• agri-environment schemes;
  • buntings;
  • farmland biodiversity;
  • finches;
  • overwinter survival;
  • population declines;
  • sparrows


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    Low winter food availability is probably critical in the declines of many farmland bird species in Europe, leading to the implementation of ameliorative agri-environment scheme options. To date, however, there has been no experimental test of the effectiveness of such options.
  • 2
    We report the results of two large-scale, 3-year, controlled experiments investigating the effects of supplementary winter seed provision on breeding farmland bird abundance. In each experiment, the use of winter feeding sites by birds was monitored and the availability of alternative, seed-rich habitat in the surrounding area was measured. The Winter Food for Birds (WFFB) project also included variable levels of food provision. Breeding bird abundance was then monitored in experimental and control areas. The Bird Aid project targeted yellowhammer Emberiza citrinella L., corn bunting Emberiza calandra L. and tree sparrow Passer montanus L., while WFFB considered 11 species that used supplementary winter food.
  • 3
    Comparisons of trends in breeding abundance between experimental and control areas revealed little evidence for positive effects of feeding, but there was great variation in the use of feeding sites by each species, and therefore in the seed quantity birds received.
  • 4
    Declines for yellowhammer, robin and dunnock were less steep where more food was provided in WFFB areas (a fourfold difference in seed provision across 1·5 times the land area).
  • 5
    Analysing trends with respect to weight-of-use of winter food revealed significant, positive relationships for yellowhammer (both projects) and up to five other species, depending on the control terms applied. Thus, positive effects of feeding on population change depend on the effective supply of seed to the species of interest. The hypothesis that winter food is currently limiting the populations concerned is also supported.
  • 6
    Synthesis and applications. Effective winter food provision to farmland bird populations has the potential to halt, and perhaps to reverse, declines in abundance. In practice, this means that agri-environment measures supplying significant quantities of winter food, such as stubbles preceded by low-input cereals, should succeed in changing population trends if they provide resources at the times of greatest need and if there is sufficient uptake.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Farmland bird population declines (Siriwardena et al. 1998) linked to major changes in agricultural practices (Campbell et al. 1997; Chamberlain et al. 2000; Anderson, Bradbury & Evans 2001; Robinson & Sutherland 2002) have become an important conservation issue across Europe (Donald, Green & Heath 2001). Considerable evidence implicates increased annual mortality, most probably resulting from reductions in winter food resources such as cereal stubble fields, as the most common demographic mechanism (Donald & Aebischer 1997; Siriwardena, Baillie & Wilson 1998; Peach, Siriwardena & Gregory 1999; Bradbury & Stoate 2000; Siriwardena et al. 2000; Miller, Stemler & Blankenship 2001; Hole et al. 2002; Gillings et al. 2005). Many European governments have made policy commitments to undo the effects of agricultural intensification on farmland biodiversity. One key approach in the United Kingdom is to enhance winter food availability for farmland birds via the national agri-environment programme (Smallshire, Robertson & Thompson 2004; Vickery et al. 2004). Although there is evidence from several habitats that increased winter food resources produce positive effects on passerine survival and/or abundance at a local scale (Newton 1994; Peach et al. 2001; Hole et al. 2002), evidence of their effectiveness for widely distributed species in farmland is lacking.

In this paper, we describe two novel, landscape-scale experiments measuring the effect of winter food enhancement on breeding farmland bird populations. We ask (i) whether trends in abundance differed between areas with and without supplementary food, (ii) whether trends differed with respect to a fourfold difference in seed provision and (iii) whether species’ trends were related to the extent to which they used, or were effectively supplied with, supplementary food. ‘Bird Aid’ was established in 2000, initially as a conservation measure providing supplementary winter seed to maintain local bird populations until appropriate agri-environment measures were implemented. Sites were chosen to target tree sparrow Passer montanus L., yellowhammer Emberiza citrinella L. or corn bunting E. calandra L.; seed food was provided in winter and breeding numbers were monitored in spring. The Winter Food for Farmland Birds (WFFB) project investigated both responses of breeding populations and the implications of the spatial distribution of food resources (Siriwardena et al. 2006). Seed was provided in winter and its use monitored; local populations were then surveyed in the breeding season.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

bird aid

Site selection and supplementary feeding

Study areas (2 × 2 km) in lowland farmland with recorded sightings of 20 or more tree sparrows, yellowhammers or corn buntings within 5 years and 5 km were selected (n = 109 across the United Kingdom, Fig. 1; Siriwardena & Stevens 2004). Areas were stratified by target species and UK government region and then assigned randomly to a treatment (fed or control), precluding systematic differences in features such as breeding habitat availability and areas of non-farmland habitat between the samples. Breeding ranges determined sample sizes for each species: tree sparrow 47 sites (25 fed, 22 control), yellowhammer 96 (46, 50) and corn bunting 46 (22, 24). In fed areas, farmers themselves chose to provide either wheat seed or grain-rich ‘tailings’ (waste from post-harvest grain drying) from October to March each winter from 2000/01 to 2002/03. Cereal grain provided an energy-rich, preferred food resource (e.g. Perkins, Anderson & Wilson 2007) that was readily available and familiar to the target species. Seed was provided weekly to maintain constant availability: 25 kg was provided initially, but this was altered subsequently at each site, according to depletion (assessed by eye), to prevent a build-up of seed. Seed was provided on a 10 × 1 m area of bare ground or sparse vegetation, close to low cover and near the centre of the 2 × 2 km area, chosen to minimize human disturbance.


Figure 1. Map showing locations of study areas. Open symbols show fed areas and filled symbols controls; squares denote Bird Aid sites and circles WFFB sites (control 1 km2 BBS squares and fed 2 × 2 km or 3 × 2 km areas; Fig. 2).

