Improving the value of field margins as foraging habitat for farmland birds

Authors

  • David J. T. Douglas,

    Corresponding author
    1. Institute of Integrative and Comparative Biology, University of Leeds, LS2 9JT, UK
    Search for more papers by this author
    • Present address: BTO Scotland, Cottrell Building, University of Stirling, FK9 4LA, UK.

  • Juliet A. Vickery,

    1. BTO, The Nunnery, Thetford, Norfolk IP24 2PU, UK
    Search for more papers by this author
    • Present address: RSPB, The Lodge, Sandy, Bedfordshire, SG19 2DL, UK.

  • Tim G. Benton

    1. Institute of Integrative and Comparative Biology, University of Leeds, LS2 9JT, UK
    Search for more papers by this author

*Correspondence author. E-mail: david.douglas@bto.org

Summary

  • 1Uncropped field margins are important foraging habitats on farmland for many declining bird species and are a key component of Agri-Environment Schemes across Europe. Maximizing the value of foraging habitats requires detailed knowledge of the factors influencing habitat selection and food availability.
  • 2We firstly conducted an observational study of foraging habitat selection by breeding yellowhammers Emberiza citrinella L. on lowland mixed farmland, in relation to underlying invertebrate and vegetation characteristics.
  • 3There was a clear seasonal shift in the relative use of field margins and cereal crops. Margins were used less than crops in late summer, despite supporting higher invertebrate abundance relative to cereals. Seasonal increases in vegetation height were most marked in margins, suggesting the seasonal decline in margin use may reflect reduced food accessibility.
  • 4In the second phase of the study, field margins were cut experimentally to create short, open patches within taller margin swards. The use of cut patches by foraging yellowhammers increased significantly between early and late summer, and patches were used more frequently with increasing height of adjacent uncut margin. These findings strongly support the theory that tall vegetation reduces margin accessibility in late summer.
  • 5Synthesis and applications. Provision of invertebrate-rich field margins is a core component of Agri-Environment Schemes, but current prescriptions may result in them having limited value in late summer. More effective management, such as more frequent cutting, may be required to maximize the benefits for foraging birds by creating short, open vegetation patches. Measures to increase accessibility to invertebrates on farmland are likely to benefit a range of bird species across a variety of crop types.

Introduction

A key impact of changing farming practices on bird populations in Europe and North America has been a reduction in the availability of food across the farmed landscape. Declining food resources have been implicated in the severe population declines of a wide range of species (Robinson & Sutherland 2002) and enhancing food availability is widely advocated as a key means of promoting population recovery (Newton 2004; Vickery et al. 2004). In Europe, this is achieved mainly through Agri-Environment Schemes (AES) providing payments to landowners for sympathetic management of foraging habitats (Vickery et al. 2004). An important component of such measures is the provision of invertebrate-rich foraging habitat during the breeding season, essential in ensuring nestling growth and productivity for a range of species (e.g. Potts 1986; Brickle et al. 2000; Macleod et al. 2005). One popular option across a number of European AES is the provision of uncropped field margins supporting high invertebrate resources relative to cropped fields (Kleijn et al. 1998; Vickery, Carter & Fuller 2002).

However, deriving optimum benefits from such measures requires detailed knowledge of the factors influencing their use by foraging birds. We investigated foraging habitat selection of the yellowhammer, a species sharing similar breeding season foraging and dietary requirements with other declining farmland passerines such as corn bunting Miliaria calandra L. and skylark Alauda arvensis L. (Brickle et al. 2000; Wilson 2001; Hart et al. 2006).

