Agriculture, Ecosystems and Environment 112 (2006) 21–40

www.elsevier.com/locate/agee

Effects of agricultural management on the use of

lowland grassland by foraging birds
Dave L. Buckingham *, Will J. Peach, Derren S. Fox
Royal Society for the Protection of Birds, Conservation Science, The Lodge, Sandy, Bedfordshire SG19 2DL, UK
Received 17 November 2004; received in revised form 9 June 2005; accepted 28 June 2005
Available online 15 September 2005

Abstract

A field-scale correlative study was used to identify which factors had the greatest influence on the usage of agricultural grassland by
foraging birds in the English West Midlands. The study extended previous work by directly comparing a more complete range of lowland
grassland management practises, bird species and seasons. Sward structure had more influence on bird usage than botanical composition. Bird
species fell into two groups based on their sward structure preferences, which closely reflected where they obtained their food. Species that
feed on soil-dwelling invertebrates selected short swards, while species that feed on sward-dwelling invertebrates or seeds selected taller
swards with greater spatial heterogeneity. Grazing had a greater influence on grassland usage than sward age and other management practices.
Birds mainly responded positively to grazing, especially by cattle. Weed control reduced the usage of grass fields by granivorous birds during
summer and winter. Intensive grazing systems create and maintain short, uniform swards that favour bird species foraging for soil-dwelling
invertebrates, but not those reliant on seeds or sward-dwelling invertebrates. It is proposed that excessive defoliation of agricultural grasslands
(associated with intensive grazing and mowing regimes) impacts granivorous birds by reducing prey abundance. Reductions in grazing
intensity and the avoidance of weed control should increase food availability for granivorous and insectivorous birds on grass fields.
# 2005 Elsevier B.V. All rights reserved.

Keywords: Farmland birds; Grassland management; Sward structure; Grazing; Mixed models

1. Introduction farming systems. Pastoral farming has increased in

Permanent grassland accounts for a large proportion of
all agricultural land in Britain (66% in 2001) and Europe
(40% in the EU15 countries in 2001) (UN Food and
Agriculture Organisation FAOSTAT data, updated Feb-
ruary 2004, www.apps.fao.org/default.asp). Since the mid-
twentieth century, grassland management has intensified to
increase livestock forage and fodder production. The
resulting habitat changes have been widespread and
pervasive. Grasslands of nature conservation value, mainly
defined by botanical communities, now comprise less than
2% of the area of lowland grassland in England and Wales
(Blackstock et al., 1999). Farms have also specialised into
either arable or livestock production, abandoning mixed
* Corresponding author. Tel.: +44 1767 680551; fax: +44 1767 692365.
E-mail address: david.buckingham@rspb.org.uk (D.L. Buckingham).
western Britain, which now holds over 60% of the area
of grassland in Britain, compared to just 12% in eastern
arable areas (MAFF, 1997).
Farmland bird populations have declined in the UK since
the 1970s (Siriwardena et al., 1998). Most research into bird
declines has focussed on arable systems, where there is
increasing evidence of links to intensification (e.g. Brickle
et al., 2000). A general intensification of pastoral farming
practice coincided with relatively severe declines in the
abundance and range of farmland birds, particularly seed-
eating species, in western pastoral regions of Britain
(Chamberlain and Fuller, 2000). The presence of arable
crops in pastoral areas is associated with increased densities
of granivorous farmland birds (Robinson et al., 2001), so the
loss of arable cropping from many livestock farms may
account in part for the declines in western Britain. The
intensification of grassland management has probably

