackdaws Corvus monedula (Linnaeus)] are the key nest predators in agricultural landscapes (Luginbuhl et al. 2001) and the main nest predators of the yellowhammer in the UK (Bradbury et al. 2000). Corvid activity shows considerable temporal variation (Luginbuhl et al. 2001) so we used both transects and point count surveys; transects were used to provide a ‘farm-scale’ level of corvid abundance and integrate over small scale (temporal and spatial) variability in activity. Corvid abundance was therefore used to investigate the average effect of corvids on chick condition and growth, whereas point counts were used to assess the temporally relevant corvid activity for investigating how short term changes in corvids near the nests influenced parental behaviour at that time.

Transects consisted of two 1 km transects walked on each farm on three separate occasions between May and July during both 2007 and 2008, according to standard methodology (Marchant et al. 1990). Each 1 km transect was walked at a steady pace during a 1-h period, and all birds seen and heard within 250 m either side of the transect path were recorded. Point count surveys were carried out for 20 min immediately prior to each adult foraging watch to determine the total abundance of corvids within 100 m of the nest.

Statistical analysis

Linear mixed-effects models (LMMs) from the nlme library (Pinheiro et al. 2009) in R (R Core Development Team 2006) were fitted to data using maximum likelihood fits to allow model comparisons and subsequent simplification; data were transformed where necessary to meet the assumption of normality of random errors (Crawley 2007). Random effects of Chick ID within Nest within Farm, or a subset thereof, were included where appropriate to control for non-independence of repeated measures and effects of individual differences in parental quality. Model comparisons using AIC values were used to determine whether terms significantly improved the fit of the model; those that did not were removed in a stepwise fashion until only those terms that improved the fit of the model at < 0·05 remained. Following model simplification, each non-significant main-effect was reinserted into the minimum adequate model (MAM) in turn and compared with the MAM using AIC comparisons and likelihood ratio tests to ensure lack of association with the response variable and lack of influence on the fit of the model. Although model simplification by stepwise-deletion has been criticised in the literature (Whittingham et al. 2006), a recent study validated stepwise deletion as a method of model selection and established that it performed just as well as other methods of producing predictive models (Murtaugh 2009). Statistics presented throughout are mean ± 1 SE unless stated otherwise.

Parental provisioning rate

Provisioning rate was designated as the response variable in an LMM to determine whether it was influenced by corvid activity and invertebrate abundance. Corvid activity, invertebrate abundance, brood size, farm management, year, chick age and quadratic terms for temperature and time of day were included as main effects and all two-way interactions between brood size, corvid abundance and invertebrate abundance were also included in the maximal model. To determine whether any effect on provisioning rate was temporally adapted to reduce the risk of nest predation (i.e. a response to the abundance of corvids in the immediate region of the nest at that time rather than to the numbers estimated to be present in the wider landscape), farm-scale corvid abundance was substituted into the model above as a main effect in place of corvid activity and the model was rerun with the same model structure and main effects.

Chick condition and growth

Chick mass was used as a surrogate for chick condition, statistically controlling for the size of the chick (wing length) using measurement from chicks of 5–7 days old. Measurements from chicks of these ages were used because external factors were thought likely to have had time to exert an influence on mass by this age (the condition of younger chicks may be a reflection of maternal effects such as egg weight). Wing length was used, because wing length at fledging is crucial to fledging success and thus is less variable than other measures of size (Nilsson & Svensson 1996). Growth was assessed by tarsus measurements, as tarsus growth is likely to be more variable than wing growth (Nilsson & Svensson 1996). For the condition model, chick mass was designated as the response variable, with wing length as a main effect to control for chick size. For the growth model, second measurement was designated as the response variable, with first measurement and hours between measurements as main effects to control for the stage of growth and the length of time in between measurements.

In both models, corvid abundance, invertebrate abundance, brood size, parental provisioning rate, year, chick age, Julian day, farm management and time of measurement were designated as main effects; all two-way interactions between brood size, corvid abundance and invertebrate abundance, along with the interaction between chick age and provisioning rate, between invertebrate abundance and provisioning rate and between invertebrate abundance and farm management were also included in the maximal model. Examination of the raw growth data indicated a possible quadratic relationship with second tarsus length; however the fit of the model containing the linear term to the data was better than that containing the quadratic terms, so the model containing the linear term was used (AIC linear model: −321·05; AIC with quadratic terms: −318·30).