The study was conducted in two phases: firstly, an observational study determined how selection of broad-scale foraging habitats (crop, margin, etc.) varied in relation to (i) invertebrate abundance, (ii) vegetation characteristics and (iii) stage in the breeding season. Secondly, we aimed to investigate experimentally whether a seasonal decline in margin use by foraging yellowhammers was due to vegetation growth reducing prey accessibility. This has been proposed because agricultural intensification is known to have resulted in a general decline in sward structural heterogeneity within and between habitats (Benton, Vickery & Wilson 2003) and this will, in turn, reduce the availability of invertebrate prey for foraging birds (Wilson, Whittingham & Bradbury 2005). Enhancing sward heterogeneity has shown positive results for ‘in-field’ foragers such as skylarks (Morris et al. 2004) but the benefits of manipulating margins to enhance food accessibility remain largely untested (although see Potts et al. 2007). Shorter swards may enhance prey accessibility and mobility for foraging birds (Butler & Gillings 2004; Devereux et al. 2004; Stillman & Simmons 2006), whilst lowering perceived predation risk (Whittingham & Evans 2004). However, short swards are generally associated with lower abundance and diversity of invertebrates compared to taller more complex vegetation, creating a trade-off between prey abundance and accessibility (Vickery et al. 2001; McCracken & Tallowin 2004). We therefore manipulated accessibility within margins by creating a mosaic of short and long vegetation and predicted that this would enhance margin use by foraging birds, especially late in the breeding season when uncut vegetation was tallest. Given the importance of field margins as foraging habitat for a wide range of bird species during the summer (e.g. Rands 1986; Brickle et al. 2000; Wilson 2001; Perkins et al. 2002), measures to improve their value could have considerable benefits for farmland bird conservation. We consider the implications of our findings for the cost-effectiveness and development of margin options within AES and in relation to the wider issue of food accessibility for foraging birds on farmland.

Methods

study sites

Fieldwork was conducted from May–August 2004–2006 on five farms in Aberdeenshire (56°54′ N, 2°16′ W), ranging in size from 65 to 247 ha. All sites were lowland mixed farmland typical of the region and the dominant habitat types by area were cereals [39%; spring barley, winter barley and winter wheat (comprising 59%, 27% and 14% of cereal area respectively)], grass (30%; grazed, silage and hay), set-aside (11%) and uncropped field margins (6%). Field sizes (8·4 ± 0·7 ha, n = 79) were consistent with mixed farming systems elsewhere (e.g. central-southern England: 9·3 ± 0·6 ha, n = 87; D. Gabriel, unpublished), although the ratio of winter- to spring-sown crops may be lower than in other regions (Robinson & Sutherland 2002, but see Discussion). Field margins were defined as the strip of uncropped semi-natural boundary habitat separating cropped fields. On the study sites, these comprised mainly permanent or semi-permanent grass margins adjacent to the crop edge (varying in width from 1–12 m), mostly established by natural regeneration, with a small number sown with grass seed within agri-environment schemes. Additional margin habitat along some field boundaries comprised small areas of ditches, hedgerow and roadside verge. All margin manipulations described below were conducted on grass margins adjacent to the crop edge.

invertebrate and vegetation sampling

Previous studies indicated that field margins and cereal crops are two of the most frequently used foraging habitats by yellowhammers on arable and mixed farmland (Stoate, Moreby & Szczur 1998; Morris et al. 2001; Perkins et al. 2002). Invertebrate and vegetation characteristics were therefore sampled from five fields each of spring- and winter-sown barley (the dominant cereal crops on the study sites) and the adjacent margins, at monthly intervals (range, 9th–21st per month) from May–August 2004. Fields were selected randomly from a subset of those which could feasibly be sampled (i.e. with a width < 300 m) at the required time intervals, distributed so that two and three fields respectively per crop type were selected from each of two different farms. Sampling was conducted along parallel transects across the entire width of each field, perpendicular to the longest boundary of the field and spaced 100 m apart. Transects ranged from 2–5 in number and 82–253 m in length, reflecting differences in field size. Sampling was conducted at the same points along each transect on each visit. Each sampling point was treated as a 1-m2 patch, marked with coloured tape around several crop stems. Sample points were located in the margin, 1 m into the crop, then at every 30-m interval across the field and again in the opposite margin. Invertebrates were sampled using a modified petrol-motor leaf vacuum (Ryobi RGBV-3100, Marlow, UK) with a 15-cm diameter aperture. Suction sampling was considered to be the single most efficient method of sampling the range of arthropod taxa important in yellowhammer nestling diet (Thomas & Marshall 1999). One invertebrate sample was taken at each sampling point on each visit, comprising 5 × 5 s sucks (as per Hart et al. 2006), one from each corner and one in the centre of the sampling patch, combined into a single sample and frozen the same day. Invertebrates were later identified as to order. Vegetation height was used as a surrogate for sward accessibility (see Discussion) and was recorded at each sampling point on each visit, taken as the mean of five measurements, each recording maximum height to the nearest 5 cm within a 5-cm radius of the measuring stick.

