Content uploaded by William J Sutherland
Author content
All content in this area was uploaded by William J Sutherland on May 05, 2018
Content may be subject to copyright.
Journal of Applied
Ecology
2006
43
, 628–639
© 2006 The Authors.
Journal compilation
© 2006 British
Ecological Society
Blackwell Publishing Ltd
The effect of the spatial distribution of winter seed food
resources on their use by farmland birds
GAVIN M. SIRIWARDENA,* NEIL A. CALBRADE,* JULIET A. VICKERY*
and WILLIAM J. SUTHERLAND†
*
British Trust for Ornithology, The Nunnery, Thetford, Norfolk IP24 2PU, UK; and
†
School of Biological Sciences,
University of East Anglia, Norwich NR4 7TJ, UK
Summary
1.
Agri-environment measures providing winter seed are central to current manage-
ment activities aiming to reverse granivorous farmland bird declines. Previous research
has considered the effectiveness of particular agri-environment options, but the
influence of their distribution in the landscape, an important factor in determining
cost-effectiveness, has received less attention. We used a large-scale field experiment in
eastern England, featuring 10 replicates of a spatial arrangement of seven artificial
feeding stations separated by different distances, to investigate bird movements between
discrete winter food resources.
2.
Feeding sites were established and bird use monitored at least weekly over two winters
(November–March). Habitat type in the areas surrounding the feeding sites was also
recorded. Two measures of feeding site use were analysed as functions of the distance
from the nearest alternative feeding site, controlling for ambient resource availability,
for 11 passerines (including chaffinch
Fringilla coelebs
, reed bunting
Emberiza schoeniclus
and yellowhammer
Emberiza citrinella
and non-granivores such as blue tit
Parus caeruleus
).
3.
Feeding site use varied significantly with separation distance for nine species. For
most, both maximum counts and bird minutes were higher at more isolated sites, but the
opposite was true for yellowhammer and reed bunting. Most relationships incorporated
significant thresholds at separations of around 500 m, with low site use below the
threshold and high above it (or vice versa).
4.
The results suggested that species such as chaffinch and blue tit made disproportion-
ately more use of food at isolated sites than at clumped ones, but that reed buntings and
yellowhammers used food in proportion to its availability, moving freely between more
clumped patches. In both cases, birds tended to share resources separated by 500 m or
less. At greater separations, resources tended to be used by discrete groups of birds.
Radio-tracking and colour-ringing work produced corroborative evidence.
5.
Synthesis and applications.
The response of birds to the experimental food resource
distribution implies that creating resource patches more than 1 km apart should be
most cost-effective, a recommendation that can inform agri-environment scheme
planning directly. This distance may represent a minimum, however, because birds
might locate standing crops more readily than inconspicuous artificial food patches.
Key-words
:agri-environment schemes, buntings, finches, overwinter survival, sparrows,
supplementary feeding
Journal of Applied Ecology
(2006)
43
, 628–639
doi: 10.1111/j.1365-2664.2006.01170.x
Introduction
Continuing population declines among farmland birds
(Siriwardena
et al
. 1998; Baillie
et al
. 2005) are a key
conservation issue throughout Europe (Krebs
et al
.
1999; Donald, Green & Heath 2001). There is strong
Correspondence: Gavin M. Siriwardena, British Trust for
Ornithology, The Nunnery, Thetford, Norfolk IP24 2PU, UK
(fax +44 1842750030; e-mail gavin.siriwardena@bto.org).
629
Spatial patterns of
winter food use by
farmland birds
© 2006 The Authors.
Journal compilation
© 2006 British
Ecological Society,
Journal of Applied
Ecology
,
43
,
628–639
evidence to link these declines to agricultural change
and, although there are probably several causes
(Chamberlain
et al
. 2000; Newton 2004), the shift from
spring to autumn sowing of cereals and consequent
loss of seed-rich stubbles may have been critical.
Changes in survival have driven the declines of many
species, probably reflecting overwinter resource
availability (Siriwardena, Baillie & Wilson 1998), and
stubbles are both preferred feeding habitats in winter
and a major source of seed (Wilson, Taylor & Muirhead
1996; Buckingham
et al
. 1999; Bradbury & Stoate 2000;
Gillings & Fuller 2001; Peach
et al
. 2001; Moorcroft
et al
. 2002). Seed-eater (finches, buntings and skylark
Alauda arvensis
) population trends are particularly
strongly associated with areas of spring barley in the
preceding year (measuring stubble availability indirectly;
Chamberlain
et al
. 2000) and have been shown to be
positively correlated with the area of winter stubble
itself at the 1
×
1 km-square scale (Gillings
et al
. 2005).
Further, the reintroduction of stubbles has been critical
among the environmental enhancements that have
reversed the UK population decline of the cirl bunting
Emberiza cirlus
(Peach
et al
. 2001; Wotton
et al
. 2004).
Winter food resource creation is now advocated
widely as an option in agri-environment schemes.
However, two key questions need to be addressed if its
cost-effectiveness is to be maximized: (i) how should
additional seed resources be provided and (ii) where should
they be established? Several studies have addressed the
first question and have recommended the retention of
stubbles and introduction of seed-bearing cover
crops as conservation measures for seed-eating birds
(Henderson
et al
. 2000; Moorcroft
et al
. 2002; Bradbury
et al
. 2004; Henderson, Vickery & Carter 2004); such
measures are included in current UK agri-environment
schemes (Vickery
et al
. 2004). The second question has
yet to be considered: if seed resource creation is to be
cost-effective, resource patches should be distributed
optimally. Specifically, within regions, how far apart
should food resources be placed in order to reach the
maximum number of individuals of the target species?
Should resource patches be far apart, reducing the
distance an average bird has to travel to the nearest
patch, or should they be clumped, so that wintering
flocks have a choice of feeding sites and alternatives
when disturbed by predators, for example?
