Mapping to inform conservation: A case study of changes in semi-natural habitats and their connectivity over 70 years

Biological Conservation 145 (2012) 30–38

Contents lists available at SciVerse ScienceDirect

Biological Conservation journal homepage: www.elsevier.com/locate/biocon

Mapping to inform conservation: A case study of changes in semi-natural habitats and their connectivity over 70 years D.A.P. Hooftman ⇑, J.M. Bullock Centre for Ecology and Hydrology, Wallingford, United Kingdom

a r t i c l e

i n f o

Article history: Received 2 March 2011 Received in revised form 23 September 2011 Accepted 29 September 2011 Available online 6 December 2011 Keywords: Grassland Habitat loss Habitat mapping Land-use change Restoration Woodland

a b s t r a c t Intensification of human activities has caused drastic losses in semi-natural habitats, resulting as well in declining connectivity between remaining fragments. Successful future restoration should therefore increase both habitat area and connectivity. The first steps in a framework for doing so are addressed here, which involve the mapping of past habitat change. We present a method which is unique in: the large area covered (2500 km2), the high resolution of the data (25  25 m), the long period assessed (70 years), and a system for translation of land use maps into Broad Habitat Types using soil surveys. We digitised land use maps from the 1930s for the county of Dorset in southern England. The resulting map was compared to the UK Land Cover Map of 2000. For our example area, land use shifted dramatically to more intensive agriculture: 97% of all semi-natural grasslands were converted into agriculturally-improved grassland or arable land as were large proportions of the heathlands and rough grasslands (57%). The other important driver of change was afforestation (+25%). The larger habitat areas became fragmented, with average fragment size of different habitats falling by 31–94%. Furthermore, the connectivity between fragments dropped drastically, by up to 98%. Analyses such as those presented here not only quantify the scale and pattern of habitat loss, but are important to inform land-use planning to restore biodiversity by both increasing the available habitat and facilitating dispersal among habitat fragments. We discuss the possible steps for such a framework. Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction During the second half of the 20th century, the Western European landscape became more intensively used, affecting especially semi-natural habitats which are the principal hotspots for biodiversity (Benton et al., 2003; Fuller, 1987; Henle et al., 2008). In particular, rural areas were altered dramatically in the pursuit of increased production, reflecting a higher demand for food and feed by a growing human population (‘‘intensification’’). Many previously low productivity areas were reclaimed, drained and/or fertilised for grass production or converted to arable land (Fuller, 1987; Hodgson et al., 2005; Swetnam, 2007). Further severe losses of habitat were caused by urbanisation and building (Feranec et al., 2010; Gerard et al., 2010; Thomson et al., 2007), as well as afforestation for timber production (Mason, 2007). In addition to severe declines in the area of such habitats, these processes decreased the connectivity of remaining fragments, which became more isolated (Fahrig, 2003; Stoate et al., 2001). Decreased connectivity reduces dispersal among plant and animal populations (Soons et al., 2005), leading to an elevated extinction risk at ⇑ Corresponding author. Address: Centre for Ecology and Hydrology, Benson Lane, Wallingford OX10 8BB, United Kingdom. Tel.: +44 1491 692700. E-mail address: [email protected] (D.A.P. Hooftman). 0006-3207/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.biocon.2011.09.015

population and regional scales (Fahrig, 2003; Keller and Waller, 2002). While loss of semi-natural habitats continues (Feranec et al., 2010; Howard et al., 2003; UK National Ecosystem Assessment, 2011), many restoration efforts are taking place (Lawton et al., 2010; Wade et al., 2008). Successful future restoration of semi-natural habitat not only requires selection of sites which are amenable for restoration to the target habitat (Walker et al., 2004), but should ideally also aim to enhance habitat connectivity (Brudvig, 2011; Bullock et al., 2002). Despite increasing interest in the restoration of such ‘‘ecological networks’’ (Lawton et al., 2010), the process is not straightforward. The necessary first step is to map the long-term changes in habitat cover in a region, on an as fine a scale as possible. Such maps will give baseline information on earlier habitat configurations and will allow planning of ecological networks based on ease of restoration (Jackson and Hobbs, 2009; Wade et al., 2008), and increasing connectivity of a variety of habitat types (Guisan and Zimmerman, 2000). We report here on a habitat mapping project covering an area of ca. 2500 km2 at the detailed resolution of 25  25 m, and comprising over 4-million data-points. We compared land use in the 1930s with that in 2000. The 1930s maps represent the period before the massive intensification of land use in the UK after World War II, and can be taken as an ‘‘ideal’’ situation for semi-natural habitats,

D.A.P. Hooftman, J.M. Bullock / Biological Conservation 145 (2012) 30–38

such as heathlands and low productivity grasslands, and the species inhabiting them (UK National Ecosystem Assessment, 2011). Previous studies using map and vegetation survey data indicate a very large loss of semi-natural habitat over the last 75 years (Fuller, 1987; Hodgson et al., 2005). However, while researchers have produced maps of loss of particular habitat types over long timescales (e.g. heathlands; Webb and Haskins, 1980; Webb, 1990) or whole landscape changes over shorter, more recent, periods (e.g. Burnside et al., 2003), little has been done to determine overall patterns of habitat change for whole socio-political regions over long periods. One problem in mapping land cover change, especially over long intervals, is the difficulty of matching land cover types between different surveys (Dallimer et al., 2009). Here, we overcome this difficulty by matching maps according to habitat types. We utilise the first Land Utilisation Survey done in the UK in the 1930s (Stamp, 1931) and after digitising these maps, combine them with soil maps to derive a habitat classification. The Dudley Stamp maps are unique in their level of detail, being an early example of volunteer recording. Only 10% is available of the second Land Utilization Survey in the 1960s and so subsequent detailed maps are not available until the UK Land Cover Map in 1990 (Fuller et al., 1994). In this study we use the UK Land Cover Map of 2000 (LCM2000), and develop a method to make the habitat types of the Dudley Stamp maps comparable to the Broad Habitat Types of the LCM2000 as described in Fuller et al. (2002a, 2002b). The model study area of Dorset, a county bordering the south coast of England, is a prime example of a pre-dominantly rural region which underwent rapid, but with gradual speed, intensification during the mid to late 20th century (Keymer and Leach, 1990; Webb and Haskins, 1980). Dorset, however, was and remains a major UK biodiversity hotspot (Prendergast and Eversham, 1995; Preston et al., 2002; Williams et al., 1996), which indicates the potential for positive responses to the restoration of ecological networks. In this paper we report on a habitat change mapping project using Dorset as a case study and present habitat maps for the 1930s and 2000. The objectives include (i) a test of the accuracy of the constructed 1930s habitat map using contemporaneous, but independent vegetation surveys. We then investigate the changes in semi-natural habitats that occurred between the 1930s and 2000 and their apparent drivers in terms of: (ii) the total available habitat; (iii) the size of fragments of semi-natural habitat types; and (iv) the connectivity among fragments. We discuss the use such mapping efforts as a first step in developing plans for ecological network restoration.

