Are hedgerows the route to increased farmland small mammal density? Use of hedgerows in British pastoral habitats

RESEARCH ARTICLE

Are hedgerows the route to increased farmland smallmammal density? Use of hedgerows in British pastoralhabitats

Merryl Gelling Æ David W. Macdonald ÆFiona Mathews

Received: 19 October 2005 / Accepted: 21 February 2007 / Published online: 16 March 2007� Springer Science+Business Media B.V. 2007

Abstract Linear habitats are becoming increas-

ingly common as a consequence of habitat frag-

mentation, and may provide the sole habitat for

some species. Hedgerows are linear features that

can vary substantially in structure and quality.

Having surveyed 180 hedgerows, in four loca-

tions, and sampled their small mammal commu-

nities we examined the effect of physical

hedgerow attributes on the abundance of small

mammal species. Using three elements of land-

scape structure, we explored whether variation

was best explained by the Random Sample

Hypothesis (that small islands represent a random

sample of those species populating larger areas),

or by the Fragmentation Hypothesis (that species

abundance will decrease with a loss of habitat

area). We tested the relationship between the

relative abundance of small mammals and 1.

hedgerow connectivity; 2. total habitat availability

and 3. local habitat complexity. We then explored

the predictive power of combinations of these

habitat variables. Connectivity was a positive

predictor of wood mice Apodemus sylvaticus,

and hedgerow gappiness was a negative predictor

of bank voles Clethrionomys glareolus. The total

amount of habitat available (hedgerow width,

height and length) was a positive indicator of total

small mammal biomass. These results support the

Fragmentation Hypothesis that species abun-

dance and distribution decrease with a loss of

habitat area. The preservation of linear and

associated habitats may therefore be important

in maintaining metapopulations of the species we

studied.

Keywords Apodemus flavicollis � Apodemus

sylvaticus � Clethrionomys glareolus �Fragmentation �Habitat corridors �Linear habitat �Population density �Microtus agrestis

Introduction

Habitat fragmentation is both a consequence of,

and may be defined in terms of reductions in

patch size, and increases in patch isolation

(Andren 1994). Both factors result in habitat loss

to wildlife, and can be associated with reduced

population sizes and increased risk of extinction

(Wilcox 1980; Wilcox and Murphy 1985). In

landscapes where suitable habitat remains, effects

of fragmentation on population size and distribu-

tion are expected to result primarily from loss of

habitat area (Andren 1994). This expectation has

been explored in relation to the decline in species

richness with decreasing island size: a simple

M. Gelling (&) � D. W. Macdonald � F. MathewsWildlife Conservation Research Unit, Departmentof Zoology, University of Oxford, Tubney House,Tubney, Oxon OX13 5QL, UKe-mail: [email protected]

123

Landscape Ecol (2007) 22:1019–1032

DOI 10.1007/s10980-007-9088-4

model being that the species found on small

islands constitute a random sample of those

populating larger areas (the Random Sample

Hypothesis, Connor and McCoy 1979; Haila

1983). The Fragmentation Hypothesis however,

further predicts that as available habitat de-

creases, so the spatial arrangement of remaining

habitat becomes increasingly important due to an

exponential increase in the distance between

patches (Saunders et al. 1991). Individual patch

size, the degree of isolation, the impact of edge

effects and interactions between patch size and

isolation therefore play an increasing role, and so

population sizes decrease more than that pre-

dicted by the Random Sample Hypothesis

(Andren 1994).

In Britain, habitat fragmentation is thought to

be a limiting factor for the distribution of some

species and a threat to the survival of others

(Bright 1993). Agricultural intensification, and

the associated loss of habitat, has resulted in

hedgerows becoming regarded as important

contributors to biodiversity conservation in the

agricultural landscape. These biological corri-

dors may alleviate negative impacts of habitat

fragmentation by allowing for movement be-

tween larges areas of habitat (Soule and Terb-

ough 1999). In some areas, historical corridors

originally linking larger areas of habitat which

has since been removed may be the only

remaining habitat (Bennett 1987). In the UK,

the term ‘hedgerow’ may encompass any field

boundary which incorporates shrubby vegeta-

tion, many of which have been planted histor-

ically to enclose parcels of land (Rackham

1997). For this study we regard a hedgerow as

a linear habitat; ‘‘a line or narrow belt of

closely-spaced woody shrubs, retained and/or

managed so as to form a more or less contin-

uous barrier’’ (Clements and Tofts 1992).

Hedgerows have been rapidly lost over the last

century (Rackham 1997) with over 50% having

disappeared since 1947 (CPRE 1999). This

results primarily from changes in management

(for example mechanical cutting or flailing)

within a system of subsidies now widely re-

garded as outdated (Barr and Gillespie 2000;

Macdonald and Johnson 2000). However, re-

forms of the common agricultural policy (CAP)

are now encouraging farmers to adopt enhanced

hedgerow management strategies (DEFRA

2005a), and to regenerate lost hedgerows (DE-

FRA 2005b).

Hedgerows have become fewer, and more

‘gappy’ (defined as a break in the living canopy

of the hedge (Clements and Tofts 1992)), and thus

patch size has decreased on a scale probably

relevant to small mammals (Haila 1990). Gaps

may occur within a hedgerow for many reasons,

including fallen trees or mismanagement of the

hedge over time. Poor-quality, gappy hedges are

known to be detrimental to several farmland bird

species (Hinsley and Bellamy 2000; Macdonald

and Johnson 1995), and could have had similar

effects on other species.

Not only may hedgerows be linear habitat

corridors for dispersal, but they also host

resident populations of species of appropriate

scale, such as small mammals, and are impor-

tant for the maintenance of many wildlife

species (Fitzgibbon 1997; Haddad et al. 2003;

Pollard et al. 1974). Many farmland species are

now marginalised to field boundaries (Pollard

and Relton 1970; Tattersall et al. 2002), with

small mammals in the pastoral landscape being

restricted largely to hedgerows and woodland

patches (Hinsley and Bellamy 2000). Recapture

studies on wood mice Apodemus sylvaticus

Linnaeus and bank voles Clethrionomys glareo-

lus Schreber indicate that the hedgerow some-

times becomes the sole habitat for both species

(Boone and Tinklin 1988). Hedgerows are often

regarded as inferior to adjacent woodland in

habitat quality (Hinsley and Bellamy 2000), and

therefore house what may be ‘sink’ populations

of small mammals (Hanski 1999; Tattersall et al.