0167-8809/$ – see front matter # 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.agee.2005.06.019
22 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40
caused a reduction in seed and invertebrate densities and
possibly their availability to foraging birds (Vickery et al.,
2001). However, the key management practices and the
mechanisms involved are poorly understood.
The UK government has committed to reversing the long-
term decline in the numbers of farmland birds by 2020
(Defra, 2002). Meeting this Public Service Agreement will
require a better understanding of avian ecology on grass-
lands, both to establish the mechanisms involved and to
design remedial measures. The development of potential
conservation measures on arable farmland is much further
advanced (e.g. Boatman et al., 2000).
The intensification of grassland management has
involved several interrelated changes in farming practice.
The principal changes that may have impacted bird
populations are increased use of fertilisers (especially
inorganic nitrogen), the switch from hay to silage, increased
stocking densities, reseeding of grasslands and drainage
(Vickery et al., 2001). All of these actions promote
structurally uniform, dense swards dominated by competi-
tive ryegrasses Lolium spp. and such swards now dominate
agricultural landscapes. Grassland management has devel-
oped along the same lines throughout the parts of western
Europe where farmland birds have declined (Lazenby, 1988;
Donald et al., 2001b).
Behavioural studies have described foraging site selec-
tion by birds as a trade-off between energy intake rates and
perceived predation risk. Vegetation structure influences this
trade-off via three main factors: food abundance, predation
risk and food accessibility. Food intake rates are higher
where food is more abundant (e.g. Brodman et al., 1997).
But food intake rates are lower where visibility is impaired:
for instance, deeper swards force birds to spend a greater
proportion of their time watching for predators, rather than
feeding (Devereux et al., 2004; Whittingham and Evans,
2004). Deeper swards also obstruct access to the bird and
conceal food items, further reducing intake rates (Whitting-
ham and Evans, 2004). Numerous studies have demonstrated
the importance of minimising the predation risk, but risk
avoidance strategies and consequently selection responses to
sward structure vary substantially between species (Watts,
1991; Devereux et al., 2004). No behavioural studies have
explicitly considered situations where the effects of
vegetation structure on predation risk conflicted with those
on food abundance. Grasslands are one such habitat as
certain key prey items (e.g. herbivorous invertebrates,
Morris, 2000) are more abundant in taller swards, where
predation risk is relatively high and accessibility relatively
poor. This behavioural framework needs to be applied to
explaining the effects of grassland management on foraging
site selection in order to identify the limiting factors before
practical solutions can be developed.
The aim of this study was to identify the key components
of lowland grassland management that influence the
suitability of grass fields as feeding habitats. A correlative
approach was used to generate hypotheses for more focussed
future work. A large sample of fields was employed to assess
the generality of foraging responses across the full suite of
management practises, bird species and seasons. Previous
studies have been restricted to winter and have considered
narrower ranges of management practices or bird groups
(Tucker, 1992; Perkins et al., 2000; Barnett et al., 2004). It is
particularly important to understand factors affecting the
utility of grasslands for birds during summer because the
abundance and quality of grassland can be a key factor
influencing the breeding performance of farmland birds (e.g.
Evans et al., 1997).
2. Methods
2.1. Study areas
A sample of 23 farms representing a wide range of
pastoral businesses was selected for study in a mixed
farming region in the English West Midlands (Shropshire,
Staffordshire and Cheshire). The farms comprised approxi-
mately 70–90% grassland, with the remaining area in arable
crops. Farms were selected to stratify the sample of fields
across the farming practices and management gradients
(such as fertiliser inputs, Fig. 1) occurring in the region.
Thus, the sample maximised sensitivity to management
effects but did not sample each management type in
proportion to its area in the region. The same fields were
surveyed in both summer and winter, though some fields
were lost or subdivided between seasons (1392 ha of grass
fields, n = 388 in winter and n = 373 in summer).
Thirteen farms had dairy herds and ten raised beef cattle.
Sheep grazed nine farms during summer and 15 during
winter. In summer, almost all grazed fields were grazed by
cattle and one-third by sheep. In winter, about three quarters
of grazed fields were sheep grazed and half cattle grazed.
Horses were kept on five farms, including two large livery
stables and a racehorse stud, totalling 40 fields. Six farms
were managed as organic dairies. Agri-environment
management agreements were in place on 15 farms
(Countryside Stewardship Scheme, CSS): 12 farms with
whole-field management options and five with grass margin
options. During the 5 years prior to fieldwork, hay was
harvested on 17 of the farms and silage on 18. Zero-grazing
(involving daily mowing of grass for consumption by cattle
housed elsewhere on the farm) was practised on two
conventional dairies. Sward ages ranged from newly
established leys to permanent pastures over 160 years old
(25% fields
5 years old, 25% fields 50 years old). Sward
types ranged from high-input ryegrass Lolium monocultures
to low-input species-rich meadows. The predominant NVC
communities were MG6 (L. perenne-Cynosurus cristatus
grassland) and MG7 (L. perenne leys), with a smaller sample
of MG4 (Alopecurus pratensis-Sanguisorba officinalis
grassland) and MG5 (C. cristatus-Centaurea nigra unim-
proved grassland) (Rodwell, 1992).
 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40
Fig. 1. Nitrogen inputs to fields in the study area. (a) Distribution of annual nitrogen inputs from agricultural sources including fixation of atmospheric nitrogen
by clover 1995–1999 (XNALL/5) and (b) estimated percentage of total nitrogen inputs fixed by clover in 1999.
2.2. Bird usage of grass fields
Birds were counted by walking parallel transects 50 m
apart after first walking the entire field perimeter and
scanning the field with binoculars. At the discretion of the
surveyor, closer transect spacings were used to avoid
overlooking birds in taller swards. All birds feeding on or
over the field were counted and care was taken to avoid
double counting of flushed birds. Only grass fields were
searched and the routes taken around each farm were
randomised between visits to control for time of day. Each
farm survey took one whole day to complete. Two
fieldworkers did all the surveys, covering alternate areas
on each visit. Bird counts did not take place within one hour
of dawn or dusk, during rain, or in winds greater than
Beaufort force 4.
Bird usage was measured four times during the winter of
1999/2000, and four times during the summer of 2000. This
sampling rate was relatively low (cf. Perkins et al., 2000),
allowing a larger sample of fields. The winter bird counts
took place between mid-November 1999 and mid-March
2000, thereby avoiding peak migration periods. Summer
counts took place from April to early September. Only birds
of breeding age were counted during the summer and birds
that were not foraging were excluded from analyses (e.g.
displaying skylarks Alauda arvensis L. and loafing gulls
Larus).
2.3. Field characteristics
A range of characteristics that might have influenced bird
usage was recorded for each field (Appendix A). An index of
field enclosure was calculated following Wilson et al. (1997)
23
but with weighting scores adjusted to reflect regional hedge
management practice:
1  L1 þ 2  L2 þ 3  L3 þ 4  L4
HEDGEIND ¼ P
Li
where L1 is the length of hedges or fences up to 1 m tall, L2
hedges between 1 and 3 m tall, L3 hedges over 3 m tall and
L4 the combined length of tree-lines over 8 m tall, woods and
buildings. The index reflects the degree of enclosure by
vertical structures. An index of field shape (AREAIND) was
derived by dividing the field area by the area of a circular
field with the same perimeter length. Low AREAIND scores
identified narrow fields with a higher proportion of their area
close to the boundary; high scores indicated round fields
with a proportionally larger field centre.
Farmers were asked to classify the soils in each field as
clay, clay-loam, loam, peat and sandy. Soil bulk density was
scored from light to heavy by ease of cultivation. The pH-
buffering capacity of soils was graded by the ease of
maintaining near-neutral soil pH (acid, slightly acid or
neutral). Calcareous soils were absent from the study sites.
Sward structure was assessed using sward height
measurements. An HFRO sward stick (Hodgson et al.,
1986) was used to measure sward height at 20 points evenly
distributed along a fixed ‘‘M’’-shaped transect. Each transect
started at one end of a field, crossing from side to side to
cover the whole field. The sward surface was defined as the
height at which the descending Perspex window first
touched live foliage of any plant, excluding the leafless
upper parts of flowering stems.
Sward composition was quantified using a 0.7 m  0.7 m
square quadrat, subdivided into 49 10 cm  10 cm squares.
24 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40
Twenty quadrat measurements were collected from each
field, at the same points as the sward height measurements.
Three sward components were measured: bare ground,
agricultural clovers (red clover Trifolium pratense L. and
white clover T. repens L.) and forbs, excluding the
agricultural clovers. The number of grid squares occupied
by each sward component was counted. A grid square was
considered occupied if the sward component covered at least
one quarter of the square.
Sward measurements were collected once in each half of
the winter. During summer, sward heights were measured
immediately after every bird count, as plant growth was
rapid and defoliations were frequent. Quadrat measurements
were collected after the first and third summer visits and on
other visits if the sward had changed significantly (usually
because of mowing). Seasonal means of sward measure-
ments were used in the models (Appendix A). Standard
deviations of measurements were calculated first for each
visit and then averaged.
Management histories for each field were obtained by
interviewing farmers. Information was collated on manage-
ment during the current season, the preceding season and the
preceding 5 years. The management variables are sum-
marised in Appendix B.
Nitrogen inputs from fertilisers were subdivided into
organic and chemical sources. The nitrogen content of organic
fertilisers (farmyard manure, slurry and dirty water) was
calculated using standard conversion tables (MAFF, 1994).
Atmospheric nitrogen fixation by clover was estimated using
a model based on the percentage of clover in the sward and the
sward age (Kristensen et al., 1995). Average clover cover
scores in summer 2000 were used in the models and clover
cover was assumed to remain constant between years.
Variables measuring total nitrogen inputs combined inputs
from fertilisers and clovers (Fig. 1); 114 fields received
average annual nitrogen inputs below 50 kg N ha1.
Detailed records of the intensity and duration of grazing
were not generally available. Fields were classed as grazed
by cattle, sheep or horses in summer or winter (after
October). The number of survey visits when recent grazing
was evident (by the presence of animals or fresh dung) was
used to measure grazing frequency, but could not quantify
grazing intensity or duration. The area of large patches of
bare ground (1 m2) was measured during the bird counts.
Livestock trampling around hedges, gateways and feeding
stations produced almost all of the large bare patches, so
their combined area was used as an indirect measurement of
grazing intensity. Few large bare patches were produced by
machinery. Quadrat measurements of fine-scale bare ground
cover were not correlated with the areas of large bare patches
(rs < 0.2 in both seasons).
Fodder crops were harvested using four methods: hay,
silage, haylage and zero-grazing. Fields were often mown
more than once in a growing season, sometimes using more
than one harvesting technique. To allow a direct comparison
between hay and silage, variables contrasted years when any
hay was cut with years when only silage was cut. To reflect
the likelihood of grass setting seed and intervals between
successive cuts, haylage crops were treated as hay, and each
zero-grazing cutting cycle was treated as a single silage cut.
Weed control was subdivided into two categories:
herbicides (mainly reactive spot-spaying) and mechanical
methods (topping and hand weeding of ragwort Senecio
jacobea L.). Treatments for endoparasitic worms or
ectoparasites were only counted when the active ingredient
could reach the field (i.e. the animals were treated on the
field or grazing by animals treated with avermectin boluses).
Insecticides were not used on the study sites.
Fields with CSS options were subdivided into those with
whole-field or margin options. Whole-field options included
extensive grazing, hay cuts and reseeding with customised
seed mixtures or hay from species-rich meadows. Margin
options allowed a fallow period during the growing season.
Mowing was not permitted on margins if grass was cut in the
centres of fields, but grazing was allowed under some
agreements. Nutrient inputs were restricted in all cases.
2.4. Data analysis
Bird usage on individual grass fields was modelled using
general linear mixed models (GLMMs) in which ‘farm’ was
declared as a random variable. The GLMMs modelled non-
independence of field usage within farm units, unlike GLMs,
which would assume independence (McCulloch and Searle,
2001; Milsom et al., 2000). Not all treatments (management
practices) took place on each farm, resulting in an
incomplete blocks design. GLMMs are better suited to
incomplete blocks designs than GLMs, because they recover
information from the between-farms error stratum, reducing
between-farms bias (Brown and Prescott, 1999). In addition,
because between-farm effects were controlled in the
GLMMs, the estimated fixed effects were more general
(wider inference space), applying to any field, not just the
fields within the specific farm groupings of the study.
GLMMs were fitted using the macro GLIMMIX supplied
with SAS 8.02 (Littell et al., 2002).
Frequency of occurrence on each field was modelled
using GLMMs with a binomial error distribution and a logit
link function (a ‘binomial trials’ model). Frequency models
readily meet statistical requirements but discard information
on the numbers of birds present. They may still be
representative of total bird usage if frequencies and total
counts are strongly correlated (Perkins et al., 2000). If this
was not the case, count data may hold additional information
on habitat preferences, so maximum counts on each field
were also modelled, using GLMMs with a Poisson error
distribution and a log link function. Low Spearman
correlation coefficients (rs < 0.4) between frequencies of
occurrence and total winter counts on each field identified
four species where this was necessary: woodpigeon
Columba palumbus L., yellowhammer Emberiza citrinella
L., fieldfare Turdus pilaris L. and goldfinch Carduelis
 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40
carduelis L. (winter only, rs = 0.25–0.31). The GLIMMIX
macro automatically rescaled the model deviance to correct
over/under-dispersion. Highly under-dispersed models (dis-
persion factor <0.3) were rejected.
Field usage by each species was modelled separately in
winter and summer. Within each season, two models were
produced using alternative sets of explanatory variables,
describing sward condition or management. Most of the
variables had been shown to be important in previous bird
usage studies. Each model was built in two stages. In the first
stage of model development, significant predictors of field
usage were identified from a suite of 26 variables describing
field characteristics and adjacent land use (Appendix A).
The influence of these variables is not interpreted here, but
their influence was controlled for before modelling other
explanatory variables. In the second stage of model
development, either sward condition or management
variables (6 and 39 variables, respectively) were added to
the set of field characteristics variables. Potential predictor
variables were incorporated into GLMMs as fixed effects
using a manual step-up procedure. At each step, the variable
with the lowest P value, based on the Wald statistic (Littell
et al., 2002), was added to the model and the process
continued until no further variables had a significant
influence on the dependent variable. Additional fixed effects
were only included in the model if all previously included
fixed effects remained significant (at P < 0.05) in Type 3
Wald tests. Quadratic terms were fitted to check whether
selected continuous variables had non-linear effects on bird
usage. Information theoretic approaches are recommended
for model selection in correlative studies (Rushton et al.,
2004), but they cannot be used to select fixed effects in
mixed models (Welham and Thompson, 1997). Each time
the fixed effects are changed, the random effects are re-
estimated, changing the error structure (residual space) and
preventing comparisons of information criteria between
models. In practice, most of the observed changes in the
random variables were small, so the change in the Akaike
Information Criterion (AIC) associated with the inclusion of
each predictor variable was retained as an approximate (not
absolute) measure of explanatory strength.
Many of the explanatory variables were correlated and this
problem was addressed by variable substitution and principal
components analysis. Correlations between pairs of predictor
variables were assessed using Spearman rank correlation
coefficients, Cramér’s V (for pairs of categorical variables) or
Kruskal–Wallis tests (for combinations of continuous and
categorical variables). Factors with more than two levels were
compared using the test statistic chi-squared divided by
degrees of freedom (Table 2). As tests of model robustness,
each sward or management variable in final models was
replaced with the two most strongly correlated sward and
management variables. In practise, replacing variables
generally resulted in one or more of the other variables
ceasing to be significant, in which case the alternative model
was rejected. Principal components analysis (PCA) was used
25
to simplify a group of 13 inter-correlated variables, describing
long-term patterns of mowing and grazing. The four resulting
axes (Appendix C) independently described the major
gradients of cropping practice across fields, thereby aiding
interpretation. Sward measurements were strongly inter-
correlated, but a PCA did not provide interpretable axes.
Nitrogen inputs were also strongly correlated, but the
variables were retained so that non-linear relationships with
bird usage could be modelled explicitly.
Models were constructed for species that occurred on a
minimum of about 20 fields. Most granivorous birds were
scarce on grassland so datasets were combined to look for
generic effects on usage. Winter data were combined for all
granivorous passerines excluding skylarks, which mainly
use field centres (rather than margins) and, unlike other
granivores, can supplement their diet by grazing (Donald
et al., 2001a). Summer data for granivorous passerines were
grouped according to chick diet. Species that feed their
chicks invertebrates (skylark, buntings, sparrows) formed
the mixed diet seedeaters group while species feeding their
young only seeds (Cardueline finches) formed the obligate
seedeaters group.
Model fit was examined graphically, using plots of
observed and predicted values (Milsom et al., 2000).
Predicted values were calculated for the full range of values
of the variable of interest and mean values of the other
variables in the model (or modal values for categorical
variables). For categorical variables, observed probabilities
of occurrence were calculated as the average frequency
across all fields falling into each category, divided by the
number of visits. For continuous variables, the fields were
grouped into five bins, based on the variable of interest, with
roughly equal numbers of fields in each bin. Observed
probabilities of occurrence were then calculated across fields
in each bin and assigned to the mean value of the continuous
variable within each bin.
3. Results
Birds used most fields on at least one survey visit (3.6%
of fields were unused in winter, 6.7% in summer) and
feeding birds were detected on two-thirds of all field-visits.
Numbers of birds and species were generally low: on 90% of
field-visits, there were
4 species and
32 birds in winter
(
3 species and
10 birds in summer). Blackbird T. merula
L. was the most ubiquitous species, utilising 65% of fields.
Carrion crow Corvus corone L., woodpigeon and magpie
Pica pica L. were the only other species to occur on over a
third of fields. In winter, 21 species were sufficiently
ubiquitous to produce models and 19 species in summer.
3.1. Sward condition
Sward height was the most consistent predictor of grass
field usage with all species except skylark (Fig. 2a) avoiding
26 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40

