Simon J. Butler, Edward Grey Institute of Field Ornithology, Zoology Department, South Parks Road, Oxford OX1 3PS, UK (fax +44 1865 271221; e-mail firstname.lastname@example.org).
1In the UK, over-wintered stubbles are an important foraging habitat for farmland birds, many of which are of current conservation concern. Maximizing the value of stubbles for farmland birds has largely focused on increasing food abundance but has ignored other aspects of foraging behaviour, such as predator avoidance and vigilance.
2In this study we investigated the effects of stubble height reduction on seed depletion and the within-field distribution of farmland birds. Stubble on one half of each of 20 fields was topped to lower vegetation height, while the other half was left untouched to act as a control area.
3The abundance of granivorous passerines and invertebrate feeders was higher on treatment plots, which had undergone stubble height reduction, while the abundance of skylark and partridges was higher on control plots. Topping had no effect on the within-field distribution of either corvids or pigeons.
4Seed densities on both plot types fell between November and March, but there was no significant difference in levels of depletion between treatment and control plots.
5For many species, food accessibility and detectability are likely to have been higher, and thermo-energetic and mobility costs lower, on treatment plots. Stubble height reduction could have led to either an increase or a decrease in associated predation risk, depending on the predator escape strategy of a species. Granivorous passerines, which flee to cover, are likely to have benefited from the reduced visual obstruction on treatment plots, while partridges, which rely on crypsis, will have benefited from the greater protection offered by the longer vegetation in control plots.
6Synthesis and applications. Increasing the structural heterogeneity of over-wintered stubble is likely to increase its value as a foraging habitat, making it better suited to the foraging requirements and predator escape strategies of a greater diversity of species. Incorporating explicit management options into agri-environment schemes, such as the new Environmental Stewardship Scheme, may represent a cost-effective strategy to achieve this.
The potential energetic gain and predation risk associated with a foraging patch is likely to be determined by an animal's foraging requirements and predator escape strategy (Watts 1990; Suhonen, Norrdahl & Korpimäki 1994). To maximize fitness, many animals must trade-off the conflicting demands of energetic gain and predator avoidance when choosing where to forage (Sih 1980). They are therefore likely to select patches with characteristics that allow them to optimize this trade-off.
For some species, vegetation is perceived as being largely protective, reducing the risk of predator detection, and patches with more complex vegetation structure are likely to have a lower associated predation risk (Ekman 1987; Högstad 1988). Conversely, some species perceive vegetation as largely obstructive, reducing the likelihood of early predator detection, and patches with increased vegetation structure are likely to have a higher associated predation risk (Metcalfe 1984; Sharpe & van Horne 1998). It follows that habitats with more heterogeneous vegetation structure will contain patches that suit the foraging requirements and predator escape strategies of a broader range of species and will therefore probably support a greater species richness.
However, aviary experiments have shown that granivorous passerines alter their foraging and vigilance behaviour in response to vegetation height manipulation (Whittingham et al. 2004). Chaffinches Fringilla coelebs L. foraging in short vegetation (3 cm) had shorter vigilance periods and faster peck rates than those foraging in long vegetation (13 cm). Individuals foraging in the short vegetation also responded significantly faster to simulated attack by a model sparrowhawk Accipiter nisus L. than those foraging in the long vegetation (Whittingham et al. 2004). These findings suggest that both the starvation risk and predation risk an individual of this species associates with a foraging patch could be reduced by lowering vegetation height in that patch. Butler et al. (2005) quantified this relationship between risk and vegetation structure in terms of food abundance. Using a similar experimental set-up, they showed that there needed to be approximately 2·5 times the density of food in a patch of long stubble before individual chaffinches showed parity of use between it and a patch of short stubble (Butler et al. 2005).
These aviary studies (Whittingham et al. 2004; Butler et al. 2005) suggest that stubble height manipulation could be used as a simple management option for increasing the value of stubbles as a foraging habitat for farmland birds. We tested this prediction under natural conditions on over-wintered wheat stubbles. Using a split-field design, we focused on the effects of stubble height reduction on bird abundance and distribution.