observational foraging habitat use

Foraging behaviour of breeding yellowhammers was studied from May–August in 2005. Nests were located by mapping territorial males across study sites and watching for signs of breeding behaviour. One 3-h foraging watch was conducted at 30 nests with nestlings aged 6–11 days. Observations were carried out from 07·00–11·00 h, coinciding with the period of greatest daily provisioning activity (D.J.T. Douglas, unpublished). Periods of adverse weather (e.g. heavy rain), were avoided. The locations of all adult foraging flights were recorded on sketch maps using prominent features such as tall weeds or fence posts to accurately re-locate foraging sites following watches. A separate foraging site was defined as one with an ‘entry point’ at least 20 m from another foraging site (following Morris, Bradbury & Wilson 2002). If a parent visited multiple locations prior to returning to the nest (< 1% of flights), the last visited was recorded as the foraging site (as per Morris et al. 2001) and multiple visits to the same site were recorded. Flights were discounted if the exact location could not be determined (c. 11% of flights per watch). However, based on the general direction of these missed flights, 79% were almost certainly only visited once during a watch, and it is unlikely that flights to any particular habitat type(s) were missed more frequently than others. Following a watch, each foraging site was visited and the distance from the nest measured using a hand-held GPS (accuracy ±5 m).

experimental margin manipulation

In 2006, field margins surrounding yellowhammer nests were manipulated to create cut areas within the foraging radius. Margin cutting was conducted at 30 nests containing nestlings aged 0–8 days old, selected arbitrarily using the first 15 with nestlings in early and late summer respectively (see Table 1 for dates). Ten patches each measuring 15 × 1 m (five patches in each of two separate margins selected at random) were cut around each nest. Within each margin, patches were cut at the following distances from the nest; 30–45 m, 60–75, 90–105, 120–135 and 150–165. Where cutting at the required distance was not possible due to features such as gates or tracks crossing the margin, the patch was cut in the nearest available length of margin. Vegetation in patches was cut down to the soil–vegetation interface using a petrol-motor brush-cutter (Stihl KM85R/KM-MB, Camberley, UK), and the cut vegetation was raked off. Each patch was marked with a uniquely coloured 2-m long cane at either end, to identify patches during foraging watches. The height of uncut vegetation in the margin immediately adjacent to each patch was measured (mean of six measurements, each recording maximum height to the nearest 5 cm within a 5 cm radius of the measuring stick; one in each corner and at each midpoint of the longest edges).

Table 1.  Structure of GLMMs used in analyses
Response variableFixed effects (italics denote factor levels)InteractionRandom termError structureLink functionOffset
  • *

    to control for non-independence of repeated visits. Including farm as an additional random term (with field nested within farm) did not significantly improve model fit.

  • Coleoptera, Diptera, Lepidoptera, Araneae (following Stoate et al. 1998; Macleod et al. 2005).

  • set-aside, grass fields, oilseed rape, potatoes, woodland/scrub, houses/farmyards, tracks.

  • §

    each nest assigned to either early or late season according to the midpoint between the earliest and latest foraging watches in 2005 (early, 20 May–2 July; late, 3 July–14 Aug).

  • nest identity used in place of individual bird ID, as not always possible to distinguish between the sexes within each pair at a nest.

  • **

    the 95th percentile of all foraging flights in 2005.

  • ††

    use of the offset expressed the number of visits to each habitat type as foraging densities (after Morris et al. 2001).

  • ‡‡

    the radius of cut patches at a nest.