Results

Twenty-nine nests were monitored on thirteen farms during 2007 and 2008. Provisioning data were collected on between one and four occasions for each of the seventeen of these nests on twelve farms that hatched successfully (2·71 ± 0·24 foraging watches per nest). Condition data were collected from 45 chicks within sixteen nests and growth data from 42 chicks within fifteen nests. These farms, under both organic and conventional management, showed high variation in both corvid and invertebrate abundances (Table S1, Supporting Information).

Parental provisioning behaviour

Corvid point counts prior to provisioning watches were undertaken on 46 occasions for 17 nests. Corvid activity significantly influenced parental provisioning rate (Table 1) with a decrease in provisioning rate with increasing corvid activity (Fig. 1). Chick age and year both significantly influenced provisioning rate (Table 1), with higher provisioning rates to older chicks (Table 1) and in 2008 compared to 2007 (2007: 5·45 ± 0·56; 2008: 8·26 ± 1·25 trips per hour). When substituted into the previous model in place of corvid activity, corvid abundance had no effect upon provisioning rate (GLMM, F1,24 = 0·30, P = 0·62).

Table 1.    (a) Results of a linear mixed-effects model determining which factors influence the rate at which yellowhammer parents provision their chicks. Back-transformed parameter estimates are presented for significant terms. (b) Statistics for non-significant terms follow reinsertion of the term of interest into the minimum adequate model (MAM)
(a) VariabledfFPEstimateSE
Insect abundance1, 240·2950·592 0·0110·005
Corvid activity1, 248·4590·008−0·0190·008
Brood size1, 242·2110·150 0·7100·200
Chick age1, 245·8200·024 0·0880·048
Year (2008)1, 47·9410·048 0·7130·193
Insect abundance × Brood size1, 243·6840·067*−0·0040·002
(b) VariabledfLRTP
  1. All models contain random effects of Nest ID within Farm ID to control for effects of parental quality and localised geographical variation. *Model comparisons indicated that this term significantly improved the fit of the model (LRT1 = 4·152, P = 0·042) and thus this term remains in the MAM but is not considered to significantly influence the response variable. Two-way interactions of invertebrate abundance × corvid activity (LRT1 = 0·051, P = 0·822) and corvid activity × brood size (LRT1 = 0·258, P = 0·612) did not significantly improve the fit of the model or influence the response variable and thus were removed from the model.

Temperature211·3680·242
Time of day210·3930·531
Management11·9700·160
Figure 1.

 Parental provisioning rate decreases with increasing corvid activity. Points show raw data; solid line is predicted provisioning rate from a MAM (Table 1) with mean chick age (4·76 days), brood size (2·85 chicks) and invertebrate abundance (85·70 invertebrates) during 2008; dashed lines show 95% CIs. Note log axes, and that values for both axes have been increased by 1 to avoid omission of zero values.

Chick mass and growth

Chick growth was influenced by all two-way interactions between corvid abundance, invertebrate abundance and brood size (Table 3 and Fig. 3), and chick mass was influenced by two-way interactions between corvid abundance and brood size, and invertebrate abundance and corvid abundance (Table 2 and Fig. 2). Large broods showed reduced mass and size growth with increasing corvid abundance, and increased growth with increasing invertebrate abundance, whilst small broods showed the opposite trends (Tables 2 and 3; Figs 2a and 3a,b). Chick mass and growth decreased with increasing corvid abundance; however, this effect was reduced in areas of high invertebrate abundance (Tables 2 and 3; Figs 2 and 3). Chick mass was also influenced by interactions between chick age and provisioning rate, and invertebrate abundance and provisioning rate, and by the time of day at which chicks were measured (Table 2). Chick growth was also influenced by an interaction between invertebrate abundance and farm management (Table 3).