We addressed this question through a novel, large-
scale field experiment, using artificial feeding stations
consisting of seed patches on the ground. Although
qualitatively different from growing crops, these seed
patches were relatively easy to establish and manipulate,
and they provided a model scenario through which to
investigate the implications of spatial separation of
food resources. Patterns of use of feeding stations by
different bird species were analysed in relation to
isolation of the food resource (i.e. distance to the next
nearest artificial patch), controlling for the ambient,
alternative resource availability (the extent of seed-rich
habitat, i.e. stubble and wild bird cover, nearby). Many
factors, in addition to relative isolation, are likely to
influence the use of individual artificial food patches by
any given species in practice, but the combination of
standardized food provision and a replicated experi-
mental design should minimize this variation and
ensure that what remains constitutes noise rather than
bias. We considered the use of these patches by a range
of granivorous species, as well as by other species
(including blackbird
Turdus merula
and blue tit
Parus
caeruleus
) that use and may benefit from supple-
mentary seed. The target granivorous species considered
included declining (e.g. yellowhammer
Emberiza
citrinella
) and stable/increasing (e.g. chaffinch
Fringilla
coelebs
) species. The latter are of less conservation
interest but they inform about the generality of the
results, thus allowing inferences to be drawn about
farmland birds in general. The implications of the
results for the cost-effective creation and targeting of
seed-rich winter habitat for birds are discussed in the
context of agri-environment schemes and wider farm
management practices.
Methods
Seed food was provided through the winters of 2002–
03 and 2003–04, at 70 feeding stations (in 10 replicates
of seven), following a pilot in 2001–02. Each replicate
consisted of a central feeding site, a second one at a dis-
tance of 100 m away, a third at 500 m from the nearest
site, a fourth at 1 km from the nearest site and a further
three sites at minima of 2, 5 and 10 km from the nearest
other site (Fig. 1). All feeding stations were located in
eastern England, in or near the East Anglian Plain Natural
Area, which describes a homogeneous arable landscape
typical of much lowland farmland in southern and
eastern England (www.english-nature.org.uk/science/
natural/NA_search.asp, accessed 14 April 2006).
Feeding stations were positioned on tracks or on
field margins, with a hedge or other vegetation nearby
to provide cover and perches. Food patch location was
standardized in this way to control non-experimental
variation in the factors likely to influence food use, but
some variation in patch context was inevitable. To
check for systematic biases, feeding station context
was measured using 15 quantitative and categorical
variables [describing the nature and structure of
adjacent cover (height, depth, type, etc.), adjacent crop
type, presence of a ditch, presence of water, distance to
human habitat and type of substrate] and appropriate
statistical tests were conducted to look for variation
with respect to feeding site separation.
At each feeding station, birds were fed 5 kg of millet
and 5 kg of sunflower hearts each week from Novem-
ber until the end of March each year. Both seed types
are highly palatable, millet providing relatively high
carbohydrate content and sunflower hearts relatively
high fat and protein (Díaz 1990). Sunflower hearts are
630
G. M. Siriwardena
et al.
© 2006 The Authors.
Journal compilation
© 2006 British
Ecological Society,
Journal of Applied
Ecology
,
43
,
628–639
an ‘unnatural’ food that would be unavailable to many
species, even where sunflowers are grown, but they are
often preferred when supplied (e.g. Perkins & Anderson
2002), so represent an efficient means of supplying food.
However, seed preferences are species-specific (Díaz
1990; Boatman & Stoate 2002), so the use of a mixture
should have helped to attract a wide range of species.
Seed was spread on the ground at each site, across an
area of about 2 m
2
, so that it was conspicuous amongst
any surrounding vegetation but sufficiently dispersed
to prevent monopolization by a few individuals.
Some sites, especially in 2002–03, were affected by use
by non-target animals, principally deer (fallow
Dama
dama
, roe
Capreolus capreolus
and Reeve’s muntjac
Muntiacus reevesi
), brown rat
Rattus rattus
, ring-necked
pheasant
Phasianus colchicus
and red-legged partridge
Alectoris rufa
. To reduce these effects, such sites were
fenced using 0·5–1-m high wire mesh (5-cm diameter:
large enough for finches and buntings to pass through
unhindered) and bamboo canes. These fences were
successful in excluding gamebirds and had little effect
on passerines; indeed, fences were commonly used
as perches en route between cover and food patches.
Gamebird use also declined through the winter as
shooting reduced numbers. Deer problems affected
10–15% of feeding sites and peaked in mid-winter;
the worst affected sites were moved slightly, where possible,
to reduce deer access. Rats were seen more frequently
later in winter, especially in 2002– 03 (up to a maximum
of 31% of sites compared with 8·5% in 2003–04) but
rarely caused complete seed depletion.
Bird use of the food patches was monitored regularly
just prior to the weekly replenishment. The order in
(and therefore time of day at) which sites were visited
was varied weekly. Additional mid-week visits were
made when time allowed during 2002–03, and proved
essential for the accurate recording of bird use at sites
where seed depletion was high. Mid-week visits were
therefore made at least bi-weekly during 2003– 04.
Bird use was recorded in 10-min observation
periods, during which the food patch and the adjacent
hedge/cover were scanned continuously using bino-
culars. The maximum number of birds of each species on
and near the patch was recorded. ‘Near to the patch’
(hereafter in its ‘vicinity’) was defined as a volume of
approximately 10 m long, 10 m wide and 5 m high,
centred on the food patch. Analyses of counts of birds
observed actually using the patch itself, i.e. definitely
using the food, or ‘approach counts’, also considering
birds that were subsequently flushed from the vicinity
of the patch, did not change any conclusions so they are
not considered further here.
Where seed patches were not completely or almost
completely depleted, the 10-min observation periods
provided an unbiased index of feeding site use. Where
complete depletion occurred in under a week and
was principally caused by birds, the weekly feeding
observations will have underestimated site use: this was
taken into account analytically (see below).
The 10-min counts did not provide an estimate of the
absolute, total number of birds using each food patch,
because flock turnover was unknown. Counts could
not be made simultaneously, so birds could have used
one patch and then moved to a second, therefore
being counted twice. Field experience suggested such
movement was only likely between centre, 100-m and,
perhaps, 500-m sites, and even here opportunities were
limited because these sites were usually observed in
quick succession. Nevertheless, following pilot work in
2002–03, this issue was addressed during 2003–04
using additional data collection at feeding sites. From
January to March, on as many visits as possible,
observation periods at each patch were extended to
20 min. The second 10-min period was spent focused
on the patch itself, rather than the vicinity. During this
time, the number of birds of each species present was
recorded at the end of each minute between minute 11
and minute 20. The species-specific sums of these
minute-by-minute counts then provided an estimate of
bird use in terms of bird-minutes, rather than a simple
maximum number of individuals.