2004).

Here, we describe the small mammal com-

munities of hedgerows on pastoral land in

Britain, and test whether variation in their

population abundancies can be explained solely

by the Random Sample Hypothesis, or whether

there is evidence in support of the Fragmenta-

tion Hypothesis. Demonstration that as habitat

availability decreases, so population densities

within remaining patches are further reduced

will provide support for the Fragmentation

Hypothesis.

1020 Landscape Ecol (2007) 22:1019–1032

123

Methods

Sites

Animals were surveyed on 12 dairy farms in four

geographical areas within Britain—Staffordshire/

Derbyshire (Central England, 53o4¢ N; 1o51¢ W),

Carmarthenshire (SW Wales, 51o49¢ N; 4o46¢ W),

Gwent (S Wales/England border, 51o44¢ N;

2o54¢ W), and North Somerset (SE England,

51o20¢ N; 2o46¢ W). Farms were randomly se-

lected as part of a larger study (Mathews et al.

2006), within which triplets of farms in a 20 km

area were identified, ensuring that local variation

in species distribution between farms in the same

geographical area would have a similar effect. All

farms selected were representative in terms of

habitat of those within the local area. The

hedgerows that we investigated on each study

farm were randomly selected from all hedgerows

on each individual farm, and were flanked by

improved or semi-improved grassland for the

grazing of dairy cattle and production of silage.

See Fig. 1 for an example map of a farm,

complete with trap line locations. In total 180

hedgerows were studied between June 2000 and

June 2003 (with a nine-month gap due to a Foot

and Mouth Disease epidemic which led to statu-

tory restrictions on access to the countryside).

Trapping and handling procedure

Preliminary trapping completed within pastoral

fields at distances of 5 m, 10 m, 15 m, and 20 m

from the hedgerow rarely caught small mammals,

suggesting that unlike the situation with arable

fields, little use was made of pastoral land by

small mammals. Spool and line tracking (Boon-

stra and Craine 1996) also suggested that small

mammals largely remained within hedgerows. We

therefore treated hedgerows as linear habitats.

Forty Longworth traps (Penlon Ltd, Abingdon,

UK) were placed along a 100 m section of each

hedgerow, with pairs of traps being sited at 5 m

intervals. These were pre-baited for a minimum

of 24 h with mixed grain. Traps were provisioned

with hay, mixed grain, apple and castors (blowfly

pupae). They were set at dusk and checked early

the following day (Gurnell and Flowerdew 1990).

A single trapping session was conducted for

3 days at each hedge section. To reduce the

likelihood of catching pygmy shrews Sorex min-

utes Linnaeus following licensing guidelines of

English Nature and the Mammal Society, and

small juveniles of all species with a high risk of

mortality, the traps were calibrated to trap only

animals’ ‡4 g. Each animal was uniquely identi-

fied using fur clips. Sex, breeding condition,

weight and length were recorded. All animals

were released at the point of capture.

Analysis

All hedgerows on each farm were evaluated using

the flow-chart procedure outlined in the Hedge-

row Evaluation and Grading System (HEGS)

(Clements and Tofts 1992). The HEGS identifies

four groups of hedgerow features (structure,

connectivity, diversity and associated features),

recording each group separately (Tofts and

Clements 1994). Although HEGS produces a

single score for each hedgerow, within this inves-

tigation we analysed the effect of each feature

recorded individually. One hundred metre sub-

sections were randomly selected for trapping

from all hedgerows more than 100 m in length.

Some hedgerow characteristics (cross-section, end

connections and ditch) were scored by reference

to a series of standard diagrams. Some additional

variables were measured, including whether or

not the hedge was flailed (mechanically cut), and

proximity to woodland.

Following inspection of the data and prelimin-

ary exploratory analyses, we examined the rela-

tionships between small mammal density and

hedgerow characteristics using General Linear

Modelling (GLM; SPSS v. 11.5). The dependent

variables we investigated were the density of each

small mammal species individually (calculated by

Minimum Number Known Alive; MNKA), the

total density of all small mammals and the total

biomass of all animals trapped per 100 m of

hedgerow. We focused on bank voles, wood mice,

field voles Microtus agrestis Linnaeus and on

yellow-necked mice Apodemus flavicollis Melch-

oir where they occurred; we had few captures of

common shrews Sorex araneus Linnaeus, however

these data were included in dependent variables

Landscape Ecol (2007) 22:1019–1032 1021

123

that included all animals, (biomass and total

abundance of all species). The total numbers of

individuals of each species trapped during the 3-

day period was used as an index of the relative

density of small mammals in the hedgerows.

The predictor variables considered were as

follows (for measurement types see Table 2):

flailed; height, width; cross-section; end connec-

tions; proximity to woodland; standards; total

gaps; number of plant species dominant; total

number of plant species; hedgebank; ditch; and

conservation buffer strip. The total available

hedgerow length ‘available length’ (total

length – total gap length) was calculated.

The associations between our predictor and

dependent variables were initially investigated

separately by the construction of univariate step-

wise GLMs. We took account of the repeated

measurements within farms (multiple hedgerows

per study site) by including the variable ‘farm’ as

a fixed blocking factor in all models. Because of

the well-documented seasonal changes in small

mammal abundance (Alibhai and Gipps 1985;

Flowerdew 1985) we also include ‘‘season’’ as a

fixed factor in all models, even where it was not a

significant predictor with our data. ‘‘Year’’ was

not a significant predictor in any model and was

therefore excluded from the analysis. Other

variables were added to the model in turn and

their effects evaluated. Residuals were examined

to ensure the data fulfilled the assumptions of the

model. Where appropriate, dependent variables

were transformed to stabilise the variance or fulfil

the assumptions of normality of errors.

