do landscape factors affect brownfield carabid assemblages?

18
Do landscape factors affect brownfield carabid assemblages? Emma Small a , Jon P. Sadler b, * , Mark Telfer c a Forestry Commission Wales, Victoria House, Aberystwyth, Ceredigion, SY23 2DQ, UK b School of Geography, Earth and Environmental Sciences, The University of Birmingham, Birmingham, B15 2TT, UK c UK Headquarters, The RSPB, The Lodge, Sandy, Bedfordshire, SG19 2DL, UK Available online 7 November 2005 Abstract The carabid fauna of 28 derelict sites in the West Midlands (England) were sampled over the course of one growing season (April–October, 1999). The study aimed to investigate the relationship between carabid assemblages and five measures of landscape structure pertinent to derelict habitat. At each site measurements of landscape features pertinent to derelict habitat were made: (i) the proximity of habitat corridors; (ii) the density of surrounding derelict land; (iii) the distance between the site and the rural fringe; and (iv) the size of the site. Concurrent surveys of the soil characteristics, vegetation type, and land use history were conducted. The data were analysed using a combination of ordination (DCA, RDA), variance partitioning (using pRDA) and binary linear regression. The results suggest that: D 2005 Elsevier B.V. All rights reserved. Keywords: Brownfield; Derelict; Carabidae; Landscape ecology; Conservation; URGENT; West Midlands 1. Introduction Urban environments are complex and subjected to intensive disturbance pressures from development cycles and pollutants (McDonnell and Pickett, 1990). As a result, they are characterised by a mosaic of different habitats, often juxtaposed in unlikely combi- nations, with a variety of former land uses. Concerns over the sustainability of urban living (Douglas, 1992), coupled with an increasing awareness of the value of urban areas for nature conservation (Breuste et al., 0048-9697/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.scitotenv.2005.08.051 * Corresponding author. E-mail address: [email protected] (J.P. Sadler). 1. There is very little evidence that the carabid assemblages of derelict sites were affected by landscape structure, with assemblages instead being principally related to within-site habitat variables, such as site age (since last disturbance), substrate type and vegetation community. 2. No evidence was found to support the hypothesis that sites away from railway corridors are more impoverished in their carabid fauna than sites on corridors. 3. There are some suggestions from this study that rarer and non-flying specialist species may be affected by isolation, taking longer to reach sites. We infer from this that older sites with retarded succession, and sites in higher densities of surrounding derelict land may eventually become more species rich and that these sites may be important for maintaining populations of rarer and flightless species. 4. Conservation efforts to maintain populations of these species should focus principally on habitat quality issues, such as maintaining early successional habitats that have a diversity of seed producing annuals and perennial plants and enhancing substrate variability rather than landscape issues. Science of the Total Environment 360 (2006) 205– 222 www.elsevier.com/locate/scitotenv

Upload: emma-small

Post on 12-Sep-2016

219 views

Category:

Documents


5 download

TRANSCRIPT

Page 1: Do landscape factors affect brownfield carabid assemblages?

www.elsevier.com/locate/scitotenv

Science of the Total Environm

Do landscape factors affect brownfield carabid assemblages?

Emma Small a, Jon P. Sadler b,*, Mark Telfer c

a Forestry Commission Wales, Victoria House, Aberystwyth, Ceredigion, SY23 2DQ, UKb School of Geography, Earth and Environmental Sciences, The University of Birmingham, Birmingham, B15 2TT, UK

c UK Headquarters, The RSPB, The Lodge, Sandy, Bedfordshire, SG19 2DL, UK

Available online 7 November 2005

Abstract

The carabid fauna of 28 derelict sites in the West Midlands (England) were sampled over the course of one growing season

(April–October, 1999). The study aimed to investigate the relationship between carabid assemblages and five measures of

landscape structure pertinent to derelict habitat. At each site measurements of landscape features pertinent to derelict habitat

were made: (i) the proximity of habitat corridors; (ii) the density of surrounding derelict land; (iii) the distance between the site and

the rural fringe; and (iv) the size of the site. Concurrent surveys of the soil characteristics, vegetation type, and land use history

were conducted. The data were analysed using a combination of ordination (DCA, RDA), variance partitioning (using pRDA) and

binary linear regression. The results suggest that:

1. There is very little evidence that the carabid assemblages of derelict sites were affected by landscape structure, with

assemblages instead being principally related to within-site habitat variables, such as site age (since last

disturbance), substrate type and vegetation community.

2. No evidence was found to support the hypothesis that sites away from railway corridors are more impoverished in

their carabid fauna than sites on corridors.

3. There are some suggestions from this study that rarer and non-flying specialist species may be affected by isolation,

taking longer to reach sites. We infer from this that older sites with retarded succession, and sites in higher densities

of surrounding derelict land may eventually become more species rich and that these sites may be important for

maintaining populations of rarer and flightless species.

4. Conservation efforts to maintain populations of these species should focus principally on habitat quality issues,

such as maintaining early successional habitats that have a diversity of seed producing annuals and perennial plants

D 2005 Elsevier B.V. All rights reserved.

Keywords: Brownfield; Derelict; Carabidae; Landscape ecology; Conservation; URGENT; West Midlands

and enhancing substrate variability rather than landscape issues.

1. Introduction

Urban environments are complex and subjected to

intensive disturbance pressures from development

0048-9697/$ - see front matter D 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.scitotenv.2005.08.051

* Corresponding author.

E-mail address: [email protected] (J.P. Sadler).

cycles and pollutants (McDonnell and Pickett, 1990).

As a result, they are characterised by a mosaic of

different habitats, often juxtaposed in unlikely combi-

nations, with a variety of former land uses. Concerns

over the sustainability of urban living (Douglas, 1992),

coupled with an increasing awareness of the value of

urban areas for nature conservation (Breuste et al.,

ent 360 (2006) 205–222

Page 2: Do landscape factors affect brownfield carabid assemblages?

E. Small et al. / Science of the Total Environment 360 (2006) 205–222206

1998), have led to resurgence of interest in urban

ecology (Pickett et al., 2001). In the UK and Europe,

this momentum has been generated by the potential

conservation significance of urban habitats (Gibson,

1998).

A number of entomological papers have highlighted

the potential of brownfield (or derelict) sites in urban

areas as important habitats for locally rare (Gruttke and

Weigmann, 1990; Kegel, 1990), and in some case

nationally rare (Eversham et al., 1996) invertebrates

species. In a few studies, species originating from

heathland (Andersen, 2000; Gruttke and Weigmann,

1990) or coastal habitats (Andersen, 2000; Lazenby,

1983) have been recorded, illustrating the importance

of brownfields as analogues for natural habitats (Ever-

sham et al., 1996). Indeed, it has been estimated that

over 50% of British rare aculeate Hymenoptera and

over 35% of British rare and scarce carabid species

have been recorded from anthropogenic habitats (Gib-

son, 1998), which highlights the need for targeted

research on urban environments (McIntyre, 2000).

More recent work has shown that carabid beetle diver-

0 5 10 15 20

D24

D25

D23D22

D20

D19 D18D17

D15

D16D11D10 D08

D09

D13

D06 D

D

D14A&B

D21A&B

D12A&B

25 Kilom

D01

D07

Fig. 1. Map of the 28 derelict survey sites in the West Midlands, show

sity is strongly influenced by habitat variables, such as

the type and successional stage of the vegetation on

brownfield sites (Schwerk, 2000; Small et al., 2003).

However, the spatial structure of the urban environment

can also have significant consequences for species

assemblages. Landscape ecology addresses issues sur-

rounding the heterogeneity and structuring of habitats

and its effect on species dynamics (Forman and God-

ron, 1981).

Outside of urban areas, numerous studies have been

conducted to investigate the effects of patch size and

isolation on carabid species in terrestrial fragmented

systems (for a review, see Niemela, 2001), utilising

the theoretical insights from islands biogeography the-

ory (MacArthur and Wilson, 1967) and metapopulation

dynamics (Hanski, 1998). Patch area and patch connec-

tivity were not found to be positively related to total

carabid species richness in forests (Abildsnes and Tom-

meras, 2000; Halme and Niemela, 1993; Hanski, 1998;

Magura et al., 2001), farm woodlands (Usher et al.,

1993), heathland (De Vries et al., 1996; Webb, 1989),

or limestone outcrops (Bauer, 1989). Other studies have

LEGEND

Derelict Survey Sites

Railways

W. Midlands Boundary

Density of Derelict Land

0 - 21.433

21.433 - 42.866

42.866 - 64.299

64.299 - 85.731

85.731 - 107.164

107.164 - 128.597

No Data

D02

04 D03

05

eters

N

S

W E

ing the density of derelict land in the conurbation (Source: JDT).

Page 3: Do landscape factors affect brownfield carabid assemblages?

(a)

16.812.88.84.8.8

10

8

6

4

2

0

Std. Dev = 5.22Mean = 9.6N = 28.00

(b)

616.7500.0383.3266.7150.033.3

10

8

6

4

2

0

Std. Dev = 211.32Mean = 264.7N = 28.00

(c)

3.002.502.001.501.00.500.00

14

12

10

8

6

4

2

0

Std. Dev = .78Mean = .67N = 28.00

Distance to nearest railway (km)

Num

ber

of s

ites

Density of derelict land (hectares in 5km)

Num

ber

of s

ites

Shortest distance to rural edge (km)

Num

ber

of s

ites

Fig. 2. Histograms of the selected survey sites in terms of (a) position

on rural–urban gradient (km to rural edge); (b) density of derelict land

(hectares within 5 km) and (c) distance to railway (km).

