Influence of slope and aspect on long-term vegetation change in British chalk grasslands



    1. Institute of Ecosystem Science, School of Biological and Biomedical Sciences, Durham University, South Road, Durham DH1 3LE, UK, and CEH Monks Wood, Abbots Ripton, Huntingdon, Cambridgeshire PE28 2LS, UK
    Search for more papers by this author

    1. Institute of Ecosystem Science, School of Biological and Biomedical Sciences, Durham University, South Road, Durham DH1 3LE, UK, and CEH Monks Wood, Abbots Ripton, Huntingdon, Cambridgeshire PE28 2LS, UK
    Search for more papers by this author

    1. Institute of Ecosystem Science, School of Biological and Biomedical Sciences, Durham University, South Road, Durham DH1 3LE, UK, and CEH Monks Wood, Abbots Ripton, Huntingdon, Cambridgeshire PE28 2LS, UK
    Search for more papers by this author

    1. Institute of Ecosystem Science, School of Biological and Biomedical Sciences, Durham University, South Road, Durham DH1 3LE, UK, and CEH Monks Wood, Abbots Ripton, Huntingdon, Cambridgeshire PE28 2LS, UK
    Search for more papers by this author

Jonathan Bennie (tel. +44 1913341252; fax +44 1913341201; e-mail


  • 1 The species composition of fragmented semi-natural grasslands may change over time due to stochastic local extinction and colonization events, successional change and/or as a response to changing management or abiotic conditions. The resistance of vegetation to change may be mediated through the effects of topography (slope and aspect) on soils and microclimate.
  • 2 To assess long-term vegetation change in British chalk grasslands, 92 plots first surveyed by F. H. Perring in 1952–53, and distributed across four climatic regions, were re-surveyed during 2001–03. Changes in vegetation since the original survey were assessed by comparing local colonization and extinction rates at the plot scale, and changes in species frequency at the subplot scale. Vegetation change was quantified using indirect ordination (Detrended Correspondence Analysis; DCA) and Ellenberg indicator values.
  • 3 Across all four regions, there was a significant decrease in species number and a marked decline in stress-tolerant species typical of species-rich calcareous grasslands, both in terms of decreased plot occupancy and decreased frequency within occupied plots. More competitive species typical of mesotrophic grasslands had colonized plots they had not previously occupied, but had not increased significantly in frequency within occupied plots.
  • 4 A significant increase in Ellenberg fertility values, which was highly correlated with the first DCA axis, was found across all regions. The magnitude of change of fertility and moisture values was found to decrease with angle of slope and with a topographic solar radiation index derived from slope and aspect.
  • 5 The observed shift from calcareous grassland towards more mesotrophic grassland communities is consistent with the predicted effects of both habitat fragmentation and nutrient enrichment. It is hypothesized that chalk grassland swards on steeply sloping ground are more resistant to invasion by competitive grass species than those on flatter sites due to phosphorus limitation in shallow minerogenic rendzina soils, and that those with a southerly aspect are more resistant due to increased magnitude and frequency of drought events.


Traditionally managed calcareous grasslands are among the most species-rich habitats in Western Europe. Their species composition is sensitive to changes in grassland management (Kahmen et al. 2002; Moog et al. 2002; Pykäläet al. 2005), eutrophication (Willems et al. 1993; Jacquemyn et al. 2003) and habitat fragmentation (Fischer & Stöcklin 1997). During the second half of the twentieth century calcareous grasslands, along with other semi-natural habitats associated with pre-World War II farming methods, were progressively fragmented and reduced in area (Poschlod & WallisDeVries 2002; Dutoit et al. 2003; Hodgson et al. 2005). Meanwhile, management practices intended to increase the productivity of grasslands produced species-poor swards dominated by productive plant species of low nature conservation value (Fuller 1987; Blackstock 1999). In the case of British chalk grasslands, widespread changes in management over this period have included abandonment of grazing or changes in grazing regime, application of fertiliser and nutrient enrichment from winter feeding of stock (Keymer & Leach 1990). The dramatic decline in rabbit numbers following the myxomatosis outbreak of 1954 was also a driver of vegetation change (Thomas 1960, 1963). In the Netherlands, vegetation changes at chalk grassland sites have been linked to the effects of increased nitrogen deposition (Bobbink 1987; Willems et al. 1993). However, evidence of similar effects in British calcareous grasslands, where nitrogen deposition rates are typically less than half those in the Netherlands, has been lacking (Wilson et al. 1995) although several other semi-natural habitat types in Britain have shown a loss of species richness and changes in species composition consistent with an increase in substrate fertility over this period (McCollin et al. 2000; Haines-Young et al. 2003; Smart et al. 2003). Several authors have concluded that phosphorus limitation in calcareous soils often prevents a short-term response to nitrogen addition (Morecroft et al. 1994; Wilson et al. 1995; Grime et al. 2000). The importance of water limitation in free-draining chalk soils has also been noted; extreme summer drought events suppress dominant plant species and ‘reset’ plant distributions (Hopkins 1978; Watt 1981; Buckland et al. 1997).