2. Material and methods 2.1. Example area and maps The county of Dorset on the south coast of England is currently 2653 km2 in area, but, because of boundary changes, was ca. 2500 km2 in the 1930s. Dorset’s population roughly doubled between 1931 and 2001 from ca. 198,000 to ca. 391,000 inhabitants, excluding the urban centres of Poole and Bournemouth (Southall et al., 2010). More importantly, the county is a diversity hotspot for species and habitats. The ‘‘New Atlas of the British and Irish Flora’’ shows that the most species-rich 100 km2 square lies in Dorset (1107 vascular plant species: Preston et al., 2002). Dorset also contains some of the 5% most species-rich 100 km2 squares in Britain for butterflies (Prendergast and Eversham, 1995), birds (Williams et al., 1996) and freshwater plants and invertebrates (Palmer, 1999). About 9% (223 km2) of the terrestrial area comprises EU Natura 2000 sites (Natura 2000, 2011).

31

2.1.1. 1930s habitat map The First Land Utilisation Survey of Great Britain was accomplished during the 1930s (Stamp, 1931). Volunteers recorded land uses onto the 1904 UK Ordnance Survey maps (1:10,560) between 1931 and 1934 (Southall et al., 2007). The merging of those maps into one-inch-to-the-mile sheets (1:63,360) created the ‘‘Dudley Stamp maps’’ (DSM). The main land-use categories were: (i) Forest and woodland; (ii) Arable land; (iii) Meadowland and permanent grassland – which we will call ‘‘managed grasslands’’; (iv) Heaths, moorlands, commons and rough hill pasture – using soil data we will split these into ‘‘heaths’’ and the ‘‘rough grasslands’’ (see next paragraph); (v) Gardens, including allotments, orchards and nurseries; (vi) Agriculturally unproductive land, which included buildings, mines, railways and suchlike; (vii) Inland water and sea; and (viii) Littoral features, shingles and saltmarshes. The maps have been scanned and can be viewed on-line (Southall et al., 2010). We manually digitised six of the 300 dpi scanned DSMs (sheets 121, 129, 130, 131, 140 and 141), which encompassed Dorset. These map sheets had been previously georeferenced to the Ordnance Survey of Great Britain 1936 national grid (OSGB-36) (Fuller et al., 2002a). Digitisation was done in Adobe Photoshop, identifying features as small as 2  2 pixels (10.8  10.8 m). The six resulting maps were imported into ArcGIS 9.3 and the eight land-use categories were vectorised into a GISlayer. Subsequently, these maps were combined with the 1:250,000 vector-based National Soil Map of England and Wales (Thompson, 2007), using the 27 soil map units of the Soilscape format. The resulting Soilscape  DSM land use matrix was then translated into 15 Broad Habitat Types (BHT; Supplementary Materials Table S1; Fig. 1), which were compatible with the BHTs of the Land Cover Map described in Section 2.1.2. In this way, we separated neutral, calcareous and acid grasslands into ‘‘managed’’ (DSM category iii) and ‘‘rough’’ (DSM category iv, excluding heaths). ‘‘Rough grassland’’ is used in the Land Cover Map to describe the tall, tussocky and scrubby grassland resulting from minimal management, and Stamp (1931) used a similar criterion to distinguish ‘‘moorlands, commons and rough hill pasture’’. For simplicity, we refer to the resulting habitat map (Fig. 1a) as the Dudley Stamp map (DSM). The DSM does not include an agriculturally-improved grassland category. However, there is evidence that almost all English grasslands at that time were unimproved, in the sense that they were not receiving inorganic fertilisers (Fuller, 1987). We confirmed the low cover of improved grassland in 1930s Dorset using the Good vegetation survey data, which we describe in Section 2.3, in the context of validating the DSM. A best fit National Vegetation Classification (NVC) category (Rodwell, 1992) was assigned to each of Good’s grassland survey sites using Tablefit (Hill, 1996). Of the 2163 grassland survey sites, all corresponded best to NVC unimproved grassland categories, except nine which were classified as MG7 (Lolium perenne leys) and 42 as MG6 (Lolium perenne – Cynosurus cristatus grassland), indicating only 2.3% of grasslands were possibly improved. 2.1.2. Land Cover Map 2000 To indicate habitat coverage in 2000 we used the Land Cover Map 2000 (LCM2000), which is the definitive land cover map for the UK, and has been widely used in conservation research (e.g., Oliver et al., 2009). The LCM2000 is derived from satellite spectral data (Fuller et al., 2002b) in which assignments of 25  25 m pixels to BHTs used a maximum likelihood algorithm (Fuller et al., 2002a). The minimum parcel size is 8 pixels (5000 m2). These pixels were converted into a vectorised GIS layer. Semi-natural grasslands were divided into acidity categories using an acidsensitivity projection, with a pH threshold of 4.5 for acid grasslands. We refer to Fuller et al. (2002a, 2002b) for technical

32

D.A.P. Hooftman, J.M. Bullock / Biological Conservation 145 (2012) 30–38

Fig. 1. Habitat maps of Dorset indicating 15 Broad Habitats (BHT) for: (a) 1930s, based on Dudley Stamp maps combined with the UK National Soil map; and (b) in 2000 from the Land Cover Map (LCM2000). The total area is 251,422 ha. Colours refer to the BHT classification manuals with adaptations. High resolution maps are available in the Supplementary Materials.