Fig. 1. Schematic of the spatial arrangement of feeding sites
for each replicate, together with the focal squares in which
habitat recording was carried out.
631
Spatial patterns of
winter food use by
farmland birds
© 2006 The Authors.
Journal compilation
© 2006 British
Ecological Society,
Journal of Applied
Ecology
,
43
,
628–639
- -
Monitoring bird use by observation of numbers or bird-
minutes at individual feeding sites provided indirect
evidence for patterns of movement; direct measurement
required individual birds’ movements to be followed.
To this end, 20 yellowhammers, spread across sites in
three replicates, were radio-tagged and tracked from
regular sampling points for approximately 1 month
from January 2003. In addition, a total of 160 yellow-
hammers, 47 chaffinches, 49 goldfinches (
Carduelis
cannabina
) and 118 greenfinches (
Carduelis chloris
)
were colour-ringed in either 2002 –03 or 2003– 04.
Resighting was then conducted during and immediately
after regular feeding site monitoring. The radio-tracking
and colour-ring resighting work was supplementary to
the main experiment and is described in full in Appen-
dices S1 and S2 (see the supplementary material).
Experimental food patches were at sites chosen for
their locations relative to one another, not for the
nearby habitat (apart from it being predominantly
agricultural). There could therefore have been signi-
ficant variations in ambient resource availability between
study tetrads and between classifications of tetrads
with respect to feeding site separation, which could
confound the experimental design. The potential influ-
ence of these key alternative resources for granivorous
farmland birds was controlled by calculating the area
of each field (or part of field) under the main seed-rich
habitat types (cereal stubble, game cover and permanent
set-aside) in each study area, using digitized maps
and the ArcView 3.3 geographical information systems
package (Environmental Systems Research Institute
Inc., Redlands, California, USA). Areas of each key
habitat type were then summed for each study area in
each winter.
Data were analysed for the 11 species that used food
patches most commonly. These were blackbird, robin
Erithacus rubecula
, dunnock
Prunella modularis
, blue
tit, great tit
Parus major
, chaffinch, goldfinch, greenfinch,
reed bunting
Emberiza schoeniclus
, yellowhammer and
house sparrow
Passer domesticus
. Bird count (vicinity)
and bird minute data were analysed using generalized
linear models in the GENMOD procedure of SAS
(SAS Institute Inc. 2001), using a log-link and Poisson
errors. Counts that were likely to be biased downwards
by disturbance or total depletion of the food patch
were excluded from all analyses. Because multiple
(weekly) counts were available from each site, they
were analysed in an auto-regressive repeated measures
framework to prevent the inflation of apparent pre-
cision through pseudoreplication. Models were fitted
to test the effect of food patch separation on count,
treating separation both as a six-level categorical (site
type central/100 m, 500 m, etc.) and as a continuous
(linear on the log scale) variable, i.e. separation
distance from the nearest other experimental patch.
Site separation (in metres) was transformed by taking
logarithms prior to analysis because bird responses to
distance are more likely to vary geometrically than
arithmetically. While the continuous variable provided
more statistical power, the categorical results revealed
whether relationships with respect to distance were
actually non-linear (curvilinear or ‘stepped’). These
shapes were revealed by inspection of plots of count
effects against separation. A control for a quadratic
effect of the number of days between feeding and
observation (i.e. both linear and squared terms)
was added to the models. These variables revealed a
significant quadratic pattern for most species: counts
were typically low within a day of replenishment at
previously depleted sites, then tended to rise to peak
3–4 days after replenishment before falling again,
presumably because of further depletion of the food
available. Including this control allowed the inclusion
of mid-week visits and took account of any variation
between site types in the number or timing of these counts.
The analyses of counts included data from both
2002–03 and 2003 –04; all models fitted incorporat-
ing a two-level factor (year) to allow for interannual
changes. The data on bird use per unit time (bird-
minutes) were analysed for 2003– 04 only (the 2002–03
pilot data were too sparse).
Additional analyses were run including the areas of
the three key seed-rich habitats as controls, accounting
for variation in the alternative resources available to
species whose winter farmland populations depend,
largely, on seeds in open field habitats: chaffinch, gold-
finch, greenfinch, reed bunting and yellowhammer. A
further set of analyses used field survey results from the
seed-rich habitats throughout the winter to consider
actual bird numbers in these fields, but the results did
not differ substantially from those using field areas, so
they are omitted for brevity.
The significance of model component effects such
as separation was assessed using Wald tests in SAS,
comparing full models with ones with the parameter
of interest removed.
Results
A total of 28 species was observed using feeding sites,
but the level of use varied hugely between species and
sites, in terms of both flock sizes and the number of
weeks during the winter in which birds were recorded.
Table 1 summarizes where and how frequently birds
were seen and in what order of abundance, when they
were seen, for six granivorous farmland passerines and
five other species that used at least 65 sites in each
632
G. M. Siriwardena
et al.
© 2006 The Authors.
Journal compilation
© 2006 British
Ecological Society,
Journal of Applied
Ecology
,
43
,
628–639
winter. Despite the variation in intensity of use, most
sites were used to some extent by the target species,
often in large numbers (e.g. greenfinch and chaffinch
maximum counts of more than 40). Some important
farmland granivores, namely skylark, linnet
Carduelis
cannabina
, corn bunting
Miliaria calandra
and tree
sparrow
Passer montanus
, made too little use of sup-
plementary food for their responses to be analysed.
There were significant or near-significant linear
relationships between count and separation for seven
species (Table 2a). Five relationships (blue tit, chaffinch,
great tit, house sparrow and robin) were positive and
two (reed bunting and yellowhammer) were negative.