E. Small et al. / Science of the Total Environment 360 (2006) 205–222 207

shown that habitat specialist (stenotopic) carabid spe-

cies are more vulnerable to habitat fragmentation than

habitat generalist (eurytopic) carabid species, for exam-

ple in heathland (De Vries et al., 1996) and forests

(Halme and Niemela, 1993; Magura et al., 2001). Ad-

ditionally, carabid species with low powers of dispersal

have been shown to be more vulnerable than good

dispersers (De Vries et al., 1996; Den Boer, 1970,

1977; Den Boer and Den Boer-Daanje, 1990; Den

Boer et al., 1980).

A study of Coleoptera in experimental forest frag-

ments (Davies and Margules, 1998) indicated that spe-

cies which occur naturally at low abundance were more

likely to decline as a result of fragmentation than

abundant species, and that predator species were more

negatively affected than species lower down the food

chain. Unfortunately, little research has examined how

important landscape structure and landscape ecology

are in defining carabid assemblages at a city scale,

although comparative data on the pervasive effects of

habitat fragmentation and disturbance caused by urban-

isation has been provided by the GLOBENET project

(Alaruikka et al., 2002; Ishitani et al., 2003; Niemela et

al., 2002).

The aim of this study was to investigate the relation-

ship between carabid assemblages and five measures of

landscape structure pertinent to derelict habitat: (i) dis-

tance from (greyway, Austin, 2002) habitat corridors,

(ii) the density of derelict land, (iii) site location on the

rural–urban gradient, (iv) site size, and (v) site age. In

order to fulfil the aim the following related hypotheses

are addressed:

H1. Landscape structure variables will be able to

explain a significant amount of variation in the as-

semblages that is not explained by habitat quality

variables.

H2. The species richness and rarefied richness of all

species, stenotopic species, species with low powers of

dispersal, large-bodied species and rare species will be

greatest at sites (i) on habitat corridors; (ii) in areas of

high density of derelict land; (iii) at the rural end of the

urban–rural gradient; (iv) in larger sites; and (v) older

sites that have not undergone succession.

H3. Some species, particularly those that are stenotop-

ic, poor dispersers, large-bodied or rare species, will

be found more frequently at sites (i) on habitat corri-

dors; (ii) in areas of high density of derelict land; (iii)

at the rural end of the urban–rural gradient; (iv) in

larger sites; and (v) older sites that have not under-

gone succession.

2. Methods

2.1. Study sites

Field surveys were undertaken at 28 derelict land

sites between mid-April and mid-October 1999 (Fig.

1). The study sites, all between 2 and 20 years old,

were selected from a pool of potential sites in the

West Midlands conurbation (England) to provide a

spread along the rural–urban gradient (Fig. 2a), in

areas with high and low density of derelict land

(Fig. 2b), and on or off railway corridors (Fig. 2c)

(Table 1). Vegetation development ranged from early

successional communities (c. 2 years old) to later

successional communities characterised by grassland

with shrub woodland. Site substrate varied between

graded and mixed rubble to rubble dressed with top-

soil and nutrient enriched topsoil. Former land use

ranged from old buildings and factories, railway land

to arable set aside and the sites varied between b1 ha

to N30 ha (Table 1).

Page 4: Do landscape factors affect brownfield carabid assemblages?

Table 1

The twenty-eight derelict sites surveyed, their previous use, age, substrate and vegetation type and habitat variables (vegetation classification: 1. bare/ruderal; 2. tall herb; 3. transient grass; 4. grassland/shrub)

Site

id

Site Previous use Substrate Disturbance since

bbirthQAge

(years)

Age since

disturbance

Vegetation

classification

(1–4)

% soil

moisture

% organic

matter

Impenetrability

(kgf/cm2)

Litter

depth

(cm)

% grass

cover

% bare

ground

pH Vegetation

density

D01 Tyseley Wharf Factories Top-soil dumped

on rubble

Little 5 5 1 18.9 4.0 2.58 2.2 16.0 83.0 7.8 96.6

D02 Minworth Sewage works Nutrient rich soil/clay Soil dumping and

turning

20 3 1 18.6 7.0 3.96 0.7 41.0 55.0 8.2 61.0

D03 Erdington1 House Graded brick rubble Flytipping, fire and

garden waste

4 4 1 21.4 10.1 2.84 1.6 14.2 85.3 7.9 91.5

D04 Erdington2 Garden Graded brick rubble Disturbed during demolition 20 4 1 21.3 11.0 2.85 2.4 38.5 79.5 7.9 91.8

D05 Reservoir Rd Housing Graded brick rubble Flytipping, fire and

garden waste

15 15 4 26.4 9.8 2.16 5.8 33.5 66.5 8.2 97.9

D06 Ashted Circus Factory/yard Compacted rubble Compaction by

vehicles, tipping

7 7 1 16.9 5.7 4.74 1.3 11.5 117.5 7.8 72.2

D07 Soho Loop Railway siding Compacted ballast Compaction by vehicles 12 12 3 20.0 6.0 2.85 1.2 17.0 81.0 7.8 74.2

D08 Heath Street Housing Graded brick rubble Disturbed during

nearby demolition

20 4 2 19.5 6.9 2.54 1.8 2.0 98.0 8.0 91.1

D09 Florence Rd Housing Graded brick rubble Flytipping, fire and

garden waste

8 8 3 25.4 9.1 2.61 1.1 12.5 81.0 7.9 91.8

D10 Cape Hill Housing Graded brick rubble Flytipping and

garden waste

8 8 4 22.2 6.1 2.68 0.7 7.3 70.7 7.7 65.0

D11 Woodlands1 Temporary

housing

Compacted rubble Flytipping, fire and

trampling

4 4 1 16.3 3.7 3.18 1.3 10.0 98.5 7.7 90.5

D12 Woodlands2 Temporary

housing

Graded brick rubble Flytipping, fire and

trampling

4 4 2 19.5 4.4 3.96 0.9 11.0 89.0 7.9 65.5

D13 Institute Rd Swimming

baths

Graded brick rubble Trampling, flytipping,

garden waste

12 12 3 19.0 3.5 1.82 1.1 30.0 80.0 8.3 77.3

D14 Vincent Drive1 Factories Compacted rubble Trampling 15 15 1 18.2 1.9 1.12 0.2 18.0 79.0 7.9 62.8

D15 Vincent Drive2 Factories Raw brick Little 15 15 2 21.7 8.1 3.51 1.5 37.0 121.0 7.8 97.0

D16 Foxyards Rd Empty Graded rubble Disturbed during

roadworks, tipping

20 5 2 20.9 8.2 2.67 0 14.0 86.0 7.9 86.6

D17 Landfill Landfill site Top-soil dumped

on rubble

Compaction by

trucks, tipping

14 4 3 18.4 9.0 2.34 1.4 22.5 76.5 7.7 80.4

D18 Tunnel Street House/garden Compacted rubble Compaction by

vehicles, tipping

10 10 1 22.2 11.3 3.41 0.4 12.1 85.4 8.0 48.8

D19 Hall Green Rd Landfill site Top-soil dumped

on rubble

Trampling, tipping and oil 10 10 4 25.6 11.3 1.83 2.4 16.5 49.0 7.5 95.3

D20 Mounts Rd Blocks of flats Graded brick rubble Garden waste and flytipping 2 2 2 18.9 8.6 3.66 0.4 3.2 96.8 8.0 67.8

D21 Old Park Rd School Compacted rubble Flytipping, fire and

garden waste

10 10 1 26.4 7.4 3.39 0.3 46.1 46.7 8.0 66.0

D22 Bentley Mill 1 Sports ground Raw brick Flytipping, fire and

garden waste

14 4 1 26.0 7.9 3.01 2.0 22.5 77.5 7.6 40.5

D23 Bentley Mill 2 Sports ground Graded brick rubble Flytipping, fire and

garden waste

14 14 4 29.7 17.8 3.76 2.4 28.5 59.5 8.0 98.5

D24 M6 Empty Graded rubble Mound created during works 6 6 2 26.1 6.7 1.80 1.9 13.5 86.5 7.8 44.5

D25 Walsall Factory/yard Compacted rubble Flytipping 10 10 1 21.4 5.2 2.43 1.5 18.0 81.0 7.9 93.7

D26 Brownhills Railway line Compacted ballast Trampling 20 20 1 16.8 5.7 4.12 0.8 15.3 69.7 8.2 45.7

D27 Set1 Arable Rich Soil Ploughing 20 4 3 21.1 4.4 2.37 2.5 36.0 46.0 7.8 99.8

D28 Set2 Arable Rich Soil Ploughing 20 4 1 17.3 4.7 1.70 0.3 33.5 53.5 7.8 70.4

E.Smallet

al./Scien

ceoftheTotalEnviro

nment360(2006)205–222

208

Page 5: Do landscape factors affect brownfield carabid assemblages?

E. Small et al. / Science of the Total Environment 360 (2006) 205–222 209

2.2. Carabid data and habitat variables

Standardised pitfall trapping and hand search techni-

ques applied over a full season (April–October 1999)

were used to draw up a list of carabid species present.

Nine pitfall traps (7 cm diameter plastic cups), part filled

with propylene glycol, were installed at each site. Where

possible, the traps were placed in the centre of homog-

enous stands of vegetation at each site. To gain as full a

species complement as possible, pitfall trapping was

supplemented by 30-min day time hand searches (Ander-

sen, 1995), conducted in late May, June and August.