the tallest swards in summer and winter (Table 1,
Appendix D.1). The strongest avoidance of taller swards
was exhibited by thrushes Turdus, starling Sturnus vulgaris
L. and corvids in both seasons and by pied wagtail Motacilla
alba L. during summer (Fig. 2b,c and e). All the thrushes and
starling were observed less frequently on the shortest swards
(Fig. 2b and c), although the models did not always reflect
this. Field usage by yellowhammer (winter peak 10.7 cm,
Fig. 2d) and linnet C. cannabina L. (summer peak 18.5 cm,
Fig. 2f) peaked at sward heights above the respective
seasonal averages. Granivorous passerines only showed
preferences for tall or above average sward heights.
Fields with more bare ground (or its positively correlated
standard deviation) were preferred by five species during
winter and three species plus the mixed-diet seedeater guild
in summer. Granivorous passerines only showed positive
associations with bare ground cover.
Fields with high forb cover tended to be avoided during
winter (Table 1, Appendix D.1, Fig. 2g); only redwing T.
iliacus L. preferred fields with higher forb densities (peak

Fig. 2. Effects of sward condition on field usage. Predicted responses from the models are plotted as lines and observed values, averaged in bins, are plotted
1S.E. MHFRO is the mean sward height (graphs a–f); MFORB is the mean forb cover (graphs g–i).
 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40 27

Fig. 2. (Continued ).

cover score 4.1). Linnets preferred forb-rich swards in
summer, but yellowhammers and other mixed-diet seedea-
ters did not (Fig. 2h and i). Clover-rich fields were preferred
by redwings and seedeaters during winter and by obligate
seedeaters and four other species during summer (peak cover
scores in quadratic models were all greater than the mean).
3.2. Long term grazing and mowing patterns
and pheasant Phasianus colchicus L. and were not avoided
by any species (Table 2, Appendix D.2). Blackbird and robin
Erithacus rubecula L. preferred summer cattle grazing over
hayfields and horse grazing (CROP4), while pied wagtails
preferred grazing to mowing (CROP1). Long-term cropping
patterns influenced few species in summer (Table 2). CROP3
contrasted cattle grazing (preferred by pied wagtails) with
hay (preferred by yellowhammers).

Fields with a history of silage cutting followed by winter 3.3. Recent grazing and mowing
cattle grazing (as reflected by a high CROP3 PCA score,
Appendix C) were strongly preferred in winter by thrushes, Recent grazing was one of the strongest predictors of
starlings, granivorous passerines, jackdaw C. monedula L. grass field usage for a wide range of bird species in summer
28 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40

However, woodpigeon and stock dove avoided frequently

mown fields during summer. Stock dove was the only
species to select hay fields (Appendices D.2 and D.3).

3.4. Fertilisers

Effects of fertilisers on field usage were weaker than

those of grazing (smaller AIC reductions) and lacked
consistency across species (Table 2). Several species
(including jackdaw and woodpigeon, Fig. 4a) preferred
fields with intermediate nitrogen inputs. Negative responses
to nitrogen inputs were largely confined to small insectivores
(Appendices D.2 and D.3, Fig. 4b). Granivorous passerines
only showed weak responses to fertiliser use, including a
negative effect for obligate seedeaters in summer (Fig. 4c)
and a positive effect of chemical fertilisers for wintering
skylarks. There was no consistent preference between
organic or chemical fertilisers (Table 2).

3.5. Other management factors

Fig. 2. (Continued ).

There were few consistent effects of other management

and winter (Table 2). Cattle grazing in the preceding summer factors on field usage (Table 2). Regular chemical weed
and the area of bare ground patches (mainly caused by control reduced field usage (Table 2) by granivorous
grazing cattle) had a strong and consistent positive effect on passerines in winter (Fig. 5a) and by mixed-diet seedeaters
winter field usage by thrushes, starlings, corvids, small in summer, while mechanical weed control reduced field
insectivores and seedeaters (Appendix D.2, Fig. 3a and b). usage by wintering skylarks (Fig. 5b). Herbicide applica-
Wintering yellowhammers avoided fields grazed by cattle in tions were also associated with increased field usage by
the preceding summer; the only species to do so (Fig. 3c). swallow and magpie in summer. Rolling and harrowing were
However, a Poisson regression model showed that winter strongly inter-correlated, and mainly positively associated
grazing frequency (NGWIN99) had a positive effect on with field usage by birds (Table 2, Appendices D.2 and D.3).
winter yellowhammer counts. Sward age was not an important predictor of grassland
Frequent summer grazing and large bare patches usage, but was correlated with other influential variables
promoted summer field usage amongst a similarly wide such as field surface topography (Kruskall–Wallis test
range of species (Table 2, Appendix D.3). No granivorous between TOPOG and LAGE, x2/d.f. = 49.3, P < 0.0001),
passerines avoided intensively grazed fields, but chaffinch which was treated as a ‘nuisance’ variable in the models.
Fringilla coelebs L. was the only one to prefer them Surface topography had a positive influence on summer field
(positive response to BAREBITS). There was no consistent usage by chaffinches, rooks and mixed seedeaters (DAIC
preference for particular grazing animals in summer range 6.6–10.8), but a weaker negative influence on
(Table 2). yellowhammers (DAIC = 2.0). Substitution of LAGE for
Mowing affected fewer species (Table 2). Wintering TOPOG did not yield significant models. Woodpigeon and
mistle thrushes T. viscivorous L. and woodpigeons preferred yellowhammer (winter Poisson regression model) preferred
fields managed for silage during the previous summer. younger swards. Skylark was the only species to respond to
Table 1
Summary of sward condition variables in bird usage models
Winter Summer Most closely correlated variables
+  NL DAIC +  NL DAIC
MHFRO 1 4 5 136 0 7 2 128 MSDHFRO (rs 0.87), MBARE (rs 0.35)
MBARE 5 1 1 80 1 0 0 2 MSDBARE (rs 0.43), MHFRO (rs 0.35)
MFORB 1 6 1 52 1 1 3 30 MCLOV (rs 0.26), MHFRO (rs 0.15)
MCLOV 3 0 0 12 0 0 5 49 MFORB (rs 0.26), MBARE (rs 0.09)
MSDHFRO 0 0 0 0 2 2 0 19 MHFRO (rs 0.87), MBARE (rs 0.27)
MSDBARE 0 1 0 2 3 0 0 11 MBARE (rs 0.43), MHFRO (rs 0.14)
The numbers of models incorporating each variable are tabulated according to the direction of the relationship: positive, negative or non-linear (+//NL) and
DAIC is the sum of the AIC reductions caused when the variable entered the models. The most closely correlated management variables are appended, with
correlation coefficients.
 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40 29