Separate GLMs were constructed for each of

the following dependent variables; wood mouse

abundance; bank vole abundance; field vole

abundance, yellow-necked mouse abundance;

wood mouse plus yellow-necked mouse abun-

dance (‘all mice’); total captures of individuals (of

Fig. 1 An example mapof a typical farm, showinghedgerows with trap linesin bold

1022 Landscape Ecol (2007) 22:1019–1032

123

all species); and total biomass of all the small

mammals trapped on each hedgerow. To assist in

interpretation of the system each landscape ele-

ment was initially investigated separately; we

investigated associations between each of these

dependent variables and three different elements

of landscape structure as suggested by the HEGS

system (Clements and Tofts 1992): the connectiv-

ity of the habitat (predictor variables were ‘end

connections’, ‘woodland proximity’ and ‘gaps’);

habitat availability (‘width’, ‘height’, ‘cross-sec-

tion’, ‘length’ and ‘conservation buffer strip’); and

localised habitat structure (the presence of a

ditch, hedgebank and standard trees). Predictor

variables not included within these models were

found during initial examination of the data to be

not significant. We then investigated the com-

bined effects of the different landscape ele-

ments—connectivity, habitat availability, and

structure. We did this separately for each depen-

dent variable and to avoid model instability we

compared the models for all possible combina-

tions of predictors and selected the most parsi-

monious model with the greatest explanatory

power (adjusted R2). Finally, the effect of small

mammal species on each other was also investi-

gated.

Gwent was the only area where yellow-necked

mice were caught in addition to wood mice.

Where there was evidence of an association

between yellow-necked mouse density and a

habitat variable, the data for wood mice from

this region were analysed separately.

Results

We captured 3,048 individual small mammals of

five species during 21,600 trap nights. 2,524

individuals (82.8%) were recaptured at least

once.

The most abundant species were wood mice

and bank voles, together comprising 80.1% of the

individual animals captured. Yellow-necked mice

were predominantly caught in only one geograph-

ical area, Gwent (Table 1). Age class (adult—-

males with scrotal testes and females with

perforate vaginas; or juvenile—males with

abdominal testes and females with imperforate

vaginas) was assigned to 2,805 (92%) of animals

in total; 54.5% of wood mice, 49.5% of bank

voles, 42.2% field voles and 42.9% yellow-necked

mice were adult at the time of first capture.

Gender was assigned to 2,867 (95.6%) of all

animals captured; 57.5% of wood mice were male,

41.8% female, 47.6% of bank voles were male,

51.5% female, 54.3% of yellow necked mice were

male, 44.1% female, and 55.9% of field voles

were male, 44.1% female.

The average total availability of hedgerow on

the 12 farms investigated was 9,370 m

(SD = 4,583 m). See Table 2 for average

Table 1 The breakdown by species of all small mammalstrapped, and percentage of hedgerows found to house eachspecies. The Gwent region has been calculated separately

to account for the regional abundance of yellow-neckedmice, and the potential impact of this on other species

Yellow-necked mouseApodemus flavicollis

Wood mouseApodemussylvaticus

Bank voleClethrionomysglareolus

Field voleMicrotusagrestis

Commonshrew Sorexaraneus

Allspecies

Total N trappedthroughoutstudy

200 1,291 1,149 247 161 3,048

Percentage of total 6.6 42.3 37.7 8.1 5.3 100Percentage

hedgerows (allareas)

n/a 93.9 92.2 42.2 37.8 –

Percentagehedgerows(Gwent only)

75 100 96 52 21 –

Landscape Ecol (2007) 22:1019–1032 1023

123

hedgerow characteristics. Average farm size was

103.6 ha (SD = 49.9 ha), with 11.1 m hedgerow

per hectare, and an average hedgerow width of 2–

3 m.

Connectivity

In analyses examining each habitat variable relat-

ing to hedgerow connectivity separately (adjusting

only for farm and season), the number of end

connections was positively associated with the

density of wood mice (F1, 164 = 4.539; P = 0.035);

all species (F1, 164 = 4.654; P = 0.032) and all mice

(F1, 164 = 4.055; P = 0.046). The number of end

connections was not associated with any other

dependent variable (P ‡ 0.146 in each case). An

increase in the gappiness of the hedge was associ-

ated with a decrease in bank vole density (F1,

164 = 5.113; P = 0.025), but was not a significant

predictor of any other dependent variable

(P ‡ 0.108). In simultaneous analysis, which also

adjusted hedgerow length similar results were

obtained for gappiness and the number of end

connections (Table 3). Woodland proximity was

not associated with any dependent variable

(P ‡ 0.207 in all cases).

Habitat availability

The height and width of hedgerows were highly

correlated (rs = 0.625, P < 0.001). We used the

variable ‘width’ rather than ‘height’ in the model

because it was a slightly stronger predictor for all

dependent variables except abundance of all mice

and wood mice (excluding Gwent): in these last

two cases, the results were not materially affected

by the use of width rather than height.

The width of the hedge was a strong positive

predictor of all dependent variables in initial

analyses (adjusting for farm and season) (Ta-

ble 4). Estimates indicate that for each one meter

increase in hedgerow width per 100 m hedgerow

length there is a mean increase of 1.2 (SE = 0.8)

bank voles, 1.1 (SE = 0.7) wood mice and 56.2 g

(SE = 29.3) biomass.

Due to the associations between hedgerow

width and yellow-necked mouse density, and the

possibility of this association confounding the

relationship between wood mouse density and

hedge width, the data presented for wood mice

exclude those individuals trapped in Gwent.

After adjusting for width, the available length

of the hedgerow (excluding gaps) was not a

significant predictor of small mammal abundance

(Table 5).

Replacing the number of gaps with the percent-

age of gaps, or total length of gap gave slightly

improved explanatory power for bank vole density

(adj. R2 = 0.182 and 0.163 respectively, compared

with 0.150 (Table 5)). There was no difference in

explanatory power for the density of wood mice, all

mice or all species. Combinations of predictor

values as measures of habitat patch were investi-

gated (width · real length; width · height; wid-

Table 2 Hedgerow characteristics (scored on a categorical basis unless otherwise indicated)

Hedgerow characteristic Measurement description Mean SD

Width Mean canopy width in m 2.7 0.83End connections Count (hedgerows = 1; woodland = 2) 3.0 1.15Woodland proximity Score 0–2, proximity of woodland patches connected via

not more than 200 m adjacent hedgerow0.5 0.70

Ditch Scored by reference to standard diagram 0.69 1.24Standard trees Count per 100 m 3.48 1.70Flailed Yes = 1/no = 0 0.31 0.46Height Mean canopy height in m 2.82 0.95Cross-section Scored from diagram 2.92 0.83Percentage gaps Per 100 m, derived from initial count of gaps <20 m in length 2.91 0.36No. species dominant 1–2 species = 1 or mixed = 3 3.33 0.95No. shrubby species Count 8.37 3.42Hedgebank Topographical features associated with hedge; yes = 1/no = 0 2.51 0.77Conservation buffer zone 2 m+ unimproved grass verge adjacent to hedge