(a) Invertebrate abundance and vegetation height in foraging habitats
  Vegetation height at individual sample pointsHabitatHabitat × monthField identity*NormalIdentityNone
Margin     
Spring barley     
Winter barley     
Month     
May     
June     
July     
August     
  Total invertebrate abundance at individual sample pointsHabitatHabitat × monthField identity*PoissonLogNone
Month     
  Chick-food abundance at individual sample pointsHabitatHabitat × monthField identity*PoissonLogNone
Month     
(b) Foraging habitat selection
  Number of visits to each habitat per nestHabitatHabitat × seasonNest identityPoissonLogLoge (area) per habitat within 333 m** foraging radius per nest††
Margins    
Cereals    
Other    
Season§    
Early summer    
Late summer    
(c) Experimental margin use
  Number of visits to each margin type per nestMargin typeMargin type × seasonNest identityPoissonLogLoge (area) per margin type within 165 m‡‡ radius per nest††
Cut    
Uncut    
Season§    
Early summer    
Late summer    

A minimum interval of 3 days (mean 4·3 ± 0·3 days, range 3–7) was left between cutting patches and conducting foraging observations at each nest. This ensured any selection of cut patches was not due to exploitation of temporary insect flushes following cutting (Vickery et al. 2001). A single foraging watch was conducted on each nest containing nestlings aged 4–11 days using the protocols described above. Foraging sites in margins were recorded as cut or uncut.

areas of foraging habitats

For the observational study, habitat types surrounding each nest were mapped onto Ordnance Survey 1:10 000 maps in three categories: field margins, cereals and ‘other’ (see Table 1 for definitions). The area of each habitat type within a 333-m foraging radius of each nest (the 95th percentile of all mapped flights in 2005) was calculated. For the experimental margin study, the area of cut and uncut margin within a radius corresponding to the distance of the farthest cut patch from each nest (mean 162 ± 8 m) was calculated.

analysis

A number of Generalized Linear Mixed Models (GLMMs) were constructed in r version 2·3·1 (r Development Team 2006) to examine seasonal variation in (i) invertebrate abundance and vegetation height in margins and cereals in 2004; (ii) observational foraging habitat use in 2005; and (iii) experimental margin use in 2006. Model structures are summarized in Table 1. The significance of terms was assessed by removal one at a time and assessing the change in deviance as chi-square. The GLMM of vegetation height was assessed for normality of the residuals. The fit of GLMMs assuming a Poisson distribution was assessed using the ratio of residual deviance : residual d.f.; no adjustments for overdispersion were required (Crawley 2002).

All models described above revealed significant interactions between fixed effects (‘habitat’ and either ‘month’ or ‘season’) (see Tables 2–4), complicating the interpretation of differences between levels of the respective habitat factor. We therefore developed post hoc contrasts based on rigorous non-parametric resampling methods as follows. For the models of invertebrate abundance and vegetation height (Table 1a), the primary aim lay in determining differences between margins and cereal crops within each month. We constructed bootstrapped 95% confidence intervals (CI; employing 10 000 permutations) around the mean differences between margins and either spring or winter barley, at each month. Significant differences (at α = 0·05) were indicated by CI not overlapping zero (Nakagawa & Cuthill 2007). For observational habitat use (Table 1b), the primary interest concerned differences in usage of each habitat between early and late summer. As the total number of foraging visits recorded differed between nests, data were converted to proportional usage of each habitat per nest. Mean differences between proportional use of each habitat in early and late summer were assessed using bootstrapped CI as above. For the model of experimental margin use (Table 1c), the interest was in determining how the use of cut margins differed between early and late summer. The number of visits to cut patches was converted to the proportion of total margin visits per nest, and the mean difference between early and late summer assessed by bootstrapping CI as above.