Table 2
Summary of management variables in bird usage models
Winter Summer Most closely correlated variables
+  NL DAIC +  NL DAIC
Long-term grazing and mowing PCA
CROP1 1 0 9 0 0 0 LAGE (rs 0.50), MECHANCL (rs 0.40)
CROP2 0 1 4 0 1 2 NALLORG (rs 0.29), MECHANCL (rs 0.25)
CROP3 8 0 51 1 1 7 PARASITE (rs 0.38), BAREBITS (rs 0.24)
CROP4 0 2 13 3 0 21 MECHANCL (rs 0.27), BAREBITS (rs 0.20)
Recent grazing
BAREBITS 4 0 23 5 0 31 NGSUM2K (rs 0.58), NMOW2K (rs 0.44)
NGWIN99 2 0 19 PARASITE (rs 0.47), NALLORG (rs 0.35)
GR99C0W 8 1 125 LAGE (x2 37.5), MECHANCL (x2 30.7)
GR99SHE 0 0 0 PARASITE (x2 64.9), NGWIN99 (x2 44.0)
GR99HOR 0 1 3 NALLORG (x2 66.7), NGWIN99 (x2 47.5)
NGSUM2K 8 0 78 NMOW2K (rs 0.70), BAREBITS (rs 0.58)
GR2KCOW 2 1 18 NGSUM2K (x2 84.9), HAY2K (V 0.41)
GR2KSHE 1 0 11 PARASITE (x2 58.7), PARA2K (V 0.34)
GR2KHOR 0 0 0 NALLORG (x2 54.4), GR2KCOW (V 0.33)
Recent mowing
HAY99 0 0 0 CSS (x2 19.3), WEEDCONT (x2 15.9)
SIL99 2 1 18 NALLORG (x2 58.6), XN99 (x2 58.0)
HAY2K 1 0 4 NGSUM2K (x2 104.2), NMOW2K (x2 70.7)
SIL2K 1 0 1 NMOW2K (x2 172.3), LAGE (x2 102.02)
NMOW2K 0 2 10 NGSUM2K (rs 0.70), MECHANCL (rs 0.52)
Nitrogen inputs
XNALL 1 1 2 33 1 1 1 11 XN99 (rs 0.84), XN2K (rs 0.79)
XN99 0 0 1 2 XNALL (rs 0.84), NALLORG (rs 0.59)
XN2K 2 1 1 8 XNALL (rs 0.79), N2KORG (rs 0.67)
NALLCHEM 1 2 19 XNALL (rs 0.70), N2KCHEM (rs 0.76)
NALLORG 0 1 4 XNALL (rs 0.69), N2KORG (rs 0.73)
N2KCHEM 1 1 12 NALLCHEM (rs 0.76), XN2K (rs 0.62)
N2KORG 0 0 0 NALLORG (rs 0.73), XN2K (rs 0.67)
DUNGED 1 1 4 1 0 4 SIL2K (V 0.32), SIL99 (V 0.32)
Other management variables
LAGE 0 0 0 1 1 4 MECHANCL (rs 0.37), NALLORG (rs 0.29)
CSS 0 1 5 0 0 0 XNALL (x2 45.6), XN99 (x2 40.8)
HERBICID 0 2 16 0 1 2 NALLCHEM (rs 0.40), PARASITE (rs 0.25)
MECHANCL 0 2 15 1 0 5 NALLORG (rs 0.37), LAGE (rs 0.37)
PARASITE 2 1 21 0 0 0 NGWIN99 (rs 0.47), HERBICID (rs 0.25)
HARROW 2 1 11 0 0 0 ROLL (rs 0.60), NGWIN99 (rs 0.20)
ROLL 1 0 5 2 0 22 HARROW (rs 0.60), NGWIN99 (rs 0.14)
FOOD 2 0 4 NGWIN99 (x2 46.1), GR99SHE (V 0.34)
HERB2K 1 1 28 HERBICID (x2 55.7), NALLCHEM (x2 19.3)
MECH2K 0 0 0 MECHANCL (x2 210), NMOW2K (x2 53.7)
PARA2K 0 0 0 PARASITE (x2 148), GR2KSHE (V 0.34)
HARROW2K 2 0 4 HARROW (x2 235), ROLL2K (V 0.59)
ROLL2K 0 1 5 ROLL (x2 185), HARROW2K (V 0.59)
The numbers of models incorporating each variable are presented according to the direction of the relationship: positive, negative or non-linear (+//NL) and
DAIC is the sum of the AIC reductions as each variable entered the models. The most closely correlated management variables are appended with appropriate
correlation coefficients.

CSS options, avoiding fields during winter where the field
centre was subject to CSS agreements.
4. Discussion
4.1. Generality of results
The study aimed to achieve generality by sampling a
more complete range of farming practice than previous
studies. The stratification of fields maximised exposure to
any potential management effects, in order to identify the
most influential ones. Mixed modelling techniques also
extended the inference space of factors modelled as fixed
effects to a notional population of all farms (Littell et al.,
2002). However, the region and the study sample were
dominated by fields grazed by cattle in summer and sheep
in winter, so the results may not be readily extended to
regions with year-round sheep grazing (e.g. Wales). The
30 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40

Fig. 3. Effects of summer cattle grazing on field usage in the following winter (a–c). Predicted responses from the models are plotted as bars and average
observed values are plotted 1S.E. GR99COWS measures the presence or absence of summer cattle grazing.

study area was in mixed farmland, not a grassland
dominated area, where bird declines were more severe.
This was a pragmatic decision to ensure adequate samples
of declining bird species, such as buntings. Grasslands can
be valuable habitats for such species, but do not readily
support populations in the absence of arable crops
(Robinson et al., 2001). Additionally, the study only
spanned one year and could not reflect inter-annual
variation in farming practice, sward condition or bird
responses.
4.2. Sward models
Sward structure had the greatest influence on the use of
grasslands by foraging birds. Sward height and bare ground
(mainly in winter) influenced the widest range of species and
had the greatest explanatory power in the models. Individual
bird species fell into two main groups, based on the nature of
their responses to sward structure. These groupings were
closely related to the diet of the birds. This division has not
been explicitly demonstrated in previous studies.
 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40 31

Fig. 4. Effects of total nitrogen inputs on field usage by birds. Predicted responses from the models are plotted as lines and observed values, averaged in bins, are
plotted 1S.E. XNALL (a and c) is presented as the average annual nitrogen input from 1995 to 1999 and XN2K (b) is the total of nitrogen inputs in 2000.

Thrushes, starlings and corvids all strongly preferred
short swards and, to a lesser extent, patchy bare ground
within the sward. Previous studies also found this pattern
(Whitehead et al., 1995; Milsom et al., 1998, 2000; Perkins
et al., 2000; Fuller et al., 2003). These species principally
feed upon soil-dwelling invertebrates like earthworms
(Lumbricidae) and insect larvae, particularly Diptera
(Tucker, 1992). Behavioural studies show that species in
this guild forage more efficiently on short swards, due to
increased visibility (reducing the time spent watching for
predators), prey accessibility and mobility (Devereux et al.,
2004). Reduced usage of the shortest swards (e.g. Fig. 2b
and c) may have been a response to reduced prey abundance.
The shortest swards occurred on intensively grazed fields,
where trampling can have a marked negative effect on soil-
dwelling invertebrates (Curry, 1994; Morris, 2000).
The second group included larks, sparrows, finches and
buntings. These species are granivorous but most also
require invertebrate prey to rear chicks (Wilson et al., 1999).
Some insectivorous passerines, notably meadow pipit,
showed similar responses to sward structure. All these
species obtain their food from the sward or the soil surface.
This guild was infrequent on grass fields, in keeping with
previous studies (e.g. Wilson et al., 1996; Buckingham et al.,
1999; Atkinson et al., 2002). None of these species showed a
negative relationship between usage and sward height.
32 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40

Fig. 5. Effects of weed control on winter field usage by birds. Predicted responses from the models are plotted as bars and average observed values are plotted
1S.E. (a) HERBICID is the number of years when herbicides were used and (b) MECHANCL is the number of years of mechanical weed control, both
measured over the preceding 5 years. The winter seed-eaters guild excludes skylarks.