These results remained largely unchanged with the
addition of controls for the local availability of
seed-rich habitat (Table 2b; exceptions were chaffinch
and reed bunting) and similar patterns were generally
revealed when categorical site type was used instead of
linear separation (exceptions were reed bunting and
dunnock). Each of the eight species named above
showed variation significant at
P
= 0·05 or less in one or
both of the continuous and categorical tests (Table 2),
providing evidence of variation in use with respect to
site separation.
Figure 2 shows the form of the relationship for
all eight species for which one or other analytical
approach suggested that an effect of separation might
Table 1. Summary of feeding site use. Numbers of sites (out of 70) where species were observed in the vicinity of the food patch
in each winter are shown, with summary statistics for species and sites, i.e. median, minimum and maximum numbers of weeks
observed (out of 21 in 2002 –03 and 19 in 2003– 04) and mean and maximum counts for the non-zero observations, the latter
describing the group sizes that tended to be observed
Species
2002–03 2003–04
No. of
sites
No. of weeks Counts
No. of
sites
No. of weeks Counts
Median Min. Max. Mean Max. Median Min. Max. Mean Max.
Blackbird 70 19 18 21 2·1 13 70 17 16 19 1·8 9
Blue tit 70 19·5 17 21 1·9 15 68 17 15 18 1·6 7
Chaffinch 70 20 19 21 4·7 40 70 17 17 19 4·0 25
Dunnock 70 20 19 21 1·8 10 70 17 16 19 1·3 5
Goldfinch 21 2 1 19 4·0 19 29 5 1 17 2·5 13
Great tit 68 20 16 20 1·5 5 66 16·5 14 19 1·5 9
Greenfinch 67 16 9 20 5·0 40 70 17 17 19 6·3 81
House sparrow 7 10 2 16 6·2 30 14 12 3 17 3·6 18
Reed bunting 20 6·5 1 14 4·5 26 26 6 4 16 2·1 11
Robin 69 19 17 20 1·1 4 70 17 15 18 1·0 3
Yellowhammer 65 6 1 16 5·6 35 63 17 5 18 4·7 28
Table 2. Summary of effects of site separation results on maximum count. Continuous parameter estimates refer to slopes
(on the log scale) of relationships between maximum count and log(separation): (a) results for all species with no additional
controls; (b) results for farmland-dependent seed-eaters controlling for areas of seed-rich habitat (see text for details)
Species
Continuous (linear): separation Categorical: site type
Parameter estimate SE Wald PWald P
(a)
Blackbird 0·069 0·045 2·37 0·124 5·20 0·392
Blue tit 0·129 0·042 9·25 0·002 16·91 0·005
Chaffinch 0·087 0·036 5·92 0·015 11·37 0·045
Dunnock 0·001 0·047 0·00 0·991 12·38 0·030
Goldfinch −0·006 0·176 0·00 0·971 8·84 0·116
Greenfinch −0·022 0·054 0·16 0·686 1·62 0·898
Great tit 0·204 0·057 12·81 < 0·001 21·52 < 0·001
House sparrow 0·602 0·339 3·16 0·075 43·52 < 0·001
Robin 0·083 0·029 8·00 0·005 13·79 0·017
Reed bunting −0·567 0·289 3·84 0·050 7·05 0·217
Yellowhammer −0·161 0·080 4·10 0·043 9·58 0·088
(b)
Chaffinch 0·071 0·037 3·80 0·051 5·97 0·309
Goldfinch −0·014 0·190 0·01 0·942 7·15 0·209
Greenfinch −0·041 0·061 0·46 0·497 4·91 0·427
Reed bunting −0·615 0·306 4·04 0·044 12·83 0·025
Yellowhammer −0·145 0·081 3·21 0·073 12·45 0·029
χ1
2
χ5
2
633
Spatial patterns of
winter food use by
farmland birds
© 2006 The Authors.
Journal compilation
© 2006 British
Ecological Society,
Journal of Applied
Ecology
,
43
,
628–639
exist (
P
< 0·1). Dunnock counts tended to peak at
500 m
−
1 km separation (Fig. 2a), showing why a sim-
ple linear effect did not identify a significant slope.
Among the remaining species, none showed smooth
linear changes in count with separation (Fig. 2) and
several showed apparent thresholds at which counts
changed rapidly. There were two such thresholds for
blue tit, great tit, robin and reed bunting and one
for chaffinch, house sparrow and yellowhammer
(Fig. 2).
Post-hoc tests were used to investigate the signifi-
cance of the apparent thresholds in the effect of site
separation by combining counts from sites above
and below the threshold into single categories. This
approach maximized statistical power by estimating
two parameters (a two-level above/below threshold
variable) rather than the six in the analysis of the
categorical site variable, but the tests were otherwise
similar. These tests examined whether there was any
statistical support for the possible thresholds suggested
by the graphs in Fig. 2; they were made as objective as
possible by considering every apparently possible
threshold, rather than just the more obvious ones. All
species for which there was significant variation in
count (with separation as a categorical or a continuous
variable) were considered and separate tests were
conducted where there was more than one possible
threshold (Fig. 2). All single, significant threshold
distances, and the more significant ones where there
were two candidates, were at site separations of either
100 m or 500 m (Table 3).
-
Considering bird use of food patches in terms of bird-
minutes (2003–04) resulted in significant or near-
significant effects of site separation, as a linear variable,
for six of the 11 species tested (Table 4). All of these
relationships were positive, suggesting higher use of
isolated food patches, and all of these species, except
Fig. 2. Variation, with respect to site separation, in the use of feeding sites by all species for which this variation was significant,
in terms of maximum counts in the vicinity of food patches: (a) dunnock, (b) robin, (c) blue tit, (d) great tit, (e) chaffinch, (f) reed
bunting, (g) yellowhammer and (h) house sparrow. ‘Count Effect’ denotes parameter estimates (on the log scale) for a categorical
site type variable from models of count as a function of this variable (and controls; see text). Dashed lines show 95% confidence
intervals. Data shown were taken from models omitting controls for seed-rich habitat availability.
634
G. M. Siriwardena
et al.
© 2006 The Authors.