details of mapping and for cover type descriptions. To create a habitat map matching the DSM, we derived BHTs from different levels of land cover classification within the LCM2000, starting from the level-3 ‘‘cover-components’’ or ‘‘variants’’ (Fuller et al., 2002a). Aggregation levels used were: the third, component, level for unimproved calcareous grasslands; ‘‘subclass’’ level 2 for acid grasslands and mesotrophic grasslands at; and ‘‘target class’’ level 1 for heathlands, rough mesotrophic grasslands, improved grassland, build-up, arable and woodlands. Broadleaved and coniferous woodlands were combined for some of the analyses as these were not distinguished in the DSM. The resulting LCM2000 map for Dorset is depicted in Fig. 1b. 2.1.3. Broad habitat types excluded from analyses On the LCM2000 map, the built-up areas – houses, gardens, roads and commercial estates etc. are not well represented because of the 0.5 ha detection threshold. To address this issue we merged the layer of buildings and roads of the DSM with the LCM2000. As a result changes from built-up into other land cover types could not be detected; we therefore make the reasonable assumption that the conversion from built-up areas to other land covers was minimal over this period. In contrast changes into built-up were identified, although the area converted could have been underestimated. Similarly, we excluded inland water features from our analyses. As with built-up areas the DSM layer of water features was merged into the LCM2000 map because the former described these with better resolution. Furthermore, littoral features (BHTs 19.1, 21.1 and 21.2) were incorporated in the ‘‘Other BHTs’’ class (Supplementary Materials Table S1). They occur on very small scales; hence their assignments could be more inaccurate. The broad habitats Inland Bare Ground (16.1) and Fen, Marsh and Swamp (BHT 11.1) were represented on the LCM2000, but could not be identified on the DSM. For the latter we incorporated these into the Other BHTs category. The total size of this Other BHT class is small; 1118 ha in the 1930s and 1204 ha in 2000 (0.05% of total the area), with a turn-over of 612 ha (Table 2).

calculated, with a minimum fragment size of 625 m2, which is the size of a gridcell in the LCM2000. Subsequently, both maps were converted into 25  25 m (625 m2) point entries, characterised by BHT, with each map containing 4022,750 entries. Additional datasets were created for each BHT. All subsequent data processing was done in Matlab v2009a; supporting codes are available from the authors upon request. Habitat change was calculated in ArcGIS, using the tabulate function in the spatial analyst toolbox, and summing the changes across all cells. The distance among all the fragments was calculated in two steps. First, we reduced the dataset to fragment outlines only. Second, the minimum distance between each outline was calculated per BHT. We performed the following calculations per BHT for both maps. (i) Distance to 5 ha of habitat: the mean across all fragments of the distance from the fragment to the closest fragments with a summed area of at least 5 ha; (ii) Area within 1 km: the mean across all fragments of the summed total area of all fragments, which fall fully or partly within 1 km of a given focal fragment; (iii) Area within 5 km: as for ‘Area within 1 km’, but with a 5 km radius. Following Soons et al. (2005) we calculated the probability Oi(r) that an organism which disperses distance r from a habitat type i lands in the same habitat type, which we subsequently refer to as connectivity.

Oi ðrÞ ¼

h¼n 1 X N hi ðrÞ  n h¼1 Nh ðrÞ

where h is the focal 25  25 m grid cell of BHT i, n is the number of focal grid cells of BHT i sampled, and r is the radius of a circle centred on the focal grid cell. Nhi(r) is the number of grid cells of BHT i intersected by a circle with radius r, and Nh(r) is the total number of grid cells intersected by circle r. We included values of r from 25 to 750 m at intervals of 25 m. These connectivity calculations were made using the full dataset, employing a random sub-set of 100,000 focal grid cells for each value of r.

2.2. Analytical tools

2.3. Map validation

For each semi-natural BHT, habitat fragments less than 25 m apart in the DSM were aggregated. The size of each fragment was

We carried out a validation of the DSM using an independent dataset of vegetation surveys carried out at 7575 sites in Dorset

33

D.A.P. Hooftman, J.M. Bullock / Biological Conservation 145 (2012) 30–38

heathlands on less productive soils (29,000 ha). Descriptive statistics for all Broad Habitat Types (BHT) are given in Table 1. Land use reflected the distribution of soil types. Arable fields were situated adjacent to calcareous grasslands running in bands from south-west to north-east of the county. The lowland heaths and associated acid grasslands dominated in the less fertile south-east of the county (Isle of Purbeck) and on its eastern border; these are among the most important lowland heath areas in the UK. In contrast, the north of the county, traditionally a dairy farming region, was dominated by mesotrophic grasslands on the more fertile soils. By 2000 (Fig. 1b), large areas had been converted to intensive agriculture (Tables 1 and 2) and the dominant land covers were fertilised, agriculturally-improved grasslands (98,000 ha) and arable (76,000 ha).

in the 1930s (Good, 1948). An electronic version of this database, with over 285,000 entries, is kept by the Dorset Environmental Records Centre (DERC). The location of each survey site has been digitised, rectified and vectorised by DERC to an Ordnance Survey Great Britain 1936 projection (see also Keith et al., 2009), including an estimation of the BHT based on its plant species composition. From this dataset, we selected the 2670 sites that were grassland (aggregating all types), woodland, or heathland. The dominant BHT according to the DSM for each site was then compared with the one assigned from the Good surveys, using the tabulate function in ArcGIS. 3. Results 3.1. Habitat map of the 1930s

3.2. Validation of the Dudley Stamp BHT classification The habitat map of Dorset in the 1930s (DSM) is depicted in Fig. 1a. High resolution maps are in the Supplementary Materials. Dorset in the 1930s was dominated by semi-natural (unimproved) grassland pastures (136,000 ha) with rough grasslands and

The BHTs assigned to Good’s survey sites showed a 90.4% concordance with the DSM assignments over all habitat types. For individual BHTs: 94.7% of the Good survey sites labelled as

Table 1 General statistics of habitat loss, fragment size and connectivity of seven semi-natural Broad Habitat Types in Dorset between the 1930s and 2000. Broad habitat type

Period Habitat statistics Total area (ha)a Area Loss (%)a Mean fragment size (ha)b Mean fragment size reduction (%)b Coefficient of Variation of fragment sizeb Connectivity – mean values Distance to 5 ha habitat (m)b Area within 1 km (ha)b Area within 5 km (ha)b 50% connectivity-distance (m)a,c a b c d e

Woodlands

Mesotrophic grasslands

Calcareous grasslands

Rough

Rough

Managed

Acid grasslands

Heathlands

Managed

1930s

2000

1930s

2000

1930s

2000

1930s

2000

1930s

2000

1930se

2000

1930s

2000

20,872 +25% 7.5 31% 4.4

25,358d

3012

0

2.6

1.2

1.6

1.7

13,722 56% 52.6 88% 3.1

6004

2.5

4334 61% 9.1 72% 2.7

1705

7.1

41,738 89% 40.4 94% 3.4

4552

1.1

6329 43% 14.0 82% 2.6

3630

2.4

90,182 100% 64.5

4.1

9036 67% 10.1 76% 2.7

5.2

216 98 811 157

200 110 1019 121

587 41 246 159

459 10 114 66

83 2302 6865 >750

372 59 456 177

477 11 122 72.1

82 813 3175 375

465 13 160 66

153 78 479 157

329 23 208 71

128 500 2402 708

520 90 475 187

5.1

2.6

6.3

Calculated based on 625-m2 gridcells. Calculated using factual area of all fragments with a minimum size of 625 m2 per fragment; The distance at which Oi(r) = 50%. Broadleaved/mixed: 17,468; coniferous: 7891. Distinction not possible in the 1930s. Rough: 244; managed: 4190. Distinction not possible in 2000.