Download figure to PowerPoint

Data collection

Each feeding site was visited twice during each winter, once in November or December and once in January or February, to measure the use of supplementary food. Sites were watched from a clear vantage point, avoiding disturbance to feeding birds, for 1 h between 1000 and 1200, recording maximum species-specific counts. Means of the six counts over the three winters of the experiment were used as indices of feeding site use.

Supplementary food use is likely to be affected by food availability in the surrounding area. This was assessed by mapping all farmland habitats within each fed and control 2 × 2 km study area in early winter in 2001/02 and 2002/03. All ‘seed-rich’, potential foraging habitats for granivorous passerines (crop stubbles, sown cover crops, weedy root crops, weedy permanent set-aside and farmyards, e.g. Hancock & Wilson 2003) were then surveyed on feeding site count days. Mean densities of each species recorded (across all sites and counts) were used to adjust the area of each habitat to a ‘cereal-stubble-equivalent’ area within each tetrad, in each winter. Summing these values produced a species- and tetrad-specific seed-rich habitat index.

To measure breeding abundance, fed and control tetrads were surveyed in 2000 (baseline year), 2002 and 2003. Access restrictions arising from the foot-and-mouth disease outbreak in the United Kingdom prevented any surveys in 2001. Each tetrad was surveyed twice, once in April–May and once in June–July, between 0700 h and 1400 h, following a constant survey route that came within 250 m of all agricultural areas present. Woodland, urban, intertidal and marine habitats were not surveyed. Standard codes were used to record the age, sex and behaviour of all birds encountered. Constant rain, strong wind (> force 4) and poor visibility were avoided. Maximum counts of territorial adult males were extracted for analysis; for the colonial tree sparrow, counts from before 15 April and after 15 June were discarded to minimize the inclusion of prebreeding feeding flocks or significant numbers of juveniles.

winter food for farmland birds

Study sites and experimental design

Ten replicates of a spatial arrangement of feeding sites at set distances from one another (Fig. 2) were established in arable farmland in East Anglia, UK in 2002 (Fig. 1). Central 3 × 2 km blocks in each replicate contained four feeding sites; the other three each contained one (Fig. 2). At feeding sites, located on tracks or field margins with adjacent hedges or other vegetation as cover, 5 kg of millet and 5 kg of sunflower hearts (chosen for high palatability) were spread on the ground weekly from November to March in 2002/03 and 2003/04. Although food provision was standardized, differential exploitation by deer, rats and gamebirds made availability to passerines variable. These effects were reduced by fencing feeding sites with 5 cm mesh wire, which did not affect passerine use visibly.


Figure 2. Schematic of the spatial arrangement of a replicate of WFFB study areas, showing feeding sites and squares in which bird surveys and habitat recording were conducted.

Download figure to PowerPoint

winter bird recording

Bird use of feeding sites was monitored weekly at varied times of day between 0900 h and 1600 h. Additional, mid-week visits were made at least once every 2 weeks during 2003–04. Bird use was recorded in 10-min periods, during which the food patch and adjacent cover (within 5 m) were observed, noting maximum counts. Mean species-specific counts at each feeding site during each winter, omitting data that could have been biased by high depletion or disturbance, provided an index of site-use. These values were summed across the four feeding sites in central blocks (Fig. 2) to provide block-specific indices and the latter were averaged across the two winters.

Habitat recording

As in Bird Aid, habitat mapping was used to quantify ambient resource availability in each block, habitat areas being converted into ‘cereal-stubble-equivalents’ and summed to give a seed-rich habitat index, for each of yellowhammer, chaffinch Fringilla coelebs L., greenfinch Carduelis chloris L., goldfinch C. carduelis L. and reed bunting Emberiza schoeniclus L., using average densities from field surveys conducted three times each winter in all blocks. Habitats other than seed-rich farmland probably determined food availability for other species, so alternative resource availability could not be controlled for. Note that greenfinches and chaffinches probably also used non-farmland habitats frequently, so further variation in their resources may have added noise to the data.

Breeding bird surveys

All 1 × 1 km squares in each block (Fig. 2) were surveyed annually during 2002–04 using methods elaborated from those of the BTO/JNCC/RSPB Breeding Bird Survey (BBS: Baillie et al. 2005). Squares were visited once during 1 April–15 May and once during 16 May–30 June, avoiding poor weather. Three squares in central blocks or a full tetrad were surveyed each day, between 0600 h and 1100 h, all squares being covered before 0900 h at least once. Visits were date-matched as far as possible between years and observers only changed for four replicates in one year. Fixed transects totalling 2 km in length were walked in each square on each visit, recording all adult birds encountered on a field map. Species-specific maximum counts across the two visits formed a square-specific count variable.

The random sample of 1 × 1 km BBS squares in Cambridgeshire, Essex, Norfolk and Suffolk surveyed annually in spring 2002–04 by skilled observers were used as controls. Note that winter bird and habitat data were not available for these squares.

Data were analysed for the 11 species that used winter food patches most: six granivores, chaffinch, goldfinch, greenfinch, house sparrow Passer domesticus L., reed bunting and yellowhammer and five insectivores/generalists, blackbird Turdus merula L., dunnock Prunella modularis L., robin Erithacus rubecula L., blue tit Cyanistes caeruleus L. and great tit Parus major L. (Siriwardena et al. 2006; see Table 1 for sample sizes). Other species were too uncommon to allow analysis.

Table 1.  Sample sizes of experimental and control survey squares in the WFFB experiment, 2002–04. The maximum sample from experimental squares was 180 (10 replicates of the design incorporating 18 1 × 1 km squares: Fig. 2); 198 BBS squares were surveyed in the region considered
SpeciesExperimental (fed) squaresControl (BBS) squares
Blue tit180186
Great tit179170
House sparrow129137
Reed bunting 25 69

statistical analysis

Data from Bird Aid and WFFB were analysed independently, but with analogous methods, using three approaches. First, breeding population changes were compared between experimental and control sites. Second, because there was a fourfold difference in seed provision within c. 1 km between central and other blocks (Fig. 2), WFFB population trends could be analysed with respect to level of food supplementation. Third, breeding population trends in both WFFB and Bird Aid fed areas were modelled as a function of winter supplementary food use, controlling for alternative resource availability. All analyses were conducted using the genmod procedure or the glimmix macro of sas (SAS Institute, Inc. 2001).