(present = 1/absent = 0)0.12 0.51

1024 Landscape Ecol (2007) 22:1019–1032

123

Table 3 Summary statistics from connectivity model using number of gaps and end connections as predictor variables.Length is a covariate. Regression coefficients (bs) are given for continuous predictor variables only

Model Variables F P b Adj. R2

All species Farm F11, 162 = 3.277 P < 0.001 0.324Season F3, 162 = 1.046 P = 0.374Length F1, 162 < 0.001 P = 0.998 1.331 E-05Number of gaps F1, 162 = 2.669 P = 0.104 –0.441End connections F1, 162 = 4.574 P = 0.031* 1.330

Wood mice Farm F11, 162 = 1.822 P = 0.054 0.247Season F3, 162 = 1.078 P = 0.360Length F1, 162 = 0.032 P = 0.858 –0.001Number of gaps F1, 162 = 0.278 P = 0.599 –0.084End connections F1, 162 = 4.567 P = 0.034* 0.769

Bank voles Farm F11, 162 = 3.711 P < 0.001 0.139Season F3, 162 = 0.736 P = 0.532Length F1, 162 = 0.110 P = 0.741 –0.001Number of gaps F1, 162 = 4.569 P = 0.034* –0.326End connections F1, 162 = 0.191 P = 0.663 0.151

Field volesa Farm F11, 162 = 4.899 P < 0.001 0.249Season F3, 162 = 0.120 P = 0.949Length F1, 162 = 2.234 P = 0.137 0.001Number of gaps F1, 162 = 5.151 P = 0.025* –0.049End connections F1, 162 = 1.251 P = 0.265 0.054

All mice Farm F11, 162 = 1.573 P = 0.111 0.339Season F3, 162 = 1.806 P = 0.148Length F1, 162 = 0.115 P = 0.735 –0.001Number of gaps F1, 162 = 0.047 P = 0.828 –0.041End connections F1, 162 = 4.127 P = 0.044* 0.856

* Significant at the 95% confidence levela Field vole data were log-transformed

Table 4 Summary statistics from habitat availability model using width as the predictor variable

Model Variables F P b Adj. R2

All species Farm F11, 164 = 4.176 P = 0.000 0.335Season F3, 164 = 0.938 P = 0.424Width F1, 16 = 8.084 P = 0.005** 2.334

Biomass Farm F11, 164 = 3.198 P = 0.001 0.251Season F3, 164 = 0.571 P = 0.635Width F1, 16 = 9.941 P = 0.002** 56.175

Wood mice Farm F11, 164 = 1.872 P = 0.046 0.255Season F3, 164 = 0.926 P = 0.430Width F1, 16 = 4.606 P = 0.033* 1.406

Bank voles Farm F11, 164 = 3.945 P = 0.000 0.154Season F3, 164 = 0.967 P = 0.410Width F1, 16 = 6.384 P = 0.012* 1.175

Yellow-necked mice Farm F42, 2 = 0.248 P = 0.782 0.148Season F42, 2 = 0.989 P = 0.380Width F42, 1 = 0.637 P = 0.014* 1.877

All mice Farm F11, 164 = 2.013 P = 0.030 0.369Season F3, 164 = 1.666 P = 0.176Width F1, 16 = 10.109 P = 0.002** 1.783

* Significant at the 95% confidence level

** Significant at the 99% confidence level

Landscape Ecol (2007) 22:1019–1032 1025

123

th · real length · height). Height · length was a

significant predictor of wood mouse abundance (F1,

164 = 8.654; P = 0.004), but no other measure of

habitat patch size was a significant predictor of any

other dependent variable (P ‡ 0.140).

Separate univariate GLMs revealed that pres-

ence of an unimproved grass conservation buffer

strip fenced off from the pastoral field was a

significant positive predictor of field vole density

(F1, 164 = 6.001; P = 0.015), but not for any other

dependent variable (P ‡ 0.271 in each case).

Structure

The presence of a hedgebank, ditch, the number

of standards, and the effects of flailing were

investigated individually (all models adjusted for

farm and season). The presence of a hedgebank

was not associated with any dependent variable

(P ‡ 0.109 in each case), except for a marginal

relationship with all species (F1, 164 = 4.003;

P = 0.047).

Ditch was positively associated with yellow-

necked mice density (F1, 42 = 18.077; P < 0.001);

bank vole density (F1, 164 = 5.590; P = 0.019); all

species (F1, 164 = 15.137; P < 0.001); all mice (F1,

164 = 16.747; P < 0.001) and biomass (F1,

164 = 21.269; P < 0.001). The presence of standard

trees was associated with wood mice density (F1,

119 = 17.492) and all mice (F1, 164 = 8.176;

P = 0.005). When ditch and standards were

entered simultaneously into a GLM similar

results were obtained (Table 6).

Hedgerows that had been flailed had a lower

bank vole density (F1, 164 = 4.237; P = 0.041), but

a higher field vole density (F1, 164 = 5.751;

Table 5 Summary statistics from habitat availability model using width and real length (total length minus gap length) aspredictor variables

Model Variables F P b Adj. R2

All species Farm F11, 163 = 4.147 P < 0.001 0.331Season F3, 163 = 0.932 P = 0.427Real length F1, 163 = 0.003 P = 0.958 0.000Width F1, 163 = 8.038 P = 0.005* –0.462

Biomass Farm F11, 163 = 3.179 P = 0.001 0.246Season F3, 163 = 0.569 P = 0.636Real length F1, 163 = 0.016 P = 0.900 –0.013Width F1, 163 = 9.895 P = 0.002* 56.227

Wood micea Farm F8, 132 = 1.892 P = 0.067 0.299Season F3, 132 = 0.966 P = 0.411Real length F1, 132 = 0.279 P = 0.598 –0.002Width F1, 132 = 7.154 P = 0.009* 1.659

