L. C. Maskell, Centre for Ecology and Hydrology, Lancaster Environment Centre, Library Avenue, Bailrigg, LA1 4AP, UK (e-mail: firstname.lastname@example.org).
1We investigated the role of non-native species (neophytes) in common British plant communities using botanical data from two stratified random surveys carried out in 1990 and 1998.
2We found that from 16 851 plots surveyed in 1998 there were 123 non-native species found mostly in arable, tall grass/herb and fertile grassland habitats. Invasive non-native species, e.g. Fallopia japonica, Impatiens glandulifera and Rhododendron ponticum, were uncommon in this survey.
3Between 1990 and 1998 the total number of non-native species increased but the mean number of species per sample plot decreased. The mean cover of non-natives increased from 1.2% to 1.9%.
4There were positive spatial and temporal relationships between non-native and native species diversity. However, there was a weak negative relationship between changes in non-native cover and native diversity.
5The species composition and ecological traits of communities containing non-natives were very different from those that did not contain them.
6In the British countryside non-native species were mainly found in habitats with anthropogenic associations, high fertility, high number of ruderal species and high diversity. There is also an indication that successional shifts where competitive invasive species dominate involve non-native species.
7National-scale changes in plant community composition are likely to be closely correlated with external land-use impacts. Changes such as eutrophication, nitrogen deposition and increased fertility in infertile habitats are likely to benefit both native and non-native invasive species; however, currently these trends benefit native species much more often than non-natives.
8Non-native species are known to have significant effects on native species at local scales in many countries; however, at the landscape scale in Great Britain they are best considered as symptoms of disturbance and land-use change rather than a direct threat to biodiversity.
The expansion of non-native species is likely to accelerate in response to climate change, global transport and habitat modification. Invasion by non-native species has been described as one of the major threats to native communities (Sala et al. 2000) and indeed they have transformed habitats by outcompeting native inhabitants (Pimm et al. 1995). In this study we ask whether the loss of biodiversity associated with non-native species is a local, non-native specific process or whether it is possible to link non-native increases to biodiversity loss at national scales. Many studies have focused upon the devastating local scale impacts; for example, Fallopia japonica forms dense stands that prevent other species from growing (Pysek et al. 1995; Bimova et al. 2004). However, this does not necessarily mean that these effects can be extrapolated to the national scale. Perception of the large-scale abundance of non-natives can be distorted. The most notorious invaders into Britain are a small subset of non-native species that are currently casual or naturalized (Williamson & Fitter 1996a; Williamson & Fitter 1996b). Some of these are particularly visible, such as the purple flowers of Impatiens glandulifera on riverbanks or non-natives in built-up areas where there has been anthropogenic disturbance.
The total number of alien plant species recorded in the British flora has increased over at least the last 100 years (Preston et al. 2002). At the same time human activities have changed the species composition and diversity of many plant communities as a result of the non-random selection of certain trait combinations (Smart et al. 2005). It is therefore possible that the non-native species in the British flora have grown in number in response to recent environmental modifications, which have increased their cover and richness at the expense of native species. Alternatively, anthropogenic influences such as eutrophication could impact upon natives and non-natives in a similar fashion, favouring species with certain traits whether native or not.
In countries such as Great Britain, where the landscape is heterogeneous and agricultural, and population density is high, there have been few studies of sufficient scale and longevity to estimate the scale and impact of the threat from non-native species. Aside from closely managed changes in abundance of crops, little is known about the current large-scale impact of non-native species in common British plant communities: whether some habitats are more likely to contain non-native species and whether spatial or temporal changes in the diversity of non-native species are negatively correlated with native species diversity. There are even fewer data on temporal variation. Studies on central European vegetation tend to focus on urban areas (Pysek & Prach 1995; Pysek 1998; Pysek et al. 2002, 2003; Pysek et al. 2004). Spyreas et al.'s (2004) study of the vegetation of Illinois found that non-natives had permeated many habitats and replaced native plant communities, but also that some well-known exotics were not as common as presupposed.