Breeding population changes in experimental and control areas

Breeding season counts for Bird Aid were analysed using generalized linear mixed models (GLMMs) with a log-link function, assuming a Poisson error distribution and correcting for overdispersion. Count was modelled as a function of fixed year and treatment (feeding or control) effects, with site (tetrad) identity included as a random effect, controlling for geographical variation in abundance. Models for WFFB were similar, but site was fitted as a categorical, fixed effect [i.e. a generalized linear model (GLM)]. This accounted for spatial variation between fed and control squares, making a main effect for treatment unnecessary. ‘Sites’ in this analysis were individual 1 × 1 km squares, matching BBS sampling units. Squares within an experimental block (Fig. 2) will nominally have contained local populations responding to the same feeding site(s), so constitute repeated measures of the effect of those sites. Thus, a repeated measures framework was used to account for potential correlations between squares in individual blocks in each year. Both categorical and continuous-year effects were fitted for both data sets. The former avoided constraining population changes to a particular form; the latter summarized them as linear trends, providing greater statistical power and more interpretable interactions. A year × treatment interaction allowed independent population changes in each treatment: its significance, assessed using likelihood-ratio or Wald tests, showed whether experimental and control trends differed.

Effects of quantity of food supplied

GLMs comprising categorical square and continuous year effects, and a block type × year interaction allowing different trends in ‘central’ and ‘outer’ WFFB blocks (Fig. 2), tested the effect of seed quantity. Repeated measures models with log link and Poisson errors were again used. The square effect accounted for spatial variation between block types, so no block type main effect was required. A seed-rich habitat index × year interaction was added to the yellowhammer, chaffinch, greenfinch, goldfinch and reed bunting models as a control to account for variation in background resource availability. Again, no main effect was required because of the categorical square effect. Significance was assessed using Wald tests.

Breeding population changes in relation to feeding site use in winter

Counts were modelled as a function of a site-use index × year interaction, testing for variation in population trend with feeding site use. For Bird Aid, GLMMs were fitted with log link and Poisson errors, incorporating a random site effect and fixed, continuous effects for year and site-use. Likelihood-ratio tests determined significance. For WFFB, repeated measures GLMs used log link and Poisson errors, incorporating categorical site and continuous year effects. No main effect of site-use was needed because of the categorical square effect. Significance was assessed using Wald tests.

Better habitat could have led both to larger local populations and more positive trends, making greater site-use a consequence of higher local abundance (given low seasonal immigration and emigration), rather than a cause of population increases. An interaction between year and the 3-year mean breeding survey count, as a control, accounted for variations in trend with local abundance. Alternative seed resource availability could have confounded food supplementation, so a seed-rich habitat index × year interaction was introduced as another control. For WFFB sites, much of the variation in site-use values could have reflected the difference between central and outer blocks. A block type × year interaction controlled for this.

Models were run including all controls individually and together, and omitting them completely.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

population trends in experimental and control squares

There were no significant effects, using either continuous or categorical year effects, of Bird Aid feeding on breeding populations between 2000 and 2003 (Table 2). There was, however, a non-significant trend for more positive population changes on fed sites vs. controls for tree sparrow (Fig. 3); this fed area trend was also the only positive, individually significant population change (Fig. 3).

Table 2.  Likelihood-ratio test results for the significance of the year × treatment interaction for each species, comparing fed and control squares in both Bird Aid and WFFB. Results are shown for models in which temporal variation (‘year’) was considered as both a categorical (three-level factor) and a continuous variable
ExperimentSpeciesCategorical year effectContinuous year effect
inline imagePinline imageP
Bird AidTree sparrow 0·59  0·558 0·88  0·352
Yellowhammer 0·64  0·530 0·93  0·337
Corn bunting 0·50  0·607 0·33  0·718
WFFBBlackbird 8·9  0·012 8·31  0·004
Blue tit15·05  0·00115·86< 0·001
Chaffinch 1·69  0·430 1·58  0·209
Dunnock 2·05  0·359 0·6  0·439
Goldfinch 1·96  0·374 0·65  0·421
Greenfinch 1·11  0·574 0·5  0·478
Great tit 9·9  0·007 7·62  0·006
House Sparrow 0·41  0·814 0·04  0·845
Robin31·42< 0·00122·49< 0·001
Reed bunting 4·96  0·084 0·03  0·868
Yellowhammer 0·89  0·639 0·99  0·319

Figure 3. Changes in breeding abundance in Bird Aid experimental and control sites, 2000–03, as estimated by the slope of a linear year effect (on the log scale). The y-axis shows the slope of linear trends between 2000 and 2003 for each data set. X-axis species codes: TS, tree sparrow; Y, yellowhammer; CB, corn bunting. Filled diamonds denote slopes for experimental squares and open squares slopes for control squares. Error bars show 95% confidence intervals.

Download figure to PowerPoint

Divergent trends between fed and control WFFB squares were found for blackbird, blue tit, great tit and robin, and categorical year effect tests provided no evidence for more complex patterns (Table 2). All significant differences revealed more positive trends in control than in fed squares (Fig. 4), i.e. fed sites saw steeper declines. There were, however, no such negative effects on the granivores (Table 2, Fig. 4).


Figure 4. Changes in breeding abundance in WFFB experimental and control squares, 2002–04, as estimated by the slope of a linear year effect (on the log scale). Graph format is as Fig. 3. X-axis species codes: B, blackbird; BT, blue tit; CH, chaffinch; GO, goldfinch; GR, greenfinch; GT, great tit; HS, house sparrow; R, robin; RB, reed bunting; Y, yellowhammer.