Table 2 Changes to semi-natural Broad Habitat Types in Dorset from the 1930s to 2000, in terms of conversion into the major semi-natural habitats (with the leading diagonal giving the figure for unconverted habitat) and to other habitats. Broad habitat type in 1930s

a b

Woodlands

Mesotrophic grasslands

Calcareous grasslands

Acid grasslands

Rough

Rough

Managed

Rough

Managed

Managed

Heathlands

Total area in 1930s (ha) 20,872 9036 90,182 Converted into semi-natural habitats in 2000 (including the area remaining as original habitat) Woodlands (broad-leaved/mixed) 6620 1221 4239 Woodlands (coniferous) 3341 408 511 Mesotrophic grasslands Rough 335 96 1216 Managed 0 0 0 Calcareous grasslands Managed 327 210 1329 Rough 176 159 1491 Acid grasslands 202 170 195 Heathlands 601 856 531

6329

41,738

244

4190

13,722

515 198 182 0 551 183 4 14

1620 488 512 0 1157 867 17 251

69 12 2 0 5 0.2 2 4

599 95 60 0 32 20 166 62

1398 2636 112 0 42 25 733 3445

Converted into highly anthropogenic, aquatic or other habitats in 2000 Open ground (–)a 160 159 Arable 4055 2316 Built-up 646 411 a Improved grass (–) 4340 2884 b Other BHT 18 138 Inland water and sea 53 9

229 2075 129 2147 85 19

1054 20,638 1024 14,047 31 32

0.3 65 13 57 15 0

26 991 629 1486 23 3

248 1361 2109 1478 114 23

621 39,353 2543 37,937 161 57

Category not present/identifiable on the 1930s Dudley Stamp map, but note our calculation in the text that virtually none of the grassland was improved in the 1930s. Includes: littoral sediments (BHT 19.1 & 21.1); saltmarshes (21.2); fens, marsh and swamps (11.1).

34

D.A.P. Hooftman, J.M. Bullock / Biological Conservation 145 (2012) 30–38

grassland (97%; Table 2) was converted to intensive agriculture, becoming improved grassland (53,470 ha) or arable (60,980 ha). In contrast, the rough grassland and heathland areas were affected to a much lesser extent (a combined 57% loss). Consequently, the proportion of such rough areas among all non-woodland semi-natural habitats increased from 10% to 65%. The area of woodland increased by 25%, mostly at the expense of heathland and rough grasslands on the less fertile soils. Considering woodland type (conifer vs. broad-leaved), conversion into conifer plantations accounted for 64% of the afforestation of the 1930s heathland, but only 32% of the afforestation of other seminatural habitats. A similar bias between heathlands and other semi-natural habitat types characterised the conversion to builtup, such as new housing estates (15% vs. 3% respectively). The data suggest various other transitions from semi-natural habitats (Table 2), but these all comprise small areas. We do not consider these transitions further as the small areas suggest that they are minor drivers of change and that their influence might be confounded with mapping inaccuracies; the latter issue is illustrated by unlikely transitions, such as from calcareous to acid grassland.

woodlands were identified as predominantly woodland on the DSM; for grasslands this figure was 97.2%; and for heathlands 72.5%. For the last, a further 14% of Good survey sites labelled as heathland was identified as rough grassland on the DSM. Providing further validation, we found that transitions from the 1930s to 2000 between pH categories were low (410 ha: 0.3% of converted area of grasslands and heathlands); such transitions might indicate erroneous assignments. In detail, 1.3% by area of the acid grasslands was calculated as converting to calcareous, and 0.04% vice versa. Similarly, only 0.55% of the calcareous grasslands were assessed as converting to heathlands, with 0.49% vice versa. Furthermore, conversions into and among the ‘‘Other BHTs’’ involved a very small area (612 ha), compared to the large changes caused by conversion into highly anthropogenic land uses. 3.3. Area loss of semi-natural habitats The total area occupied by semi-natural vegetation in Dorset decreased by 74% between the 1930s and 2000 to 48,300 ha (139,450 ha). More than half of this remaining semi-natural habitat was woodland (Table 1). Most semi-natural managed

% of all connected areas

A

50%

Heathlands 1930s (N=242) Heathlands 2000 (N=952)

B

Woodlands 1930s (N=2783) Woodlands 2000 (N=4875)

40%

30%

20%

10%

0%

% of all connected areas

C

50%

Managed calcareous grasslands1930s (N=1027) Managed calcareous grasslands 2000 (N=1827)

D

Rough calcareous grasslands 1930s (N=455) Rough calcareous grasslands 2000 (N=1411)

40%

30%

20%

10%

0% 50%

Rough mesotrophic grasslands 1930s (N=896) Managed mesotrophic grasslands 1930s (N=1347)

Acid grasslands 1930s (N=460)

F

Acid grasslands 2000 (N=644)

Rough mesotrophic grasslands 2000 (N=1656)

40%

30%

20%

Field site size classes (hectares)

> 100

75-100

50-75

25-50

10-25

7.5-10

5-7.5

2.5-5

1-2.5

0.75-1

0.5-0.75

0.25-0.75

0.06-0.25

> 100

75-100

50-75

25-50

10-25

7.5-10

5-7.5

2.5-5

1-2.5

0.75-1

0.5-0.75

0%

0.25-0.75

10%

0.06-0.25

% of all connected areas

E

Field site size classes (hectares)

Fig. 2. Size distribution of habitat fragments of semi-natural habitat types (A–F) in the 1930s and 2000. A minimum fragment size of 625 m2 was employed. N refers to the number of fragments.