Bank voles Farm F11, 163 = 3.906 P < 0.001 0.150Season F3, 163 = 0.910 P = 0.438Real length F1, 163 = 0.318 P = 0.574 –0.001Width F1, 163 = 6.420 P = 0.012* 1.182

Field voles Farm F11, 163 = 5.019 P < 0.001 0.276Season F3, 163 = 0.827 P = 0.481Real length F1, 163 = 3.444 P = 0.065 0.002Width F1, 163 = 4.382 P = 0.038* –0.462

Yellow-necked mice Farm F2, 41 = 0.133 P = 0.876 0.132Season F2, 41 = 0.959 P = 0.392Real length F1, 41 = 0.209 P = 0.650 –0.002Width F1, 41 = 6.676 P = 0.013* 1.908

All mice Farm F11, 163 = 2.011 P = 0.030 0.356Season F3, 163 = 1.664 P = 0.177Real length F1, 163 = 0.094 P = 0.760 –0.002Width F1, 163 = 10.093 P = 0.002* 1.659

* Significant at the 95% confidence level

** Significant at the 99% confidence levela Wood mice results exclude Gwent data due to the presence of yellow-necked mice confounding results

1026 Landscape Ecol (2007) 22:1019–1032

123

P = 0.018). (P ‡ 0.477 for all other dependent

variables).

We then investigated the combined effects of

the different landscape elements on each depen-

dent variable. Habitat availability and structure

remained significant predictors of wood mice

density (width F1, 118 = 5.156; P = 0.025; stan-

dards F1, 118 = 15.141; P < 0.001; adj.

R2 = 0.377) but the minimum sufficient model

excluded connectivity. The density of bank

voles was associated with connectivity and

structure (number of gaps F1, 163 = 6.015;

P = 0.015; ditch F1, 163 = 6.492; P = 0.012; adj.

R2 = 0.175) with the minimum sufficient model

excluding habitat availability. Field vole density

was associated with connectivity and availability

(number of gaps F1, 163 = 5.781; P = 0.017;

width F1, 163 = 5.381; P = 0.022). The explana-

tory power of this model was slightly reduced

when compared to the habitat availability model

(adj. R2 = 0.260 and 0.276 respectively). Bio-

mass was associated with availability and struc-

ture (width F1, 163 = 4.822; P = 0.030; ditch

F1,163 = 15.751; P < 0.001). This combined mod-

el for biomass had the best explanatory power

(adj. R2 = 0.312). Two separate models, with

approximately equal explanatory power were

associated with all species; connectivity and

structure (end connections F1, 163 = 4.781;

P = 0.030; ditch F1, 163 = 15.211; P < 0.001; adj.

R2 = 0.375) and availability and structure (width

F1, 163 = 4.042; P = 0.046; ditch F1, 163 = 10.889;

P = 0.001; adj. R2 = 0.373). Similarly for the

density of all mice, connectivity and structure

(end connections F1, 162 = 5.703; P = 0.018;

standards F1, 162 = 5.971; P = 0.016; ditch F1,

162 = 12.340; P = 0.001; adj. R2 = 0.421) and

habitat availability and structure (width F1,

162 = 0.029; P = 0.029; ditch F1, 162 = 9.060;

P = 0.003; standards F1, 162 = 3.912; P = 0.050;

adj. R2 = 0.418) produced models with similar

explanatory power. In the combined model of

habitat availability and structure for yellow-

necked mice density, width was no longer a

significant variable (P = 0.324) but ditch re-

mained significant (F1, 41 = 10.877; P = 0.002).

Table 6 Summary statistics from structure model using ditch and standards as the predictor variables

Model Variables F P b Adj. R2

All species Farm F11, 163 = 4.662 P < 0.001 0.357Season F3, 163 = 0.448 P = 0.719Ditch F1, 163 = 13.844 P < 0.001* 2.001Standards F1, 163 = 0.089 P = 0.765 0.149

Biomass Farm F11, 163 = 3.748 P < 0.001 0.292Season F3, 163 = 0.332 P = 0.802Ditch F1, 163 = 19.635 P < 0.001* 50.924Standards F1, 163 = 0.067 P = 0.797 2.750

Wood micea Farm F8, 118 = 2.804 P = 0.007 0.352Season F3, 118 = 1.600 P = 0.193Ditch F1, 118 = 0.503 P = 0.480 0.265Standards F1, 118 = 14.731 P < 0.001* 1.140

Bank voles Farm F11, 163 = 3.664 P < 0.001 0.150Season F3, 163 = 0.745 P = 0.527Ditch F1, 163 = 6.341 P = 0.013* 0.780Standards F1, 163 = 0.935 P = 0.335 –0.278

Yellow-necked mice Farm F2, 41 = 0.507 P = 0.606 0.303Season F2, 41 = 0.067 P = 0.935Ditch F1, 41 = 18.013 P < 0.001* 2.070Standards F1, 41 = 0.548 P = 0.463 –0.780

All mice Farm F11, 163 = 3.273 P < 0.001 0.405Season F3, 163 = 1.356 P = 0.258Ditch F1, 163 = 12.777 P < 0.001* 1.292Standards F1, 163 = 4.442 P = 0.037* 0.707

* Significant at the 95% confidence levela Wood mice data exclude Gwent results for consistency with earlier analysis

Landscape Ecol (2007) 22:1019–1032 1027

123

Discussion

Our results suggest that hedgerow connectivity,

habitat availability and local structure may be

important predictors of the density of small

mammal populations on lowland British farm-

land. These results did not support the Random

Sample Hypothesis (Connor and McCoy 1979;

Haila 1983; Andren 1994) that small islands

represent random samples of the species found

within larger areas. For each of our dependent

variables (except yellow-necked mice) we were

able to create models to predict small mammal

density with greater explanatory power when two

classes of landscape structure, operating at dif-

ferent scales were investigated simultaneously.

This suggests that where the proportion of suit-

able habitat is low, so patch size (i.e. habitat

availability), isolation (i.e. connectivity) and local

structure of the habitat play an increasing role.

These findings are in line with the Fragmentation

Hypothesis, suggesting that, even for small,

localised populations, landscape-scale habitat

availability and connectivity is vital to maintain

viable population sizes of each species (Saunders

et al. 1991).