Journal compilation
© 2006 British
Ecological Society,
Journal of Applied
Ecology
,
43
,
628–639
blackbird, had also shown a significant increase in
use with separation when bird use was measured as
maximum count. The same six species also showed
significant or near-significant increases in site-use
with separation as a categorical variable (Table 4), the
exceptions being house sparrow (for which bird minute
data were too sparse to fit a model) and dunnock (for
which only the categorical variable was significant).
Note that the results for house sparrow and robin did
not reach significance at
P
= 0·05, but that all the other
species showed variation significant at
P
< 0·01 in one
or both tests.
No significant variation in bird-minutes was found
for yellowhammer or reed bunting (Table 4), con-
trasting with the maximum count results, although
there was weak evidence for a negative, linear effect
for reed bunting when controls for seed-rich habitat
availability were included (Table 4b). These controls
also altered the categorical site results for chaffinch and
goldfinch, as with maximum count data (chaffinch was
no longer significant and goldfinch became significant;
Table 4b).
Relationships between bird use and distance were
again often non-linear, stepped functions (Fig. 3). The
significance of apparent threshold separations (as
linear or categorical variables) was tested as before,
revealing significant threshold steps in bird use at 100
m or 500 m for four species: blackbird, blue tit,
chaffinch (marginally more significant at the 100 m
than the 500 m threshold) and great tit (Table 5). All of
the changes were positive, suggesting greater use of
more isolated food patches. There was only weak
evidence for a significant threshold for robin and the
pattern for goldfinch (significant only after controlling
for alternative resource availability; Table 4b) was com-
plex, probably reflecting the patchy distribution of the
species rather than any real effect of feeding (Fig. 3g).
- -
Full details of the radio-tracking and ring resighting
results are presented in Appendices S1 and S2, respec-
tively (see the supplementary material). Individual
Table 3. Results of tests investigating possible thresholds in the effects of site separation on maximum counts. The effect direction
column shows whether counts were higher (+) or lower (–) above the distance threshold(s) shown. The putative threshold
distances listed were tested individually: where two are shown for a single species, test results are given, respectively
Species Possible threshold(s) Direction Wald P
Blue tit 500 m, 2 km + 8·28, 6·02 0·004, 0·014
Chaffinch 500 m + 5·35 0·021
Great tit 500 m, 2 km + 13·67, 9·11 < 0·001, 0·003
House sparrow 100 m + 30·69 < 0·001
Robin 100 m, 1 km + 5·52, 0·97 0·019, 0·326
Reed bunting 100 m, 1 km – 3·41, 5·86 0·065, 0·016
Yellowhammer 500 m – 7·23 0·007
χ1
2
Table 4. Summary of effects of site separation results on bird use of feeding sites (bird-minute data: (a) results for all species with
no additional controls; (b) results for farmland-dependent seed-eaters controlling for areas of seed-rich habitat). Continuous
parameter estimates refer to slopes (on the log scale) of relationships between bird-minutes and log(separation). Note that the
data for house sparrow were too sparse to allow the fitting of a six-level site type effect
Species
Continuous (linear): separation Categorical: site type
Parameter estimate SE Wald PWald P
(a)
Blackbird 0·148 0·079 3·53 0·060 17·39 0·004
Blue tit 0·296 0·062 22·94 < 0·001 27·24 < 0·001
Chaffinch 0·174 0·059 8·80 0·003 10·95 0·052
Dunnock 0·023 0·060 0·14 0·705 11·75 0·038
Goldfinch 0·163 0·273 0·35 0·552 8·90 0·113
Greenfinch 0·120 0·098 1·50 0·221 7·50 0·186
Great tit 0·322 0·084 14·86 < 0·001 26·96 < 0·001
House sparrow 0·499 0·270 3·41 0·065 – –
Robin 0·105 0·093 0·61 0·059 10·16 0·071
Reed bunting −0·236 0·056 3·57 0·123 6·34 0·275
Yellowhammer −0·121 0·153 2·38 0·207 3·85 0·572
(b)
Chaffinch 0·118 0·055 4·54 0·033 6·17 0·290
Goldfinch 0·224 0·339 0·43 0·510 70·39 < 0·001
Greenfinch 0·075 0·093 0·65 0·419 3·48 0·626
Reed bunting −0·235 0·137 2·95 0·086 7·99 0·157
Yellowhammer −0·075 0·097 0·60 0·439 5·48 0·360
χ1
2
χ5
2
635
Spatial patterns of
winter food use by
farmland birds
© 2006 The Authors.
Journal compilation
© 2006 British
Ecological Society,
Journal of Applied
Ecology
,
43
,
628–639
radio-tagged birds were located at mean distances of
25–850 m from ringing sites. Minimum distances of,
effectively, zero were common, birds being detected at
the site of trapping. Maximum distances were generally
of the order of several hundred metres (median
629·3 m) and the greatest distance recorded was 1200 m
from the site of release. Considering individual birds’
mean distances from sites of ringing, they were located
230 m (SE 68 m) away, on average, which rose to 607 m
(SE 75 m) when individuals’ maximum distances were
considered.
No birds colour-ringed at sites in central blocks were
ever resighted at sites outside the block where they had
been ringed and none of the birds ringed in outer
blocks was ever seen away from their ‘home’ food
patches. Within central blocks, 30–40% of chaffinches
and greenfinches that were resighted were seen at least
once at sites up to around 1 km away, while these
figures rose to 40– 50% for yellowhammer and 55 –75%
for goldfinch. Many birds were therefore only recorded
where they had been ringed and the mean distances
individuals moved between feeding sites tended to be
small: even using the maximum distances recorded for
Fig. 3. Variation in the use of feeding sites, in terms of bird-minutes derived from point-count observations of birds feeding on
seed patches, by all species with respect to site separation for which this variation was significant: (a) blackbird, (b) robin, (c)
dunnock, (d) blue tit, (e) great tit, (f) chaffinch and (g) goldfinch. ‘Bird-minutes effect’ denotes parameter estimates (on the log
scale) for a categorical site type variable from models of the number of bird-minutes as a function of this variable as well as several
controls (see text). Dashed lines show 95% confidence intervals. All data were taken from models omitting controls for seed-rich
habitat availability, except for goldfinch (Table 4b).