35

D.A.P. Hooftman, J.M. Bullock / Biological Conservation 145 (2012) 30–38

3.4. Fragment size and equalisation

3.5. Loss of connectivity

The large changes in land use led to dramatic decreases not only in the area of semi-natural habitat but also in habitat fragment sizes. Except for the woodlands, the average size of individual fragments of semi-natural habitat dropped by 72–94% (Fig. 2; Table 1). This decrease was caused by the fragmentation of almost all large (P10 ha) heath or grassland fragments (Fig. 2). Moreover, fragment size structure of heaths and semi-natural grasslands became much more uniform. Aggregated fragment size in 2000 was mostly 1–2.5 ha, and the variation in fragment size diminished substantially (Table 1). Although woodland increased in area as a whole, fragment size decreased between the 1930s and 2000, although by much less (31%) than the grass and heath habitats (Table 1).

The connectivity between remaining fragments also decreased dramatically (Fig. 3), with the exception of woodlands, which showed little change in any of the measures. For the other seminatural habitats, connectivity expressed as Oi(r) – the proportion of the same habitat type at distance r – decreased more sharply with distance in 2000 (Fig. 3). In Table 1 we report the 50% connectivity distance for these curves, which decreased more than twofold between the 1930s and 2000 for each semi-natural BHT. Combining these effects of decreased fragment size and increased isolation, we found that the area of the same BHT to which a grid cell was connected to within 1 and 5 km radii decreased even more strongly, by up to 98% (Table 1).

Connectivity: Oij(r)

A

100%

Heathlands 1930s

Woodlands 1930s

Heathlands 2000

Woodlands 2000

B

80%

60%

40%

20%

0%

Connectivity: Oij(r)

C

100%

Managed calcareous grasslands 1930s

Rough calcareous grasslands 1930s

Managed calcareous grasslands 2000

Rough calcareous grasslands 2000

D

80%

60%

40%

20%

0%

Connectivity: Oij(r)

E

100%

Rough mesotrophic grasslands 1930s

Acid grasslands (managed and rough) 1930s

Managed mesotrophic grasslands 1930s

F

Acid grasslands rough 2000

Rough mesotrophic grasslands 2000

80%

60%

40%

Distance from source

700

600

500

400

300

200

100

0

700

600

500

400

300

200

100

0%

0

20%

Distance from source

Fig. 3. Connectivity among habitat cells of semi-natural habitat types (A–F) in the 1930s and 2000. Shown is Oi(r), the probability that seeds that disperse over distance r from a habitat patch end up at another patch of the same BHT class.

36

D.A.P. Hooftman, J.M. Bullock / Biological Conservation 145 (2012) 30–38

4. Discussion We have described an approach for constructing detailed and comparable habitat maps, and using these to calculate changes over a long time period in habitat area, fragment size and connectivity for the major semi-natural habitats. For habitat distribution mapping of individual species such high resolution maps will become increasingly important, since the dominant parcel size we found in our study was about 1 ha. This is beyond the resolution of the publicly available CORINE Land Cover maps for Europe (250  250 m; CLC, 2000); obtaining higher CORINE-based resolutions is very demanding, as shown by Gerard et al. (2010) and Thomson et al. (2007). The main strength of the mapping and comparisons presented here is the fine-scale resolution (25  25 m) combined with a large area (2500 km2) and the use of soil data to make comparable maps 70 years apart. To illustrate the mapping method we have analysed the losses of semi-natural habitats between the 1930s and 2000 in a rural landscape which has undergone a large decline of semi-natural habitats over the last century, but which remains a biodiversity hotspot in the UK (Prendergast and Eversham, 1995; Preston et al., 2002; Williams et al., 1996).

4.1. Dramatic changes in land use and connectivity in the study area Over the 70 year period spanned by the two maps we calculated that 74% by area of semi-natural habitat was lost in Dorset. This may be an under-estimate as it excludes several important small land covers that were not specified on the DSM, such as mires, bogs and fens. The majority of these losses impacted on managed grasslands. Losses included calcareous (83% area loss), mesotrophic (97%) and acid grasslands (61%) as well as heathlands (56%). Our calculation of change between two points in time might be misleading if these changes were non-linear over the 70 years. However, other studies in this region using multiple time points have shown the declines in specific habitats to be roughly linear over the last century (Supplementary Materials). In addition, our calculated decreases in grasslands and heathlands are similar to those reported for various parts of the UK (e.g., Fuller, 1987; Hodgson et al., 2005; Swetnam, 2007), Sweden (e.g., Cousins and Eriksson, 2008; Johansson et al., 2008) and The Netherlands (e.g., Soons et al., 2005; Tamis et al., 2005). These comparisons suggest our example region and the temporal changes described are representative of similar regions in North-Western Europe. For comparison, data from these sources are provided in the Supplementary Materials. By far the biggest losses of all grassland types were due to intensification (see Feranec et al., 2010 for terminology); either grassland improvement or arable cultivation in roughly equal measure (39% and 43% respectively). This is not surprising given our starting period of the 1930s. In the years following World War II, agriculture underwent massive changes in Western Europe (Henle et al., 2008; Stoate et al., 2001), including the UK (Thomson et al., 2007, UK National Ecosystem Assessment, 2011). Analyses of the 1990–2000 period illustrate that these losses through intensification continued latterly in the UK (Howard et al., 2003) and across Europe (Feranec et al., 2010). Afforestation was a second major factor in grassland decrease (7% lost to woodland), and was also a major cause of heathland losses (29%). Our data suggest the majority of new woodland in Dorset consisted of broadleaved trees, but with a bias according to initial habitat. Woods appearing on grasslands were mostly broadleaved, but coniferous woods dominated on the heathlands (see also Rose et al., 2000). Our data add important information; underlying this overall increase in area was a large turnover in woodlands. Only 48% of the area covered by woodland in the