Table 2.  Outputs of GLMMs examining invertebrate abundance and vegetation height between margins, spring barley and winter barley sampled at monthly intervals from May–August. Response variable was insect abundance or vegetation height at individual sample points at each visit with field identity (n = 10) as random term
Termd.f.DevianceMean devianceDeviance ratioP
(a) Total invertebrate abundance
  Habitat24291·302145·651663·29< 0·001
  Month33248·901082·97 839·51< 0·001
  Habitat × month61272·70212·12 164·43< 0·001
  Residual17242223·961·29  
  Total173511 036·86   
(b) Chick-food abundance
  Habitat21641·40820·70 661·85< 0·001
  Month32454·80818·27 659·89< 0·001
  Habitat × month6893·20148·87 120·05< 0·001
  Residual17242137·761·24  
  Total17357127·16   
(c) Vegetation height
  Termd.f.Wald statisticP  
  Habitat2250·33< 0·001  
  Month3651·16< 0·001  
  Habitat × month62808·50< 0·001  
  Residual1724    
  Total1735    
Table 3.  Deviance table for GLMM examining seasonal variation (early or late summer) in foraging habitat use by yellowhammers provisioning nestlings. Response variable was the number of visits to each habitat (margins, cereals and ‘other’) per nest, with nest identity (n = 30) as random term
Termd.f.DevianceMean devianceDeviance ratioP
Habitat 2187·7693·8871·66< 0·001
Season 1  0·70 0·70 0·530·404
Habitat × season 2108·7954·4041·52< 0·001
Residual84110·04 1·31  
Total89407·29   
Table 4.  Deviance table for GLMM examining seasonal variation (early or late summer) in foraging use of experimentally manipulated field margins by yellowhammers provisioning nestlings. Response variable was the number of visits to each margin type (cut or uncut) per nest, with nest identity (n = 30) as random term
Termd.f.DevianceMean devianceDeviance ratioP
Margin type 1149·48149·48134·67< 0·001
Season 1  4·70  4·70  4·230·030
Margin type × season 1 54·76 54·76 49·33< 0·001
Residual56 62·16  1·11  
Total59271·10   

An additional analysis of experimental margin use was conducted to examine the frequency of visits to cut patches in relation to the height of adjacent uncut margin. Mean vegetation heights adjacent to each patch were assigned to five categories (0–30 cm, 31–60 cm, 61–90, 91–120 cm, 121–150 cm). The expected frequency of visits (the number of cut patches in each height category) was compared to the observed frequency (number of visits to patches in each height category) using chi-square. Means are presented as mean ± 1SE unless otherwise stated.

Results

invertebrates and vegetation height in cereals and margins

There was a trend for total invertebrate abundance to increase in all habitats from May–July and then decline in August (Fig. 1a). Differences in total abundance between habitats varied seasonally (significant habitat × month interaction, Table 2a). A consistent pattern was that field margins supported significantly higher total abundance than either spring or winter barley across the entire summer (Fig. 1a). Total invertebrate abundance peaked in all habitats in July, when the mean count per sample in margins (45·3 ± 2·1 invertebrates) was nearly twice that of either cereal crop (spring barley = 27·5 ± 0·9 invertebrates; winter barley = 23·1 ± 0·8 invertebrates). A similar pattern was found when considering important chick-food arthropods, with significantly higher abundance in margins relative to both cereal crops in every month (Fig. 1b; Table 2b). These differences in chick-food abundance between habitats were lowest in July (mean difference between margins and spring barley = 3·8 invertebrates, 95% CI = 0·9–6·9 invertebrates; difference between margins and winter barley = 6·8, 95% CI = 3·8–9·8 invertebrates), although these were due mostly to a peak in Coleoptera in cereals, in particular spring barley (Fig. 1c). Abundances of the other arthropod orders comprising chick-food also varied seasonally between habitats (Fig. 1d–f).

Figure 1.

Invertebrate abundance and vegetation height in primary foraging habitats of breeding yellowhammers. Plots c–f represent the four arthropod orders comprising yellowhammer chick-food. Data were collected from five fields of each crop type and adjacent margins at monthly intervals during the breeding season. Shown are means with bootstrapped 95% confidence intervals. Significance tests between habitats at each month were conducted by bootstrapping differences between means (*, significant difference between margins and spring barley; +, margins vs. winter barley).