Here, we investigate the role of non-native species in national patterns of within-community change using botanical data from two large-scale surveys of common British plant communities (Firbank 2003; Firbank et al. 2003; Smart et al. 2003a). Plant species composition was recorded in 1990 and 1998 from 9596 small, fixed plots located within a stratified random sample of 1-km squares covering Britain. These data represent within-community patterns but the large geographical scale of the surveys means inferences can be made about the entire extent of common vegetation types in Britain. First, we summarize spatial and temporal patterns in non-native distribution. Then we go on to ask the following questions. (i) Are changes in the number and abundance of non-natives correlated with native diversity? (ii) Is there a relationship between non-native species and the ecological traits and composition of associated native species?
The Countryside Survey (CS) sample design is based on a series of stratified, randomly selected 1-km squares. Stratification of sample squares was based on predefined strata referred to as ITE land classes. These have been derived from a classification of all 1-km squares in Britain, based on their topographic, climatic and geological attributes obtained from published maps (Bunce et al. 1996; Firbank et al. 2003).
The methods used for vegetation monitoring have been described in detail in Smart et al. (2003a) so here we present only a brief outline. Within each 1-km Countryside Survey sample square the land cover was mapped, including physiographic features, vegetation types, forestry features, boundaries, built-up land and land use. Vegetation plots were located within each square using a restricted randomization procedure designed to avoid spatial autocorrelation. Linear features (road verges, watercourse banks, hedges, arable margins and field boundaries) and areal features (fields, unenclosed land and small semi-natural biotope patches) were sampled. Linear plots were 1 × 10 m, laid out along a feature, whilst unenclosed land and small biotopes were sampled using 2 m × 2 m plots. Larger randomly placed plots were nested 14 m2 plots with an inner nest of 2 m × 2 m. Here we analyse data from this smallest nest. The locations of all plots were mapped and permanently marked when first recorded. The same plots were then relocated in subsequent surveys by means of compass bearings, sketch maps, plot photographs and relocation of a buried metal plate. The number of plots surveyed varied between years because it was not always possible to resurvey a plot and additional plots were recorded in 1998. The number of plots recorded in 1998 was 16 851, whereas the number of replicates surveyed in both 1990 and 1998 was 9596.
In each vegetation plot a complete list of all vascular plants and a selected range of the more easily identifiable bryophytes and macro-lichens was made. Nomenclature followed Stace (1991). Cover estimates were made to the nearest 5% for all species reaching at least an estimated 5% cover.
The plots were classified into eight vegetation types (aggregate classes) and full descriptions of each class and the construction of the classification are given in Bunce et al. (1996, 1999) and Smart et al. (2003a). The vegetation types are: crops/weeds, tall grassland and herb, fertile grassland, infertile grasslands, lowland wooded, upland wooded, moorland grass and mosaics and heath/bog.
native and non-native species
The classification of species as native or non-native followed the Atlas of the British and Irish Flora (Preston et al. 2002). This defines natives as species that are thought to have colonized Britain without the aid of humans. Archaeophytes are species introduced by humans before AD 1500. Neophytes are species that first occurred in Britain after AD 1500 and all are assumed to be non-natives introduced by humans. Although other authors have analysed both archaeophytes and neophytes (Pysek et al. 2003; Pysek et al. 2004) and there is merit in partitioning the effects of both, archaeophytes are generally considered as well integrated into the British flora, and problem species and invasions tend to arise from neophytes (Manchester & Bullock 2000). Therefore for this study, we have defined non-natives as neophytes only.
Distribution of non-natives
The distribution of non-natives in 1998 was analysed using a mixed model analysis of variance (GLIMMIX procedure in SAS) with the mean number of non-natives per plot as a response variable and habitat and plot type as fixed effects. The Countryside Survey square was incorporated as a random effect to account for the non-independence of plots located within the same square. Degrees of freedom were calculated using the approximation of Satterthwaite (1946). PROC GLIMMIX was used rather than PROC MIXED (as used in other tests) because a Poisson distribution was specified. All 16 851 plots surveyed in 1998 were used.
Mixed model analysis of variance was carried out to determine the relationship between the number of non-native species and climatic variables. The number of non-natives in each plot in the 1998 CS was the response variable; the height of the highest point in the square, the average precipitation, the mean June temperature and the mean January temperature as averages for the square were explanatory variables. The number of native species in a plot was included as a covariate and site as a random factor.
Plots were selected from the data base where crop species (including arable, forestry and horticulture) were less than 30% of the total cover, i.e. we excluded arable fields and woodland plantations to remove the effects of planted aliens. Note that crop species were not excluded from the analysis; exclusion applied only where they were likely to have been planted as crops.