The habitat variables recorded at each site are detailed

in Table 1. Much of the information on the age (Age)

and previous land use of the sites was gathered from a

questionnaire sent to 30 households around each of the

sites, and from aerial photos (CityView, 1995), data from

the Birmingham and Black Country Urban Wildlife and

dated Ordinance Survey maps. Fifteen soil samples at 5–

10 cm depth were taken from each site in early October

1999. These samples were used to establish mean soil

moisture (% weight loss on drying at room temperature)

(Moist) and mean organic matter content (% weight

Table 2

The twenty-eight derelict sites and the landscape variables

Site id Site Distrail

km

Der5000

ha

Der1000

ha

De

ha

D01 Tyseley Wharf 0.4 140.4 3.2 0

D02 Minworth 0.6 53.1 3.2 0.7

D03 Erdington1 1.8 147.6 2.5 0.2

D04 Erdington2 2.0 145.5 0 0.2

D05 Reservoir Rd 0.6 198.4 0 0.5

D06 Ashted Circus 0.5 267.7 8.1 0

D07 Soho Loop 0 166.4 2.1 1.0

D08 Heath Street 0 163.4 2.6 2.1

D09 Florence Rd 1.2 111.1 0.5 0.3

D10 Cape Hill 1.4 105.2 0.1 0

D11 Woodlands1 0 221.4 1.5 1.7

D12 Woodlands2 0 221.4 1.5 1.7

D13 Institute Rd 0.6 78.4 0.7 0

D14 Vincent Drive1 0 26.4 17.7 1.5

D15 Vincent Drive2 0 26.4 17.7 1.5

D16 Foxyards Rd 0.7 615.5 9.9 0

D17 Landfill 1.5 334.4 73. 3.7

D18 Tunnel Street 0 520.1 8.0 0

D19 Hall Green Rd 0.1 434.2 17.6 3.4

D20 Mounts Rd 0 557.2 11.9 0.7

D21 Old Park Rd 1.1 667.8 17.9 0.1

D22 Bentley Mill 1 0.2 565.7 47.3 0.9

D23 Bentley Mill 2 0.2 565.7 47.3 0.9

D24 M6 0.5 557.5 46.6 0.8

D25 Walsall 0 370.1 33.2 1.8

D26 Brownhills 0 112.1 5.3 0.5

D27 Set1 3.1 18.6 0 3.3

D28 Set2 1.5 18.2 0 1.5

loss on combustion) (LOI). A mean of 30 readings of the

following habitat variables were recorded at each site:

soil impenetrability (measured in kgf/cm2 (Impenet);

percentage of bare ground (Bare); and litter depth

(measured in cm) (Litter). In addition, the plant com-

munities of each of these 26 sites were surveyed in the

summer of 1999 and 2000 (Austin, 2002). The vegeta-

tion of three 1�1 m quadrats in the trap area was

surveyed using a Braun–Blanquet scale. Each quadrat

was then matched in Tablefit to NVC communities and

used to provide a vegetation classification for the sites

(categories 1–4 in Table 1). Lastly, substrate type was

classified (as Soil, Graded, Compact, Rawbrick

(agricultural sites were considered as Soil)). As the

activities of the council and locals can affect derelict

sites, resetting important successional processes (Small

et al., 2003), we also established when the sites were last

disturbed (Agedist) and the nature of that disturbance.

2.3. Landscape variables

The proximity of each site to the nearest railway

(=nearest habitat corridor) was measured using Ordi-

r100 Urban cover

in 5 km (%)

Distedge

km

Size (ha) Logarea

ha

76.2 11.2 8.7 1.9

34.1 0 1.4 1.2

63.5 5.0 5.7 1.8

63.0 6.0 2.6 1.4

72.6 10.5 1.5 1.5

78.6 18.0 4.2 1.6

73.6 16.0 10.3 2.0

72.2 15.0 0.5 1.2

71.2 12.0 5.7 1.8

71.2 11.0 11.4 2.1

71.3 12.5 19.9 2.3

71.3 15.5 19.9 2.3

65.2 6.5 2.0 1.3

65.2 11.0 10.1 2.0

65.2 11.0 10.1 2.0

67.7 10.0 0.7 0.8

73.7 9.5 14.8 2.2

69.0 8.0 1.8 1.3

66.8 18.5 38.9 2.6

72.7 17.0 8.5 1.9

75.4 12.5 3.4 1.5

71.6 8.5 29.4 2.5

71.6 8.5 29.4 2.5

70.8 9.0 10.5 2.0

60.7 8.0 3.5 1.5

36.3 0 N30 1.3

40.5 0 8.7 1.9

29.5 0 15.5 2.2

Page 6: Do landscape factors affect brownfield carabid assemblages?

E. Small et al. / Science of the Total Environment 360 (2006) 205–222210

nance Survey maps (km, Distrail) (Table 2). The

amount of derelict land surrounding each site was

quantified within buffers at three distances from the

site boundary, 100 m, 1 km and 5 km. The land

cover within the 100 m radius from the boundary was

mapped on the ground between August and October

1999. Each map was digitised, and incorporated into an

ArcView GIS (Version 3.2). Data from the JDT (1998)

database of derelict land in the West Midlands were

also added. ArcView was then used to extract the

percentage of derelict land within each buffer

(Der100, Der1000 and Der5000) (Table 2). Site

position on the rural–urban gradient was measured in

two ways. Firstly, the boundary of the conurbation was

established using Ordinance Survey maps, and the

shortest distance between site and the urban edge was

measured in kilometres (Distedge). Secondly, the

ANALYS

(1) InvestigaterelationshipbetweenLANDSCAPE &VARIATION INASSEMBLAGE

(2) InrelatbetwLAN& SPRICH

SIMPLE TESTING: 1A: RedundancyAnalysis (RDA):Test significance ofeach landscapevariable inexplaining similarityin species matrix

2A: Curve(linear, log,exponentiasignificancerelationshiprichness / rand each la

MODELLING: 1B: Partial RDA:Partition variance inspecies matrix (Y) into thatexplained by habitatvariables (X), landscapevariables (W) &unexplained variance.

Test for normality of variables

2B: UniRegresFind bespeciesusing h(X) andvariable

Preliminaryanalyses:

(i) Single landscapevariables used

(ii) All landscapevariables used together

QUESTIONQ1: Landscape factors can explain a significant amoQ2: Species richness is greater at less isolated sitesQ3: Some species will be found more frequently at l

Fig. 3. Study questions and m

percentage of urban and suburban land cover within a

5 km buffer around the site boundary was calculated in

ArcInfo using the Institute of Terrestrial Ecology Land-

cover data (Urb5000). ArcView was also used to

provide accurate measures of site area, measured in

square metres and logged (Logsize).

2.4. Data analysis

The three study hypotheses outlined above required

the use of several different statistical techniques, which

are illustrated in Fig. 3. All of the species, landscape

and habitat variables were tested for normality using

Kolgorov–Smirnov exact tests in SPSS (SPSS, 2000).

All conformed to normal distributions except (i) each

of the substrate types, which are binary variables

( p b0.001) and (ii) Urb5000 ( p=0.038).

IS ROUTES

vestigateionshipeenDSCAPEECIESNESS

(3) Investigaterelationship betweenLANDSCAPE &INDIVIDUALSPECIESDISTRIBUTIONS

estimation power orl): Test of the between speciesarefied richnessndscape variable

3A: Logistic binaryregression:Test significance of eachlandscape variable inexplaining thepresence/absence of speciesin the first step of regression

and for covariation between variables

variate Linearsion:st predictors of richness (Y)abitat variables landscapes (W).

3B: Univariate LogisticBinary Regression:Find best predictors ofspecies presence/absence(Y) using habitatvariables (X) andlandscape variables (W).

Sunt of the variation in the assemblage; or those with retarded successioness isolated sites or those with retarded succession

ethods of data analysis.

Page 7: Do landscape factors affect brownfield carabid assemblages?

E. Small et al. / Science of the Total Environment 360 (2006) 205–222 211

All habitat and landscape variables were then tested

for autocorrelation using non-parametric two-tailed

Spearman’s Rank Correlation tests in SPSS (Table 3).

Correlations between landscape variables showed that

there was a strong association between Distedge and

Urb5000 (rs=0.734, n =28, p b0.001) which are both

measures of site position on the rural–urban gradient.

As Urb5000 was not normally distributed, it was

removed from further analysis. Sites on railway corri-

dors were positively associated with a greater density of

derelict land within 1 km (Distrail and Der1000,

rs=�0.418, n =28, p b0.05). Urban sites had a greater

density of derelict land within 5 km (Urb5000 and

Der5000, rs=0.520, n =28, p b0.005). Larger sites

tended to have more derelict habitat in the immediate

vicinity (Logsize and Der100, rs=0.491, n =28,

p b0.01). Sites at the edge of the conurbation tended

to be older (Age and Urb5000, rs=�0.460, n =28,p b0.005; Age and Distedge, rs=�0.500, n =28,

p b0.005), though this was unrelated to age since dis-

turbance. Some significant correlations between land-

scape and habitat variables were also found (Table 3),

showing that some differences in the habitat quality of

sites are themselves spatially structured. In particular,

tests showed that sites on corridors tended to be com-

pacted (Distrail and Compact, rs=�0.476, n =28,p b0.05) and bare (Distrail and Bare, rs=�0.530,n =28, p b0.01) rather than grassy (Distrail and

Grass, rs=0.381, n =28, p b0.05) or with a soil sub-

strate (Distrail and Soil, rs=0.482, n =28, p b0.05).

Sites away from the rural edge also tended to be less

grassy and more bare (Distedge and Grass,

rs=�0.441, n =28, p b0.05; Distedge and Bare,

rs=0.418, n =28, p b0.05).

Detrended Correspondence Analysis (DCA) was

carried out on the total species data set and specialist

species only. The resulting gradient lengths were fairly

short (b2), suggesting that linear analytical techniques

were more appropriate. Accordingly, Redundancy

Analysis (RDA) (rather than Canonical Correspon-

dence Analysis) was the preferred method of analysis.