Vegetation height increased in all three habitats over the summer (Fig. 1g). Winter barley was harvested in mid-August. The difference in height between habitats varied significantly by month (habitat × month interaction, Table 2c). In early summer (May and June), vegetation height was significantly greater in margins than spring barley, but lower than or comparable with winter barley (May and June, respectively, Fig. 1g). By July, margins were significantly taller than either cereal crop (mean difference between margins and spring barley = 28·9 cm, 95% CI = 23·4–34·4 cm; difference between margins and winter barley = 20·2 cm, 95% CI = 14·6–25·7 cm). This pattern was mirrored in August, in particular for margins relative to winter barley, which was harvested in mid-August (Fig. 1g).

observational foraging habitat use

A total of 739 foraging flights to known destinations were recorded in 2005 (range 15–45 flights to 7–21 different sites per watch). In early summer (20 May–2 July; 233 flights, n = 10 nests), field margins were used heavily (32·4 ± 7·0% of visits per nest), with little use of cereal crops (7·9 ± 3·9%; Fig. 2a). In late summer (3 July–14 August; 506 flights, n = 20 nests), use of field margins declined markedly (15·4 ± 3·4% of visits per nest), whilst use of cereal fields increased (55·8 ± 7·2%; Fig. 2b). This change in overall habitat use between early and late season was significant (habitat × season interaction, Table 3); with a significant decrease in field margin use (mean difference between proportional use of margins per nest in early and late summer 17·0 with 95% CI = 4·1–31·2) and a significant increase in cereal use (mean difference 47·9 with 95% CI = 31·7–63·0). There was also a significant seasonal decrease in use of ‘other’ habitat types (mean difference 30·9 with 95% CI = 10·3–51·5), but as this category consists of numerous habitat types each visited relatively infrequently the pattern is difficult to interpret and we do not consider it further.

Figure 2.

Foraging habitat use by breeding yellowhammers. Data shown are percentage of visits to the two primary foraging habitats (grey bars) relative to area available (black bars), both calculated as mean ± 1SE per nest within a 333-m radius of a nest (the 95th percentile of all foraging flights).

experimental margin use

A total of 841 foraging flights were recorded in 2006, of which 336 (40·0%) were to margins (mean 11·2 ± 1·7 margin visits per nest, range 0–37). Thirty-one of these margin visits were to cut patches (range 0–5 per foraging watch). As predicted, the overall use of cut and uncut margins differed significantly between early and late summer (Table 4; Fig. 3), with a significant increase in the use of cut patches in late summer (mean difference between proportional use of cut patches per nest in early and late summer 30·7 with 95% CI = 15·4–47·0). In early summer, margin vegetation surrounding nests was generally short in height (Fig. 4) and the difference in height between cut patches (0 cm) and the adjacent uncut margin rarely exceeded 60 cm (Fig. 4). During this period 2·9 ± 2·7% of margin visits per nest (n = 5 visits to cut patches across all nests) were to cut patches, with cut patches comprising 2·3 ± 0·6% of the total margin area within the foraging radius (Fig. 3). In late summer, margin height increased, leading to a greater difference in height between cut and uncut margin (generally > 60 cm, Fig. 4). The use of cut patches increased to 33·6 ± 8·1% of margin visits per nest (n = 26 visits to cut patches across all nests), with cut patches comprising 2·4 ± 0·7% of margin area (Fig. 3). The frequency of visits to cut patches in relation to the height of adjacent uncut margin differed significantly from that expected under the null hypothesis that patch use was not influenced by vegetation height (χ2 = 53·26, d.f. = 4, P < 0·001). Across all nests, there was a trend for greater use of cut patches in tall margins (Fig. 4), with 84% of visits to cut patches recorded in margins with swards > 60 cm tall.

Figure 3.

Seasonal use of experimentally cut margins by foraging yellowhammers. Black bars show area of cut margins (mean ± 1SE per nest) as a percentage of total margin area available within the radius of cut patches at a nest (162 ± 8 m). Grey bars show frequency of visits to cut patches expressed as a percentage of total margin visits within the same radius. Early summer (May–June): n = 15 nests; Late summer (July–August): n = 15 nests.

Figure 4.

Percentage use of cut margin patches (grey bars) by foraging yellowhammers relative to frequency of cut patches in different heights of adjacent uncut margin (black bars). Percentages in each sum to 100% across (a) Early summer (May–June) and (b) Late summer (July–August).