Table 3.    (a) Results of a linear mixed-effects model determining which terms influence chick growth. Back-transformed parameter estimates are presented for significant terms. (b) Statistics for non-significant terms follow reinsertion of the term of interest into the minimum adequate model (MAM)
(a) VariabledfFPEstimateSE
1st tarsus measurement1, 105646·31<0·0010·7480·054
Management (Organic)1, 252·59<0·0011·3860·704
Hours between measurements1, 10539·450·0240·0570·017
Brood size1, 1052·020·1581·0580·441
Corvid abundance1, 90·010·9220·0290·007
Invertebrate abundance1, 29·790·002−0·0540·030
Day1, 1050·32  0·574*0·0230·008
Management × Invertebrate abundance1, 1058·210·005−0·0150·002
Brood size × Corvid abundance1, 10529·59<0·0010·0190·007
Corvid abundance × Invertebrate abundance1, 1056·050·0160·0010·001
Invertebrate abundance × Brood size 1, 1055·590·020−0·0480·016
(b) VariabledfLRTP
  1. All models contain random effects of Nest ID within Farm ID to control for effects of parental quality and localised geographical variation. *Model comparisons indicated that this term significantly improved the fit of the model (LRT1 = 8·38, < 0·01) and thus this term remains in the MAM but is not considered to significantly influence the response variable. Two-way interactions of invertebrate abundance × provisioning rate (LRT1 = 0·06, P = 0·81) and chick age × provisioning rate (LRT1 = 0·01, P = 0·92) did not significantly improve the fit of the model or influence the response variable and thus were removed from the model.

Year10·1030·748
Chick age10·1690·681
Provisioning rate10·1050·746
Time of measurement10·8710·351
Figure 3.

 Chick growth was influenced by interactions between brood size and (a) corvid abundance and (b) invertebrate abundance, as well as between (c) corvid abundance and invertebrate abundance. Points represent brood averages of raw data ± 95% CI; randomly generated values between 0 and 10 have been added to x values in order to distinguish between points with the same x value. Lines are predicted from a MAM (Table 3) with mean values for 1st tarsus length (10·13 mm), hours between measurements (46·11), invertebrate abundance (46·04 invertebrates), corvid abundance (93·32 corvids), and brood size (3·15 chicks) on farms under conventional management. Dashed lines represent 95% CI.

Table 2.    (a) Results of a linear mixed-effects model determining which terms influence chick mass. Back-transformed parameter estimates are presented for significant terms. (b) Statistics for non-significant terms follow reinsertion of the term of interest into the minimum adequate model (MAM)
(a) VariabledfFPEstimateSE
Wing length1, 501283·01<0·001−0·2370·081
Provisioning rate1, 505·370·0220·4320·107
Chick age at measurement1, 50362·90<0·0012·8200·195
Time of measurement1, 50126·93<0·0010·1850·017
Invertebrate abundance1, 506·630·005−0·1560·063
Brood size1, 80·790·409−0·0821·200
Corvid abundance1, 85·970·0430·0360·021
Age × Provisioning rate1, 5021·99<0·001−0·0670·017
Invertebrate abundance × Brood size1, 503·14  0·081*0·0550·019
Invertebrate abundance × Provisioning rate1, 508·460·012−0·0010·001
Corvid abundance × Brood size1, 86·840·034−0·0240·008
Invertebrate abundance × Corvid abundance1, 504·420·0480·0010·001
(b) VariabledfLRTP
  1. All models contain random effects of Chick ID within Nest ID to control for effects of parental quality. *Model comparisons indicated that this term significantly improved the fit of the model (LRT1 = 8·38, < 0·01) and thus this term remains in the MAM but is not considered to significantly influence the response variable. The invertebrate abundance × farm management interaction (LRT1 = 2·85, P = 0·09) did not significantly improve the fit of the model or influence the response variable and thus was removed from the model.

Year10·4400·507
Julian day10·8420·359
Management13·0280·082
Figure 2.

 Interactions between (a) brood size and corvid abundance and (b) corvid abundance and invertebrate abundance both influenced chick mass when controlling for size. Points show brood averages of raw data ± 95% CI; randomly generated values between 0 and 10 have been added to x values in order to distinguish between points with the same x value. Lines are predicted from a MAM (Table 2) with mean values for wing length (14·96 mm), provisioning rate (8·46 trips per hour), chick age (5·67 days), time of day (13:41) (a) invertebrate abundance (51·0 invertebrates) and (b) corvid abundance (93·1 corvids). Dashed lines represent 95% CI.