Redundancy Analysis (RDA) was performed on the

same data using CANOCO for Windows version 4.0

(ter Braak and Smilauer, 1998). Presence/absence data

rather than abundance data were used in order to focus

the ordinations on species’ distribution patterns. In the

first step, single landscape variables were used individ-

ually as explanatory environmental variables (1A in

Fig. 3) and their significance tested using the Monte

Carlo technique with 999 permutations (Manly, 1994).

Thereafter, Partial Redundancy Analysis was used (1B

in Fig. 3), in which the variation in Y (the response

variables, i.e. the species matrix) is partitioned out

according to X (the set of explanatory environmental

variables, i.e. habitat factors) and W (the set of explan-

atory spatial variables, i.e. landscape factors) (Borcard

et al., 1992; Legendre and Legendre, 1998). Two sets of

runs were performed using this technique. Firstly single

landscape variables were again used individually as

explanatory environmental variables, but this time all

habitat variables were in each case used as covariates to

constrain the ordination (1Bi in Fig. 3). Secondly, all

landscape variables were used together, again using

habitat variables as covariates (1Bii in Fig. 3).

Data on the known habitat preference, dispersal

ability and mean body size (in mm) of all species

found were drawn from the literature (Table 4). Species

reported to have a strong preference for open, dry

habitats were identified as derelict specialists. Species

reported to be brachypterous or otherwise of doubtful

flight ability were identified as poor dispersers. Data on

the British rarity of each species were derived from the

Ground Beetle Recording Scheme (GBRS) database

(Luff, 1998). Species with fewer than 150 post-1970

GB 10 km2 records were classified as buncommonQ.The published conservation status of species was also

used (Hyman, 1992).

Total species richness (Totrich); the number of

specialist (derelict) species (Specrich); the total num-

ber of non-flying species (Notflytot); the number of

non-flying specialist species (Notflyspec); the num-

ber of uncommon species (Rare); and the maximum

body length of the total (Meansiztot) and specialist

(Meansizespec) species. In acknowledgement of the

biases in the data due to unequal sample size, rarefied

richness was also calculated (Hurlbert, 1971; Sanders,

1968). Rarefaction computes the expected number of

species in a standardised sampling unit. In this case the

standardised sampling unit was 60 individuals when all

species were considered (Rarefiedtot).

Curve estimation, performed in SPSS, was used as

a simple test for the relationships between landscape

factors and species richness or rarefied richness (2A

in Fig. 3). In the next step, a univariate linear re-

gression approach was taken, with both landscape

and habitat variables in the pool of explanatory vari-

ables (2B in Fig. 3). Stepwise selection ( p b0.05 for

inclusion, p N0.10 for removal) was used to find the

best combination of predictors of the species richness

measures.

The relationship between landscape factors and the

presence/absence of individual species was investigated

using logistic binary regression in SPSS (SPSS, 2000).

Only species occurring at 20–80% of sites were tested

Page 8: Do landscape factors affect brownfield carabid assemblages?

Table 3

Two-tailed Spearman’s rank correlations between the habitat and landscape variables measured at the derelict sites (n.s. not significant, yp b0.1, *p b0.05, **p b0.01, ***p b0.005)

Distrail

Distrail 1.000 DER5000

DER5000 –0.189n.s.

1.000 DER1000

DER1000 –0.418*

0.618***

1.000 DER100

DER100 –0.178n.s.

-0.065n.s.

0.111n.s.

1.000 URB5000

URB5000 –0.211n.s.

0.520**

0.203n.s.

–0.122n.s.

1.000 Distedge

Distedge †0.346†

0.339†

0.179n.s.

0.109n.s.

0.734***

1.000 Logsize

Logsize –0.019n.s.

0.112n.s.

0.181n.s.

0.491**

0.218n.s.

0.267n.s.

1.000 Age

Age 0.186n.s.

–0.341†

–0.139n.s.

0.170n.s.

–0.460***

–0.500***

–0.375†

1.000 Agedist

Agedist –0.298n.s.

–0.035n.s.

0.228n.s.

–0.150n.s.

–0.003n.s.

0.064n.s.

–0.066n.s.

0.128n.s.

1.000 VEG

VEG 0.248n.s.

–0.095n.s.

–0.287n.s.

0.235n.s.

0.106n.s.

0.143n.s.

0.336†

0.105n.s.

0.186†

1.000 Moist

Moist 0.197n.s.

0.414*

0.258n.s.

–0.106n.s.

0.122n.s.

0.014n.s.

0.117n.s.

–0.082n.s.

0.294n.s.

0.325†

1.000 LOI

LOI 0.256n.s.

0.434*

0.189n.s.

–0.095n.s.

0.096n.s.

–0.014n.s.

–0.061n.s.

0.064n.s.

0.011n.s.

0.212n.s.

0.607***

1.000 Impenet

Impenet –0.335†

0.232n.s.

0.185n.s.

–0.218n.s.

0.191n.s.

0.115n.s.

–0.111n.s.

–0.168n.s.

–0.087n.s.

–0.379*

–0.096n.s.

0.185n.s.

1.000 Litter

Litter 0.161n.s.

0.054n.s.

–0.042n.s.

0.278n.s.

0.121n.s.

–0.037n.s.

0.252n.s.

0.023n.s.

–0.002n.s.

0.217n.s.

0.402*

0.259n.s.

–0.148n.s.

1.000 Grass

Grass 0.381*

–0.230n.s.

0.010n.s.

0.045n.s.

–0.272n.s.

–0.441*

–0.057n.s.

0.542**

0.183n.s.

0.014n.s.

0.162n.s.

0.052n.s.

–0.110n.s.

0.213n.s.

1.000

Bare –0.530**

0.153n.s.

0.131n.s.

–0.109n.s.

0.288n.s.

0.418*

–0.103n.s.

–0.446*

–0.140n.s.

–0.508**

–0.237n.s.

–0.115n.s.

0.267n.s.

–0.090n.s.

–0.596***

(a)

(b)

(c)

(a) Landscape vs. landscape; (b) landscape vs. habitat; (c) habitat vs. habitat variables.

E.Smallet

al./Scien

ceoftheTotalEnviro

nment360(2006)205–222

212

Page 9: Do landscape factors affect brownfield carabid assemblages?

E. Small et al. / Science of the Total Environment 360 (2006) 205–222 213

as the assumptions of logistic binary regression are

violated at more extreme values. Initial tests examined

the significance of each landscape variable at the first

step of the regression (3A in Fig. 3). In the next step,

habitat factors were added to the pool of explanatory

variables (3B in Fig. 3), and stepwise selection

( p b0.05 for inclusion, pN0.10 for removal) was used

to find the best combination of predictors. The results

were interpreted in the context of the known habitat

preferences, dispersal ability, body size and rarity of the

species.

3. Results

3.1. Carabid assemblage variation and landscape

factors

Exploratory analyses, where single landscape vari-

ables were used on their own to constrain the ordina-

tions, showed that for the total species data set,

Der1000, Distedge and Agedist had a significant

effect (Table 5a, all with p b0.05), while for the spe-

cialist species data set, only Agedist was significant

(Table 5a, p=0.024). Both Distedge and Agedist

showed autocorrelation with some habitat variables

(Table 3). Therefore, these results were reanalysed

using Partial Redundancy Analysis (Table 5b and

Table 5b). Once habitat variables were used as covari-

ates, only Agedist ( p =0.036) and Der1000

( p =0.018) remain significant for the total data set.

None of the landscape variables remains significant

for the specialist data set. This reduction in significance

after accounting for autocorrelation with the habitat

variables indicates that the effect of the landscape vari-

ables is to some extent dexplained awayT by spatially

structured habitat factors [e], leaving [ f] as non-signif-

icant and in most cases attributable to random variation

(see Legendre and Legendre, 1998, p. 776). Only the

weakly significant effects of Agedist and Der1000

on the total data set can be considered as potentially

btrueQ spatial effects in this analysis.

Table 5c shows the results of Partial Redundancy

Analysis, which was used to analyse the total effect of

the measured landscape factors on the ordinations.

The fraction of variance explained by the landscape

variables [b +c] was large and significant in the total

data set (40.5% of variance, p =0.004, but not signif-

icant for the derelict specialist data set (37.6% of

variance, p =0.097). When [b +c] was partitioned

out, [c] (the spatially structured variation of Y that

is not explained by the habitat variables) was still

substantial (33.4% and 33.0% of the variance in the

total species and specialist species data sets respec-

tively). A substantial fraction [c] can be a result of

spatially structured habitat variables that have not

been included in the model, or can be due to a true

response to landscape. However, [c] was not found to

be significant in either case, and therefore must be

interpreted as random variation. In addition, the

results showed that the specialist data set showed no

more alignment to the landscape variables than the

total species data set.

3.2. Species richness measures and environmental

variables

Most of the species richness measures were best

modelled using habitat variables only. The number of

specialist species and the number of rare species and the

rarefied richness of specialist species were most strong-

ly related to the successional stage of the vegetation

(Specrich and Veg, B =�2.59 p =0.016; Rare and

Veg, B =�2.17, p =0.039; Specrare and Veg,

B =�3.00, p =0.006). The number of non-flying spe-

cies was greatest at impenetrable sites (Notflyspec

and Impene, B =2.83, p =0.009). The total species

richness at a site was negatively related to site age

since disturbance (Totrich and Agedist, B =�2.86,p =0.008).