Download figure to PowerPoint

effects of quantity of supplementary food

Within the pattern for species to decline (seven significantly) in WFFB fed areas (Fig. 4), dunnock, robin and yellowhammer declined significantly less steeply in central blocks, where more food had been supplied in winter (Table 3, Fig. 5). Dunnock and yellowhammer abundance was stable in squares in central blocks and declined elsewhere, whereas robin abundance declined everywhere, but less steeply in central blocks (Fig. 5). No further differences appeared after controlling for background resource availability, but the yellowhammer pattern remained (Table 3).

Table 3.  Results of tests of the dependence of trends in breeding abundance between 2002 and 2004 on winter feeding activity in the local survey square block. Tests considered the effects of block type × year interaction, where ‘block type’ could be either ‘central’ or ‘outer’ (Fig. 2). Parameter estimates for block type × year interactions show slopes for central blocks relative to outer blocks. Results of models describing temporal variation as simple linear trends (on the log scale) are shown to provide context. Significant results are shown in bold type
SpeciesLinear trend with timeInteraction with block typeInteraction with block type, controlling for interaction with area of seed-rich habitat
Parameter estimateSEWald inline imagePParameter estimateSEWald inline imagePParameter estimateSEWald inline imageP
Blue tit–0·0600·027 4·890·027–0·0230·0560·160·688    
Chaffinch–0·0230·017 1·830·1770·0160·0330·240·6210·0190·0340·230·630
Greenfinch–0·0160·031 0·250·6170·0480·0660·540·4640·0340·0690·710·400
Great tit–0·0510·030 2·830·0930·0330·0630·270·602    
House sparrow–0·0250·037 0·460·497–0·0270·0760·130·722    
Reed bunting–0·0910·182 0·250·6150·0230·3870·000·952–0·0370·4230·000·949
Yellowhammer–0·0570·030 3·520·0610·1190·0623·770·0520·0960·0623·630·057

Figure 5. Trends in breeding abundance (linear on the log scale) in squares in WFFB central and outer blocks for species showing significant differences (Table 3): (a) dunnock, (b) robin and (c) yellowhammer.

Download figure to PowerPoint

relationships with feeding site use

Results are presented in Table 4 and summarized, for species showing significant relationships, in Figs 5 and 6 (dividing sites into groups according to the variation in site-use – low, intermediate or high – and averaging the most significant model predictions for each species, including all control terms, to illustrate variations in trend). Note that these graphs are intended only to illustrate the effects of feeding site use on population trends revealed by the continuous interaction results in Table 4: they do not provide definitive descriptions. In Bird Aid-fed areas, there was a significant positive relationship between winter site-use and breeding population trajectory for yellowhammer, but not tree sparrow or corn bunting (Table 4, Fig. 6a). Adding control terms did not affect the yellowhammer result, but a more negative effect appeared for corn bunting when the average breeding abundance control was included (Table 4). However, omitting this control and including that for alternative seed rich habitats produced a significantly positive result (Table 4). Figure 6b clarifies these confounded patterns: high site-use was associated with an increasing trend for corn bunting, as opposed to stability with low use, but the pattern for declines with intermediate use was stronger after accounting for the confounding variation in breeding abundance.

Table 4.  Results of tests of the effect of variation in site use on trends in abundance in the Bird Aid (2000–03) and WFFB (2002–04) study areas. Tests refer to models incorporating site-use × year interaction terms, together with controls as indicated. Parameter estimates (PE) are shown for the interaction between the continuous site-use and year variables: positive values reflect more positive trends in abundance where site-use was greater. Test statistics (TS) are likelihood ratio test T (Bird Aid) or Wald χ2 (WFFB; 1df). Significant or near-significant results are shown in bold type
SpeciesNo controlsInteraction with site use, contolling for:
Interaction with breeding abundanceInteraction with block typeinteraction with block typeAll controls
Bird Aid
 Tree sparrow–0·0050·005–1·050·298–0·0020·006–0·440·664    –0·0050·005–1·140·260–0·0030·006–2·830·663
 Yellowhammer0·0110·0034·14<0·0010·0120·0034·41<0·001    0·0100·0033·46<0·0010·0110·0033·78<0·001
 Corn bunting0·0030·0021·580·119–0·0210·003–6·38<0·001    0·0040·0022·300·025–0·0170·004–4·34<0·001
 Blue tit–0·0230·0162·010·156–0·0240·0162·350·125–0·0400·0272·270·132        
 Great tit–0·0570·0352·570·109–0·0630·0353·380·066–0·1030·0406·730·009        
 House sparrow0·0260·0172·220·1360·0320·0173·410·0650·0250·0181·930·164        
 Reed bunting0·0640·0237·930·0050·0660·0238·100·0040·0750·02014·36<0·0010·0840·0711·390·2380·1700·0765·030·025

Figure 6. Variation in breeding population trend, 2000–03, with respect to the use of the Bird Aid experimental feeding sites (significant results from Table 4 only). ‘Low’, ‘intermediate’ and ‘high’ use reflect the range of mean counts per winter seed patch monitoring occasion for each species (see graph legends) and were chosen to maintain a reasonable sample size in each category. Error bars show 95% confidence intervals from the samples of predicted values. (a) Yellowhammer, (b) corn bunting.