Although hedgerows can act as corridors link-

ing woodland habitat fragments, and so provide a

route for migration (Soule and Terbough 1999),

within a British pastoral landscape linear hedge-

rows are often the sole habitat available for small

mammals (Fitzgibbon 1997; Pollard et al. 1974).

Assart hedgerows originating from and connected

to woodland patches show greater diversity of

woody species (Edwards et al. 2006), although

our investigation found that the proximity and

connectivity of hedgerows to woodland had no

impact on the relative abundance of small mam-

mals. The ratio of adults: juveniles captured

during different seasons were entirely consistent

with population age structure for each species in

static populations as found in woodland habitats

(rough grassland habitats for field voles) (Alibhai

and Gipps 1991a, b; Flowerdew 1991; Montgom-

ery 1991). This indicates that animals in this study

are resident in hedgerows, rather than merely

using them for migration purposes. Differing

ecological requirements may explain some of

the variation in the numbers trapped of each

species; of the two species found most prolifically,

bank voles are active burrowers, creating runs

and pathways through the ground vegetation in

deciduous habitats (Morris 1982; Alibhai and

Gipps 1985), whereas wood mice are a generalist

species occupying a wide variety of habitat

(Flowerdew 1993) and consuming as a wild range

of food sources depending upon season and

availability (Montgomery 1978; Flowerdew

1993). Of the species trapped less frequently,

yellow-necked mice are patchily distributed

throughout the UK, but where they are found

their primary habitat is mature deciduous wood-

land (Montgomery 1978; Flowerdew 1993). Field

voles are specialists, depending upon rough,

ungrazed grassland, although marginal woodland

and hedgerows with long grass are also used

(Alibhai and Gipps 1991b). Our results suggest

that field voles are more likely to be present in

hedgerows adjacent to areas of rough grassland;

this is in line with Schweiger et al. (2000), who

found distribution of another Microtine species,

the Prairie vole (Microtus ochrogaster) in succes-

sional phases of an old field plant community to

be closely juxtaposed with their preferred grass-

land habitat found in interstitial areas between

experimental blocks.

In arable environments wood mice (but no

other small mammal species) have been shown to

make substantial use of the field at certain times

of the year (Tattersall et al. 2001; Tew et al. 2000;

Todd et al. 2000). No small mammal species have

been shown to make use of agriculturally im-

proved pastoral fields at any time of year (Mont-

gomery and Dowie 1993).

A high level of connectivity to adjoining

hedgerows typically maintains a higher density

of wood mice and bank voles (Fitzgibbon 1997).

The localised differences revealed between wood

mice and bank voles in this study may be

explained by their different natural histories.

Mice have larger home ranges than do voles

(Flowerdew 1993; Alibhai and Gipps 1991a),

which in a linear habitat may give rise to more

individual mice utilising end connections, and

may mean gaps within the hedgerow are of less

consequence. In contrast, gaps within the smaller

vole territories may occupy a larger proportion of

the total available habitat, therefore voles utilis-

1028 Landscape Ecol (2007) 22:1019–1032

123

ing territories without gaps or with a larger

number of smaller gaps would maximise food

and resource availability.

The overall abundance of birds has been shown

to increase with increasing hedgerow size and the

presence or abundance of trees (Green et al.

1994; Hinsley and Bellamy 2000). As mammals

and birds all require food and shelter we expected

to find similar results for small mammal habitat

requirements in hedgerows. A reduction in hedge

width will reduce the total amount of hedgerow

habitat available and at a scale relevant to small

mammal species the habitat may be subjected to

an increase in edge-effect, thereby reducing the

quality of the core habitat. Our study shows that

all small mammal species increase in relative

abundance with increased hedge width, an unsur-

prising result when considering that wider hedge-

rows contain more physical habitat than narrow

ones at a scale relevant to small mammals. In

addition, wider hedgerows may be indirectly

preferred by small mammals due to an increase

in microhabitat complexity and therefore in-

creased refuge opportunity from potential preda-

tors (Orrock et al. 2004).

This study found the presence of a conserva-

tion buffer strip to be positively associated only

with the relative abundance of field voles. Hinsley

and Bellamy (2000) found that birds preferred

hedgerow types that were similar to their normal

non-hedgerow habitat. These results concord with

our findings that field voles may be found in

atypical improved or semi-improved pastoral

hedgerow habitat when it is associated with a

fenced-off grassland conservation buffer strip.

The presence of a buffer strip may provide the

field voles’ preferred habitat of rough, ungrazed

grassland vegetation (Alibhai and Gipps 1991b).

We found the presence of a ditch a strong

positive predictor of all dependent variables

except wood mice and field voles, and the

presence of a hedgebank to be associated with

the relative density of all species. Bank voles have

previously been found to be more numerous in

boundaries with ditches than those without (Tew

et al. 1994). They create extensive burrow sys-

tems below ground within which are found nests

and food caches (Flowerdew 1993). A ditch may

increase the amount of habitat available to create

a burrow system, thereby increasing habitat

availability at a scale relevant to small mammals.

A ditch may also encourage the growth of more

ground vegetation, which has previously been

associated with bank vole density (Boone and

Tinklin 1988; Gurnell 1985; Pollard and Relton

1970), and localised soil drainage may help

maintain dryer burrow systems and food caches’.

Our results showed a strong positive associa-

tion between yellow-necked mice and the pres-

ence of a ditch, and although the effects of a ditch

have not previously been studied in relation to

yellow-necked mice, Kotzageorgis and Mason

(1997) found hedgerow condition to be the most

significant feature predicting the presence of this

species, which showed a preference for well-

established hedgerows with few gaps.