However, three of the species richness measures

showed significant relationships with the landscape

variables. Firstly, although total rarefied species rich-

ness was primarily related to the successional age of the

vegetation (Rarefiedtot and Veg B =5.39,

p =0.001), it was higher at sites with more surrounding

derelict land in the surrounding 1000 m (Rarefied-

tot and Der100, B =3.33, p =0.003). However, this

was not corroborated by the results for total species

richness (Totrich), specialised species richness (Spe-

crich) or rarefied specialist species richness (Spe-

crare). Further investigation of the data showed that

most abundant and dominant species in the survey,

Pterostichus madidus, was also strongly related to

Der1000 (p.m.c.c.=�0.572, p b0.001), resulting in a

higher rarefied total species richness (Rarefiedtot).

Secondly, the mean body size of the carabids species

was negatively related to the density of land within 5

km of the sites (Meansizetot and Der5000,

r2=0.348 p =0.001) and also to urban cover (Mean-

sizetot and Urb5sq, r2=0.289, p=0.003). Finally,

the mean body size of the derelict specialist species at a

site was primarily positively related to the amount of

bare ground (Meansizetot and Logbare, B =3.26,

p =0.009), but was also significantly lower at sites with

Page 10: Do landscape factors affect brownfield carabid assemblages?

Table 4

Carabid species captured by pitfall and hand search at the 28 derelict survey sites; their habitat preference (derived from Lindroth, 1974; Luff, 1998)

and their national rarity (**** recorded in b100 GB 10 km2, ***b200 squares, **b300 squares, *b400 squares)

Species Sites Total Habitat Rarity Species Sites Total Habitat Rarity

Pterostichus madidus

(Fabricius 1775)

26 6002 Generalist Notiophilus aquaticus

(Linnaeus 1758)

5 51 Open

Harpalus affinis (Schrank) 25 605 Open Pterostichus niger

(Schaller 1783)

5 30 Wood

Amara eurynota (Panzer) 23 464 Open *** Pterostichus strenuus

(Panzer 1796)

5 28 Generalist

Nebria brevicollis

(Fabricius 1792)

22 581 Generalist Amara praetermissa

(Sahlberg)

5 9 Open **** Nb

Amara lunicollis Schiodte 22 523 Generalist * Amara apricaria (Paykull) 4 24 Open *

Bembidion lampros

(Herbst 1784)

21 686 Open Harpalus tardus (Panzer) 4 22 Open **

Amara communis (Panzer) 21 145 Generalist * Asaphidion flavipes

(Linnaeus 1761)

4 13 Open

Trechus quadristriatus

(Schrank 1781)

20 218 Generalist Metabletus foveatus

(Fourcroy)

4 12 Open *

Notiophilus biguttatus

(Fabricius 1779)

19 198 Generalist Olisthopus rotundatus

(Paykull 1790)

4 9 Open

Notiophilus substriatus

Waterhouse1833

19 140 Open * Pterostichus melanarius

(Illiger 1798)

4 6 Generalist

Harpalus rufipes (Degeer) 18 507 Open Calathus piceus

(Marsham 1802)

3 18 Wood

Calathus fuscipes

(Goeze 1777)

18 231 Generalist Bembidion quadrimaculatum

(L. 1761)

3 8 Open

Calathus melanocephalus

(L. 1758)

18 142 Open Synuchus nivalis

(Panzer 1797)

3 8 Generalist **

Bradycellus verbasci Duft 17 92 Open Carabus nemoralis

Muller O.F. 1764

3 6 Generalist

Trechus obtusus Erichson 1837 16 95 Generalist Acupalpus meridianus

(Linnaeus)

3 5 Open ***

Amara aulica (Panzer) 15 51 Open Agonum muelleri

(Herbst 1784)

3 4 Generalist

Amara ovata (Fabricius) 14 284 Generalist * Anisodactylus binotatus

(Fabricius)

2 6 Generalist ***

Loricera pilicornis

(Fabricius 1775)

14 177 Generalist Amara plejeba (Gyll.) 2 3 Generalist

Amara aenea (Degeer) 14 75 Open Bembidion guttula

(Fabricius 1792)

2 2 Generalist

Amara bifrons (Gyll.) 13 93 Open ** Amara anthobia Villa

and Villa

1 30 Open ****

Harpalus rubripes

(Duftschmid)

13 56 Open ** Pterostichus cupreus

(Linnaeus 1758)

1 6 Open *

Dromius linearis (Olivier) 13 17 Open Notiophilus palustris

(Duftschmid 1812)

1 4 Generalist

Carabus violaceus Linnaeus

1758

12 87 Generalist Bembidion properans

Stephens 1828

1 3 Open **

Amara familiaris (Duft) 11 41 Open Agonum dorsale

(Pontoppidan 1763)

1 1 Open

Badister bipustulatus Sturm. 11 24 Open Amara convexior

Stephens

1 1 Open ***

Amara similata (Gyll.) 10 78 Generalist * Bembidion tetracolum

Say 1823

1 1 Generalist

Bradycellus harpalinus

(Serv)

10 44 Open Clivina fossor

(Linnaeus 1758)

1 1 Generalist

Platyderus ruficollis

(Marsham 1802)

10 41 Open *** Nb Dromius melanocephalus

Dejean

1 1 Open

Amara tibialis (Paykull) 7 36 Open ** Pterostichus diligens

(Sturm 1824)

1 1 Wet

E. Small et al. / Science of the Total Environment 360 (2006) 205–222214

Page 11: Do landscape factors affect brownfield carabid assemblages?

Species Sites Total Habitat Rarity Species Sites Total Habitat Rarity

Bembidion obtusum Serville

1821

6 44 Open * Pterostichus vernalis

(Panzer 1795)

1 1 Wet

Leistus ferrugineus

(Linnaeus 1758)

6 15 Generalist Pterostichus versicolor

(Sturm 1824)

1 1 Generalist *

Harpalus rufibarbis (Fab.) 6 9 Open ** Amara convexiuscula

(Marsham)

1 1 Coast ***

bNbQ denotes Notable B status from Hyman and Parsons (1992). Species ranked by number of sites were recorded.

Table 4 (continued)

E. Small et al. / Science of the Total Environment 360 (2006) 205–222 215

more derelict habitat within 100 m (Meansizespec

and Logder100, B =�2.3, p=0.03). However, this

result is difficult to interpret as Der100 and Logsite-

size were found to covary ( p b0.001); Table 3) and

Logsitesize was strongly related to Meansizespec

(r2=�0.156, p =0.037).

3.3. Individual species distributions

The results of Binary Logistic Regression analysis of

species presence/absence, performed on all species

occurring at between 20% and 80% of sites and using

landscape and habitat factors as potential explanatory

variables, are given in Table 6. The significance of all

landscape variables in the first step is also reported,

highlighting those that are significant after corrections

for multiple testing.

Very few of the seventeen specialist and twelve non-

specialist species tested showed significant relation-

ships to the landscape factors. The exceptions were

Bembidion obtusum (a derelict specialist; Table 6a),

Pterostichus niger and P. melanarius (both large gen-

eralist species), which were related to Distedge or

Urbsq.

Two derelict specialists, Harpalus rubripes (at youn-

ger sites) and Badister bipustulatus (at older sites), were

significantly more frequent on railway corridors. Five

specialist (Calathus melanocephalus (Der100), Dro-

mius linearis (Logder100), B. obtusum (Der100),

Amara praetermissa (Logder100), Notiophilus aqua-

ticus (Der100)) and two generalist species (Trechus

obtusus (Der100) and Harpalus rufibarbis (Der1000)

were significantly related to the amount of derelict land

in the local area.

Overall, site age appeared to be more important.

Four specialist species were significantly less common

at older sites (C. melanocephalus, Notiophilus substria-

tus, Amara bifrons and H. rubripes). Five non-special-

ist species (T. obtusus, Loricera pilicornis, Amara

similata, H. rufibarbis and Pterostichus melanarius)

also had age as a significant variable in their linear

models (Table 6b).

4. Discussion

4.1. Variation in the assemblage and relation to land-

scape factors

The central hypothesis of this study was that land-

scape structure variables would be able to explain a

significant amount of variation in the assemblages that

was not explained by habitat quality variables. Con-

trary to expectations, no landscape factors were able

to explain a significant amount of variation in the

assemblages of derelict specialist species (Table 5).

However, landscape variables did explain a significant

proportion of the variation in the total species data set,

notably the density of derelict land within 1 km of the

site (Der1000; Table 5). However, later analyses

showed a strong relationship between P. madidus

and Der100, which may have contributed to this

result. More important was the age of the site which

showed significant relationships in the ordination

results (Table 5). Unfortunately, this variable can be

viewed as both a landscape (relating to the amount of

time over which a site was available for colonising

individuals) and habitat (relating to the progress of

vegetational succession), making the result difficult to

interpret.

4.2. The species richness of the assemblage and the

representation of various traits

Hypothesis (H2) was that species richness and rare-

fied richness of all species, stenotopic species, species

with low powers of dispersal, large-bodied species and

rare species would be greatest at sites (i) on habitat

corridors; (ii) in areas of high density of derelict land;

(iii) at the rural end of the urban–rural gradient; (iv) in

larger sites; and (v) less on older sites that have under-

gone succession. None of the species richness metrics

was related to the landscape variables in a systematic

manner, although site age and in particular site age

since the last disturbance event did appear to be impor-

tant (Small et al., 2003), as total species richness de-

Page 12: Do landscape factors affect brownfield carabid assemblages?

Table 5

Results of RDA of presence/absence data using (i) total species data set; and (ii) derelict specialist species data set

(a) Single landscape variables used to constrain the ordination

Fraction of

variance

Landscape

variable

Total species data set Derelict species data set

Trace

(Eig)

Proportion of

total variation

Probability (999

permutations)

Trace

(Eig)

Proportion of

total variation

Probability (999

permutations)

[e + f] Distrail 0.039 3.9% 0.357 n.s. 0.035 3.5% 0.516 n.s.