Discussion

habitat selection

The importance of grass field margins for foraging yellowhammers, as well as other passerines, is well known (Brickle et al. 2000; Wilson 2001; Perkins et al. 2002). This has been attributed to higher arthropod abundance than in adjacent cereal crops (Thomas & Marshall 1999; Hart et al. 2006), a finding supported by the present study. In our initial observations, yellowhammers showed a clear preference for margins in early summer, with little use of cereals. In late summer, habitat use appeared more varied, with a marked decline in margin use and greater use of cereals, despite margins supporting significantly higher abundance of invertebrate prey.

Two plausible mechanisms could explain the seasonal shift in habitat selection. Firstly, a reduction in insect accessibility in margins associated with seasonal vegetation growth, which was most pronounced within margins, a mechanism that has been previously suggested (Hart et al. 2006) but not quantified. Secondly, a progressive increase in arthropod abundance in cereal crops during the summer, as they migrate into fields from overwinter refuges in boundaries (Coombes & Sotherton 1986). Although chick-food abundance was significantly higher in margins than either cereal crop across the whole season, in July this difference in relative abundance was small. These two mechanisms are not necessarily competing hypotheses, indeed they may operate synergistically to influence seasonal patterns of habitat selection by altering the relative availability of arthropod prey between habitats. Although yellowhammers may also visit cereal fields to collect semi-ripe grain, nestling diet comprises mainly of arthropods (Stoate et al. 1998; Wilson et al. 1999), suggesting that cereal use was driven primarily by arthropod availability. The hypothesis that birds responded to declining accessibility of prey in margins is strongly supported by our experimental margin manipulations. These demonstrated a significant increase in the use of cut patches between early and late summer, and greater use of cut areas in taller uncut margins. The low rate of use of cut patches in early summer suggests that when the height of uncut margins is low (e.g. < 60 cm), food accessibility for foraging birds is less constrained by vegetation height. In late summer, however, when uncut margins are tall, birds are likely to become more dependent on open patches to access food, and hence, the observed increase in the use of cut patches. A previous study of yellowhammers foraging in experimentally manipulated field margins found no difference in use of cut and uncut areas; however, small sample sizes may have reduced the power to detect a difference (Perkins et al. 2002).

Short, sparse swards have been shown to enhance foraging efficiency for birds, through increasing prey accessibility and detectability (Butler & Gillings 2004; Stillman & Simmons 2006), facilitating forager mobility (Devereux et al. 2004) and reducing perceived predation risk (Whittingham & Evans 2004). Numerous studies of passerines in arable and grassland habitats have demonstrated foraging preferences for shorter swards consistent with the results of the present study (see Wilson et al. 2005 and references therein). Short swards are, however, generally associated with lower abundance and diversity of invertebrates (Vickery et al. 2001; McCracken & Tallowin 2004). Open patches may increase food availability at the interface between long vegetation, which serves as a reservoir of invertebrates, and short vegetation, where prey becomes more accessible. In the present experimental study, the majority of the sward (and hence foliar arthropods) was removed from within the cut patches but limited visibility of birds within these patches meant it was not possible to determine whether they foraged along the uncut edges or the soil surface.

The vegetation structure of margins, and their height relative to cereal crops, could vary across sites for a number of reasons, potentially influencing foraging patterns: firstly, through the method of margin establishment (natural regeneration or sown grass mixes), and secondly, the ratio of spring- to winter-sown crops (and hence, the relative timing of crop and margin development). However, as a general rule, margin vegetation will undoubtedly continue to become taller and denser throughout the summer, unless management to improve accessibility is implemented. We suggest that seasonal changes in food accessibility across a range of habitats may be a more important driver of foraging patterns than previously thought. Current margin management practices may provide sub-optimal foraging habitats for birds in late summer, reducing their value in agri-environment schemes (see Management Recommendations).

We have used vegetation height as a surrogate for sward accessibility, an approach used successfully in previous studies of habitat selection (e.g. Butler & Gillings 2004; Stillman & Simmons 2006). Whilst accessibility for foraging birds may be determined by a complex range of factors such as sward height, density and structural complexity, many of these measures are confounded (Devereux et al. 2006), and further work is required to disentangle their relative influence on avian foraging behaviour. Although yellowhammers are multi-brooded (Bradbury et al. 2000), brood sizes during the foraging watches differed little between early and late summer (3·04 ± 0·15 and 2·95 ± 0·14, respectively), suggesting that seasonal variation in foraging patterns was unlikely to have been driven by varying nestling demand. In addition, most breeding pairs (c. 78%) remained in the same locality all season (judged by individually identifying pairs using colour ringing or plumage characteristics), suggesting little effect of within-season territory shifts on foraging preferences. Hence, any seasonal variations in the availability of habitats within the foraging radius of nests reflected the fact that watches were conducted at a small number of different territories in early and late summer, respectively.