Download figure to PowerPoint

In WFFB fed areas, trends in yellowhammer breeding abundance were highly significantly more positive with greater use of winter supplementary food, a result that was robust to the addition of controls (Table 4). There were other consistent, positive effects of food use on trends for goldfinch, reed bunting and dunnock (Table 4). A similar effect for chaffinch was only detected after controlling for block type, and a weak pattern suggested another, for house sparrow, when the breeding abundance control was added (Table 4). The results for great tit suggested more negative trends with greater use of winter food, but this pattern was only significant after controlling for block type (Table 4), reflecting declines where both site-use and abundance were relatively high, but stability elsewhere (Fig. 7a). For goldfinch and dunnock (Fig. 7b,c), breeding abundance was stable where site-use was high and declined where it was lower, although high use occurred at lower average abundance for goldfinch and higher abundance for dunnock (graph intercepts, Fig. 7b,c). The results for chaffinch and reed bunting were less clear-cut, with stable abundance at high site-use but slight declines with intermediate and low use (Fig. 7d,e). There were stronger effects for house sparrow and yellowhammer: declines with lower site-use and increases with higher use, although high use occurred at different relative abundances (Fig. 7f,g). Note that, for yellowhammer, although low site use occurred where average abundance was also relatively low, areas with high and intermediate site-use had similar levels of abundance (Fig. 7g). The larger error bars for house sparrow and reed bunting (Fig. 7e,f) reflect these species’ patchy distributions: zero winter site use was common and breeding birds were found in relatively few squares (Table 1).


Figure 7. Variation in breeding population trend, 2002–04, with respect to the use of the local WFFB winter food patches (significant results from Table 4 only). Graph formats are as Fig. 6. (a) Great tit, (b) dunnock, (c) goldfinch, (d) chaffinch, (e) reed bunting, (f) house sparrow, (g) yellowhammer.

Download figure to PowerPoint


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Comparisons of trends in breeding abundance in fed and control areas revealed no significant, positive effects of supplementary winter feeding in either Bird Aid or WFFB, indicating that resources provided in the relevant quantities and contexts is unlikely to reverse population declines. However, this result masks important associations between winter feeding and population trends in fed areas. More positive trends were observed for yellowhammer, robin and dunnock in WFFB central blocks, where more food was provided, although the effects were not sufficiently large to produce increases. Additional effects of supplementary feeding were revealed post hoc when relationships between trends in abundance and the level of use of winter food were examined. In Bird Aid, there was a significant positive association between food use and population trends for yellowhammers, but not for tree sparrows (non-significant) or corn buntings (confounded patterns). In WFFB, up to five species (depending on the controls included) declined less steeply with higher food use: chaffinch, dunnock, yellowhammer, house sparrow and reed bunting. The results provide evidence for winter resource limitation in these populations. They suggest that supplementing winter food can produce landscape-scale, positive effects, but that effective supply to enough individuals is critical: increasing general availability may not increase the seed that target species actually consume. Nevertheless, the results constitute the first experimental evidence that appropriate enhancements of food availability, such as through agri-environment measures providing seed resources when they are most needed during the winter, have realistic potential to halt, or at least to slow, population declines at the landscape scale.

This study does not show that population increases result from food supplementation. Removing a winter food constraint could possibly introduce another density-dependent limit (e.g. saturation of available breeding habitat: Desrochers, Hannon & Nordin 1988) that would prevent population increases. This seems unlikely, however, because many species’ densities are now far lower than were found in the recent past and breeding performance has often increased as abundance has fallen, suggesting that it is not currently constrained (Siriwardena et al. 2000). Further, previous experimental studies of winter food supplementation in passerines have demonstrated increases in abundance (Samson & Lewis 1979; van Balen 1980; Jansson, Ekman & von Bromsson 1981; Källander 1981). Other studies have shown enhanced survival (Smith et al. 1980; Jansson, Ekman & von Bromsson 1981; Brittingham & Temple 1988; Desrochers, Hannon & Nordin 1988; Newton 1994; Lahti 1998; Doherty & Grubb 2002) or condition (Grubb & Cimprich 1990; Rogers & Heath-Coss 2003), both of which probably indicate positive population effects where food limits abundance. Failures of winter food supplementation to generate detectable increases in passerine populations have usually been attributed to constraints on nesting opportunities (nest sites or territory vacancies) or superabundant natural food (e.g. Krebs 1971; Brittingham & Temple 1988; Newton 1994). Neither possibility is likely to apply to declining British farmland birds.

The latter scenarios are among the potential explanations for trends in abundance in fed areas not being more positive than those in control areas, despite the widespread use of feeding sites by declining species. Possible methodological explanations also included the duration of the experiments (perhaps too short for effects to be seen), the quantity of food supplied (potentially too small genuinely to benefit the flocks attracted, especially after depletion by competitors), the quality of food supplied (perhaps nutritionally inadequate, although energy is probably the key requirement), sample sizes (perhaps providing insufficient statistical power) and abundance monitoring methods (perhaps too coarse to detect small changes in abundance). Biologically, the populations concerned might only be limited by winter food in harsher conditions than in the study winters [maximum and minimum (respectively) January daily temperatures (°C), annual means for 2001–04: Durham 7·0 (range 5·7–7·6), 1·5 (–0·1 to 2·4); Cambridge 7·8 (6·7–8·9), 2·1 (1·2–2·7); Lyneham 7·3 (5·9–8·5), 1·8 (0·7–2·8); Anon 2001–2004, see Fig. 1 for locations]. Wintering and breeding populations might also have been insufficiently linked for a change to be detectable, if within-winter movements encompassed large areas (but see Siriwardena et al. 2006) or dispersal from wintering to breeding locations was too extensive. In addition, impacts of predators such as sparrowhawks Accipiter nisus L. attracted to feeding sites could have overridden positive effects of feeding (e.g. Bro et al. 2004). However, although fed areas supplied more resources than controls, neither experiment established a clear dichotomy, in practice, between zero and significant food supplementation for individual species; this was accounted for in analyses incorporating site-use.