[e + f] Der5000 0.046 4.6% 0.131 n.s. 0.041 4.1% 0.322 n.s.

[e + f] Der1000 0.057 5.7% 0.024* 0.051 5.1% 0.114 n.s.

[e + f] Der100 0.049 4.9% 0.067 n.s. 0.042 4.2% 0.271 n.s.

[e + f] Urb5sq 0.062 6.2% 0.009** 0.041 4.1% 0.316 n.s.

[e + f] Distedge 0.057 5.7% 0.023* 0.045 4.5% 0.181 n.s.

[e + f] Logsize 0.050 5.0% 0.067 n.s. 0.053 5.3% 0.075 n.s.

[e + f] Age 0.051 5.1% 0.054 n.s. 0.044 4.4% 0.238 n.s.

[e + f] Agedist 0.067 6.7% 0.002* 0.056 5.6% 0.040*

(b) Single landscape variables used to constrain the ordination with habitat variables used as covariates

Fraction of

variance

Landscape

variable

Total species data set Derelict species data set

Trace

(Eig)

Proportion of

total variation

Probability (999

permutations)

Trace

(Eig)

Proportion of

total variation

Probability (999

permutations)

[ f] Distrail 0.047 4.7% 0.067 n.s. 0.053 5.3% 0.064 n.s.

[ f] Der5000 0.027 2.7% 0.779 n.s. 0.026 2.6% 0.769 n.s.

[ f] Der1000 0.054 5.4% 0.018* 0.049 4.9% 0.100 n.s.

[ f] Der100 0.042 4.2% 0.181 n.s. 0.040 4.0% 0.302 n.s.

[ f] Urb5sq 0.042 4.2% 0.181 n.s. 0.040 4.0% 0.318 n.s.

[ f] Distedge 0.048 4.8% 0.065 n.s. 0.048 4.8% 0.129 n.s.

[ f] Logsize 0.034 3.4% 0.497 n.s. 0.037 3.7% 0.461 n.s.

[ f] Age 0.032 3.2% 0.054 n.s. 0.035 3.5% 0.538 n.s.

[ f] Agedist 0.052 5.2% 0.036* 0.046 4.6% 0.171 n.s.

(c) Partition of variance between habitat and landscape variables

Fraction of

variance

Description Total species data set Derelict species data set

Trace

(Eig)

Proportion

of variation

Probability (999

permutations)

Canonical

E1Probability (999

permutations)

Trace

(Eig)

Proportion

of variation

Probability (999

permutations)

Canonical

E1Probability (999

permutations)

[a +b] Habitat 0.320 32.0% 0.004 0.086 0.061 0.293 29.3% 0.104 0.085 0.266

[b +c] Landscape 0.405 40.5% 0.004 0.104 0.009 0.376 37.6% 0.097 0.110 0.092

[a +b +c] Habitat+Land 0.654 65.4% 0.048 0.124 0.762 0.623 62.3% 0.240 0.131 0.864

[a] Habitat only 0.249 24.9% 0.262 0.058 0.948 0.247 24.7% 0.454 0.073 0.813

[b] Autocorrelation 0.071 7.1% – – – 0.046 4.6% – – –

[c] Landscape only 0.334 33.4% 0.182 0.079 0.790 0.330 33.0% 0.377 0.094 0.657

[d] Unexplained 0.346 34.6% – – – 0.337 33.7% – – –

[a +b +c +d] Total 1.000 100% – – – 1.000 100% – – –

E.Smallet

al./Scien

ceoftheTotalEnviro

nment360(2006)205–222

216

Page 13: Do landscape factors affect brownfield carabid assemblages?

E. Small et al. / Science of the Total Environment 360 (2006) 205–222 217

clined on older sites (Agedist, Table 5). It seems

likely that this is a non-causal relationship relating to

habitat variables not measured in this study that change

over the course of succession, for example, a decline

in habitat heterogeneity, or more likely, vegetation

succession.

4.3. Individual species distributions

Hypothesis (H3) was that some species, particularly

those that are stenotopic, poor dispersers, large-bodied

or rare species, would be found more frequently at sites

(i) on habitat corridors; (ii) in areas of high density of

derelict land; (iii) at the rural end of the urban–rural

gradient; (iv) in larger sites; and (v) older sites. Con-

trary to this hypothesis, only 2 of the 17 derelict spe-

cialist species were found to be positively associated

with habitat corridors. However, a few specialist and

non-specialist species did occur significantly more fre-

quently at sites surrounded by a greater density of

derelict land. At the wider scale, Notiophilus aquatius

(a specialist) and H. rufibarbis (non-specialist) were

positively associated with greater derelict land within

1 km and 5 km respectively, although the results indi-

cated that this could be a spurious non-causal relation-

ship in the case of H. rufibarbis. A. praetermissa and B.

obtusum (specialists), and T. obtusus, Carabus viola-

ceus and P. niger (non-specialists) were positively re-

lated to greater density of derelict land at the local 100

m scale. Two derelict specialists were found at older

sites (B. bipustulatus and Platyderus ruficollis). B.

bipustulatus, while preferring dry, open habitats is ac-

tually fairly eurytopic, being found in shadier habitats

too (Lindroth, 1974). The result for P. ruficollis is more

interesting. It is a nationally scarce (Nb) species, pre-

ferring dry, sandy or chalky soils in open situations, and

has reduced wings.

4.4. Two possible interpretations of the results

These findings, indicating the very limited effect of

measured landscape variables on the derelict assem-

blages, have two possible interpretations.

4.5. Interpretation 1: Derelict assemblages are affected

by isolation, but these were not measured

If it is assumed that the isolation of a derelict site is

an important causal factor determining the composition

of derelict assemblages, then it must also be assumed

that this study was unable to measure the landscape

variables relevant to carabid dispersal in this city land-

scape. There are a number of reasons why this indeed

may have been the case.

4.5.1. Habitat corridors

In this study, twelve of the twenty-eight study sites

were located within 200 m of a railway, but distance

from railway was shown to have no significant effect on

assemblage or distributions of individual species. It is

possible that urban railways are not effective corridors,

and therefore have no bearing on the degree of habitat

isolation, perhaps because they have too many breaks

(e.g. bridges, stations), narrow areas or patches of

unsuitable habitat, which in turn create barriers, retard

movement, reduce reproduction and cause species to

exit the corridor.

4.5.2. Source pools within the conurbation

Hot spots of high density of derelict land occur in

the West Midlands (Fig. 1), particularly in the old

industrial areas such as Tipton and Walsall, while less

industrialised areas such as Birmingham city centre and

Solihull have very low densities. In this study, density

was measured at three scales (100 m, 1 km and 5 km

from the site) and high density was assumed to indicate

a higher concentration of source pools from which

dispersal could occur, relating to Gibson’s (1998) con-

ceptual model (c) where colonisation of a site princi-

pally occurs from sources within the urban area. It is

possible however that densities at the 1 km and 5 km

scales are irrelevant to carabid colonisation, if for ex-

ample carabid species find much smaller distances on

the scale of a few hundred metres unbridgeable. It is

also possible that the many tiny fragments of dderelictThabitat surrounding sites (for example in unkempt gar-

dens, walls, pavement cracks, road verges) rendered the

rather coarse measurements of derelict density at the 1

km and 5 km rather inaccurate. In these respects derelict

density at the local, 100 m scale (Der100) was a much

more reliable estimate as all fragments, however tiny, of

derelict land were mapped. Indeed, Der100 was more

strongly related to individual species distributions than

either Der1000 or Der5000.

4.5.3. Source pools in the rural area

Another possible source pool from which the dis-

persal of specialists might occur is the rural area beyond

the urban fringe, relating to Gibson’s (1998) conceptual

model (a). In this study the distance between sites and

the rural edge was not significantly related to any aspect

of the assemblage except for the distribution of a single

specialist, B. obtusum. However, it is possible that this

measure too was unable to dcaptureT a meaningful mea-

Page 14: Do landscape factors affect brownfield carabid assemblages?

Table 6

Binary logistic regression of individual species presence/absence, using both habitat and landscape variables and stepwise selection to find the best combination of predictors ( p b0.05 to include, p N0.10 to remove)

Species Sites Significance of landscape variables in first step Binary logistic regression model

Distrail Der5000 Der1000 Der100 Distedge Urb5000sq Logsize Age Agedist Variables in model, their coefficient

(B), and significance if removed

% Correctly predicted Sig of

modelAbsence Presence Total

(a) Specialist species

Harpalus rufipes 20 n.s. n.s n.s. n.s n.s. n.s n.s. n.s 0.044 Moist (B =�35.50, p =0.007) 62.5% 90.0% 82.1% 0.007

Calathus melanocephalus 19 n.s n.s n.s n.s n.s n.s n.s 0.012 n.s Age (B =�0.21, p =0.007) 44.4% 84.2% 71.4% 0.008

Notiophilus substriatus 19 n.s n.s n.s n.s n.s n.s n.s n.s n.s Veg (B =�1.5, p =0.001);Logder5000 (B =+2.6, p =0.029)

88.9% 89.5% 89.3% 0.001

Bradycellus verbasci 18 n.s n.s n.s n.s n.s n.s n.s n.s n.s Veg (B =�1.0, p =0.007) 70.0% 83.3% 78.6% 0.008

Amara aulica 16 n.s n.s n.s n.s n.s n.s n.s n.s n.s No significant variables – – – –

Amara aenea 15 n.s n.s. n.s n.s. n.s n.s. n.s n.s. n.s. No significant variables – – – –

Amara bifrons 14 n.s n.s n.s n.s n.s n.s n.s n.s 0.031 Logagedist (B =�3.7, p =0.022) 71.4% 71.4% 71.4% 0.022