Instead, they preferred combinations of taller swards and
greater spatial heterogeneity, in the form of patchy bare
ground. Other studies found this pattern for yellowhammer
(winter only), skylark and meadow pipit (Wakeham-Dawson
and Aebischer, 1999; Milsom et al., 2000; Perkins et al.,
2000; Donald et al., 2001a). Avoidance of short swards that
facilitate accessibility and vigilance suggests that these
factors were less influential than they were for soil
invertebrate feeding species. Food abundance is proposed
to be the limiting factor for this bird guild. Experiments have
shown that the sward height preferences of chaffinches are
reversed where food abundance in tall swards exceeds that in
shorter swards by an appropriate threshold factor (Butler
et al., 2005). Food abundance is likely to be greater in taller
grass swards. Taller swards support a greater abundance and
diversity of invertebrates (Morris, 2000) and most grassland
plants attain maximum height at seeding. Perkins et al.
(2000) confirmed that grass seed heads were more frequent
in tall swards. The taller swards also exhibited greater
heterogeneity in sward height (means and standard
deviations were strongly correlated), potentially improving
accessibility. However, most bird species preferred inter-
mediate sward heights, suggesting that the tallest swards
impeded foraging efficiency. Patchy bare ground was a
better predictor of sward accessibility to foraging birds than
sward height heterogeneity.
Sward composition was less influential than sward
structure. Bird usage generally decreased as forb cover
increased, particularly in winter. Forbs were more likely to
have inhibited prey accessibility, than abundance as
invertebrate diversity and abundance increase with broad-
leaved plant diversity (Fuller et al., 2003). Linnet, which
mainly feeds on the seeds of broad-leaved plants (Wilson
et al., 1999), was the only species to prefer high forb cover.
The direction of responses may have been dependent on the
forb species present, as clovers were associated with
increased bird usage across feeding guilds.
4.3. Management effects on birds with soil-dwelling
prey
Grazing was the most influential management practice,
strongly increasing field usage by soil-dwelling invertebrate
feeders. Summer grazing increased field usage in both the
summer and the following winter. Mown fields were not
selected in the summer, but silage fields were strongly
selected in the winter if they were subjected to aftermath
grazing by cattle. Consequently, a wider range of grass fields
became suitable in winter. Wilson et al. (1996) also detected
a preference for recently grazed pastures in winter. These
patterns probably reflect the availability of short swards
under the different management regimes. Grazing during the
summer growing season maintains short swards, but swards
are too tall for birds in growing hay or silage crops. The bird
usage measurements were too widely spaced to detect the
fleeting burst of usage of short swards immediately after
 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40
mowing, but they did reflect the lack of usage over the
greater part of the summer. Grass growth largely ceases in
autumn, so short swards persist through the winter once all
the summer growth is removed. This was achieved where
summer grazing continued until growth ceased or where
aftermath grazing removed any growth following the last
hay or silage cut of the summer.
Summer cattle grazing was preferred to sheep or horse
grazing, particularly during the following winter. Cattle
were the dominant livestock on summer, but sheep were
sufficiently widespread to suggest that this preference is
correct. Cattle may promote the availability of soil
invertebrates by their grazing and dunging patterns. Cattle
initially reject grass around dung patches, which grows tall,
excluding birds. Bare patches develop once grazing
resumes, exposing weathered dung or bare soil where
coprophagous insects and earthworms buried dung. Copro-
phagous larvae (Diptera and Coleoptera) and earthworms
are also concentrated where dung enhances levels of organic
matter in the soil (Curry, 1994). In contrast, horses do not
graze their habitual latrine areas, which develop rank
vegetation, and sheep dung is finely distributed across the
field, only accumulating where sheep spend the night
(Ausden and Treweek, 1995).
The remaining management practices affected fewer
species and lacked consistency across the guild. Positive
relationships between usage and nitrogen inputs detected by
Fuller et al. (2003) were probably masked in these models by
stronger grazing effects, because graziers adjust grazing
pressure to fully utilise herbage. Tucker (1992) found that this
guild preferred frequent manure applications and older
permanent pastures, where earthworms were more abundant.
The response to old swards may also be related to the benefits
of summer cattle grazing, which was more frequent on older
swards in our study area. Barnett et al. (2004) showed that
unimproved hayfields were preferred to improved silage fields
and pastures by wintering soil invertebrate feeders, but did not
distinguish the effects of sward improvement from differences
between hay and silage. Seasonal drying of soil surface layers
is also known to limit prey availability (Peach et al., 2004) and
this is, in turn, exacerbated by drainage, fertiliser use and soil
compaction (Lazenby, 1988).
4.4. Management effects on birds with food
sources in the sward
The sward models showed that this guild preferred fields
with higher average sward heights compared to other bird
species. This relationship was probably caused by manage-
ment practises that lengthened defoliation intervals (on all or
part of the sward), allowing the replenishment and survival
of key food taxa and incidentally allowing swards to grow
taller. Research on the effects of management on grassland
invertebrates (reviewed in Curry, 1994) supports this
assertion. Defoliations remove the niches required by foliar
invertebrates and the primary production that supports their
33
populations. Short intervals between defoliations only favour
r-selected phytophagous species that are multivoltine, mobile,
or fecund (e.g. certain Auchennorhycha and Dipteran grass
leaf miners). Such taxa may be abundant in intensively
managed grasslands, but they tend to be small-bodied and do
not figure highly in the diet of farmland birds. Important
dietary items, including seeds and large invertebrates, are less
tolerant of intensive management. Key invertebrate taxa tend
to be long-lived, less mobile and reproduce slowly (K-
selected species) and these are characteristic of unmanaged
grasslands (e.g. orthoptera and spiders). Frequent defoliations
prevent plants from flowering, limiting populations of flower-
feeding invertebrates (e.g. the heteropteran bug Leptoterna
dolobrata, McNeill, 1973) as well as stopping seed
production. Invertebrates that winter in the sward may also
be eradicated by defoliations in the autumn or winter (Brown
et al., 1990; Fry and Lonsdale, 1991).
Grazing affected winter field usage by the largest number
of granivorous species. Fields grazed during winter were
favoured by several species. Autumn aftermath grazing also
allowed wintering granivores to use silage fields, which were
avoided at other times of year. However, summer grazing
reduced field usage in the following winter by skylarks and
yellowhammers, probably by preventing seeding. Grazing
had a more limited effect on field usage during summer, only
benefiting a few granivores and small insectivores.
Regular weed control, both by herbicides and mechanical
methods, had a consistent negative effect on field usage by
granivorous birds. This result is analogous to the indirect
effects of pesticides demonstrated in arable crops, where
herbicides kill the plants that produce the seed and
invertebrate food of birds (Boatman et al., 2004).
Granivores did not use mown fields in summer, with the
exception of yellowhammer, which showed a weak
preference for hayfields. The timing of hay cuts allows
grasses to set seed prior to mowing, unlike silage fields and
pastures. Few granivorous passerines took advantage of the
enhanced seed abundance on hayfields during winter,
possibly because the seed was inaccessible in the relatively
dense swards. Fields with the tallest swards were avoided by
this guild, apparently contradicting the thesis that bird usage
increased with sward height because sward height was
positively related to defoliation intervals. The positive
relationship between average sward height and defoliation
interval was probably qualitatively correct amongst grazed
fields and amongst silage fields, but quantitatively different
between the two practices. The tallest average sward heights
(and, similarly, the highest clover contents) occurred on
silage fields, but the cutting intervals were still too short to
allow prey production and the dense swards would also have
limited access and visibility.
4.5. Conservation implications
This study showed that grazing, through its effects on
sward structure, had the greatest influence on bird usage of
34 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40
agricultural grasslands. Increased grazing intensity since the
1970s coincided with population declines of a range of
farmland birds (Chamberlain et al., 2000). Intensive grazing
systems aim to maximise the efficiency of grass utilisation
by maintaining short, uniform swards (Milne and Fisher,
1993). Our models showed that the resulting swards were
not attractive to granivorous passerines and this was likely to
be because food abundance was low. The loss of feeding
opportunities due to intensive grazing is likely to have been
an important mechanism in the decline of granivorous
passerines in pastoral areas. In contrast, the sward structure
requirements of the soil invertebrate-feeding guild were
more easily satisfied on intensively grazed fields, providing
the swards were not over-grazed.
It is likely that severe, frequent or badly timed
defoliations have a disproportionate negative effect on the
abundance of large sward-dwelling invertebrate prey, which,
in turn, limits breeding success in declining granivorous bird
species. For example, breeding productivity in cirl buntings
E. cirlus L. is limited by grasshopper availability (Evans
et al., 1997), which is reduced by intensive grassland
management (van Wingerden et al., 1992). This is also likely
to affect large, soft-bodied larvae, particularly Lepidoptera,
the availability of which is known to influence chick body
condition in skylark and corn bunting Miliaria calandrella
L. (Brickle et al., 2000; Donald et al., 2001c).
Extensification of grazing regimes has the greatest
potential to instate defoliation cycles that enhance food
abundance for granivorous passerines. Promising manage-
ment tools include reducing grazing intensity (stocking rates
and duration) and increasing rest periods between grazing
periods. Both approaches allow all or part of the sward to
escape defoliation for longer, favouring production of seeds
or invertebrates. Furthermore, both approaches can be
manipulated experimentally to confirm this mechanism and
develop improved agri-environment measures. It is likely
that benefits to birds will require larger changes to grassland
management than those available under current agri-
environment schemes. For example, defoliation cycles
greater than 1 year in duration and rest periods spanning
the whole winter may be necessary to boost large
invertebrate populations (Kruess and Tscharntke, 2002;
Fry and Lonsdale, 1991).
Accessibility requirements (including vigilance beha-
viour) are also key determinants of grassland usage by birds.
Spatial heterogeneity in sward structure, particularly the
fine-scale availability of bare areas within the sward, was
shown to be a consistent requirement of birds across guilds.
Sward height probably acted as a surrogate measurement,
inversely related to horizontal accessibility, which is less
easily quantified. Extensifying grassland management can
provide spatial heterogeneity and is a focus of current agri-
environment development work (Buckingham et al., 2004).
Although this study detected no benefits from grassland
agri-environment measures for foraging birds, there was
noticeable variation in performance between individual
management agreements. Unfenced margins on silage fields
(left uncut during mowing, but exposed to aftermath
grazing) were heavily utilised by granivorous passerines
during winter, suggesting that grass margin techniques have
potential for further development. It is questionable whether
reinstating hay cutting on improved grasslands as an agri-
environment measure will benefit foraging birds. However,
if hayfields are cut sufficiently late it may be possible to
reduce nest losses for ground-nesting birds, potentially
improving annual productivity (e.g. Brickle and Harper,
2002; Donald et al., 2002).
The lack of preferences for forb-rich grasslands suggests
that the conservation needs of birds might not be congruent
with those of plants. The only species that selected forb-rich
fields was linnet during summer. Linnets utilise seed from a
succession of different grassland herb species through the
breeding season, so greater temporal diversity ensures
continuity of food supplies (D.L. Buckingham, unpublished
data). Herbicide use reduced field usage by granivorous
passerines, suggesting that the influence of botanical
diversity may have been underestimated. Herbicide use
was surprisingly prevalent, with 46% of non-organic fields
being treated over a 5 year period. Most herbicide was
applied conservatively by spot-spraying, as allowed under
existing agri-environment schemes, but this still had a
detectable negative effect on bird usage. More careful
targeting may be necessary if herbicide use continues to be
permitted on agri-environment grasslands (e.g. weed-wipers
used in conjunction with light grazing).
Acknowledgements
We are particularly indebted to all the farmers who
allowed us access onto their farms and provided detailed
management histories for all their fields. We would also like
to thank Roy Taylor and staff from FWAG and FRCA for
helping to locate suitable study farms; Keith Chaney, Jerry
Tallowin and Richard Smith for advice on agricultural
issues; Steve Langton and David Elston for statistical
advice. Andy Evans, Simon Potts, Ken Norris and two
anonymous referees provided helpful comments to develop
the manuscript. Irene Hutson and Ellen Kelly efficiently
computerised the large sward datasets. This project was
funded by the RSPB and by a generous donation from the
Nottinghamshire RSPB Members Group.
 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40 35

Appendix A

A.1. Descriptions and summary statistics for field characteristics and sward condition variables

The number of fields in the modal category in winter follows the mode of categorical variables. Winter (W, n = 388) and
summer (S, n = 373) means are presented for the sward condition measurements.
Variable Description Mean (plus range) or mode
Field characteristics
ALT Altitude (middle of highest/lowest points, m) 103 (21–260)
AREA Field area (ha) 3.59 (0.24–18.3)
AREAIND Field area index (see Section 2) 0.63 (0.1–0.98)
ASPECT Direction of slope (highest to lowest points, nine levels) Flat (78)
BYARABLE By arable field (y/n) No (275)
BYBUILD By buildings (y/n) No (238)
BYFOOT By footpath (y/n) No (326)
BYLANE By lanes or tracks (y/n) Yes (201)
BYPLG By ploughed fields on at least one visit (y/n) No (328)
BYPOWER By powerlines (y/n) No (270)
BYSETAS By setaside (y/n) No (368)
BYSTUB By stubble on at least one visit (y/n) No (296)
BYWATER By water bodies, streams, or wet ditches (y/n) Yes (239)
BYWOOD By woods/scrub 0.25 ha (y/n) No (208)
DIPS Depressions/valleys running through field (y/n) No (278)
HEDGEIND Field enclosure index (see Section 2) 2.50 (0–4)
HILLTOPS Hilltops/ridgetops in field (y/n) No (266)
LTREES log10 (number of trees in field + 1) 0.15 (0–1.32)
MAXWIDTH Width of field at widest place (m) 288 (81–699)
PERIM Length of field perimeter (m) 810 (197–2460)
PH pH buffering capacity of soil (acid to neutral, three levels) Neutral (217)
SLOPE Slope between highest/lowest points (degrees) 1.76 (0–11.3)
SOILBULK Ease of tillage (light to heavy, four levels) Heavy (147)
SOILTYPE Soil type (five levels) Clay (128)
TOPOG Fine-scale surface topography; e.g. ridge and furrow (y/n) No (320)
WETBITS Area of large water patches (seasonal mean, m2) 56.6 (0–6630)