Discussion

We investigated whether nestlings were influenced by indirect predator–prey interactions altering the behaviour of their parents, because if such effects do exist they may be contributory to population declines. We addressed this using a declining species known to be under pressure from reductions in food availability (Morris et al. 2005; Hart et al. 2006) and subject to increased exposure to nest predators (Gregory & Marchant 1995; Eaton et al. 2008). Although this study was conducted on a single species, given the known impacts of predator behaviour on the foraging of a wide range of organisms, from grasshoppers (Beckerman 2002) to mammals (Thomson et al. 2006), it is likely that similar food-predator interactions are widespread in ecology. Corvids are generalist nest predators of a wide range of birds (passerines and waders, ground and hedge nesters alike) so it is likely that this result is generally applicable.

We show a strong negative correlation between predator activity and parental provisioning rate. As corvids are visually oriented nest predators (Eggers, Griesser & Ekman 2005), activity around the nest when corvids are abundant increases the risk of nest predation (Skutch 1949; Martin, Scott & Menge 2000). Our results indicate that adult yellowhammers are able to assess the risk of nest predation from corvids, and reduce their provisioning rate when the risk to the nest is high, as has been shown in other species (e.g. Eggers, Griesser & Ekman 2005). The lack of association between corvid abundance and parental provisioning rate adds further weight to this argument, indicating that behavioural compensation varies temporally, and that parents are able to assess the current risk to their nest and adjust their provisioning behaviour accordingly.

Invertebrate abundance estimated in the locality of the nest had no effect on provisioning rate, indicating that either resources may not have been limiting, or sampling variance obscured any underlying relationship, or that reducing the risk of nest predation takes precedence over ensuring that chicks are well fed. The latter possibility seems more plausible, as previous studies have shown invertebrate abundance to be critical in influencing chick growth, body condition and fledging success (e.g. Hart et al. 2006). Conversely, there is no evidence to suggest that nest predation rates increased concurrently with corvid population increases (Baillie et al. 2009), implying that behavioural compensation for corvid nest predation risk may be highly sensitive to changes in corvid abundance and thus independent of food availability.

Chick mass and growth were both influenced by interactions between brood size and both corvid abundance and invertebrate abundance. Large broods showed decreased condition and growth with increasing corvid abundance, and increased condition and increased growth with increasing invertebrate abundance. Conversely, small broods show the opposite trends. That corvid abundance influences nestling growth, measured over a period of days, implies that parents are unable to compensate for the predator-induced reduction in provisioning rate, either through increasing provisioning rate when corvid activity is reduced, or by increasing load size (Eggers, Griesser & Ekman 2005, 2008), at least when demand from large broods is high. This indicates that resources are limiting and is further supported by the interactive effect of invertebrate availability and corvid availability on chick growth. In areas of low food abundance, growth decreases with increased corvid abundance, indicating that food availability limits parents’ ability to increase provisioning rate when nest predation risk is low. However, in areas of high food abundance, and in small broods, growth appears to increase with corvid abundance. This result appears counterintuitive; however, it is possible that where food is abundant but access to the nest to provision chicks is mediated by corvid activity and thus unpredictable, that parents may adopt a strategy to reduce the amount of time chicks are exposed to nest predation risk. They may do this both by increasing load size when provisioning rate is reduced (Eggers, Griesser & Ekman 2008), and by increasing provisioning rate when corvid activity is low (Eggers, Griesser & Ekman 2005). This would allow the maximum possible growth of chicks in order that they fledge quickly and are thus removed from the risk of nest predation.

Food availability and predation risk have been found to interact to influence reproductive success in both mammals (Krebs et al. 1995) and birds (Clinchy et al. 2004). This study both supports this interaction at the nest level, and provides evidence for a behavioural mechanism, mediated by parent birds, by which this effect may occur. Previous studies (Krebs et al. 1995; Clinchy et al. 2004) demonstrate these effects through predator exclusion and food addition, with consequent increases in reproductive success. Here, we suggest that two distinct factors associated with changes in farmland management: a decline in invertebrate food availability and an increase in nest predator abundance; may act detrimentally, both separately and interactively, to influence chick biology in passerines.