Dromius linearis 14 n.s n.s n.s n.s n.s n.s n.s n.s n.s Loglitter (B =+10.8, p =0.001) 71.4% 78.6% 75.0% b0.001

Logder100 (B =+6.3, p =0.021)

Harpalus rubripes 13 0.034 n.s n.s n.s n.s n.s 0.027 0.014 n.s Age (B =�0.2, p =0.001) 80.0% 61.5% 71.4% 0.004

Logdistrail (B =�6.1, p =0.016)Amara familiaris 12 n.s n.s n.s n.s n.s n.s n.s n.s n.s No significant variables – – – –

Badister bipustulatus 11 n.s n.s. n.s n.s. n.s n.s. n.s n.s. 0.042 Veg (B =+1.7, p =0.001) 88.2% 72.7% 82.1% 0.002

Distrail (B =�2.00, p =0.016)Bradycellus harpalinus 10 n.s n.s n.s n.s n.s n.s n.s n.s n.s No significant variables – – – –

Platyderus ruficollis 10 n.s n.s n.s n.s n.s n.s n.s n.s n.s Impene (B =+2.5, p =0.006) 88.9% 80.0% 85.7% b0.001‘

Logage (B =+5.8, p =0.006)

Bembidion obtusum 8 n.s n.s n.s 0.039 0.010 0.017 n.s n.s n.s Distedge (B =�0.3, p =0.000) 95.0% 62.5% 85.7% 0.001

Der100 (B =+1.4, p =0.009)

Amara tibialis 7 n.s n.s n.s n.s n.s n.s n.s n.s n.s No significant variables – – – –

Amara praetermissa 5 n.s n.s n.s 0.046 n.s n.s n.s n.s n.s Logder100 (B =+8.2, p =0.004) 91.3% 40.0% 82.1% 0.009

Logbare (B =+25.6, p =0.031)

E.Smallet

al./Scien

ceoftheTotalEnviro

nment360(2006)205–222

218

Page 15: Do landscape factors affect brownfield carabid assemblages?

Notiophilus aquaticus 5 n.s n.s n.s 0.001 n.s n.s 0.005 n.s n.s Der1000 (B =0.2, p =0.007) 100.0% 80.0% 96.4% b0.001

Veg (B =2.2, p =0.014)

(b) Non-specialist species

Trechus quadristriatus 21 n.s. 0.035 n.s. n.s. n.s. n.s. n.s. 0.044 0.023 Agedist (B =�0.9, p b0.001) 85.7% 100.0% 96.4% b0.001

Distrail (B =�3.7, p b0.001)Veg (+1.2, p =0.003)

Notiophilus biguttatus 20 n.s. n.s. 0.003 n.s. n.s. n.s. n.s. n.s. n.s. Moist (B =�98.4, p b0.001) 87.5% 95.0% 92.9% b0.001

Logimpenet (B =+28.1, p =0.001)

Calathus fuscipes 19 n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. No significant variables – – – –

Trechus obtusus 17 n.s. n.s. n.s. n.s. n.s. 0.033 n.s. 0.024 n.s. Age (B =�0.3, p =0.001) 72.7% 88.2% 82.1% 0.001

Der100 (B =+1.7, p =0.011)

Litter (B =+0.9, p =0.048)

Loricera pilicornis 16 n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. 0.001 Logagedist (B =�7.5, p b0.001) 83.3% 81.3% 82.1% b0.001

Amara ovata 15 n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. Loggrass (B =�22.7, p =0.024) 53.8% 73.3% 64.3% 0.024

Carabus violaceus 14 0.049 n.s. n.s. 0.027 n.s. n.s. n.s. n.s. n.s. Der100 (B =2.1, p =0.003) 78.6% 78.6% 78.6% 0.003

Distrail (B =+2.3, p =0.009)

Amara similata 11 n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. 0.014 Agedist (B =�0.4, p =0.003) 76.5% 72.7% 75.0% 0.002

Logbare (B =25.1, p =0.017)

Pterostichus niger 7 0.024 n.s. n.s. 0.012 0.024 0.028 0.050 n.s. n.s. Veg (B =+62.5, p b0.001) 100.0% 100.0% 100.0% b0.001

Logdistedge (B =�327.8, p b0.001)Logsize (B =+243.3, p b0.001)

Harpalus rufibarbis 6 n.s. n.s. n.s. n.s. n.s. n.s. n.s. 0.043 n.s. Age (B =+0.35, p =0.002) 95.5% 50.0% 87.7% 0.008

Logder1000 (B =+2.6, p =0.014)

Pterostichus melanarius 6 n.s. n.s. n.s. n.s. 0.009 0.012 n.s. n.s. 0.020 Logdistedge (B =�6.8, p =0.028) 95.5% 83.3% 92.9% b0.001

Agedist (B =�33.8, p b0.001)Logloi (B =�10187.7, p b0.001)

Pterostichus strenuus 6 n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. No significant variables – – – –

Significance of landscape variables in the first step of regression reported.

E.Smallet

al./Scien

ceoftheTotalEnviro

nment360(2006)205–222

219

Page 16: Do landscape factors affect brownfield carabid assemblages?

E. Small et al. / Science of the Total Environment 360 (2006) 205–222220

sure of isolation, because the West Midlands conurba-

tion has grown from not one, but four original centres

(Birmingham, Wolverhampton, Dudley, the Black

Country), with areas of relatively open, almost druralTspace between (e.g. Sutton Park, Sandwell Valley).

4.6. Interpretation 2 — derelict assemblages are largely

unaffected by isolation at this city scale

Despite the limitations of the study in terms of the

difficulty in taking meaningful measurements of isola-

tion (Bastin and Thomas, 1999), there are other rea-

sons why the lack of positive results is perhaps not

surprising. Twenty-eight of the thirty-one specialist

species found in the survey are known to be capable

of flight (Andersen, 2000; Luff, 1998; Turin, 2000).

Andersen (2000), studying the origin of the carabid

fauna of dry anthropogenic habitats, concluded that the

majority originated from naturally open, dry habitats

similar to steppe. These species invaded the bare land-

scape soon after deglaciation, and may have survived

in the postglacial period by surviving in habitats such

as heaths, dunes, dtalusT (scree), dalvarT (steppe), there-after expanding their range when cultivation of the

landscape began. The ability to capitalise on these

relatively fragmented and/or short-lived habitats

would necessarily have required the species to develop

good dispersal power, thereby rendering them less

responsive to habitat isolation. There is certainly a

growing literature on grassland butterflies which illus-

trates that species presence and persistence in habitat

patches are affected by habitat quality (Collinge et al.,

2003).

Indeed, Wood and Pullin’s (2002) study of genetic

relatedness of four species of grassland butterflies in

Birmingham provides important corroboratory data

from the same conurbation. Their work suggests that

species are more limited by the availability of suitable

habitat than their ability to move through the city. It is

tempting, therefore, to draw the simple conclusion that

the derelict species found in this survey are, on the

whole, simply too efficient at dispersing to be affected

by the levels of habitat isolation present in the urban

environment. However, Denys and Schmidt (1998)

illustrated how habitat isolation along the rural–

urban gradient was important for invertebrates on

Mugwort, but one must be mindful that their experi-

ment was a short-term project, with the colonisation of

the Mugwort pots being studied over the course of a

single season (May to September). By comparison,

this present study investigated sites that were between

2 and 20 years old. The implication therefore, is that

the impacts of isolation on colonisation speed in an

urban environment are short-lived, being apparent

only over the course of a very few seasons. After

more time has elapsed, the majority of derelict species

have already been able to find their way to even the

most isolated sites.

4.7. Flightless or rare specialist species

Another interesting finding Denys and Schmidt’s

(1998) study was that rare species, particularly rare

parasitoids, were the least successful in colonising

urban habitats. This was in line with Pimm’s (1991)

assertion that any species is more likely to fail to colo-

nise an isolated habitat if their populations are small. In

this study, A. praetermissa and P. ruficollis are the only

two nationally scarce species encountered (Hyman,

1992). A. praetermissa was most frequently encoun-

tered on bare sites with a high density of derelict land

within 100 m of the site boundary (Table 6), indicating

some effect of landscape factors on the likelihood of site

colonisation. The findings in relation to P. ruficollis

were even more intriguing. P. ruficollis was significant-

ly associated with older sites, but only at those sites that

had retarded succession (Table 6). This is precisely the

result that one would expect for early-successional spe-

cies that are dispersal limited. While this was the only

species tested that showed this pattern, it is of note that

this is the only species found in the survey that is both

non-flying and uncommon and therefore theoretically

most likely to be sensitive to isolation. Aside from P.

ruficollis only two other flightless specialist derelict

species were recorded (Metabletus foveatus and C. mel-

anocephalus.).M. foveatus occurred at only 4 sites so its

distribution could not be tested against landscape vari-

ables, while C. melanocephalus was 19 of the 28 survey

sites and therefore is not thought to be sensitive to

isolation. However, it was interesting to note that 2 of

the 145 specimens of C. melanocephalus were found to

have long wings (1.6%). Long-winged C. melanocepha-

lus are known but are very rare (Lindroth, 1974).

Further studies would be needed to establish whether

an increased proportion of full-winged individuals is

a consequence of urbanisation in normally flightless

species.

5. Conclusions

The work has highlighted a number of implications

for the conservation of urban invertebrates. First, the

common derelict carabid species recorded during this

survey, do not appear to be affected by the location of

Page 17: Do landscape factors affect brownfield carabid assemblages?

E. Small et al. / Science of the Total Environment 360 (2006) 205–222 221

derelict habitat patches in the West Midlands landscape.