Many species’ use of feeding sites ranged from zero to regular large flocks (Siriwardena et al. 2006), while use by competitors, including gamebirds, deer and rats, also varied. Variation in local habitat features and local bird abundance probably also affected the propensity of each species to use each feeding site, despite every effort to standardize site context (Siriwardena & Stevens 2004). Food supplementation to each species was not therefore consistent across sites. Despite this, there was a fourfold difference in the quantity of food available to the whole granivorous community, across 1·5 times the land area (6 km2 vs. 4 km2; Fig. 2), between WFFB central and outer blocks. Radio-tracking results suggest that most individuals’ regular winter foraging movements occur within such areas (Calladine, Robertson & Wernham 2006; Siriwardena et al. 2006). Central blocks therefore probably supplied more food to local populations of each species, which seems to have changed the average yellowhammer trend from decline to stability (Table 3, Fig. 5c). Similarly positive effects occurred for dunnock and robin (Fig. 5a,b).

Data quantifying feeding site use provided a surrogate for species-specific quantities of food consumed in both experiments. Most significant relationships between site-use and population trend suggested that heavier food use resulted in more positive trends. Yellowhammers, which have recently declined rapidly (Baillie et al. 2005), responded particularly strongly (Table 4, Fig. 7). There were also significant positive effects of site-use on population trends for goldfinch, house sparrow, reed bunting and chaffinch, but not for greenfinch or tree sparrow, while apparent effects for corn bunting were inconsistent (Table 4). The differences between species may reflect differences in ecology and population status. For example, yellowhammers and house sparrows are still declining in many areas (Baillie et al. 2005), suggesting ongoing shortfalls in winter food resources. At the farm scale, house sparrow breeding abundance can respond to supplementary winter food (Hole et al. 2002); this study suggests similar patterns at the landscape scale. Reed buntings are now stable nationally, having disappeared from many farmland areas in the 1980s (Peach, Siriwardena & Gregory 1999), but a lack of winter food could be limiting re-colonization. Weaker effects might be expected for species with stable or increasing national populations because they are less likely to suffer from winter food shortages, which could explain the tree sparrow and greenfinch results. Further, there was a significant effect on chaffinch trends, but it was small and harder to detect (Fig. 7d, Table 4). The corn bunting results reflect confounded relationships between trend and each of site-use and absolute abundance (Fig. 6b): site-use was high where abundance was high and these populations increased. At low abundance, greater site-use was associated with steeper declines, perhaps because insufficient food was provided to help smaller, declining populations. Note, however, that the Bird Aid site-use measure may have been less accurate because only two counts were made per winter, compared to 20 or more in WFFB.

Although insectivorous and generalist species might be less dependent on seeds in winter, dunnock population trends were strongly affected by winter feeding (Table 4, Fig. 7b). Reduced survival is the probable mechanism for this species’ decline (Siriwardena, Baillie & Wilson 1998; Baillie et al. 2005) and dunnocks used feeding sites heavily (Siriwardena et al. 2006), so positive responses to increased winter resources are plausible. There was also a significant relationship for great tit but this showed more negative trends with greater site-use (Table 4, Fig. 7a,b). It is possible that great tits used supplementary food where food resources in hedgerows, woodland, scrub and gardens were scarce (at higher breeding abundance: Fig. 7a), but not enough was available to prevent declines. With an increasing national population (Baillie et al. 2005), declines at high local abundance are more likely to reflect density-dependent stabilization than a conservation problem.

Overall, despite the various potential sources of noise in the comparisons conducted, there was some evidence of positive responses to winter food use for seven of the 13 species considered across both experiments and an eighth (robin) responded positively to food quantity alone (Tables 3 and 4). A further positive association was found for corn bunting, but an apparently stronger negative association was also found (Table 4). All species for which no positive effect was identified are currently increasing nationally (Baillie et al. 2005), so winter food may not currently be limiting these species. Yellowhammer and house sparrow were among the declining species that appeared to be limited by winter food, although effects varied with breeding density in WFFB study areas. Yellowhammers declined, irrespective of local population density, unless feeding site use was high, but intermediate or high food patch occurred most at higher breeding densities (Fig. 7g). High food use and more positive trends occurred at relatively low densities for house sparrow (Fig. 7f). These local populations may be declining, with insufficient winter food resources, while larger concentrations persist, supporting Hole et al.'s (2002) suggestion that smaller local populations are vulnerable to extinction.

There are two caveats to the interpretation of the site-use and population trend results. First, the analyses were, strictly, correlational. However, absolute abundance was controlled for, so positive results did not merely show greater site-use and healthier populations at higher densities. More positive trends could still have arisen from high local breeding success in the previous summer causing high winter abundance, and thus high site-use, but this assumes considerable spatial variation in breeding success and contradicts relevant existing evidence on population limitation (Siriwardena et al. 1998, 2000; Peach, Siriwardena & Gregory 1999; Hole et al. 2002). Moreover, differences in trend between WFFB central and outer blocks (Fig. 5) are very unlikely to reflect variation in breeding success.

The second caveat is that increased local breeding density could reflect re-distribution in response to winter food availability. This could be tested by comparing breeding density within and just outside heavily used fed areas. One recent study suggested that yellowhammers preferentially settle to breed adjacent to seed-rich late-winter stubble or set-aside (Whittingham et al. 2005), but failed to account for these habitats’ value for breeding. Ongoing research is investigating the extent of winter-to-breeding-season movements of resident granivores and effects of feeding on body condition, over-winter survival and abundance. This should clarify the relative importance of population growth and re-distribution in this context.

Winter food provision features in several agri-environment measures that aim to restore farmland bird populations. This study shows that such resource enhancement can slow or stop landscape-scale declines. The lack of demonstrable differences between experimental and control areas shows, however, that seed delivery mechanisms must be effective: the timing and context of seed provision must fill critical resource gaps for target species. Winter food resource prescriptions, such as within the English Environmental Stewardship (ES) scheme, include sown bird seed mixtures, over-wintered cereal and root-crop stubbles and unharvested conservation headlands. Such options, particularly given sympathetic chemical management, can produce large quantities of seed and field observations show that birds use such habitats intensively. These prescriptions are therefore likely to provide appropriate resources, but they have to be taken up sufficiently frequently by farmers. For example, more than 10 ha km2 of average-quality stubble (or a smaller area managed sympathetically) is likely to be needed to stop local skylark Alauda arvensis declines (Gillings et al. 2005), but this is considerably more than current levels of relevant ES option uptake will generate (Butler, Vickery & Norris 2007).