1930s remained as woodland in 2000, the majority of losses being caused by conversion to arable and improved grassland (44%). Thus 61% of the 2000 woodland area was new since the 1930s. Some new woodlands may have arisen through natural colonisation of trees into poorly managed grasslands and heaths (Bullock, 2009; Manning et al., 2004), but the major driver is likely the extensive planting of woodlands as a consequence of post-World War II afforestation policies (Mason, 2007). These losses of older woods and planting for productive forestry suggests that, although there is now more woodland in Dorset, its value for biodiversity may be less than in the 1930s (Mason, 2007; Keith et al., 2009). Urbanisation/Development was not a major factor in our example area, with the exception that new building (e.g. housing estates; Webb, 1990) accounted for large losses of heathlands (15%). An analysis of remotely-sensed imagery showed that large areas of Europe were built upon in the period 1950–2000, especially in the north-western countries (Gerard et al., 2010). Our finding of a generally smaller urbanisation effect in Dorset may be because it remains a predominantly rural county, but the poor detection power of the LCM2000 of small buildings may also be a factor. These land use changes led to sharp declines in connectivity among remaining fragments, a process which has been less well studied than area losses in previous analyses of land use change. Decreased connectivity is likely to have reduced movement of species among habitat patches at both medium (area connected within 1 km) and large (within 5 km) scales (e.g., Soons et al., 2005; Soons and Ozinga, 2008). Organisms dispersing from the remaining fragments have now a much higher chance of landing in unsuitable habitat. The drop to almost zero connectivity beyond few hundred metres, as we found for most grassland types, suggests therefore a severe limitation on migration among patches. We determined that the amount of connected area of remaining fragments had reduced by up to 98%. Furthermore, fragments became much smaller, which would likely have lowered population sizes of a variety of species (Fahrig, 2003). These two factors could have led to an increased extinction risk of remaining species’ populations (i.e., Matthies et al., 2004; Walker and Preston, 2006). 4.2. Using mapping to inform conservation and restoration There is increasing interest in restoration at landscape and regional scales (Brudvig, 2011; Forup et al., 2008; Lawton et al., 2010; Lindborg et al., 2008); for example by guiding the placement of new semi-natural habitats to improve the connectivity among fragments. Analysing patterns of habitat changes and connectivity at a large scale is the first step towards quantitative frameworks to support such conservation planning (Guisan and Zimmerman, 2000; Burnside et al., 2003). We suggest a framework with the following steps, for which we have illustrated the first two: (i) translate high resolution maps of a more ideal historical situation and of the current situation into comparable GIS-layers; (ii) calculate the magnitude of loss of area and connectivity between the two time periods; (iii) identify the locations of semi-natural habitat to be preserved, for example based on the Natura 2000 initiative (directive 92/43/EEC, see e.g., Weber and Christophersen, 2002); (iv) decide the target connectivity level to be reached relative to the more ideal situation provided by the historic maps, and potentially based on knowledge of the requirements of key species as suggested by Sutherland et al. (2010); (v) iteratively use probability or land use mapping to assign new areas (e.g., De Wan et al., 2009; Johnson et al., 2004), recalculating connectivity after every iteration. Hence, connectivity can be optimised based on available financial options and land availability. This framework addresses the plans promoted by Lawton et al. (2010) and Sutherland et al. (2010) to enhance the resilience and coherence of ecological networks in the UK. A further reason to

D.A.P. Hooftman, J.M. Bullock / Biological Conservation 145 (2012) 30–38

use historic data is that they provide information on previous land uses and so the potential ease of restoration (Jackson and Hobbs, 2009; Wade et al., 2008; Walker et al., 2004). 4.3. Mapping caveats We consider the DSM, with added soil maps, as an adequate description of large-scale habitat patterns. Discrepancies suggested by comparison between maps and by our validation with the Good data were small. For the latter, we estimated about 90% concurrence of assigned habitat with an independent dataset of vegetation surveys from the same period. This accuracy is similar to the LCM2000 itself, which has a reported 90% agreement with terrestrial surveys (Fuller et al., 2002b), and is higher than reported validations of CORINE maps (Feranec et al., 2007). However, there are caveats to our approach, caused by the characteristics of the maps used. Different techniques were used to construct the two maps; satellite spectral data vs. ground survey by volunteers. Furthermore, each DSM sheet is a merging of ca. 36 different 1904 UK Ordnance Survey maps (Stamp, 1931), but georeferencing was done for whole sheets at once (Fuller et al., 2002a), which could have led to within-sheet variation when aligned with the current maps (N. Brown, pers. comm.). Also, the scanning of old maps from the 1930s could have led to further distortion (Swetnam, 2007). Moreover, the distinction between coniferous and broadleaved woodlands was not possible using the 300 dpi scans, which blurred subcategory labels beyond recognition. We identified several unlikely and probably incorrect habitat transitions, such as changes between calcareous and acid grasslands or heathlands, summing to 410 ha. This error may have been caused both by non-detection of small heterogeneities in acidity because of the 1:250,000 scale of the UK soil map, and the use of different methods for the maps; LCM2000 was based on a soil acid sensitivity projection (Fuller et al., 2002a). Finally, while such mapping can detect substantial habitat change, it may still under-estimate losses. The remaining habitat could be degraded in terms of an altered species composition and diversity, and this has been detected in re-surveys of extant habitat patches (Bennie et al., 2006; Keith et al., 2009; Newton et al., in press). 5. Conclusion We have developed an ecological mapping method in this paper, using Broad Habitat Types to make comparable maps constructed 70-years apart. In our example area of Dorset, UK – a biodiversity hotspot – we have shown large decreases in seminatural habitats. Habitat loss in turn severely diminished the connectivity among the remaining fragments and fragment size. The most important driver for loss of semi-natural habitat was agricultural intensification combined with afforestation of grasslands and heathlands. Looking forward, there is increasing interest in restoration at landscape and regional scales (Lawton et al., 2010; Lindborg et al., 2008), for example by guiding the placement of new seminatural habitats to improve the connectivity among fragments. Therefore, analysing patterns of habitat changes over a large scale, as we do here, is the first step towards a more quantitative framework to support conservation planning decisions (Guisan and Zimmerman, 2000). Acknowledgements We thank France Gerard for taking us through digitisation protocols and Richard Pywell for discussions and maps. Mark Hill and David Roy helped with the Tablefit analyses. Two anonymous

37

reviewers and the subject editor greatly improved the manuscript. The Dorset Environmental Records Centre is acknowledged for providing the Good data and the accompanying GIS layers. Georeferenced scans of the Dudley Stamp maps were provided under licence number 10001880 (DEFRA, 2008), NSRI soilscapes under licence agreement LC0037/041, the LCM2000 is proprietary licensed to CEH. This project was supported by SCALES EU–FP7– 226852. The funding source had no involvement in collection, analyses and interpretation of data, in writing the manuscript and the decision to submit this manuscript for publication.

Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.biocon.2011.09.015.

References Bennie, J., Hill, M.O., Baxter, R., Huntley, B., 2006. Influence of slope and aspect on long-term vegetation change in British chalk grasslands. Journal of Ecology 94, 355–368. Benton, T.G., Vickery, J.A., Wilson, J.D., 2003. Farmland biodiversity: is habitat heterogeneity the key? Trends in Ecology and Evolution 19, 182–188. Brudvig, L.A., 2011. The restoration of biodiversity: where has research been and where does it need to go? American Journal of Botany 98, 549–558. Bullock, J.M., 2009. A long-term study of the roles of competition and facilitation in the establishment of an invasive pine following heathland fires. Journal of Ecology 97, 646–656. Bullock, J.M., Moy, I.M., Pywell, R.F., Coulson, S.J., Nolan, A.M., Caswell, H., 2002. Plant dispersal and colonization processes at local and landscape scale. In: Bullock, J.M., Kenward, A., Hails, R. (Eds.), Dispersal Ecology. Blackwell Science, Oxford, pp. 279–302. Burnside, N.G., Smith, R.F., Waite, S., 2003. Recent historical land use change on the South Downs, United Kingdom. Environmental Conservation 30, 52–60. CLC, 2000. CORINE land cover technical guide. European Environmental Agency ETC./LC, Copenhagen, 105 pp. Cousins, S.A.O., Eriksson, O., 2008. After the hotspots are gone: land use history and grassland plant species diversity in a strongly transformed agricultural landscape. Applied Vegetation Science 11, 365–374. Dallimer, M., Tich, D., Acs, S., Hanley, N., Southall, H.R., Gaston, K.J., Armsworth, P.R., 2009. 100 years of change: examining agricultural trends, habitat change and stakeholder perceptions through the 20th century. Journal of Applied Ecology 46, 334–343. De Wan, A.A., Sullivan, P.J., Lembo, A.J., Smith, C.R., Maerz, J.C., Lassoie, J.P., Richmond, M.E., 2009. Using occupancy models of forest breeding birds to prioritize conservation planning. Biological Conservation 142, 982–991. Fahrig, L., 2003. Effects of habitat fragmentation on biodiversity. Annual Review of Ecology Evolution and Systematics 34, 487–515. Feranec, J., Hazeu, G., Christensen, S., Jaffrain, G., 2007. Corine land cover change detection in Europe (case studies of the Netherlands and Slovakia). Land Use Policy 24, 234–247. Feranec, J., Jaffrain, G., Soukup, T., Hazeu, G., 2010. Determining changes and flows in European landscapes 1990–2000 using CORINE land cover data. Applied Geography 30, 19–35. Forup, M.K., Henson, K.S.E., Craze, P.G., Memmott, J., 2008. The restoration of ecological interactions: plant–pollinator networks on ancient and restored heathlands. Journal of Applied Ecology 45, 742–752. Fuller, R.M., 1987. The changing extent and conservation interest of lowland grasslands in England and Wales: a review of grassland surveys 1930–84. Biological Conservation 40, 281–300. Fuller, R.M., Groom, G.B., Jones, A.R., 1994. The Land Cover Map of Great Britain: an automated classification of Landsat Thematic Mapper data. Photogrammetric Engineering and Remote Sensing 60, 553–562. Fuller, R.M., Smith, G.M., Sanderson, J.M., 2002a. Land Cover Map 2000. Country side survey report. . Last visited 05.08.11. Fuller, R.M., Smith, G.M., Sanderson, J.M., Hill, R.A., Thomson, A.G., 2002b. The UK Land Cover Map 2000: construction of a parcel-based vector map from satellite images. Cartographic Journal 39, 15–25. Gerard, F., Petit, S., Smith, G., Thomson, A., Brown, N., Manchester, S., Wadsworth, R., Bugar, G., Halada, L., Bezak, P., et al., 2010. Land cover change in Europe between 1950 and 2000 determined employing aerial photography. Progress in Physical Geography 34, 183–205. Good, R., 1948. A geographical handbook of the Dorset flora. The Dorset Natural History and Archaeological Society, Dorchester. Guisan, A., Zimmerman, N.E., 2000. Predictive habitat distribution models in ecology. Ecological Modelling 135, 147–186. Henle, K., Alard, D., Clitherow, J., Cobb, P., Firbank, L., Kull, T., McCracken, D., Moritz, R.F.A., Niemela, J., et al., 2008. Identifying and managing the conflicts between