Sward condition variables (see Section 2)

MBARE Seasonal mean bare ground cover W 0.46; S 1.16 (0–23.3)
MCLOV Seasonal mean clover cover W 0.24; S 8.81 (0–45.7)
MFORB Seasonal mean forb cover W 0.40; S 6.86 (0–41.1)
MHFRO Seasonal mean sward height (cm) W 5.79; S 13.6 (1.6–35.4)
MSDBARE Seasonal mean standard deviation of bare ground cover W 0.28; S 1.52 (0–11.0)
MSDHFRO Seasonal mean standard deviation of sward height (cm) W 2.21; S 4.84 (0.5–14.9)

Appendix B

B.1. Descriptions and summary statistics for management variables

The 13 italicised variables were combined into four PCA axes describing grass cropping (CROP1-4, Appendix C). The
number of fields in the modal category follows the mode of categorical variables. Winter means or modes are presented
(n = 388), unless variables only apply to summer (n = 373).
Variable Description Mean (plus range) or mode
Management variables
BAREBITS Area of large bare patches (seasonal mean, m2) 253 (0–1670)
CSS Type of Countryside Stewardship Scheme (three levels) None (266)
DUNGED Manure/slurry winter 1999–2000 or summer 2000 (y/n) No (342)
FOOD Supplementary food, winter 1999/2000 (y/n) None (325)
GR2KCOW Cattle grazed, summer 2000 (y/n) Yes (241)
GR2KHOR Horse grazed, summer 2000 (y/n) No (328)
GR2KSHE Sheep grazed, summer 2000 (y/n) No (317)
GR99COW Cattle grazed, summer 1999 (y/n) Yes (279)
36 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40

Appendix B (Continued )
Variable Description Mean (plus range) or mode
GR99HOR Horse grazed, summer 1999 (y/n) No (334)
GR99SHE Sheep grazed, summer 1999 (y/n) No (306)
HARROW Number of years when harrowed 1995–1999 0.86 (0–5)
HARROW2K Harrowed, summer 2000 (y/n) No (288)
HAY2K Cut for hay in 2000 (y/n) No (303)
HAY99 Cut for hay in 1999 (y/n) No (338)
HAYRATE Average Number hay cuts per year 1995–1999 0.13 (0–1.2)
HAYYRS Number of years when hay crop taken 1995–1999 0.62 (0–5)
HERB2K Sprayed with herbicide, summer 2000 (y/n) No (327)
HERBICID Number of years when herbicide used 1995–1999 0.79 (0–5)
LAGE log10 (age of sward) (range 0–2.2) 1.19 (0–2.2)
MECHANCL Number of years of mechanical weed control 1995–1999 2.89 (0–5)
MECH2K Mechanical weed control summer 2000 (y/n) No (198)
MOWRATE Average number mown crops per year, 1995–1999 0.64 (0–6)
N2KCHEM Nitrogen from chemical fertilisers 2000 (kg N ha1) 67.7 (0–666)
N2KORG Nitrogen from organic fertilisers 2000 (kg N ha1) 53.0 (0–843)
NALLCHEM Nitrogen from chemical fertilisers 1995–1999 (kg N ha1) 423 (0–5020)
NALLORG Nitrogen from organic fertilisers 1995–1999 (kg N ha1) 278 (0–2530)
NGSUM2K Number of summer visits with recent grazing 2000 2.24 (0–4)
NGWIN99 Number of visits with recent grazing winter 1999/2000 2.08 (0–5)
NMOW2K Number of forage crops taken, summer 2000 0.67 (0–6)
NSUMCOW Number of summers of cow grazing 1995–1999 3.27 (0–5)
NSUMGRZ Number of summers of grazing 1995–1999 3.15 (0–5)
NSUMHOR Number of summers of horse grazing 1995–1999 0.7 (0–5)
NSUMSHE Number of summers of sheep grazing 1995–1999 1.07 (0–5)
NWINCOW Number of winters of cow grazing 1995–1999 1.16 (0–5)
NWINGRZ Number of winters of grazing 1995–1999 3.15 (0–5)
NWINHOR Number of winters of horse grazing 1995–1999 0.68 (0–5)
NWINSHE Number of winters of sheep grazing 1995–1999 2.23 (0–5)
PARASITE Number of years of parasite treatments 1995–1999 3.11 (0–5)
PARA2K Parasite treatments in 2000 (y/n) No (292)
ROLL Number of years when sward rolled 1995–1999 1.26 (0–5)
ROLL2K Rolled, summer 2000 (y/n) No (252)
SIL2K Cut for silage/zero-grazing (no hay) in 2000 (y/n) No (281)
SIL99 Cut for silage/zero-grazing (no hay) in 1999 (y/n) No (311)
SILRATE Average number silage cuts per year 1995–1999 0.51 (0–6)
SILYRS Number of years of silage (but not hay) 1995–1999 0.7 (0–5)
XN2K Total nitrogen inputs in 2000 (kg N ha1) 164 (0–890)
XN99 Total nitrogen inputs in 1999 (kg N ha1) 180 (0–1430)
XNALL Total nitrogen inputs 1995–1999 (kg N ha1) 851 (0–5250)

Appendix C

C.1. Eigenvectors from a principal components analysis of long-term grass cropping variables

The table shows Pearson correlation coefficients between each component axis and the underlying variables for the winter
dataset. The summer PCA axes, based on a smaller sample of fields, showed similar relationships to the variables. The
eigenvectors were interpreted by correlation coefficients with magnitudes greater than 0.3 (highlighted in bold text).
Variables Eigenvectors
CROP1 CROP2 CROP3 CROP4
NSUMGRZ 0.40 0.01 0.11 0.10
NSUMCOW 0.31 0.23 0.14 S0.36
NSUMSHE 0.29 0.16 0.20 0.19
NSUMHOR 0.16 S0.40 0.11 0.46
NWINGRZ 0.36 0.05 0.28 0.19
NWINCOW 0.19 0.07 0.40 -0.16
NWINSHE 0.28 0.28 0.22 0.10
NWINHOR 0.15 S0.40 0.06 0.47
 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40 37

Appendix C (Continued )
Variables Eigenvectors
CROP1 CROP2 CROP3 CROP4
HAYYRS 0.12 0.43 S0.36 0.32
SILYRS 0.25 0.19 0.41 0.05
HAYRATE 0.12 0.43 S0.36 0.32
SILRATE S0.38 0.17 0.35 0.21
MOWRATE S0.36 0.27 0.26 0.28
Percent variance explained 28.7 18.9 15.4 13.4
Eigenvalues 3.73 2.45 2.00 1.74

Appendix D

D.1. Winter and summer bird usage models based on sward condition variables

The relationship sense (positive (no signs), negative , \- or [-shaped) and AIC reduction follow each variable. See
Appendices D.2 and D.3 for field characteristics variables and Appendices A and B for descriptions of the variables.
Winter Summer
Species/guild Dispersion factor Model Species/guild Dispersion factor Model
Blackbird 1.36 MFORB (16.3), Blackbird 0.93 MHFRO (10.7)
MHFRO (39.8)
Song thrush No sward model
Redwing 0.73 MBARE (9.1),
MCLOV (6.7),
MFORB (\ 6.6)
Fieldfare 0.71 MFORB (3.0),
MHFRO (13.4)
Mistle thrush No sward model
Starling 0.80 MHFRO (\ 5.0) Starling 0.70 MHFRO (8.0)

Magpie No sward model Magpie 0.97 MHFRO (17.4),

MSDHFRO (3.2)
Jackdaw 0.74 MBARE (3.6) Jackdaw 0.98 MCLOV ([ 3.0),
MHFRO (32.3)
Rook 1.00 MFORB (10.1), Rook 0.77 MSDBARE (9.5),
MHFRO (3.9) MSDHFRO (3.7)
Carrion crow 1.15 MHFRO (\ 21.5) Carrion crow 1.13 MHFRO (18.8)
Skylark 0.37 MFORB (3.8),
MHFRO (3.3)
Chaffinch No sward model Chaffinch No sward model
Goldfinch No sward model
Linnet 0.33 MBARE (1.8),
MFORB (4.6),
MHFRO (\ 1.3)
Yellowhammer 0.33 MBARE (13.0), Yellowhammer 0.34 MFORB (8.3)
MHFRO (\ 2.7)
Mallard 0.64 MFORB (\ 3.1)
Pheasant 0.95 MCLOV (1.0) Pheasant No sward model
Moorhen 0.75 MFORB (1.9),
MHFRO (2.6)
Snipe 0.51 MBARE (2.8),
MHFRO (\ 17.2)
Woodpigeon 1.12 MBARE (\ 6.7) Woodpigeon 0.95 MCLOV (\ 20.3),
MSDBARE (0.5),
MSDHFRO (10.0)
Stock dove No sward model
Swallow 1.11 MFORB ([ 3.1),
MHFRO ([ 0.5)
House martin No sward model
38 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40

Appendix D (Continued )
Winter Summer
Species/guild Dispersion factor Model Species/guild Dispersion factor Model

Meadow pipit 0.85 MBARE (13.6),

MHFRO (\ 9.8)
Pied wagtail 0.76 MBARE (16.3), Pied wagtail 0.43 MHFRO (35.8)
MSDBARE (2.5)
Wren No sward model
Dunnock 0.34 MCLOV (\ 12.2),
MHFRO (3.5)
Robin 1.02 MFORB (8.4) Robin 0.83 MCLOV (\ 11.5)
All seedeaters 0.94 MCLOV (4.1),
(except skylark) MFORB (1.9)
Obligate seedeaters 0.79 MCLOV (\ 2.0)
Mixed-diet seedeaters 0.98 MFORB ([ 11.3),
MSDBARE (1.2),
MSDHFRO (2.1)