Non-native and native species diversity
To examine the relationship between change in number of non-native species and the number of native species, plots were selected where non-natives had increased or decreased in plot occupancy between 1990 and 1998. Chi-square analysis was carried out to test the relationship. The above relationship between change in number of native species and change in number of non-native species was also tested using a mixed model anova (Proc mixed procedure in SAS) with site as a random factor. The relationship between cover of non-natives and native species diversity was tested using a mixed model anova with the change in number of native species as response variable and the change in cover of non-native species as an explanatory variable.
Is there a relationship between non-native species and the ecological traits and composition of associated native species?
Ordination (ter Braak & Smilauer 2002) was used to analyse changes in community composition. Detrended correspondence analysis (DCA) was used to assess overall variation in native species distribution of 8524 plots unrestricted by environmental variables. Plots were selected where the number of non-native species had increased, decreased or were absent in 1990 and 1998. These three categories were then used as supplementary variables. The ordination values were extracted and mean values plotted for these variables.
Each quadrat was given a cover-weighted mean value for each of a number of attributes based on native species only. Plant species were assigned an Ellenberg indicator value for fertility (N), light (L), moisture (F) and pH (R) taken from Hill et al. (2000). Plant strategy, following Grime's classification, was characterized by values for competitor (C), stress tolerator (S) and ruderal (R) (Grime et al. 1988).
Other traits analysed were specific leaf area (SLA), canopy height and growth form (woody, forb, grass, sedge). Data on most of these traits were obtained from Grime et al. (1988) and Stace (1991), but SLA came from an unpublished database (http://www.LEDA-traitbase.org). These traits were chosen to indicate environmental preferences (Ellenberg scores), position on disturbance and stress axes (Grime strategies), vegetative performance (canopy height and SLA) and plant type (growth form) of natives associated with non-natives.
Analyses were carried out as above using a mixed model analysis of variance with site as a random factor and the Satterthwaite approximation for degrees of freedom (Satterthwaite 1946). The traits of native species and archaeophytes in 1998 were used as the dependent variable, and whether the number of neophyte species had increased, decreased or never been there as the explanatory variable. Change in cover of non-natives was the explanatory variable in a separate analysis.
distribution of non-natives in 1998
Of 569 1-km squares surveyed, 123 contained no non-native species.
There were 123 non-native species in the survey from a total species number of 1258. The most frequent species were, in order of decreasing abundance, Acer pseudoplatanus, Triticum aestivum, Matricaria discoidea, Hordeum distichon, Veronica persica, Picea sitchensis, Brassica napus ssp. oleifera, Lolium multiflorum, Solanum tuberosum, Picea abies, Epilobium brunnescens, Impatiens glandulifera and Aesculus hippocastanum.
These comprise both plants grown as crops, which have been introduced into the wider countryside and subsequently escaped into adjacent habitats, e.g. Triticum aestivum and Brassica napus ssp. oleifera, garden escapes, e.g. Impatiens glandulifera, and species that have spread without deliberate introduction, e.g. Epilobium brunnescens.
Non-native species recorded in CS plots for the first time in 1998 included Populus canescens, Anisantha diandra, Calystegia sylvatica, Juglans regia, Doronicum pardalianches, Tragopogon porrifolius, Alnus incana, Elodea nuttallii, Leycesteria formosa, Mimulus moschatus, Senecio cineraria and Cucurbita pepo.
Invasive non-native species such as Fallopia japonica and Rhododendrum ponticum, regarded as common nuisance species, were infrequent in this survey and found in fewer than 20 plots out of 9556.
There were significant differences between habitats and plot types in the mean number of natives per plot. Tukey multiple comparisons tests showed that there were more non-native species in arable, tall grass/herb and fertile grassland habitats (Table 1) and a higher mean number of non-native species in arable and area plot types (Table 2). There were fewer non-native species in upland habitats, such as heath/bog and moorland grass, and in unenclosed and habitat plots.