Yellow-necked mice are reputed to dwell

primarily within mature deciduous woodland

(Montgomery 1978) although telemetry studies

have indicated that individuals can live wholly

within a linear hedgerow habitat (Kotzageorgis

and Mason 1996). For arboreal small mammal

species able fully to utilise a three-dimensional

habitat as a means of increasing the available

habitat (i.e. yellow-necked mice (Flowerdew

1993), and to some extent wood mice (Flowerdew

1991)) we investigated the presence of standard

trees. We found no significant relationship be-

tween yellow-necked mice density and standard

trees but there was a significant relationship

between wood mice and standards. This may be

explained by the presence of standard trees within

a hedgerow reflecting a high seed-mast potential,

thus attracting and supporting a larger number of

wood mice than hedgerows without standard

trees (Mallorie and Flowerdew 1994; Montgom-

ery and Dowie 1993). Wood mice are habitat

generalists, able to thrive in a wide range of

conditions (Tattersall et al. 2001; Tew et al. 1994;

Todd et al. 2000). However, in areas of high grass

productivity wood mice are less abundant (Mont-

gomery and Dowie 1993). This may be due to the

combination of few seed-producing plants, recur-

rent disturbance from grass management, live-

stock and the application of lime and fertilizers

within the adjacent field (Montgomery and Do-

wie 1993). Our investigation reflected this phe-

nomenon, with overall numbers of wood mice and

Landscape Ecol (2007) 22:1019–1032 1029

123

bank voles trapped being very similar (1,291

wood mice and 1,149 bank voles).

Increased agricultural mechanisation and the

reduced cost of alternative methods of boundary

fencing has led to a decline in traditional hedge-

row management (Joyce et al. 1988; Pollard et al.

1974). Many hedgerows are now either under-

managed, creating lines of trees with little or no

understorey cover, or over-managed, creating

box-like hedges (Macdonald and Johnson 1995).

Under CAP reform, the UK Government has

recently changed and simplified its environmental

subsidy schemes, creating a new single payment

scheme for which the farmer must accrue a

number of points from different management

practices. Options include the application of a

combination of basic management methods (DE-

FRA 2005a), including hedgerow management

practice options relevant to small mammals

(Macdonald and Baker 2005) such as mechani-

cally flailing hedgerows on rotation (Macdonald

et al. 2006). Our findings indicate that mechani-

cally flailing hedgerows does significantly reduce

the bank vole population, although the opposite is

true for field voles. Flailing different aspects of a

single hedgerow in rotation may therefore benefit

both species within a single linear habitat.

When suggesting ways of improving habitat

management for bats, Russ and Montgomery

(2002), stress the importance of maintaining and

enhancing connecting linear habitats. Our evi-

dence supports their view; connectivity of linear

hedgerows is important for small mammal popu-

lation density and even small gaps may have a

negative impact on some species. Habitat man-

agers should remain mindful of the relative scale

of habitat for each different species being consid-

ered. The evidence presented here suggests that

preserving hedgerows and their adjacent habitat

may be important for maintaining metapopula-

tions of small mammal species within the UK, in

light of which the authors recommend that wider

and more complex hedgerows with adjacent

habitats should be encouraged, even at the risk

of loosing some adjacent arable habitat. On an

international scale this study adds weight to the

proposal that corridors may play a vital role in the

preservation of a number of species deemed to be

‘at risk’ from the impact of habitat fragmentation.

Acknowledgements We are grateful to landowners,fieldworkers and volunteers too numerous to mention fortheir help with data collection. We would like to thankPaul Johnson for statistical advice, Amanda Lloyd for helpwith GIS, and Fran Tattersall for comments on earlierdrafts. This work was funded as part of a larger project byDEFRA (SE3009) and by grants to David Macdonaldfrom the Peoples Trust for Endangered Species. FionaMathews was supported by a Royal Society DorothyHodgkin Fellowship.

References

Alibhai SK, Gipps JHW (1985) The population dynamicsof bank voles. Symp Zool Soc Lond 55:277–305

Alibhai SK, Gipps JHW (1991a) Bank vole. In: Harris S(ed) The handbook of British mammals. BlackwellScientific Publishing, Oxford

Alibhai SK, Gipps JHW (1991b) Field vole. In: Harris S(ed) The handbook of British mammals. BlackwellScientific Publishing, Oxford

Andren H (1994) Effects of habitat fragmentation on birdsand mammals in landscapes with different proportionsof suitable habitat: a review. Oikos 71:355–366

Barr CJ, Gillespie MK (2000) Estimating hedgerow lengthand pattern characteristics in Great Britain usingcountryside survey data. J Environ Manage 60:23–32

Bennett AF (1987) Conservation of mammals within afragmented forest environment: the contributions ofinsular biogeography and autecology. In: SaundersDA, Arnold GW, Burbidge AA, Hopkins AJM (eds)Nature conservation: the role of remnants of nativevegetation. Surrey Beatty and Sons, Chipping Norton,Australia, pp 41–52

Boone GC, Tinklin R (1988) Importance of hedgerowstructure in determining the occurrence and density ofsmall mammals. Asp Appl Biol 16:73–78

Boonstra R, Craine ITM (1996) Natal nest location andsmall mammals tracking with a spool and line tech-nique. Can J Zool 64:1034–1036

Bright PW (1993) Habitat fragmentation – problems andpredictions for British mammals. Mammal Rev23:101–111

Clements DK, Tofts RJ (1992) Hedgerow evaluation andgrading system (HEGS). A methodology for theecological survey, evaluation and grading of hedge-rows. Countryside Planning and Management, UK

Connor EF, McCoy ED (1979) The statistics and biologyof the species-area relationship. Am Nat 113:791–833

CPRE (1999) Hedging your bets: is hedgerow legislationgambling with our heritage? Council for the protec-tion of rural England, UK

DEFRA (2005a) Entry level stewardship handbook.Department of Environment, Food and Rural Affairs,London

DEFRA (2005b) The countryside stewardship scheme:traditional farming in the modern environment.Department of Environment, Food and Rural Affairs,London

1030 Landscape Ecol (2007) 22:1019–1032

123

Edwards K, Leighton D, Llewellyn P (2006) Hedgerowsand the historic landscape: a case study from southGower. Br Wildl 17(4):260–269

Fitzgibbon CD (1997) Small mammals in farm woodlands:the effects of habitat, isolation and surrounding land-use patterns. J Appl Ecol 34:530–539

Flowerdew JR (1985) The population dynamics of woodmice and yellow-necked mice. Symp Zool Soc Lond55:315–332

Flowerdew JR (1991) Wood mouse. In: Harris S (ed) Thehandbook of British mammals. Blackwell ScientificPublications, Oxford

Flowerdew JR (1993) Mice and voles. Whittet Books,London

Green RE, Osborne PE, Sears EJ (1994) The distributionof passerine birds in hedgerows during the breeding-season in relation to characteristics of the hedgerowand adjacent farmland. J Appl Ecol 31:677–692