D.2. Winter bird usage models based on management variables

Each management variable is followed by the sense of the relationship (positive (no signs), negative , \- or [-shaped) and
the resulting reduction in the model AIC. Only the sense of the relationship is presented for field characteristic variables. See
Appendices A and B for descriptions of the variables.
Species/guild Dispersion factor Management variables Field characteristics
Blackbird 1.34 CROP3 (7.1), CROP4 (2.3), GR99COW (37.0) BYBUILD, BYLANE, BYWATER,
HEDGEIND (\), MAXWIDTH,
SOILTYPE
Song thrush 0.83 GR99COW (2.6) BYWATER, SOILTYPE
Redwing 0.91 CROP3 (8.7), GR99COW (8.9), HARROW (5.2) BYSETAS (), MAXWIDTH, SOILTYPE
Fieldfare 0.86 XNALL (\ 3.7) ASPECT, BYBUILD, MAXWIDTH
Mistle thrush 0.91 SIL99 (8.8) BYBUILD, MAXWIDTH
Starling 0.78 CROP3 (5.6), GR99COW (11.6) BYPLG (), BYPOWER, MAXWIDTH,
SOILTYPE
Magpie 0.93 HARROW (5.1), PARASITE (6.8) BYSETAS, PERIM
Jackdaw 0.75 CROP3 (5.3), NALLORG (4.2), XN99 (\ 1.8) AREAIND (\), BYARABLE (),
BYBUILD, BYFOOT, HEDGEIND (),
MAXWIDTH, SOILTYPE, WETBITS
Rook 0.97 DUNGED (1.0), GR99COW (22.3), AREA, BYBUILD, BYPOWER (),
NALLCHEM (6.0), NGWIN99 (6.1) DIPS, HEDGEIND ()
Carrion crow 1.15 GR99COW (21.0) AREAIND, PERIM
Skylark 0.58 CSS (5.5), DUNGED (2.6), GR99HOR (2.9), AREA, BYFOOT (), BYPLG (), BYPOWER (),
MECHANCL (9.4) NALLCHEM (5.4), BYWATER, BYWOOD (), HEDGEIND (),
PARASITE (5.0) PH, SOILBULK
Chaffinch 0.96 BAREBITS (12.8), CROP3 (4.4), FOOD (2.6), ASPECT, BYWOOD (), PERIM
GR99COW (3.8)
Yellowhammer 0.35 CROP2 (4.0), CROP3 (2.0), GR99COW (6.5), No field characteristic variables
ROLL (4.9)
All seedeaters 0.91 BAREBITS (3.7), CROP3 (2.0), HERBICID (6.1) BYBUILD, BYWOOD (), PERIM
(except skylark)
Pheasant 0.95 BAREBITS (3.1), CROP3 (15.6), FOOD (1.4) BYFOOT, BYSETAS, MAXWIDTH
Moorhen 0.62 HARROW (0.9), XNALL (13.1) BYBUILD, BYSTUB (), BYWATER,
PERIM, WETBITS
Snipe 0.51 PARASITE (9.2), SIL99 (3.5) AREAIND, BYARABLE (), BYWATER,
DIPS, PERIM
Woodpigeon 1.08 SIL99 (5.4), XNALL (\ 10.8) BYLANE, PERIM, WETBITS
Meadow pipit 0.88 GR99COW (11.1), HERBICID (10.4) AREA, BYBUILD, HEDGEIND (),
LTREES ()
Pied wagtail 0.75 BAREBITS (3.2), CROP1 (9.0), XNALL (5.5) HEDGEIND (), HILLTOPS (),
WETBITS
Wren 0.82 MECHANCL (5.9) AREAIND ()
Robin 0.92 CROP4 (11.0), NALLCHEM (7.8), NGWIN99 (13.1) BYWATER, PERIM
 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40 39

D.3. Summer bird usage models based on management variables

Each management variable is followed by the sense of the relationship (positive (no signs), negative , \- or [-shaped) and
the resulting reduction in the model AIC. Only the sense of the relationship is presented for field characteristic variables. See
Appendices A and B for descriptions of the variables.
Species/guild Dispersion factor Management variables Field characteristics
Blackbird 0.94 HARROW2K (1.0), NGSUM2K (2.8) AREAIND (), BYBUILD, BYLANE, LTREES
Starling 0.73 BAREBITS (9.1), NGSUM2K (1.5) BYBUILD, DIPS, HEDGEIND (\)
Magpie 0.93 HERBICID (2.5), MECHANCL (5.2), PERIM
NGSUM2K (20.3)
Jackdaw 1.09 NGSUM2K (18.9) AREA, BYBUILD, BYPOWER
Rook 0.80 BAREBITS (8.6) AREA, TOPOG
Carrion crow 1.16 DUNGED (4.3), NGSUM2K (4.0) AREA
Chaffinch 0.88 BAREBITS (5.6) BYFOOT, TOPOG
Goldfinch 0.74 CROP4 (13.3) AREAIND (), BYWOOD (), HILLTOPS
Linnet No management model AREA (), BYLANE, BYWATER (),
HEDGEIND (),PERIM
Yellowhammer 0.33 CROP3 (1.3), N2KCHEM (0.7), ROLL2K (5.3) BYLANE, BYWOOD (), TOPOG ()
Swallow 1.08 HERB2K (18.8), ROLL (14.1) BYBUILD, BYSTUB (), BYWATER (),
PERIM, SOILTYPE
House martin 0.78 ROLL (7.9) PERIM
Mallard 0.47 GR2KCOW (4.0), SIL2K (0.8), XNALL (\ 7.7) HEDGEIND (), WETBITS
Pheasant 0.79 XNALL (1.3) LTREES
Woodpigeon 1.00 LAGE (2.2), NMOW2K (8.8), XN2K (\ 3.6) AREA, BYBUILD, BYWOOD ()
Stock dove 0.41 CROP4 (4.0), HAY2K (3.7), NGSUM2K (3.2), BYBUILD, HEDGEIND (), MAXWIDTH, PH
NMOW2K (1.3) XN2K (0.5)
Pied wagtail 0.34 BAREBITS (3.5), CROP3 (6.2), GR2KCOW (3.6), BYBUILD, BYPLG (), PERIM
HARROW2K (2.8), LAGE (1.6), NGSUM2K (26.0),
XN2K (2.7)
Dunnock 0.31 GR2KCOW (10.5), N2KCHEM (11.5), AREAIND (), BYBUILD, BYPOWER (),
NGSUM2K (1.6), XN2K (1.6) BYWOOD (), MAXWIDTH
Robin 0.80 BAREBITS (4.6), GR2KSHE (10.7) AREA, BYBUILD
Seedeater guilds
Mixed-diet 1.03 CROP2 (2.0), HERB2K (8.9) AREA, HEDGEIND (), TOPOG
Obligate 0.82 CROP4 (3.4), XNALL (2.4) BYLANE, BYWOOD (), HEDGEIND (), PERIM