Table 1. Results from the mixed model anova (GLIMMIX procedure) with the number of non-native species as the response variable and habitat type as explanatory variable. Site was included as a random effect. The mean number of non-native species per plot in each habitat type is shown
Mean number of non-natives per plot
1. Crops and weeds
2. Tall grassland/Herb
3. Fertile grassland
4. Infertile grassland
5. Lowland wooded
6. Upland wooded
7. Moorland grass/mosaic
Table 2. Results from the mixed model anova (GLIMMIX procedure) with the number of non-native species as the response variable and plot type as explanatory variable. Site was included as a random effect. The mean number of non-native species in each plot by plot type is shown
Mean number of non-natives per plot
Arable (50 m2)
Boundary (10 m2)
Hedge diversity (30 m2)
Hedge (10 m2)
Roadside verge (10 m2)
Streamside (10 m2)
Unenclosed (4 m2)
Area (200 m2)
Habitat (4 m2)
Non-native species per plot were more prevalent in the east of the country (Fig. 1), positively related to mean June temperature (F538 = 30.31, P < 0.001) and negatively related to precipitation (F538 = 12.16, P < 0.001).
change in number of alien species 1990–98
Ninety-four non-native species were present in 1990 and 107 in 1998 in the 9596 replicate plots. However, the mean number of non-native species per plot declined from 0.23 in 1990 to 0.19 in 1998 (T = 4.17, P < 0.001). The proportion of plots containing at least one non-native species declined from 26% in 1990 to 23.5% in 1998. However, the proportional cover of non-native species was 1.2% of native cover in 1990 and 1.9% in 1998.
relationship between non-native and native diversity
A chi-square analysis demonstrated that where non-native species richness increased, native species richness also increased (Table 3), and vice versa. This positive relationship between native and non-native diversity was also demonstrated in the mixed model anova (F8915 = 40.63, P < 0.001). However, there was a weak negative relationship between changes in cover of non-native species and the number of native species (F8944 = 16.87, P < 0.001).
Table 3. Chi-squared table of observed and expected values of the number of plots where non-natives increased or decreased and native species increased, stayed the same or decreased. The numbers in bold are where the observed values exceed the expected and the values in italic are where the observed values are lower than expected. Chi-squared = 31.997, P < 0.001
Frequency of alien species (1907 plots)
Native species diversity
is there a relationship between non-native species and the ecological traits and composition of associated native species?
While there was little difference in the native species composition of plots where non-natives had increased or decreased, there were significant differences between these and plots that contained no non-natives (Fig. 2). Plots where non-natives were present contained native species associated with lowland fertile disturbed environments, e.g. Anisantha sterilis, Convolvulus arvensis, Lamium album, Glechoma hederacea and Calystegia sepium. Plots where non-natives had not been present were more associated with plant assemblages comprising species of acidic, upland habitats, e.g. Erica tetralix, Empetrum nigrum, Calluna vulgaris, Juncus squarrosus and Vaccinium myrtillus.
The biggest differences in trait values were not between plots where non-natives increased or decreased but between those that contain a non-native and those that did not (Table 4). When non-native species were present native communities had higher pH, fertility, SLA, competitor and ruderal scores, more forbs, fewer woody species and lower light, moisture and stress tolerator scores.
Table 4. Mean trait values of native species in 8524 plots where non-natives increased, decreased or were never present, GLM with traits as response, aliens increasing, decreasing or never present as variables and site as a random factor. Only traits showing statistically significant responses (P < 0.05) are shown
Higher with non-natives present
Lower with non-natives present
Grime score Stress
Other traits Woody
There was a positive relationship between change in cover of non-natives, fertility (F8417 = 11.4, P < 0.001) and canopy height (F8460 = 8.2, P < 0.01) and a negative relationship between change in non-native cover and Ellenberg light score (F8545 = 28.4, P < 0.001).
distribution of non-native species
Non-natives were present in 25% of plots, but the contribution to cover was low, although it increased from 1.2% in 1990 to 1.9% in 1998. Invasive species such as Fallopia japonica, Impatiens glandulifera and Rhododendron ponticum were not common in the survey.
We detected clear, although small, changes in the non-native flora between 1990 and 1998. This contrasted with an increase in local cover of non-natives, suggesting certain species were increasing locally. The total number of non-native species increased but apparently many of these remain rare or occasional.
Non-natives comprise nearly 10% of species found in the 1998 Countryside Survey (CS). This is similar to other national surveys, e.g. France (Heywood 1989; Chytry et al. 2005), and much lower than some countries such as New Zealand where non-natives make up nearly 50% of the flora.