Gurnell J (1985) Woodland rodent communities. SympZool Soc Lond 55:377–411

Gurnell J, Flowerdew JR (1990) Live trapping smallmammals: a practical guide. Mammal Society, London

Haddad NM, Bowne DR, Cunningham A, Danielson BJ,Levey DJ, Sargent S, Spira T (2003) Corridor use bydiverse taxa. Ecology 84:609–615

Haila Y (1983) Land birds on northern islands: a samplingmetaphor for insular colonization. Oikos 41:334–351

Haila Y (1990) Towards an ecological definition of an is-land: a north-west European perspective. J Biogeogr17:561–568

Hanski I (1999) Metapopulation ecology. Oxford Univer-sity Press, Oxford

Hinsley SA, Bellamy PE (2000) The influence of hedgestructure, management and landscape context on thevalue of hedgerows to birds: a review. J EnvironManage 60:33–49

Joyce B, Williams G, Woods A (1988) Hedgerows: still acause for concern. RSPB Conserv Rev 2:34–37

Kotzageorgis GC, Mason CF (1996) Range use, deter-mined by telemetry, of yellow-necked mice (Apode-mus flavicollis) in hedgerows. J Zool 240:773–777

Kotzageorgis GC, Mason CF (1997) Small mammal pop-ulations in relation to hedgerow structure in an arablelandscape. J Zool 242:425–434

Macdonald DW, Baker S (2005) The state of Britain’smammals 2005. Mammals Trust, UK, London

Macdonald DW, Johnson PJ (1995) The relationship be-tween bird distribution and the botanical and struc-tural characteristics of hedges. J Appl Ecol 32:492–505

Macdonald DW, Johnson PJ (2000) Farmers and the cus-tody of the countryside: trends in loss and conserva-tion of non-productive habitats 1981–1998. BiolConserv 94:221–234

Macdonald DW, Tattersall FH, Service KM, Firbank LG,Feber RE (2006) Mammals, agri-environmentschemes and set-aside – what are the putative bene-fits? Mammal Rev (in press)

Mallorie HC, Flowerdew JR (1994) Woodland smallmammal population ecology in Britain – a preliminaryreview of the mammal-society survey of wood mice

Apodemus-sylvaticus and bank voles Clethrionomys-glareolus, 1982–87. Mammal Rev 24:1–15

Mathews F, Macdonald DW, Taylor GM, Gelling M,Norman RA, Honess PE, Foster R, Gower CM,Varley S, Harris A, Palmer S, Hewinson G, WebsterJP (2006) Bovine Tuberculosis (Mycobacterium bovis)in British farmland wildlife: the importance to agri-culture. Proc R Soc B 273:357–365

Montgomery WI (1978) Studies on the distributions ofApodemus sylvaticus (L.) and A. Flavicollis (Mel-chior) in Britain. Mammal Rev 8:177–184

Montgomery WI (1991) Yellow-necked mouse. In: HarrisS (ed) The handbook of British mammals. BlackwellScientific Publishing, Oxford

Montgomery WI, Dowie M (1993) The distribution andpopulation regulation of the wood mouse Apodemus-sylvaticus on field boundaries of pastoral farmland.J Appl Ecol 30:783–791

Morris P (ed) (1982) The country life book of naturalhistory of the British Isles. Hamlyn Publishing GroupLtd., London, 6–7

Orrock JL, Danielson BJ, Brinkerhoff RJ (2004) Rodentforaging is affected by indirect, but not by direct, cuesof predation risk. Behav Ecol 15(3):433–437

Pollard E, Relton J (1970) A study of small mammalsin hedges and cultivated fields. J Appl Ecol 7:549–557

Pollard E, Hooper MD, Moore NW (1974) Hedges. Col-lins, London

Rackham O (1997) History of the countryside. PhoenixGiant, London

Russ JM, Montgomery WI (2002) Habitat associations ofbats in Northern Ireland: implications for conserva-tion. Biol Conserv 108:49–58

Saunders DA, Hobbs RJ, Margules CR (1991) Biologicalconsequences of ecosystem fragmentation: a review.Conserv Biol 5:18–32

Schweiger W, Diffendorfer J, Holt RD, Pierotti R (2000)The interaction of habitat fragmentation, plant, andsmall mammal succession in an old field. Ecol Monogr70:383–400

Soule ME, Terbough J (1999) Continental conservation:scientific foundations of regional reserve networks.Island Press, Washington, DC

Tattersall FH, Macdonald DW, Hart BJ, Johnson P,Manley W, Feber R (2002) Is habitat linearityimportant for small mammal communities on farm-land? J Appl Ecol 39:643–652

Tattersall FH, Macdonald DW, Hart BJ, Manley WJ, Fe-ber RE (2001) Habitat use by wood mice (Apodemussylvaticus) in a changeable arable landscape. J Zool255:487–494

Tattersall FH, Macdonald DW, Hart BJ, Manley WJ(2004) Balanced dispersal or source-sink – do bothmodels describe wood mice in farmed landscapes?Oikos 106:536–550

Tew TE, Todd IA, Macdonald DW (1994) Field marginsand small mammals. BCPC Monograph 58 fieldmargins-integrating agriculture and conservation,85–94

Landscape Ecol (2007) 22:1019–1032 1031

123

Tew TE, Todd IA, Macdonald DW (2000) Arable habitatuse by wood mice (Apodemus sylvaticus). 2.Microhabitat. J Zool 250:305–311

Todd IA, Tew TE, Macdonald DW (2000) Arable habitatuse by wood mice (Apodemus sylvaticus). 1. Macro-habitat. J Zool 250:299–303

Tofts RJ, Clements DK (1994) The development andtesting of HEGS, a methodology for the evaluationand grading of hedgerows. BCPC Monogr 58:277–282

Wilcox BA (1980) Insular ecology and conservation. In:Wilcox BA (ed) Conservation biology: an evolution-ary-ecological perspective. Sinauer, Sunderland, MA,pp 95–117

Wilcox BA, Murphy DD (1985) Conservation strategy: theeffects of fragmentation on extinction. Am Nat125:879–887

1032 Landscape Ecol (2007) 22:1019–1032

123