This study was carried out from October 2003 to March 2004 on 20 wheat Triticum spp. stubble fields from 12 lowland farms in central England (three farms with three fields, two farms with two fields and seven farms with a single field). Each field was either over-wintering as stubble prior to spring sowing or had been entered into a rotational set-aside scheme (Evans et al. 1998). To reduce the likelihood of double counting during bird surveys, fields on the same farm were only included in this study if they were not adjacent to each other.
In late October, the stubble on one half of each field was cut using a standard topper to lower vegetation height. Additional chaff generated by topping was left on the fields. The stubble on the other half of each field was left untouched to act as a control area. Where possible, fields were divided so that the treatment and control plots had similar boundary characteristics, slope and aspect, to reduce any confounding effects of these factors on bird abundance and distribution. The field half to receive stubble height reduction was randomly allocated.
The area of each field was obtained from the farmers (mean field size ± 1 SE, 9·6 ± 0·92 ha). The boundary characteristics of the treatment and control plots were recorded in order to calculate a boundary height index (for full details see Wilson et al. 1997).
Fields were visited on six occasions between November 2003 and February 2004, with a minimum of 1 week between each survey. Bird abundance and distribution were estimated by walking parallel transects (which ran perpendicular to the boundary of the treatment and control plots) at 30-m intervals and counting all birds that flushed, recording which plot they had flushed from. Care was taken to avoid double counting by noting where previously flushed birds landed. Counts were undertaken between 1 h after dawn and 1 h before dusk to avoid periods when birds were leaving or arriving at roost sites. Periods of wet or windy weather were avoided because of the effects of these conditions on bird activity (Robbins 1981).
Following stubble height manipulation, vegetation characteristics in the treatment and control plots were recorded along six transects, three in each field half. Each transect ran perpendicular to the field boundary and consisted of three sample points, positioned 5 m, 30 m and 60 m from the field edge (i.e. a total of 18 sample points, nine in the treatment plot and nine in the control plot). The starting point of each transect was randomly positioned around the field perimeter, provided that point was more than 60 m from a field corner.
At each sample point, the mean vegetation height and the percentage of bare earth within a 50 × 50-cm quadrat were recorded. Mean vegetation height was calculated from five height measurements, taken from the stem nearest the four corners and centre of the quadrat.
soil seed densities
Seeds were collected from surface soil scrapes (20 × 20 cm) taken at each of the sample points described above. To restrict sampling to the part of the seed bank likely to be accessible to small passerines (Green 1978; Robinson 1997), only soil on or immediately below the surface was collected. In each plot, soil from the three sample points at 5 m, 30 m and 60 m was bulked. These were then placed in resealable polythene bags and stored within 24 h at 4°C (to prevent germination) until they could be analysed.
The bulked samples were weighed and a random subsample (25% of original mass) removed. Seeds from each subsample were extracted by washing the soil through sieves of decreasing mesh size (1 mm, 500 µm and 63 µm). The contents of the sieves were then washed into a white sample tray and allowed to dry before being hand sorted, with each seed being counted and identified using an appropriate guide (Flood & Gates 1986) and reference material collected in the field. Initial checks showed that the material collected by the 63-µm mesh sieve only contained seeds of very small-seeded species (e.g. poppies Papaver spp.). Densities of such seeds are usually underestimated by hand counting (Thompson, Bakker & Bekker 1997) and they are a relatively unimportant component in the diet of the species considered here (Green 1978; Campbell et al. 1997; Wilson et al. 1999). Seed density calculations were therefore based only on the contents of the 1-mm and 500-µm mesh sieves. Soil samples were collected in November 2003 and March 2004 to allow an estimation of overwinter seed loss. Three fields were ploughed before the second set of samples could be collected, limiting estimates of seed depletion to just 17 fields.