Management recommendations

This study is, to our knowledge, the first to demonstrate non-lethal impacts of nest predation risk on nestlings resulting from predator-induced reductions in provisioning rate by parent birds: the implications of this for management are likely to be significant. Yellowhammers have undergone considerable population declines in the U.K. since the onset of agricultural intensification, coincidental with, if slightly behind, the increase in populations of corvids (Baillie et al. 2009). Whilst increasing magpie numbers showed no relationship with nest failure rates in songbirds (Gooch, Baillie & Birkhead 1991), our results indicate that the fear of nest predation influences parental behaviour with negative consequences for chick growth and condition. There is ever-increasing evidence suggesting that conditions early in life determine the life-history trajectory of an individual (Metcalfe & Monaghan 2001; Taborsky 2006), whereby individuals who experience a reduced growth rate or undergo compensatory growth in the nest suffer a reduced body size or lifespan in adulthood (reviewed in Metcalfe & Monaghan 2001). Since the beginning of the yellowhammer population decline, breeding productivity has almost doubled (Cornulier et al. 2009), suggesting that parents may be protecting the survival of their nestlings at a cost to their growth and quality, and producing more broods per year potentially impacting upon adult survival.

It is conceivable, therefore, that in areas of high corvid density, nestlings experiencing reduced growth rates and poor condition whilst in the nest have lower survival rates post-fledging and a lower reproductive success should they survive to breed, leading to a long-term reduction in recruitment to the breeding population and a consequent population sink. As yellowhammer territory numbers are restricted partially by the availability of suitable habitat (Bradbury et al. 2000), territories within high corvid areas, but where invertebrate abundance is low, may act as ecological traps resulting in long-term population declines. Changes in demography – such as those we report – may not necessarily feed through into changes in population sizes; however, in a companion study using the same farms, a negative relationship is reported between specialist farmland bird counts and corvid densities (Gabriel et al. 2010). This is at least suggestive that indirect interactions, driven by risk avoidance, contribute to population declines.

Whilst corvid control might seem like the obvious management strategy to combat the effects reported here, our results also show that where invertebrates are abundant, the impact of corvids is negated. Therefore we suggest that management should concentrate on increasing the value of available foraging habitat to allow birds to increase their foraging rate when necessary, rather than concentrating management on corvid control.

In the U.K., the agri-environment options currently available include buffer strips along hedgerows and around habitat features such as ponds and watercourses (Anon 2010). However, once established, the cutting of these buffer strips, or margins, is only permitted every two years, or twice yearly outside the avian breeding season if flowers are sown in the sward (Anon 2010). In practice, this provides suitable habitat for invertebrate populations, but the dense sward encouraged by this management denies accessibility to foraging birds. Ground-foraging bats also forage less over dense grass cover, even when invertebrate abundance is higher, as they have reduced foraging success in denser vegetation (Rainho, Augusto & Palmeirim 2010). Similarly, birds switch foraging habitats from the buffer strips to cropped habitat mid-breeding season when margins become too dense to allow efficient foraging, despite the greater abundance of invertebrates in the margins compared to cropped habitat (Douglas, Vickery & Benton 2009).

Thus, we recommend an addition to the current management prescription of field margins to encourage the creation of a mosaic of long and short patches within these margins. These patches have been shown to be utilised by foraging passerines (Douglas, Vickery & Benton 2009), and would reduce food constraints that restrict parental provisioning rate in areas of high corvid activity. This would enable birds in areas of high corvid abundance to compensate for reductions in food provisioned to nestlings during times of peak corvid activity. Increasing the accessibility of invertebrate food would benefit a wide range of bird species that rely upon invertebrates during the breeding season.

Acknowledgements

Thanks to Doreen Gabriel for help with field sites, enthusiasm during rainy spells, and the farm-scale corvid abundance data. Thanks to the RELU field assistants for their help with fieldwork and conveying their yellowhammer sightings. Thanks to Debbie Russell for her help and advice with figures. JCD was supported by BBSRC Studentship BBSSK200512132.

Ancillary