Because CS does not survey urban habitats, it is likely that the disturbed strongholds of some of these species are under represented (Roy et al. 1999). However, CS provides an unbiased assessment of the British rural landscapes and includes favourable habitats for non-natives such as roadside verges, riverbanks, strips alongside arable fields, hedgerows and small remnant patches of semi-natural vegetation (Manchester & Bullock 2000). The number of plots (16 851) and biogeographical coverage are so great that if aliens were abundant in the wider countryside they would be detected in the CS at higher frequencies.
Few countries have a national-scale randomized survey procedure like the GB Countryside Survey and most data for non-native presence tend to come from larger sampling units, e.g. county or 10 km square (Preston et al. 2002; McKinney 2004). CS combines national coverage with small plots, which should capture the dynamics of the most common species, so if non-native species do not feature in these plots it indicates that they are not important in engineering change. Therefore, although non-natives can have large impacts at the local scale, the invasive status of non-natives at the landscape scale in Britain is low. The low frequency of invasive non-natives suggests that a majority of locations in Britain are under little threat from non-native plants.
non-native and native species associations
There was a clear differentiation in native species composition and traits between plots where non-natives were present and where they were absent. There was no difference between plots where non-natives increased or decreased, which suggests that the dynamics of the non-native species are themselves having no impact on native species. This may be because communities where non-natives are found are so different from those where they are absent that external influences are more important than the non-natives themselves. Non-natives were associated with competitive species and more disturbed anthropogenically impacted communities, as typified by higher pH, fertility, SLA, competitor and ruderal scores and more forbs and fewer woody species. Higher SLA, fewer woody species and more forbs can be related to ability to respond rapidly to disturbance (Lake & Leishman 2004), a higher SLA being a critical trait in carbon fixation. Where nutrient enrichment occurs an ability to respond quickly favours success. There was also some association between non-natives and lower Ellenberg light scores and increased canopy height, which suggests that natives more tolerant to shade might persist under the canopy of some larger aliens. This may be because some non-natives may be associated with communities that are proceeding successionally towards rank, overgrown habitats where there is competitive dominance by one or more invasive (but not necessarily alien) species. Absence of non-natives was associated with native species of infertile, upland environments.
It is important to put these results into the context of more general vegetation changes. Results from CS show that there have been shifts along gradients of disturbance and fertility. Trait changes were largely consistent with the impact of increased nutrient availability across vegetation types associated with inherently low fertility, such as infertile grassland, heath, bog and moorland (Smart et al. 2005). Mesotrophic species have made minor yet consistent incursions into upland infertile habitats (Smart et al. 2003, 2005). In lowland landscapes linear features and small biotope fragments have shifted species composition to those associated with greater shade and less disturbance. It appears that these habitats are moving to a later stage of succession with taller species and lower species diversity.
An increase in fertility in upland infertile habitats and colonization by mesotrophic species may suggest greater opportunities for invasion of these habitats by non-natives. Over this same period there has also been an increase in grazing in the uplands, so these habitats are regularly disturbed, which should also favour invasion. However, we found that non-natives were less common in upland habitats, so either some other factor prevents their establishment or, more likely, they have still not reached many of these habitats.
Linear habitats and small biotopes in the lowlands are the sorts of habitats more likely to be invaded because of increased potential for dispersal, introduction by humans and disturbance. Riverine habitats are particularly likely to be invaded by propagules carried by water and waterside habitats have also shown marked signs of eutrophication and secondary succession involving increases in nutrient-demanding competitive species and loss of mesotrophic grassland species (Smart et al. 2003a; 2005). If such habitats are becoming more rank and overgrown it may be more difficult for non-natives to become established even if propagule supply is not limiting.
non-native and native diversity
This study found a positive relationship between diversity of natives and non-natives and a slight weak negative relationship between cover of non-natives and native diversity. Other studies of the relationship between non-native and native species diversity have found both negative (Elton 1958; Fox & Fox 1986; Tilman 1997; Knops et al. 1999) and positive relationships (Levine & D’Antonio 1999; Lonsdale 1999; Shea & Chesson 2002; Stohlgren et al. 2005). Lonsdale (1999) and Fridley et al. (2004) suggest that a positive relationship in large-scale survey data is simply a sampling artefact, the result of a greater habitat diversity and area, and that there is no causal relationship between non-natives and natives. This study considered small plots set within the landscape context so there would not be a sampling artefact. Pysek et al. (2002) showed a positive effect of the number of native species on the proportion of neophytes using a statistical procedure removing the effect of area. Stohlgren et al. (1999) found variation in the relationship between native and non-native species diversity with both spatial scale and habitat type, but sites with high native diversity were invasible in many landscapes. In a different study they found that the best predictor of non-native plant species density at the national scale was the density of native species (Stohlgren et al. 2005).