Most bird species were not recorded in sufficient numbers to permit statistical analysis of abundance at the species level (see the Appendix). Species were therefore assigned to functional groups, based on ecological and taxonomic characteristics (Henderson et al. 2000; Bradbury & Allen 2003), with particular emphasis on foraging requirements and predator escape strategy (see the Appendix for details of groupings). For similar reasons, the six surveys were grouped into two temporal categories, with the first three surveys on each field specified as early season and the second three surveys on each field specified as late season.
The effect of stubble height reduction on the abundance and distribution of six functional groups was tested using logistic regression in GLIM 4.0 (NAG 1993). The term field was included in all models to allow within-field comparisons, while controlling for variation resulting from unmeasured site-specific parameters. To investigate whether the relative abundance of birds on treatment and control plots changed over the course of the winter, the term season (either early or late) was incorporated into the model. The number of times each plot (treatment or control) held a higher abundance of each functional group than the other plot was specified as the response variable, and the number of surveys in each season (three) was identified as the binomial denominator (specifying a binomial error structure with logit-link function; Crawley 1993). This method of abundance analysis represents a biologically realistic approach; birds are unlikely to select foraging habitats independently of conspecifics in a flock, but variation in the number of individuals using a particular habitat is likely to indicate their relative value as foraging sites (Buckingham et al. 1999).
As expected given the random allocation of treatment plots, paired t-tests showed that there were no significant differences in area (t19 = −1·18, P= 0·25), boundary height index (t19 = −0·15, P= 0·88), initial seed density (t19 = −0·40, P= 0·70) and percentage of bare earth (t19 = −1·59, P= 0·13) between treatment and control plots, so these terms were excluded from the models. The significance of field, season and treatment terms was assessed using the change in deviance (ΔD), which is distributed asymptotically as χ2, on removal of each term from a model including all three predictors. The fit of the model to the assumptions of a binomial distribution can be approximated by comparing the ratio of residual deviance/residual degrees of freedom (Collett 1991; Crawley 1993). Ratios close to one indicate a reasonable fit to the data, whereas ratios greater than 2·5 indicate a poor, overdispered fit (Crawley 1993).
Within-field comparisons of the overall levels of seed depletion and the depletion at specific distances from the field edge on treatment and control plots were made using paired t-tests. All probabilities quoted are two-tailed. Means and standard errors are presented in the form mean ± 1 SE.
Topping resulted in a mean stubble height reduction of 58·3 ± 1·9%, with a mean stubble height on the treatment plots of 5·9 ± 0·29 cm compared with 14·1 ± 0·33 cm on the control plots. The percentage of bare earth on the treatment plots (mean = 20·9 ± 4·4%) tended to be slightly lower than that on the control plots (25·7 ± 4·4%; t19 = −1·59, P= 0·13), possibly because of increased stubble residue resulting from the topping process.
A total of 28 bird species (4941 individuals) was recorded at the study sites during the survey period (see the Appendix). Of these, seven species were recorded only in the treatment plots and one only in control plots.
When all bird species were aggregated, there was no difference in their relative abundance on treatment and control plots (χ2 = 1·78, P > 0·1). However, when the relative abundances of each of the six functional groups were analysed separately, it was evident that the various groups responded differently to the stubble height manipulation. The abundance of granivorous passerines and invertebrate feeders was higher on treatment plots than on control plots, while the abundance of skylark Alauda arvensis L. and partridges was higher on the control plots (Table 1, Fig. 1). Neither corvid nor pigeon distribution was influenced by stubble height reduction. Field was a significant predictor of relative bird abundance on treatment and control plots for all functional groups, but this result was of little biological importance for this within-field study. No functional group showed a significant change in relative abundance on treatment and control plots between early and late-season surveys (Table 1).
Table 1. Results of logistic regression analyses for six functional groups surveyed on 20 wheat stubble fields at 12 lowland farms in central England. Probability values are presented for the effects of field identity, season and treatment (stubble manipulation) on within-field distribution. Significant effects of treatment are shown in bold. Directions of any association (C, more on control plots; T, more on treatment plots) between predictor variable and abundance are provided
Model goodness-of-fit (res. dev./res. d.f)
res. dev., residual deviance; res d.f., residual degrees of freedom.