Table 5. Results of tests investigating possible thresholds
in the effects of site separation on bird-use (bird-minute
data). The effect direction column shows whether counts
were higher (+) or lower (–) above the distance threshold(s)
shown. The putative threshold distances listed were tested
individually: where two are shown for a single species, test
results are given respectively
Species
Possible
threshold(s) Direction Wald P
Blackbird 100 m + 12·69 < 0·001
Blue tit 500 m + 25·02 < 0·001
Chaffinch 100 m, 500 m + 6·95, 6·64 0·001, 0·008
Great tit 500 m + 22·55 < 0·001
Robin 100 m + 2·79 0·095
χ1
2
636
G. M. Siriwardena
et al.
© 2006 The Authors.
Journal compilation
© 2006 British
Ecological Society,
Journal of Applied
Ecology, 43,
628–639
each individual, species means were 436 m (SE 57 m)
for yellowhammer, 244 m (SE 119 m) for chaffinch, 614
m (SE 48 m) for goldfinch and 360 m (SE 100 m) for
greenfinch.
Discussion
This study represents the first comprehensive attempt to
quantify movements of farmland birds in winter through
measuring responses to clumped and isolated food
resources. Although these resources were artificial,
most were used, often extensively, both by declining
species, such as yellowhammer, and by more stable or
increasing ones, such as chaffinch (Table 1). Given this
high use it seems likely that relationships with isolation
reflected resource exploitation rather than just local
distributions. We consider below patterns of resource
use with respect to the distances separating resource
patches, and the implications of the patterns for the
design of agri-environment scheme options aiming to
meet winter foraging requirements (Vickery et al. 2004),
in the context of bird foraging ecology.
Birds are likely to respond to variations in food
resource isolation in one of two ways. First, a pool of
individuals could distribute itself among the resources
within its winter home range, so less isolated patches
are used less and more isolated ones (where local
birds congregate) are used more. Second, the greater
resource availability presented by closely grouped food
sources could attract aggregations of individuals, such
that sites are used in proportion to the total resource
base present, or even that closely grouped sites are
overused in comparison with isolated ones. The
strategy adopted is likely to depend on the general
behavioural characteristics of a species. Thus, more
gregarious and more mobile species might tend to
exhibit the second pattern, while more solitary and
sedentary species might exhibit the first, as has been
shown by comparisons of the resource use of solitary
and social bee species (Johnson & Hubbell 1975;
Steffan-Dewenter et al. 2002).
Evidence for different species’ strategies comes from
comparing maximum count data at patches (measur-
ing numbers present) and bird-minute data (measuring
‘weight-of-use’ of the food). The most common pattern
was for bird use, in terms of both maximum counts and
bird-minutes, to increase as patches became more
isolated (Figs 2 and 3). This was apparent whether or
not controls for ambient resource availability around
patches were included and was evident for chaffinch
and house sparrow (among the granivores), as well as
for blue tit, great tit, robin and, perhaps, blackbird.
This increasing pattern suggests that larger groups of
birds used more isolated food patches and made dis-
proportionately more use of the food there. Among the
other species considered, dunnock maximum counts
and bird minutes showed a complex relationship with
separation, there was weak evidence for a second
complex relationship for goldfinch and there were no
significant effects for greenfinch. Reed bunting and
yellowhammer showed the opposite pattern to the
majority of species, with maximum counts declining
with isolation (Fig. 2), but there were no strong
patterns in the bird minutes results (Table 4). This
suggests that they formed larger groups at less isolated
patches, but with no associated increase in weight-
of-use of the food, i.e. that the higher counts did not
reflect greater use in terms of birds per unit time. This
would occur if flocks moved freely between the less
isolated patches, using food at a similar intensity to that
which would have been observed had the flocks,
instead, distributed themselves among these patches in
proportion to food availability. Although it is possible
that some patterns reflected variation in local abun-
dance instead of real resource location effects, most
species considered were widely distributed and any
such variation should have been averaged out by the
experimental replication. The more patchily distrib-
uted species, goldfinch and reed bunting, were most
likely to have been affected in this way.
Most of the increasing and decreasing patterns of
use with respect to separation were non-linear (and not
smooth). Instead, there was evidence for thresholds
at which foraging behaviour changed. Interestingly,
whether use increased or decreased with separation,
the thresholds were usually at the 100–500-m separa-
tions. Thus, for species with increasing patterns, such as
chaffinch, counts and weight-of-use were relatively low
at patches 500 m or less from other patches. Similarly,
for yellowhammer and reed bunting, whose resource
use decreased with separation, counts were higher at
patches 500 m or less from other patches (but weight-
of-use did not vary). Average counts at the central three
patches for yellowhammer tended to be two to three
times those at more isolated sites. Taken together,
the central patches provided approximately three times
as much food within a 500-m radius as the more iso-
lated patches (Fig. 1), suggesting that yellowhammers
used food in proportion to its availability (probably
moving between food patches as a large flock, each
individual perhaps sampling the landscape over a
larger scale). In contrast, both maximum count and
bird-minute data showed the same increasing pattern
with site separation for chaffinch, blue tit, great tit
and robin, so use of food at the central patches was
disproportionately low in relation to its availability.
Thus, the latter species seem to have divided themselves
between clustered food patches. Regardless of the strat-
egy chosen, however, the resource separation distance
at which behaviour changed was relatively small,
around 100– 500 m, suggesting that local populations
share resources within this radius, but that more widely
spaced resources are likely to be used by different,
discrete groups of birds.
All the discussion above considers individual species
independently but, in reality, multiple species use the
same environment and share its resources, so one
species’ use of a food resource could reduce or prevent
637
Spatial patterns of
winter food use by
farmland birds
© 2006 The Authors.
Journal compilation
© 2006 British
Ecological Society,
Journal of Applied
Ecology, 43,
628–639
that by another. Competition could be important in
determining feeding site use for subordinate species,
depending on the local abundance of dominants. How-
ever, it should only be a source of statistical noise for
our experiment because there is no reason to suppose
that ambient densities varied systematically between
feeding site types. Moreover, it is important to note that
competition will be a feature of any ‘seed-delivery
mechanism’ in practice, so it is appropriate to measure
responses to a model system in the presence of such
potential interspecific interactions.