Resource dilution and obscuring vegetation (Butler et al. 2005) will make search times longer in open field habitats, even seed-rich ones, so a given quantity of seed might provide less benefit than in superabundant patches. However, interference competition from other birds will probably also be lower, as will seed depletion by rats and deer and disturbance from predator attacks (because flocks are less concentrated). It is unknown, therefore, whether resources in seed patches or open fields provide greater net benefits. Nevertheless, feeding sites were used most heavily in late winter (Siriwardena et al. 2006), when ‘natural’ food availability is probably lowest (Evans, Vickery & Shrubb 2004); open-field agri-environment options tend not to cover this period well. In particular, ‘over-wintered’ stubbles in ES can be ploughed in mid-February, unless part of a more stringent option specifying that they subsequently become a set-aside. Sowing, cutting and cultivation dates for wild bird seed mixtures are also not stipulated (Defra 2005), although doing so could increase their value in late winter. Late winter food availability may therefore still be low: research is needed into the benefits of agri-environment prescriptions in practice and possible improvements in their effectiveness in supplying late winter seed. Precise quantities of seed required for particular population outcomes are also unknown. Nevertheless, if they supplied seed effectively throughout the winter, our results indicate that agri-environment prescriptions could stop and, perhaps, reverse some farmland bird declines and hence contribute significantly to meeting UK government conservation targets (Vickery et al. 2004).