< 0·05 (T)
< 0·01 (T)
< 0·001 (C)
< 0·001 (C)
Spring seed densities were significantly lower than autumn seed densities on both treatment and control plots (t17 = 2·8, P= 0·01, t16 = 4·1, P= 0·001, respectively; Fig. 2). There was no significant difference in the overall level of depletion on treatment and control plots (t16 = −0·92, P= 0·37). In both treatment and control plots, the highest level of seed depletion was recorded in samples taken 30 m from the field edge, while the lowest level of seed depletion was recorded in samples taken 5 m from the field edge (Fig. 2). There were no significant differences in seed depletion levels between treatment and control plots at any of the distance bands measured (P > 0·4 in all cases).
This study has shown that vegetation height manipulation can bring about differential spatial use of stubble fields by a range of bird species. The abundance of granivorous passerines and invertebrate feeders was higher in treatment plots, which had undergone stubble height reduction, while skylark and partridge abundances were higher in control plots.
The potential energetic gain from a patch is likely to be dictated by food abundance, detectability and accessibility and the costs associated with foraging in that patch. While there was no difference in seed abundance between plot types at the start of the winter, it is likely that both food detectability and accessibility were higher in the shorter vegetation of the treatment plots (Whittingham & Markland 2002; Butler & Gillings 2004). Thermo-energetic and mobility costs are also likely to have been lower for foragers in the shorter vegetation of treatment plots; taller or denser vegetation is likely to retain more moisture, particularly as a result of frost and dew in winter, and to be more difficult to move through (Dawson, Carey & van't Hof 1992; Brodman, Reyer & Baer 1997; Clarke et al. 1997).
The granivorous passerines recorded in this study are likely to rely on early detection of predators to flee to protective cover (Lima & Dill 1990; Whittingham & Evans 2004), often feeding near field edges and using surrounding hedgerows and trees as refuges (Robinson & Sutherland 1999). The level of visual obstruction offered by the vegetation within a foraging patch is therefore likely to have a far greater influence on their perception of predation risk than the degree of protection it offers. The results of aviary experiments (Whittingham et al. 2004; Butler et al. 2005) suggest that the preferential selection of treatment plots by granivorous passerines in this study occurred because they are likely to have associated both a lower predation risk and greater potential energetic gain with treatment plots.
Treatment plots are also likely to have offered greater potential energetic gain and lower predation risk to invertebrate feeders. While no measure of food abundance for these species was made, it is unlikely that it differed between the two patches when stubble height reduction took place, although invertebrates may have subsequently responded to the increased stubble residue on the surface of treatment plots (Hole et al. 2005) and any consequent changes in microclimate at soil level (McCracken, Foster & Kelly 1995). However, invertebrate prey detectability and accessibility are likely to have been greater in the shorter vegetation of treatment plots (Butler & Gillings 2004), and recent research has shown that the mobility and foraging efficiency of invertebrate feeders is lower in long vegetation than in short vegetation (Devereux et al. 2004). When threatened, most of the species included in this functional group fly to nearby bushes and trees until the danger has passed (Clement & Hathway 2000), so the influence of vegetation on visual obstruction is again likely to have a greater influence on perceived predation risk than its protective value.
The effect of stubble height reduction on the predation risk of individuals that leave their foraging site and flee to protective cover in the event of predator attack may be influenced by the size of the individual. For example, the difference in visual obstruction, and therefore perceived predation risk, between treatment and control plots is likely to have been much smaller for larger-bodied species such as pigeons and corvids. This limited effect of stubble height reduction on the perceived predation risk of pigeons and corvids, as well as their broader diet (Murton, Isaacson & Westwood 1966; Snow & Perrins 1998), may explain why there was no treatment effect on their within-field distribution in this experiment.