References
Atkinson, P.W., Fuller, R.J., Vickery, J.A., 2002. Large-scale patterns of
summer and winter bird distribution in relation to farmland type on
England and Wales. Ecography 25, 466–480.
Ausden, M., Treweek, J., 1995. Grasslands. In: Sutherland, W.J., Hill, D.A.
(Eds.), Managing Habitats for Conservation. Cambridge University
Press, Cambridge, pp. 197–229.
Barnett, P.R., Whittingham, M.J., Bradbury, R.B., Wilson, J.D., 2004. Use
of unimproved and improved lowland grassland by wintering birds in
the UK. Agric. Ecosyst. Environ. 102, 49–60.
Blackstock, T.H., Rimes, C.A., Stevens, D.P., Jefferson, R.G., Robertson,
H.J., Mackintosh, J., Hopkins, J.J., 1999. The extent of semi-natural
grassland communities in lowland England and Wales: a review of
conservation surveys 1975–1996. Grass Forage Sci. 54, 1–18.
Boatman, N.D., Stoate, C., Watts, P.N., 2000. Practical management solu-
tions for birds on lowland arable farmland. In: Aebischer, N.J., Evans,
A.D., Grice, P.V., Vickery, J.A. (Eds.), Ecology and Conservation of
Lowland Farmland Birds. British Ornithologists’ Union, Tring, pp.
105–114.
Boatman, N.D., Brickle, N.W., Hart, J.D., Milsom, T.P., Morris, A.J.,
Murray, A.W.A., Murray, K.A., Robertson, P.A., 2004. Evidence for
the indirect effect of pesticides on farmland birds. In: Vickery, J.A.,
Evans, A.D. (Eds.), Ecology and Conservation of Lowland Farmland
Birds. II. The Road to Recovery. Ibis 146 (Suppl. 2), 131–143.
Brickle, N.W., Harper, D.G.C., 2002. Agricultural intensification and the
timing of breeding of Corn Buntings Miliaria calandrella. Bird Study
49, 219–228.
Brickle, N.W., Harper, D.G.C., Aebischer, N.J., Cockayne, S.H., 2000.
Effects of agricultural intensification on the breeding success of corn
buntings Miliaria calandra. J. Appl. Ecol. 37, 742–755.
Brodman, P.A., Reyer, H.-U., Baer, B., 1997. The relative importance of
habitat structure and prey characteristics for the foraging success of
Water Pipits (Anthus spinoletta). Ethology 103, 222–235.
Brown, V.K., Gibson, C.W.D., Sterling, P.H., 1990. The mechanisms con-
trolling insect diversity in calcareous grasslands. In: Hillier, S.H., Walton,
D.W.H., Wells, D.A. (Eds.), Calcareous Grasslands – Ecology and
Management. Bluntisham Books, Bluntisham, pp. 79–87.
Brown, H., Prescott, R., 1999. Applied Mixed Models in Medicine. Wiley,
Chichester, UK.
Buckingham, D.L., Evans, A.D.E., Morris, A.J., Orsman, C.J., Yaxley, R.,
1999. Use of set-aside land in winter by declining farmland bird species
in the UK. Bird Study 46, 157–169.
Buckingham, D.L., Atkinson, P.W., Rook, A.J., 2004. Testing solutions in
grass dominated landscapes. In: Vickery, J.A., Evans, A.D. (Eds.),
Ecology and Conservation of Lowland Farmland Birds. II. The Road
to Recovery. Ibis 146 (Suppl.2), 164–171.
Butler, S.J., Whittingham, M.J., Quinn, J.L., Cresswell, W., 2005. Quanti-
fying the interaction between food density and habitat structure in
determining patch selection. Anim. Behav. 69, 337–343.
40 D.L. Buckingham et al. / Agriculture, Ecosystems and Environment 112 (2006) 21–40
Chamberlain, D.E., Fuller, R.J., 2000. Contrasting patterns of change in the
distribution and abundance of farmland birds in relation to farming
system in lowland Britain. Global Ecol. Biogeogr. 10, 399–409.
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. J. Appl.
Ecol. 37, 771–788.
Curry, J.P., 1994. Grassland Invertebrates: Ecology, Influence on Soil
Fertility and Effects on Plant Growth. Chapman and Hall, London.
Defra, 2002. Achieving a better quality of life: review of progress towards
sustainable development, Government Annual Report 2001. Depart-
ment for Environment, Food and Rural Affairs, London.
Devereux, C.L., McKeever, C.U., Benton, T.G., Whittingham, M.J., 2004.
The effect of sward height and drainage on Starlings and Lapwings
foraging in grassland habitats. In: Vickery, J.A., Evans, A.D. (Eds.),
Ecology and Conservation of Lowland Farmland Birds. II. The Road to
Recovery. Ibis 146 (Suppl. 2), 115–122.
Donald, P.F., Buckingham, D.L., Moorcroft, D., Muirhead, L.B., Evans,
A.D., Kirby, W.B., 2001a. Habitat use and diet of skylarks Alauda
arvensis wintering on lowland farmland in southern Britain. J. Appl.
Ecol. 38, 536–547.
Donald, P.F., Green, R.E., Heath, M.F., 2001b. Agricultural intensification
and the collapse of Europe’s farmland bird populations. In: Proceedings
of the Royal Society of London (Series B), vol. 268, pp. 25–29.
Donald, P.F., Muirhead, L.B., Buckingham, D.L., Evans, A.D., Kirby, W.B.,
Gruar, D.J., 2001c. Body condition, growth rates and diet of Skylark
Alauda arvensis nestlings on lowland farmland. Ibis 134, 658–668.
Donald, P.F., Evans, A.D., Muirhead, L.B., Buckingham, D.L., Kirby, W.B.,
Schmitt, S.I.A., 2002. Survival rates, causes of failure and productivity of
Skylark Alauda arvensis nests on lowland farmland. Ibis 144, 652–664.
Evans, A.D., Smith, K.L., Buckingham, D.L., Evans, J., 1997. Seasonal
performance and nestling diet of Cirl Buntings Emberiza cirlus in
England. Bird Study 44, 66–79.
Fry, R., Lonsdale, D. (Eds.), 1991. Habitat Conservation for Insects – A
Neglected Green Issue. The Amateur Entomologist’s Society, Orping-
ton, Kent, UK.
Fuller, R.J., Atkinson, P.W., Asteraki, E.J., Conway, G.J., Goodyear, J.,
Haysom, K., Ings, T., Smith, R.E.N., Tallowin, J.R., Vickery, J.A., 2003.
Changes in lowland grassland management: effects on invertebrates and
birds, BTO Research Report 350. British Trust for Ornithology, Thet-
ford, UK.
Hodgson, J., Mackie, C.K., Parker, J.W.G., 1986. Sward surface heights for
efficient grazing. Grass Farmer 24, 5–10.
Kristensen, E.S., Hogh-Jensen, H., Kristensen, I.S., 1995. A simple model
for estimation of atmospherically derived nitrogen in grass-clover
systems. Biol. Agric. Horti. 12, 263–276.
Kruess, A., Tscharntke, T., 2002. Contrasting responses of plant and insect
diversity to variation in grazing intensity. Biol. Conserv. 106, 293–302.
Lazenby, A., 1988. The availability and use of nitrogen and water in
extensive legume-based systems. In: Wilkins, R.J. (Ed.), Nitrogen
and Water Use by Grassland. AFRC Institute for Grassland and Animal
Production, Hurley, UK, pp. 58–73.
Littell, R.C., Milliken, G.A., Stroup, W.W., Wolfinger, R.D., 2002. SAS1
System for Mixed Models. SAS Institute, Cary, USA.
MAFF, 1994. Making better use of livestock manures on grassland. Mana-
ging Livestock Manures, booklet 2. MAFF, London.
MAFF, The Scottish Office Agriculture, Environment and Fisheries Depart-
ment, Department of Agriculture for Northern Ireland, Welsh Office,
1997. The Digest of Agricultural Census Statistics. Stationery Office,
London, UK.
McCulloch, C.E., Searle, S.R., 2001. Generalised, Linear and Mixed
Models, John Wiley and Sons Ltd., New York.
McNeill, S., 1973. The dynamics of a population of Leptoterna dolobrata
(Heteroptera: Miridae) in relation to its food resources. J. Anim. Ecol.
42, 495–507.
Milne, J.A., Fisher, G.E.J., 1993. Sward structure with regard to production.
In: Haggar, R.J., Peel, S. (Eds.), Grassland Management and Nature
Conservation, British Grassland Society Occasional Symposium, vol.
28. British Grassland Society, Reading, UK, pp. 33–42.
Milsom, T.P., Ennis, D.C., Haskell, S.D., Langton, S.D., McKay, H.V., 1998.
Design of grassland feeding areas for waders during winter: the relative
importance of sward, landscape factors and human disturbance. Biol.
Conserv. 84, 119–129.
Milsom, T.P., Langton, S.D., Parkin, W.K., Peel, S., Bishop, J.D., Hart, J.D.,
Moore, N.P., 2000. Habitat models of bird species’ distribution: an aid to
the management of coastal grazing marshes. J. Appl. Ecol. 37, 706–727.
Morris, M.G., 2000. The effects of structure and its dynamics on the ecology
and conservation of arthropods in British Grassland. Biol. Conserv. 95,
129–142.
Peach, W.J., Denny, M., Cotton, P.A., Hill, I.F., Gruar, D., Barritt, D., Impey,
A., Mallord, J., 2004. Habitat selection by song thrushes in stable and
declining populations. J. Appl. Ecol. 41, 275–293.
Perkins, A.J., Whittingham, M.J., Bradbury, R.B., Wilson, J.D., Morris,
A.J., Barnett, P.R., 2000. Habitat characteristics affecting use of lowland
agricultural grassland by birds in winter. Biol. Conserv. 95, 279–294.
Robinson, R., Wilson, J.D., Crick, H.Q.P., 2001. The importance of arable
habitat for farmland birds in grassland landscapes. J. Appl. Ecol. 38,
1059–1069.
Rodwell, J.S. (Ed.), 1992. British Plant Communities. Grasslands and
Montane Communities, vol. 3. Cambridge University Press, Cambridge.
Rushton, S.P., Ormerod, S.J., Kerby, G., 2004. New paradigms for model-
ling species distributions? J. Appl. Ecol. 41, 193–200.
Siriwardena, G.M., Baillie, S.R., Buckland, S.T., Fewster, R.M., Marchant,
J.H., Wilson, J.D., 1998. Trends in the abundance of farmland birds: a
quantitative comparison of smoothed Common Birds Census indices. J.
Appl. Ecol. 35, 24–43.
Tucker, G.M., 1992. Effects of agricultural practises on field use by
invertebrate-feeding birds in winter. J. Appl. Ecol. 29, 779–790.
van Wingerden, W.K.R.E., van Kreveld, A.R., Bongers, W., 1992. Analysis
of species composition and abundance of grasshoppers (Orth. Acridi-
dae) in natural and fertilised grasslands. J. Appl. Entomol. 113, 138–
152.
Vickery, J.A., Tallowin, J.R., Feber, R.E., Asteraki, E.J., Atkinson, P.W.,
Fuller, R.J., Brown, V.K., 2001. The management of lowland neutral
grasslands in Britain: effects of agricultural practises on birds their food
resources. J. Appl. Ecol. 38, 647–664.
Wakeham-Dawson, A., Aebischer, N.J., 1999. Management of grassland for
skylarks Alauda arvensis in downland environmentally sensitive areas
in southern England. In: Donald, P.F., Vickery, J.A. (Eds.), The Ecology
and Conservation of Skylarks Alauda arvensis. RSPB, Sandy, UK,
pp.189–201.
Watts, B.D., 1991. Effects of predation risk on distribution within and
between habitats in Savannah Sparrows. Ecology 72, 1515–1519.
Welham, S.J., Thompson, R., 1997. Likelihood ratio tests for fixed model
terms using residual maximum likelihood. J. Roy. Stat. Soc. (Series B)
59, 701–714.
Whitehead, S.C., Wright, J., Cotton, P.A., 1995. Winter field use by the
European Starling (Sturnus vulgaris): habitat preferences and the
availability of prey. J. Avian Biol. 26, 193–202.
Whittingham, M.J., Evans, K.L., 2004. A review of the effects of habitat
structure on predation risk of birds in agricultural landscapes. In:
Vickery, J.A., Evans, A.D. (Eds.), Ecology and Conservation of Low-
land Farmland Birds. II. The Road to Recovery. J. Avian Biol. 146
(Suppl. 2), 210–220.
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.
Wilson, J.D., Evans, J., Browne, S.J., King, J.R., 1997. Territory distribution
and breeding success of skylarks Alauda arvensis on organic and
intensive farmland in southern England. J. Appl. Ecol. 34, 1462–1478.
Wilson, J.D., Morris, A.J., Arroyo, B.E., Clarke, S.C., Bradbury, R.B., 1999.
A review of the abundance and diversity of invertebrate and plant foods
of granivorous birds in northern Europe in relation to agricultural
change. Agric. Ecosyst. Environ. 75, 13–30.