Conservation efforts to maintain populations of these

species should focus principally on habitat quality

issues rather than landscape issues. Secondly, no evi-

dence was found to support the hypothesis that sites

away from railway corridors are more impoverished in

their carabid fauna than sites on corridors. This study

suggests that this is because derelict carabid species are

generally good dispersers rather than that these railways

are poor corridors. Thirdly, there are some suggestions

from this study that rarer and non-flying specialist

species, may be affected by isolation, taking longer to

reach sites. An inference from this is that older sites

with retarded succession, and sites in higher densities of

surrounding derelict land may eventually become more

species rich in rarer and flightless species, and that such

sites may be important for maintaining populations of

these species. Further studies are required to clarify this

issue.

Acknowledgements

We thank the National Environment Research Coun-

cil who funded this work as a PhD studentship (to ECS)

under the NERC Urban Regeneration and Environment

(URGENT) thematic programme (GST/02/1979). We

also thank Kevin Austin for his invaluable work on the

vegetation survey and colleagues in the University of

Birmingham, University of Helsinki, and at CEH

Monks Wood for their constructive comments and

ideas.

References

Abildsnes J, Tommeras BA. Impacts of experimental habitat frag-

mentation on ground beetles (Coleoptera, Carabidae) in a boreal

spruce forest. Ann Zool Fenn 2000;37:201–12.

Alaruikka D, Kotze DJ, Matveinen K, Niemela J. Carabid beetle and

spider assemblages along a forested urban–rural gradient in south-

ern Finland. J Insect Conserv 2002;6:195–206.

Andersen J. A comparison of pitfall trapping and quadrat sampling

of Carabidae (Coleoptera) on river banks. Entomol Fenn 1995;6:

65–77.

Andersen J. What is the origin of the carabid beetle fauna of dry,

anthropogenic habitats in western Europe? J Biogeogr 2000;

27:795–806.

Austin KC. Botanical processes in urban derelict spaces. PhD, The

University of Birmingham, Birmingham, 2002.

Bastin L, Thomas CD. The distribution of plant species in urban

vegetation fragments. Landsc Ecol 1999;14:493–507.

Bauer LJ. Moorland beetle communities on limestone dhabitat islandsT1 Isolation, invasion and local species diversity in carabids and

staphylinids. J Anim Ecol 1989;58:1077–98.

Borcard D, Legendre P, Drapaeu P. Partialling out the spatial compo-

nent of ecological variation. Ecology 1992;73:1045–55.

Breuste J, Feldmann H, Uhlmann O, editors. Urban ecology. Berlin7

Springer-Verlag; 1998.

CityView. Cities revealed: aerial photographs of Birmingham. City

View, 1995.

Collinge SK, Prudic KL, Oliver JC. Effects of local habitat character-

istics and landscape context on grassland butterfly diversity. Con-

serv Biol 2003;17:178–87.

Davies KF, Margules CR. Effects of habitat fragmentation on

carabid beetles: experimental evidence. J Anim Ecol 1998;67:

460–71.

Den Boer PJ. On the significance of dispersal power for popula-

tions of Carabid-beetles Coleoptera, Carabidae. Oecologia 1970;

4:1–28.

Den Boer PJ. Dispersal power and survival. Carabids in a cultivated

countryside (with a mathematical appendix by J. Reddingius).

Misc. Paper., LH. Wageningen, 14, Veenman, 1977.

Den Boer PJ, Den Boer-Daanje W. On life history tactics in carabid

beetles: are there only spring and autumn breeders? In: Stork NE,

editor. The role of ground beetles in ecological and environmental

studies. Andover7 Intercept; 1990. p. 247–58.

Den Boer PJ, van Huizen THP, Den Boer-Daanje W, Aukema B, Den

Biemen CFM. Wing polymorphism and dimorphism as stages in

an evolutionary process (Coleoptera, Carabidae). Entomol Gen

1980;6:107–34.

Denys C, Schmidt H. Insect communities on experimental mugwort

(Artemisia vulgaris L) plots along an urban gradient. Oecologia

1998;113:269–77.

De Vries HH, Den Boer PJ, Van Dijk TS. Ground beetle species in

heathland fragments in relation to survival, dispersal, and habitat

preference. Oecologia 1996;107:332–42.

Douglas I. The case for urban ecology. Urban Nat Mag 1992;1:

15–7.

Eversham BC, Roy DB, Telfer MG. Urban, industrial and other

manmade sites as analogs of natural habitats for carabidae. Ann

Zool Fenn 1996;33:149–56.

Forman RTT, Godron M. Patches and structural components for a

Landsc Ecol. Bioscience 1981;31:733–40.

Gibson CWD. Brownfield: red data The values artificial habitats have

for uncommon invertebrates. Peterborough7 English Nature; 1998.

Gruttke H, Weigmann G. Ecological studies on the carabid fauna

(Coleoptera) of a ruderal ecosystem in Berlin. In: Stork NE, editor.

The role of ground beetles in ecological and environmental stud-

ies. Andover7 Intercept; 1990. p. 181–9.

Halme E, Niemela J. Carabid beetles in fragments of coniferous

forest. Ann Zool Fenn 1993;30:17–30.

Hanski I. Metapopulation dynamics. Nature 1998;396:41–9.

Hurlbert SH. The non-concept of species diversity: a critique and

alternative parameters. Ecology 1971;52:577–86.

Hyman PS. (Updated by MS Parsons) A Review of the Scarce and

Threatened Coleoptera of Great Britain. Part 1 JNCC, Peterbor-

ough, 1992.

Ishitani M, Kotze DJ, Niemela J. Changes in carabid beetle assem-

blages across an urban–rural gradient in Japan. Ecography

2003;26:481–9.

Kegel B. The distribution of carabid beetles in the urban area of west

Berlin. In: Stork NE, editor. The role of ground beetles in

ecological and environmental studies. Andover7 Intercept; 1990.

p. 325–9.

Lazenby A. Ground beetles (Carabidae) and other Coleoptera on

demolition sites in Sheffield. Sorby Rec 1983;21:39–51.

Legendre L, Legendre P. Numerical ecology. Amsterdam7 Elsevier;

1998.

Page 18: Do landscape factors affect brownfield carabid assemblages?

E. Small et al. / Science of the Total Environment 360 (2006) 205–222222

Lindroth CH. Carabidae. London7 Royal Entomological Society;

1974.

Luff ML. Provisional atlas of the ground beetles (Coleoptera, Car-

abidae) of Britain Biological Records Centre, Institute of Terres-

trial Ecology, Monks Wood, Abbots Ripton, 1998.

MacArthur RH, Wilson EO. The theory of island biogeography. New

Jersey7 Princeton; 1967.

Magura T, Kodobocz V, Tothmeresz B. Effects of habitat fragmenta-

tion on carabids in forest patches. J Biogeogr 2001;28:129–38.

Manly BFJ. Multivariate statistical methods: a primer. 2nd edition.

London7 Chapman & Hall; 1994.

McDonnell MJ, Pickett STA. Ecosystem structure and function along

urban–rural gradients: an unexploited opportunity for ecology.

Ecology 1990;71:1232–7.

McIntyre NE. Ecology of urban arthropods: a review and a call to

action. Ann Entomol Soc Am 2000;93:825–35.

Niemela J. Carabid beetles (Coleoptera: Carabidae) and habitat frag-

mentation: a review. Eur J Entomol 2001;98:127–32.

Niemela J, Kotze DJ, Venn S, Penev L, Stoyanov I, Spence JR, et al.

Carabid beetles assemblages (Coleoptera, Carabidae) across

urban–rural gradients: an international comparison. Landsc Ecol

2002;17:387–401.

Pickett STA, Cadenasso ML, Grove JM, Nilon CH, Pouyat RV,

Zipperer WC, et al. Urban ecological systems: linking terrestrial

ecological, physical, and socioeconomic components of metropol-

itan areas. Ann Rev Ecolog Syst 2001;32:127–57.

Pimm SL. The balance of nature? Ecological issues in the conservation

of species and communities. Chicago7University of Chicago; 1991.

Sanders HL. Marine benthic diversity: a comparative study. Am Nat

1968;102:243–82.

Schwerk A. Ecological aspects of carabid beetle coenoses (Coleop-

tera: Carabidae) on industrial fallow grounds in the Ruhr Valley

Area. In: Brandmayr P, Lovei G, Zetto T, Brandmayr A, Casale A,

Vigna-Taglionti A, editors. Natural history and applied ecology of

carabid beetles. Sophia7 Pensoft; 2000. p. 277–87.

Small EC, Sadler JP, Telfer MG. Carabid beetle assemblages on

urban derelict sites in Birmingham, UK. J Insect Conserv 2003;

6:233–46.

SPSS MG. SPSS. Chicago7 SPSS; 2000.

ter Braak CFJ, Smilauer P. CANOCO reference manual and user’s

guide to canoco for windows: software for canonical community

ordination (Version 4). Ithaca7 Microcomputer Power; 1998.

Turin H. De Nederlandse loopevers, verspreiding en oecologie (Co-

leoptera: Carabidae). Leiden7 Nationaal Natuurhistorisch Museum

Naturalis; 2000.

Usher MB, Field JP, Bedford SE. Biogeogaphy and diversity of

ground-dwelling arthropods in farm woodlands. Biodivers Lett

1993;1:54–62.

Webb NR. Studies on the invertebrate fauna of heathland in Dorset,

UK, and implications for conservation. Biol Conserv 1989;47:

153–65.

Wood BC, Pullin AS. Persistence of species in a fragmented urban

landscape: the importance of dispersal ability and habitat avail-

ability for grassland butterflies. Biodivers Conserv 2002;11:

1451–68.