Partridges and skylark adopt different strategies for avoiding predation compared with granivorous passerines and invertebrate feeders. Instead of retreating to cover, partridges often remain still and rely on crypsis to avoid predator detection (Madge & McGowan 2002). The usual raptor avoidance behaviour of skylark is also to crouch, often not breaking cover until the last minute (Donald 2004). While the shorter stubble on treatment plots may have provided less visual obstruction and allowed earlier predator detection, it is likely to have offered less protection to partridges and skylark once a predator had been detected. The greater abundance of partridges and skylark on control plots suggests they associated lower predation risk with these plots and were trading this off against the greater potential energetic gain on treatment plots. A number of other studies, across a range of taxa, have shown that the trade-off between potential energetic gain and predation risk is often more heavily weighted in favour of minimizing predation risk (Brown 1988; Nonacs & Dill 1990; Todd & Cowie 1990; Moody, Houston & McNamara 1996; Beck & Watts 1997; Butler et al. 2005).
The discussions above have centred on the effects of stubble height manipulation on forager behaviour. It is also possible that predator behaviour may have been influenced by treatment effects. Sparrowhawks are the main predators of adult farmland birds (Götmark & Post 1996). Quinn & Cresswell (2004) showed that sparrowhawks hunt according to the vulnerability of their prey, which is likely to be affected by vegetation structure, while studies of kestrels Falco tinnunculus L. have shown that foraging activity and hunting success are higher over less densely vegetated habitats (Shrubb 1980; Toland 1987). The effect of stubble height manipulation on hunting behaviour by farmland bird predators, both avian and terrestrial, needs further investigation to allow a greater understanding of how it influences actual predation risk for farmland birds.
There was no significant difference in the level of seed depletion between the two plot types. However, as stubble height manipulation may have influenced a number of other factors that influence seed depletion levels, besides foraging birds, it is not possible to draw any firm conclusions about seed depletion from this result. For instance, although most plants only set seed prior to the start of the winter (Grime, Hodgson & Hunt 1989), topping in late October may have suppressed some over-winter seed production. Furthermore, differences in microclimate between plots may have caused differential seed germination or rotting rates (Wright 1993). It is also possible that stubble height manipulation may have affected seed depletion rates by other seed predators, such as insects and small mammals (Telleria, Santos & Diaz 1994; Jacob & Brown 2000; Thorbek & Bilde 2004). Further research should therefore aim to investigate the effects of stubble height manipulation on other taxa that play an important role in stubble field dynamics.
synthesis and applications
We have shown that stubble height manipulation can influence foraging site selection by a range of farmland bird species, many of which are of current conservation concern. It appears that the value of wheat stubble fields to granivorous passerines and invertebrate feeders can be increased simply by lowering stubble height. However, wholesale reductions in height would be detrimental to other species, such as skylark and partridges, which have different foraging requirements and predator escape strategies. Increasing the structural heterogeneity of overwinter stubble is likely to increase its value as a foraging habitat, making it better suited to the foraging requirements and predator escape strategies of a greater diversity of species. Incorporating targeted management options into agri-environment schemes such as the British government's new Environmental Stewardship Scheme may represent the most cost-effective strategy to achieve this increased structural heterogeneity. In this experiment, farmers were paid £5 per hectare cut. Structural heterogeneity could also be achieved through rolling, discing or low intensity grazing, but further research is required to determine which approach is most suitable and also the most appropriate spatial scale for such management. We predict that increasing structural heterogeneity within other farmland habitats, such as grassland and cereal crops, will also increase the value of these habitats, as they too are utilized by a range of species with differing foraging requirements and predator escape strategies.
We thank the 12 landowners for permission to work on their land and for undertaking the required stubble height manipulation. The authors would also like to thank Nik Aspey for his assistance with fieldwork. Earlier drafts of this manuscript were improved by comments from Andy Evans, Chris Elphick, Phil Atkinson and one anonymous referee. S.J. Butler was supported by the Biotechnology and Biological Sciences Research Council and RSPB. M.J. Whittingham was supported by a BBSRC David Phillips Fellowship and a BBSRC grant (43/D13408).