Tests of the variation in context between feeding
stations at different separations revealed significant
or near-significant variation (and therefore potential
bias) in only two variables, distance from cover and
distance from human habitation. Patches 2 km from
others tended to be further from cover (P = 0·080 using
a categorical variable for site separation) and more
isolated patches tended to be closer to human sites
(P = 0·011 using the continuous variable for separation
but P > 0·1 using a categorical variable). None of the
species considered is averse to feeding within a few
hundred metres of rural gardens or farmyards, but the
latter effect could bias the results for human-associated
species, such as house sparrow. However, there was a
maximum difference of only 200 m between the average
distances from human habitation of any pair of feeding
site separations, and this was influenced heavily by sites
in one replicate. Considering also the lack of general
patterns across species revealed by analyses of the
influence of local- and landscape-scale habitat context
on food patch use (Siriwardena & Stevens 2004), it is
therefore unlikely that the results here were biased by
variation in feeding site context.
Indirect evidence, inferred from count data, there-
fore suggests that several farmland birds tend to move
less than 1 km between food resource patches. This
result was supported by direct evidence from colour-
ring resighting of yellowhammers, chaffinches, gold-
finches and greenfinches, and also radio-tracking of
yellowhammers in 2002– 03 (see Appendices S1 and S2
in the supplementary material).
Colour-ring resighting was conducted during, or for
short periods after, food patch monitoring watches, so
only considered birds using the patches; although the
latter were up to 10 km from sites of ringing, they
would have been increasingly difficult for birds to find
further from a central site. Radio-tracking placed no
constraint on the locations at which birds could be
recorded but the distance over which tags could be
detected was limited (< 100 m in undulating terrain).
Both methods considered only the birds that were
detected after release and it is unknown how represent-
ative of wider populations these individuals were,
particularly given that data were collected in only a
few areas and that tracking, in particular, was only
conducted over a brief period. Both the colour-ring
resighting and radio-tracking approaches may there-
fore have underestimated movements to some extent.
Nevertheless, the results were consistent with the
pattern suggested by the food patch monitoring data,
in which food resources within 500 m−1 km of one
another tended to be used by single-source popu-
lations. All three approaches had limitations that must
be acknowledged, but the fact that they all tend to
agree is encouraging.
It is important to consider these results in context
and to note a number of caveats to their interpretation.
First, and perhaps most important, they relate to
movements of birds in a landscape where food, as
artificial patches, was abundant and reliable. It might
be predicted that home ranges would be larger where
food supplies are sparse and less predictable; work to
address this question is currently underway. Secondly,
the results relate to an arable region of lowland
Britain and may not be applicable to other landscapes.
Calladine, Robertson & Wernham (2006) used radio-
tracking and ring resighting to follow the movements
of yellowhammers, tree sparrows and chaffinches
around a 25-km2 area of lowland farmland in southern
Scotland in a single winter. The study area included
six long-term feeding sites and recorded maximum
movements far greater than the present study, for
example yellowhammer movements of up to 3·6 km
and one chaffinch movement of 2·0 km, although both
species’ average movements over the 3– 4 days between
observations were of around 600 m. There are several
possible reasons for the difference from our results.
First, the study area itself was much larger than the
experimental blocks in East Anglia, potentially
facilitating the detection of movements over greater
distances. Second, data were collected earlier in winter,
when more food resource options were available for
birds to move between (stubbles, prior to ploughing,
and other relatively undepleted seed resources). Third,
the feeding sites had existed for some years and may
therefore have had a long-term impact on the winter
foraging patterns of local birds. The locations of stubble
fields and, to a lesser extent, wild bird cover crops usually
vary between winters, a factor that may be better reflected
by the relatively new artificial food patches created in
the present study.
Artificial food patches provide a model for seed food
resources in general. Although the Royal society for the
Protection of Birds (RSPB) operates ‘Bird Aid’, in
which farmers put out seed-rich harvest waste for birds
in winter (G. Q. A. Anderson & D. K. Stevens, personal
communication), the major options for winter food for
birds within agri-environment schemes are overwinter
stubbles and wild bird cover crops (Vickery et al. 2004).
These are likely to fit into agricultural practices and the
farm landscape more readily than seed patches and will
benefit a wider range of wildlife.
Artificial seed patches represent (at least initially) a
novel resource that birds must learn to use and they are
far smaller and less conspicuous than crops or stubble
fields. It is possible, therefore, that the relative ease of
discovery and recognition of field-scale seed resources
638
G. M. Siriwardena
et al.
© 2006 The Authors.
Journal compilation
© 2006 British
Ecological Society,
Journal of Applied
Ecology, 43,
628–639
will encourage farmland granivores to move greater
distances, so effective isolation of resource patches might
occur at greater separations than the 500 m−1 km
suggested by the results of this study. Field-scale seed
resources might also receive greater use per unit weight
of seed because it is harder for dominant individuals
(or species) to monopolize a sparser and structurally
more complex resource, allowing more birds to use it.
Much qualitative information is available on the
nature of food resources required by farmland birds in
winter in the form of stubbles and wild bird cover crops
(Moorcroft et al. 2002; Henderson, Vickery & Carter
2004; Newton 2004). A major gap in current knowl-
edge is the issue of scale: how much resource is required
(Bradbury et al. 2004; Vickery et al. 2004) and whether
it should be provided as scattered patches or single
blocks of habitat (Atkinson & Robinson 2002). Recent
work suggests that providing 10–20 ha of stubble
1 × 1-km squares in lowland farmland landscapes could
halt the population declines of yellowhammers and sky-
larks (Gillings et al. 2005). The present study suggests
that local populations of several farmland bird species
mix freely or share resources within an area with a
radius of around 500 m and that individuals tend to
range over distances below 1 km, so food resources like
this essential area of stubble might be provided most
effectively in patches 1 km or more apart. It should be
possible to put this spatial pattern into practice at the
farm scale, at which agri-environment schemes are
deployed, to maximize the cost-effectiveness of seed-
provision options in terms of the number of wintering
birds supported. This could be achieved, for example,
during the Farm Environment Plan stage of an appli-
cation for Higher Level Stewardship in England and
Wales.