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

WFFB was funded by Defra (project BD1616), with additional support from English Nature. Bird Aid was funded by RSPB, with support from Defra (BD1616). We thank the RSPB and BTO staff who helped each project (Bird Aid: Richard Winspear, Dominic Coath, Emily Woodfield, Trevor Smith, Frazer MacFarlane, Kevin Mawhinney, Roger Taylor, Larry Griffin, Claire Devereux, Andy Wraithmell and Michael Coplestone; WFFB: Greg Conway, Chas Holt, David Barr, Loyd Berry, Mark Grantham, Rick Goater, Jez Blackburn, Bridget Griffin, Richard Thewlis and Mark Collier) and we are indebted to numerous landowners for allowing access. Steve Freeman provided valuable statistical advice, Bill Sutherland participated in much helpful discussion and the comments of Mark Whittingham and four referees improved the manuscript. CJ Wildbird Foods assisted with WFFB seed supplies.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Anderson, G.Q.A., Bradbury, R.B. & Evans, A.D. (2001) Evidence for the Effects of Agricultural Intensification on Wild Bird Populations in the UK. RSPB Research Report no. 3. RSPB, Sandy, UK.
  • Anon (200104) Weather logs. Weather, 56 59.
  • Baillie, S.R., Marchant, J.H., Crick, H.Q.P., Noble, D.G., Balmer, D.E., Beaven, L.P., Coombes, R.H., Downie, I.S., Freeman, S.N., Joys, A.C., Leech, D.I., Raven, M.J., Robinson, R.A. & Thewlis, R.M. (2005) Breeding Birds in the Wider Countryside: Their Conservation Status 2004. BTO Research Report no. 385. BTO, Thetford. Available at:, accessed June 2007.
  • Van Balen, J.H. (1980) Population fluctuations of the Great Tit and feeding conditions in winter. Ardea, 68, 143164.
  • Bradbury, R.B. & Stoate, C. (2000) The ecology of yellowhammers Emberiza citrinella on lowland farmland. Ecology and Conservation of Lowland Farmland Birds (eds N.J.Aebischer, A.D.Evans, P.V.Grice & J.A.Vickery), pp. 165172. British Ornithologists’ Union, Tring.
  • Brittingham, M.C. & Temple, S.A. (1988) Impacts of supplemental feeding on survival rates of black-capped chickadees. Ecology, 69, 581589.
  • Bro, E., Mayot, P., Corda, E. & Reitz, F. (2004) Impact of habitat management on grey partridge populations: assessing wildlife cover using a multisite BACI experiment. Journal of Applied Ecology, 41, 846857.
  • Butler, S.J., Vickery, J.A. & Norris, K. (2007) Farmland biodiversity and the footprint of agriculture. Science, 315, 381384.
  • Butler, S.J., Whittingham, M.J., Quinn, J.L. & Cresswell, W. (2005) Quantifying the interaction between food density and habitat structure in determining patch selection. Animal Behaviour, 69, 337343.
  • Calladine, J., Robertson, D. & Wernham, C.V. (2006) The movements of some granivorous passerines in winter on farmland. Ibis, 148, 169173.
  • Campbell, L.H., Avery, M.I., Donald, P., Evans, A.D., Green, R.E. & Wilson, J.D. (1997) A Review of the Indirect Effects of Pesticides on Birds. JNCC Report no. 227. Joint Nature Conservation Committee, Peterborough.
  • Chamberlain, D.E., Fuller, R.J., Bunce, R.G.H., Duckworth, J.C. & Shrubb, M. (2000) Changes in the abundance of farmland birds in relation to the timing of agricultural intensification in England and Wales. Journal of Applied Ecology, 37, 771788.
  • Department for the Environment, Food and Rural Affairs (DEFRA) (2005) Environmental Stewardship: Entry Level Stewardship Handbook. Rural Development Service, Defra, London.
  • Desrochers, A., Hannon, S.J. & Nordin, K.E. (1988) Winter survival and territory acquisition in a northern population of black-capped chickadees. Auk, 195, 727736.
  • Doherty, P.F. Jr & Grubb, T.C. Jr (2002) Survivorship of permanent-resident birds in a fragmented forested landscape. Ecology, 83, 844857.
  • Donald, P.F. & Aebischer, N.J., eds (1997) The Ecology and Conservation of Corn Buntings Miliaria calandra. UK Nature Conservation no. 13. Joint Nature Conservation Committee, Peterborough.
  • Donald, P.F., Green, R.E. & Heath, M.F. (2001) Agricultural intensification and the collapse of Europe's farmland bird populations. Proceedings of the Royal Society of London, Series B, Biological Sciences, 268, 2529.
  • Evans, A.D., Vickery, J.A. & Shrubb, M. (2004) Importance of over-wintered stubble for farmland bird recovery: a reply to Potts. Bird Study, 51, 9496.
  • Gillings, S., Newson, S.E., Noble, D.G. & Vickery, J.A. (2005) Winter availability of cereal stubbles attracts declining farmland birds and positively influences breeding population trends. Proceedings of the Royal Society of London, Series B, 272, 733739.
  • Grubb, T.C. & Cimprich, D.A. (1990) Supplementary food improves the nutritional condition of wintering woodland birds: evidence from ptilochronology. Ornis Scandinavica, 21, 277281.
  • Hancock, M.H. & Wilson, J.D. (2003) Winter habitat associations of seed-eating passerines on Scottish farmland. Bird Study, 50, 116130.
  • Hole, D.G., Whittingham, M.J., Bradbury, R.B., Anderson, G.Q.A., Lee, P.L.M., Wilson, J.D. & Krebs, J.R. (2002) Widespread local house sparrow extinctions. Nature, 418, 931932.
  • Jansson, C., Ekman, J. & Von Bromsson, A. (1981) Winter mortality and food supply in tits Parus spp. Oikos, 37, 313322.
  • Källander, H. (1981) The effects of provision of food in winter on a population of the great tit Parus major and the blue tit P. caeruleus. Ornis Scandinavica, 12, 244248.
  • Krebs, J.R. (1971) Territory and breeding density in the great tit, Parus major L. Ecology, 52, 222.
  • Lahti, K. (1998) Social dominance and survival in flocking passerine birds: a review with an emphasis on the willow tit Parus montanus. Ornis Fennica, 75, 117.
  • Miller, M.R., Stemler, C.L. & Blankenship, D.S. (2001) Mourning dove productivity in California during 1992–95: was it sufficient to balance mortality? Journal of Wildlife Management, 65, 300311.
  • Newton, I. (1994) Experiments on the limitation of breeding bird densities: a review. Ibis, 136, 397411.
  • Peach, W.J., Lovett, L.J., Wotton, S.R. & Jeffs, C. (2001) Countryside stewardship delivers cirl bunting in Devon. Biological Conservation, 101, 361373.
  • Peach, W.J., Siriwardena, G.M. & Gregory, R.D. (1999) Long-term changes in over-winter survival rates explain the decline of reed buntings Emberiza schoeniclus in Britain. Journal of Applied Ecology, 36, 798811.
  • Perkins, A.J., Anderson, G.Q.A. & Wilson, J.D. (2007) Seed food preferences of granivorous farmland passerines. Bird Study, 54, 4653.
  • Robinson, R.A. & Sutherland, W.J. (2002) Post-war changes in arable farming and biodiversity in Great Britain. Journal of Applied Ecology, 39, 157176.
  • Rogers, C.M. & Heath-Coss, R. (2003) Effect of experimentally altered food abundance on fat reserves of wintering birds. Journal of Animal Ecology, 72, 822830.
  • Samson, F.B. & Lewis, S.J. (1979) Experiments on population regulation in two North American parids. Wilson Bulletin, 91, 222233.
  • SAS Institute, Inc. (2001) SASOnlineDoc®, version 8. SAS Institute Inc., Cary, NC.
  • Siriwardena, G.M., Baillie, S.R., Buckland, S.T., Fewster, R.M., Marchant, J.H. & Wilson, J.D. (1998) Trends in abundance of farmland birds: a quantitative comparison of smoothed Common Birds Census indices. Journal of Applied Ecology, 35, 2443.
  • Siriwardena, G.M., Baillie, S.R., Crick, H.Q.P. & Wilson, J.D. (2000) The importance of variation in the breeding performance of seed-eating birds for their population trends on farmland. Journal of Applied Ecology, 37, 122.
  • Siriwardena, G.M., Baillie, S.R. & Wilson, J.D. (1998) Variation in the survival rates of some British passerines with respect to their population trends on farmland. Bird Study, 45, 276292.
  • Siriwardena, G.M., Calbrade, N.A., Vickery, J.A. & Sutherland, W.J. (2006) The effect of the spatial distribution of winter seed food resources on their use by farmland birds. Journal of Applied Ecology, 43, 628639.
  • Siriwardena, G.M. & Stevens, D.K. (2004) Effects of habitat on the use of supplementary food by farmland birds in winter. Ibis, 146 (Suppl. 2), 144154.
  • Smallshire, D., Robertson, P. & Thompson, P. (2004) Policy into practice: the development and delivery of agri-environment schemes and supporting advice in England. Ibis, 146 (Suppl. 2), 250258.
  • Smith, J.N.M., Montgomerie, R.D., Taitt, M.J. & Yom-Tov, Y. (1980) A winter feeding experiment on an island song sparrow population. Oecologia, 47, 164170.
  • Vickery, J.A., Bradbury, R.B., Henderson, I.G., Eaton, M.A. & Grice, P.V. (2004) The role of agri-environment schemes and farm management practices in reversing the decline of farmland birds in England. Biological Conservation, 119, 1939.
  • Whittingham, M.J., Swetnam, R.D., Wilson, J.D., Chamberlain, D.E. & Freckleton, R.P. (2005) Habitat selection by yellowhammers Emberiza citrinella on lowland farmland at two spatial scales: implications for conservation management. Journal of Applied Ecology, 42, 270280.