The positive relationship between non-native and native diversity has been shown to be positively correlated with soil fertility and water availability (Stohlgren et al. 1999; Stohlgren et al. 2002; Stohlgren et al. 2003), leading to the suggestion that invasion by non-natives is related to resource availability. An intermediate disturbance regime may interact with resource availability in that communities that receive some level of disturbance may have more resources available to invading species. Thus non-native species will not necessarily impact on native species diversity and drive change. Rather, they may be ‘passengers’ subjected to the same environmental changes influencing native species (MacDougall & Turkington 2005). It was previously thought that high native species richness would provide some defence against non-natives (Elton 1958) as all resources are being used. As in Stohlgren et al. (2002), this study demonstrates that this is not the case. It is difficult to determine why plots with non-natives are more diverse, without further study, but we know from previous work (Smart et al. 2003a) that CS tends not to sample rare and valuable habitats and the diversity may simply be due to factors such as accessibility, disturbance and resource availability.
Factors influencing non-native species establishment and richness may be different to those influencing non-native species cover and potential dominance (Stohlgren et al. 1999) and in this study we found a negative relationship between non-native cover and native species diversity. As many studies of non-native species show, it is when a non-native becomes successfully established and increases in abundance that it becomes a threat to native species.
The spread of non-natives into native communities is a global issue (Sala et al. 2000) and has been shown to result in major disruptions to communities, changes in ecological functioning and species extinctions. Although high densities of an invasive non-native are likely to depress the abundance of co-occurring natives, the regional effect of an exotic invasion is less easy to predict (Denslow & Flint Hughes 2004). Non-native species are seen generally as a threat by many authors because some possess the ability to become pernicious dominants. In fact, this may only be a small proportion (Williamson & Fitter 1996a; Williamson & Fitter 1996b) but they can still have significant impacts at a local scale. The low frequency of invasive non-natives found in CS suggests that they are not a significant influence on vegetation change at the landscape scale.
In the British countryside non-natives are mainly found in disturbed fertile habitats with anthropogenic associations, where species diversity of both natives and non-natives is high. At the same time increased cover of non-natives is associated with low light scores, higher canopy height and fertility. These results can be reconciled by viewing the relationship between native and non-native diversity and abiotic conditions as a successional process. The conditions that enable higher species diversity (disturbance and fertility) make it more likely that at some point an invasive species will be recruited. When a competitive invasive dominates it is associated with increased rankness and loss of diversity. There is no certainty that the switch from diverse subordinate species (Grime 1998) to domination by one or more species will be by a non-native. It is equally likely that invasive native species will dominate, such as Urtica dioica, Rubus fruticosus or Pteridium aquilinum (Pearman 2004). It is worth considering why the British flora has not been invaded as much as that in a country such as New Zealand; possibly the presence of native species possessing traits suitable for invasion and competitive dominance makes it less likely that non-natives will be able to establish. This would be an interesting area for further research.
Land-use changes such as eutrophication (Smart et al. 2003b), succession, expansion of mesotrophic species into upland habitats (Smart et al. 2005) and deposition of atmospheric nitrogen (Smart et al. 2004) may facilitate expansion of species tolerant of anthropogenic land use and be a threat to many habitats. This includes areas of higher conservation value. Future dispersal into infertile habitats coupled with changes in fertility and disturbance may enable non-native expansion and detrimental impacts on native communities. However, the low proportion of invasive non-natives found nationally in this study suggests that at the landscape scale, other common native invasives would be more likely to respond. The next Countryside Survey is due to take place in 2007 and it will be interesting to see if this pattern has changed.
We would like to thank all the staff in the Landuse Research Group at CEH Lancaster who have helped with organization and analysis of CS data: Rick Stuart, David Howard, Lisa Norton, Martin Rossall, Rod Scott, Colin Barr and John Watkins. We would also like to thank DEFRA for part-funding the Countryside Survey. K.T. was supported by the European Commission (Project EVR1-CT- 2002–40022 ‘LEDA’).