38

D.A.P. Hooftman, J.M. Bullock / Biological Conservation 145 (2012) 30–38

agriculture and biodiversity conservation in Europe – a review. Agriculture Ecosystems and Environment 124, 60–71. Hill, M.O., 1996 TABLEFIT, Version 10, for Identification of Vegetation Types. Centre for Ecology and Hydrology, Wallingford. UK. Hodgson, J.G., Grime, J.P., Wilson, P.J., Thompson, K., Band, S.R., 2005. The impacts of agricultural change (1963–2003) on the grassland flora of Central England: processes and prospects. Basic and Applied Ecology 6, 107–118. Howard, D.C., Watkins, J.W., Clarke, R.T., Barnett, C.L., Stark, G.J., 2003. Estimating the extent and change in Broad Habitats in Great Britain. Journal of Environmental Management 67, 219–227. Jackson, S.T., Hobbs, R.J., 2009. Ecological restoration in the light of ecological history. Science 325, 567–569. Johansson, L.J., Hall, K., Prentice, H.C., Ihse, M., Reitalu, T., Sykes, M.Y., Kindström, M., 2008. Semi-natural grassland continuity, long-term land-use change and plant species richness in an agricultural landscape on Öland, Sweden. Landscape and Urban Planning 84, 200–211. Johnson, C.J., Seip, D.R., Boyce, M.S., 2004. A quantitative approach to conservation planning: using resource selection functions to map the distribution of mountain caribou at multiple spatial scales. Journal of Applied Ecology 41, 238–251. Keith, S.A., Newton, A.C., Morecroft, M.D., Bealey, C.E., Bullock, J.M., 2009. Taxonomic homogenization of woodland plant communities over 70 years. Proceedings of the Royal Society B-Biological Sciences 276, 3539–3544. Keller, L.F., Waller, D.M., 2002. Inbreeding effects in wild populations. Trends in Ecology and Evolution 15, 230–241. Keymer, R.J., Leach, S.J., 1990. Calcareous Grasslands – A limited resource in Britain. In: Hillier, S.H., Walton, D.W.H., Wells, D.A. (Eds.), Calcareous Grasslands: Ecology and Management. Bluntisham Books, Bluntisham, UK, pp. 11–17. Lawton, J.L., Brown, V., Elphick, C., 2010. Making Space for Nature: A review of England’s Wildlife Sites and Ecological Network. DEFRA-report 2010-09. Lindborg, R., Bengtsson, J., Berg, A., Cousins, S.A.O., Eriksson, O., Gustafsson, T., Hasund, K.P., Lenoir, L., Pihlgren, A., Sjodin, E., Stenseke, M., 2008. A landscape perspective on conservation of semi-natural grasslands. Agriculture Ecosystems and Environment 125, 213–222. Manning, P., Putwain, P.D., Webb, N.R., 2004. Identifying and modelling the determinants of woody plant invasion of lowland heath. Journal of Ecology 92, 868–881. Mason, W.L., 2007. Changes in the management of British forests between 1945 and 2000 and possible future trends. Ibis 149, 41–52. Matthies, D., Bräuer, I., Maibom, W., Tscharntke, T., 2004. Population size and the risk of local extinction: empirical evidence from rare plants. Oikos 105, 481– 488. UK National Ecosystem Assessment, 2011. The UK National Ecosystem Assessment: Synthesis of the key findings. UNEP-WCMC, Cambridge. Natura 2000, 2011. GIS data: Natura 2000. . Last visited 05.08.11. Newton, A.C., Walls, R.M., Golicher, D., Keith, S.A., Diaz, A., Bullock, J.M., in press. Structure, composition and dynamics of a calcareous grassland metacommunity over a seventy year interval. Journal of Ecology. Oliver, T., Hill, J.K., Thomas, C.D., Brereton, T., Roy, D.B., 2009. Changes in habitat specificity of species at their climatic range boundaries. Ecology Letters 12, 1090–1101. Palmer, M.A., 1999. The application of biogeographical zonation and biodiversity assessment to the conservation of freshwater habitats in Great Britain. Aquatic Conservation: Marine and Freshwater Ecosystems 9, 179–208. Prendergast, J.R., Eversham, B.C., 1995. Butterfly diversity in southern Britain: hotspot losses since 1930. Biological Conservation 72, 109–114.

Preston, C.D., Pearman, D.A., Dines, T.D., 2002. New atlas of the British and Irish Flora: an atlas of the vascular plants of Britain, Ireland, the Isle of Man and the Channel Islands. Oxford University Press, UK. Rodwell, J.S., 1992. British plant communities: volume 3 Grasslands and montane communities. Cambridge University Press, Cambridge. Rose, R.J., Webb, N.R., Clarke, R.T., Traynor, C.H., 2000. Changes on the heathlands in Dorset, England, between 1987 and 1996. Biological Conservation 92, 117–125. Soons, M.B., Ozinga, W.A., 2008. How important is long-distance seed dispersal for the regional survival of plant species? Diversity and Distributions 11, 165–172. Soons, M.B., Messelink, J.H., Heil, G.W., 2005. Habitat fragmentation reduces grassland connectivity for both short-distance and long-distance winddispersed forbs. Journal of Ecology 1214, 1225. Southall, H., Bailey, B., Aucott, P., 2007. 1930s Land utilisation mapping: an improved evidence-base for policy? Environment Agency, Science Report: SC050031. Southall, H., Aucott, P., van Lunen, A., 2010. . Last visited 05.08.11. Stamp, D.L., 1931. The Land Utilisation Survey of Britain. The Geographical Journal 78, 40–47. Stoate, C., Boatman, N.D., Borralho, R., Rio Carvalho, C., de Snoo, G., Eden, P., 2001. Ecological impacts of arable intensification in Europe. Journal of Environmental Management 63, 337–365. Sutherland, W.J., Albon, S.D., Allison, H., Amstrong-Brown, S., Bailey, M.J., Brereton, T., Boyd, I.L., Carey, P., Edwards, J., et al., 2010. The identification of priority policy options for UK nature conservation. Journal of Applied Ecology 47, 955–965. Swetnam, R.D., 2007. Rural land use in England and Wales between 1930 and 1998: mapping trajectories of change with a high resolution spatio-temporal dataset. Landscape and Urban Planning 81, 91–103. Tamis, W.L.M., van’t Zelfde, M., van der Meijden, R., Groen, C.L.G., de Haes, H.A.U., 2005. Ecological interpretation of changes in the Dutch flora in the 20th century. Biological Conservation 125, 211–224. Thompson, D., 2007. The National Soil Map and Soil Classification. National Soil Resources Institute, UK. Information leaflet. . Last visited 05.08.11. Thomson, A.G., Manchester, S.J., Swetnam, R.D., Smith, G.M., Wadsworth, R.A., Petit, S., Gerard, F.F., 2007. The use of digital aerial photography and CORINE-derived methodology for monitoring recent and historic changes in land cover near UK Natura 2000 sites for the BIOPRESS project. International Journal of Remote Sensing 28, 5397–5426. Wade, M.R., Gurr, G.M., Wratten, S.D., 2008. Ecological restoration of farmland: progress and prospects. Philosophical Transactions of the Royal Society BBiological Sciences 363, 831–847. Walker, K.J., Preston, C.D., 2006. Ecological predictors of extinction in the flora of lowland England, UK. Biodiversity and Conservation 15, 1913–1942. Walker, K.J., Stevens, P.A., Stevens, D.P., Mountford, J.O., Manchester, S.J., Pywell, R.F., 2004. The restoration and re-creation of species-rich lowland grassland on land formerly managed for intensive agriculture in the UK. Biological Conservation 119, 1–18. Webb, N.R., 1990. Changes on heathlands of Dorset, England between 1978 and 1987. Biological Conservation 51, 272–286. Webb, N.R., Haskins, L.E., 1980. An ecological survey of heathlands in the Poole basin, Dorset, England in 1978. Biological Conservation 17, 281–296. Weber, N., Christophersen, T., 2002. The influence of non-governmental organisations n the creation of Natura 2000 during the European Policy process. Forest Policy and Economics 4, 1–12. Williams, P., Gibbons, D., Margules, C., Rebelo, A., Humphries, C., Pressey, R., 1996. A comparison of richness hotspots, rarity hotspots, and complementary areas for conserving diversity of British birds. Conservation Biology 10, 155–174.