Acknowledgements
This project was funded by Defra under project
BD1616, with additional support from EN. We thank
the BTO staff who helped with fieldwork (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
to their farms. Andy Wilson and Phil Atkinson helped
to set up the project, Steve Freeman provided valuable
statistical advice, Guy Anderson and Danaë Stevens
participated in much helpful discussion and three
anonymous referees provided useful comments. Thanks
to CJ Wildbird food for assistance with seed supplies
and Brian Cresswell and Sean Walls of Biotrack for
radio-tracking advice.
References
Atkinson, P.W. & Robinson, R.A. (2002) Landscape diversity
and bird populations: the need for regionally flexible
and scale-dependent agri-environment schemes. Avian
Landscape Ecology: Proceedings of the 11th IALE (UK)
Conference (eds D.E. Chamberlain & A.M. Wilson),
pp. 289–299. International Association for Landscape
Ecology, Northampton, UK.
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, UK. http://
www.bto.org/birdtrends (accessed 14 April 2006).
Boatman, N.D. & Stoate, C. (2002) Growing crops to provide
food for seed-eating birds in winter. Aspects of Applied
Biology, 67, 229– 236.
Bradbury, R.B. & Stoate, C. (2000) The ecology of yellow-
hammers Emberiza citrinella on lowland farmland. The
Ecology and Conservation of Lowland Farmland Birds (eds
N.J. Aebischer, A.D. Evans, P.V. Grice & J.A. Vickery),
pp. 165–172. British Ornithologists’ Union, Tring, UK.
Bradbury, R.B., Browne, S.J., Stevens, D.K. & Aebischer, N.J.
(2004) Five-year evaluation of the impact of the Arable
Stewardship Pilot Scheme on birds. Ecology and conserva-
tion of lowland farmland birds. II. The road to recovery.
Ibis, 146 (Supplement 2), 171–180.
Buckingham, D.L., Evans, A.D., Morris, T.J., Orsman, C.J.
& Yaxley, A. (1999) Use of set-aside in winter by declin-
ing farmland bird species in the UK. Bird Study, 46, 157–
169.
Calladine, J., Robertson, D. & Wernham, C.V. (2006) The
movements of some granivorous passerines in winter on
farmland. Ibis, 148, 169–173.
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, 771–788.
Díaz, M. (1990) Interspecific patterns of seed selection among
granivorous passerines: effects of seed size, seed nutritive
value and bird morphology. Ibis, 132, 467– 476.
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
(B), 268, 25–29.
Gillings, S. & Fuller, R.J. (2001) Distribution and habitat
preferences of skylarks Alauda arvensis wintering in Britain
in 1997/98. Bird Study, 48, 293– 307.
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 popu-
lation trends. Proceedings of the Royal Society of London
(B), 272, 733–739.
Henderson, I.G., Cooper, J., Fuller, R.J. & Vickery, J.A.
(2000) The summer abundance and distribution of birds on
set-aside and neighbouring crops on arable farms in
England. Journal of Applied Ecology, 37, 335–347.
Henderson, I.G., Vickery, J.A. & Carter, N. (2004) The use of
winter crops by farmland birds in lowland England.
Biological Conservation, 118, 21–32.
Johnson, L.K. & Hubbell, S.P. (1975) Contrasting foraging
strategies and coexistence of two bee species on a single
resource. Ecology, 56, 1398–1406.
Krebs, J.R., Wilson, J.D., Bradbury, R.B. & Siriwardena, G.M.
(1999) The second silent spring? Nature, 400, 611–612.
Moorcroft, D., Whittingham, M.J., Bradbury, R.B. & Wilson,
J.D. (2002) Stubble field prescriptions for granivorous
birds: the role of vegetation cover and food abundance.
Journal of Applied Ecology, 39, 535–547.
Newton, I. (2004) The recent declines of farmland bird
populations in Britain: an appraisal of causal factors and
conservation actions. Ibis, 146, 579– 600.
Peach, W.J., Lovett, L.J., Wooton, S.R. & Jeffs, C. (2001)
Countryside Stewardship delivers cirl buntings (Emberiza
639
Spatial patterns of
winter food use by
farmland birds
© 2006 The Authors.
Journal compilation
© 2006 British
Ecological Society,
Journal of Applied
Ecology, 43,
628–639
cirlus) in Devon, UK. Biological Conservation, 101, 361–
374.
Perkins, A.J. & Anderson, G.Q.A. (2002) Seed selection by
tree sparrows Passer montanus: determining appropriate
seeds for supplementary feeding on farmland. Aspects of
Applied Biology, 67, 213– 220.
SAS Institute Inc. (2001) Sasonlinedoc (R), Version 8. SAS
Institute Inc., Cary, NC.
Siriwardena, G.M. & Stevens, D.K. (2004) Effects of habitat
on the use of supplementary food by farmland birds in
winter. Ibis, 146 (Supplement 2), 144–154.
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, 24– 43.
Siriwardena, G.M., Baillie, S.R. & Wilson, J.D. (1998) Varia-
tion in the survival rates of some British passerines with
respect to their population trends on farmland. Bird Study,
45, 276–292.
Steffan-Dewenter, I., Münzenburg, U., Bürger, C., Thies,
C. & Tscharntke, T. (2002) Scale-dependent effects of
landscape context on three pollinator guilds. Ecology, 83,
1421–1432.
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,
19–39.
Wilson, J.D., Taylor, R. & Muirhead, L.B. (1996) Field use
by farmland birds in winter: an analysis of field type
preferences using resampling methods. Bird Study, 43,
320–332.
Wotton, S., Rylands, K., Grice, P., Smallshire, D. & Gregory,
R. (2004) The status of cirl buntings in Britain and the
Channel Islands in 2003. British Birds, 97, 376– 385.
Received 3 October 2005; final copy received 9 February 2006
Editor: Des Thompson
Supplementary material
The following supplementary material is available as
part of the online article (full text) from http://
www.blackwell-synergy.com
Appendix S1. Winter movement patterns of yellowham-
mers Emberiza citrinella: evidence from radio-tracking.
Appendix S2. Winter movement patterns of granivorous
farmland passerines: evidence from colour-ringing.