The use of vacuum sampling to measure invertebrate abundance could potentially bias comparisons between habitats, as sampling efficiency may be influenced by vegetation structure (Thomas & Marshall 1999). However, if sampling efficiency is expected to be lower in the taller, denser margin vegetation, the greater abundance of invertebrates recorded in margins relative to cereals could potentially represent an underestimate of the true difference.

management recommendations

The importance of invertebrate-rich field margins as foraging habitat for farmland birds is well documented and field margin options are central to many Agri-Environment Schemes (AES) (Vickery et al. 2002, 2004). Current guidelines for margin management usually allow cutting at intervals to control woody growth (Anon 2005), but these are often infrequent and give no consideration to the requirements of foraging birds. The evidence from this study suggests more effective management is required if these habitats are to deliver improved foraging resources in late summer, a crucial period in determining population trends for many multi-brooded farmland species (Wilson et al. 1997; Siriwardena et al. 2000; Brickle & Harper 2002).

A detailed consideration of the optimal margin management techniques was outside the scope of this study. However, approaches that promote a more heterogeneous sward structure, such as a mosaic of short and long vegetation in close proximity, are likely to provide the greatest benefits for foraging birds like the yellowhammer. AES are designed to enhance a range of wildlife and more active margin management should also consider the effects on wider margin biodiversity such as insects, forbes and nesting birds. Management during the breeding season could minimize disturbance of nests and dependent young by cutting only outer portions of margins and avoiding the inner margin (e.g. near the base of hedgerows) that tends to be favoured for nesting (Rands 1986; Bradbury et al. 2000). The use of scarification or graminicides to open up grass margins may provide alternative benefits to cutting (Potts et al. 2007). Future work should investigate the biodiversity impacts and cost-effectiveness of creating more open margin swards through a range of methods. The availability of short, sparse vegetation for foraging has been shown to enhance productivity and local population trends on farmland for a number of passerine species, for example skylark (Morris et al. 2004; Donald & Morris 2005) and northern wheatear Oenanthe oenanthe L. (Pärt 2001; Arlt et al. 2008). Researching the effectiveness of margin manipulations in enhancing productivity, as well as foraging, would therefore also be desirable.

A recent study suggests that targeting AES options at the cropped area of farmland rather than field margins may deliver greater benefits for farmland birds (Butler, Vickery & Norris 2007). However more effective management of grass margins (and other margin types) would undoubtedly increase their value for birds by extending their ‘useful lifetime’ into late summer, and may reduce the reliance on cropped areas by some species. Although more active management is likely to require higher AES payments, encouraging such management has several key benefits. Firstly, it will increase the overall cost-effectiveness of margins as an AES option, particularly important given their high uptake rates (e.g. over 16 000 ha of margin options within ELS; Grice et al. 2007). Secondly, enhancing the quality of margins will increase the biodiversity benefits without increasing the area under AES options, particularly important given rising commodity prices and increasing pressure on land for other ‘products’ such as biofuels, ecosystem services and climate change mitigation measures (Sutherland et al. 2008). Thirdly, enhancing margin management may also increase the value of adjacent hedgerow options (e.g. by providing nearby foraging habitat for hedgerow nesting species) and in-field options (e.g. skylark plots) (Morris et al. 2004; Cook et al. 2007).

Acknowledgements

We thank all those farmers who kindly allowed access to their land. Mark Lewis, Isobel Gough and Ina Gruenhagen assisted with data collection and Ian Henderson and Peter Edwards provided advice throughout the study. Comments from Jeremy Wilson, Tomas Pärt, Dave Parish and Tony Morris improved earlier drafts; D.J.T.D. was funded by a NERC/BTO CASE studentship with support from Syngenta.

Ancillary