The fragmentation of chalk grassland patches may have an effect on the species composition of the vegetation in the remaining habitat, even in the absence of changes in habitat quality. The theory of island biogeography (MacArthur & Wilson 1967) predicts that small populations in fragments of habitat experience higher rates of extinction than populations in larger fragments, and that habitat patches that are more isolated from a source of propagules are less likely to be recolonized than closer patches. Furthermore, fragmented habitat patches are likely to have high rates of input of propagules from non-habitat specialist species from the surrounding area, which may increase the probability of invasion by non-specialists. Local extinction rates in Swiss calcareous grassland sites were higher for small populations, for species with a short life cycle and for species with high habitat specificity, as well as for species with longer-lived seeds, but were independent of nutrient indicator values (Fischer & Stöcklin 1997; Stöcklin & Fischer 1999).

Archived biological records represent a potential resource for quantifying long-term vegetation change, provided that comparable recent data are available and a comparable methodology is used (McCollin et al. 2000). In the case of vegetation records from plots, a re-inspection of sites should reveal long-term trends in vegetation composition, provided that the original survey plots can be approximately re-located (Hédl 2004), and that the trend over the period of interest is not obscured by short-term variability. A potential problem is that of pseudo-turnover (Fischer & Stöcklin 1997), where, due to imprecise location of the original plots, not all species present at a site are recorded and species may, misleadingly, appear to colonize or go extinct. Re-survey produces snapshots of the state of the vegetation at points in time, and can complement observations from permanent plots and experimental manipulations, which are seldom maintained for more than a few years (Dodd et al. 1995; Kahmen et al. 2002). Although the causes of change may sometimes be obscure, environmental indicators such as Ellenberg's indicator values (Ellenberg 1988) can suggest the drivers of vegetation change from observed changes in vegetation composition (McCollin et al. 2000; Haines-Young et al. 2003; Smart et al. 2003; Hédl 2004).

Perring (1959) showed that the species composition of chalk grassland at many sites in Britain was strongly influenced by slope and aspect. Two distinct topographic gradients lie behind the observed distribution patterns, namely (i) the tendency of slopes facing the equator (and hence intercepting more direct solar radiation) to be both drier and warmer than those facing away from the equator (Geiger 1965; Påhlsson 1974; Radcliffe & Lefever 1981; Rorison et al. 1986a) and (ii) the tendency for soils on relatively steep (> c. 15°) chalk and limestone slopes to form shallow, unstable rendzina soils with a higher carbonate content and lower available phosphorus than the more stable brown calcareous earths formed on flatter ground (Balme 1953; Trudgill 1976). Measurements taken during the present study at Millington Pastures in the Yorkshire Wolds were consistent with these observations. Volumetric soil moisture on south-facing slopes was typically 10–20% lower than on north-facing slopes during the summers of 2001–03, with soil moisture at flat sites intermediate; total extractable phosphorus in July was measured at 8.2 mg kg−1 soil (n = 6, SE = 0.28) on a brown calcareous earth at a flat plateau site, and 1.5 and 1.4 mg kg−1 (n = 6, SE = 0.31 and 0.32) on minerogenic rendzinas on north and south facing 30° slopes, respectively. If phosphorus and/or water limitation help to maintain a species-rich sward in chalk grassland, it would be expected that, at sites with varied topography, steeper slopes would be more resistant to change in vegetation composition than flatter areas, and that slopes with a high radiation load (i.e. south-facing) will be more resistant to change than those facing north.

This study aimed to assess the changes in grassland vegetation over a period of approximately 50 years in sites sampled across several different regions of the English chalklands. These sites were surveyed by F. H. Perring in 1952 and 1953, and re-surveyed using the same methods over the period 2001–03. The original data set was made available for this study in the form of punched hole cards, site notes and maps. The aim of Perring's original survey was to sample chalk grassland on a range of slopes and aspects from four distinct English regions (and one region of northern France), in order to elucidate the relationships between vegetation, soils, climate and topography (Perring 1956, 1958, 1959, 1960). An initial subjective assessment suggested that, even where plots retained traditionally managed grassland of semi-natural character, often for nature conservation, chalk grassland swards are currently of a more mesotrophic nature than those recorded in the original survey.

We used Perring's data and our recent surveys of his sites to test the following hypotheses concerning vegetation change, local extinctions, topography and resistance to change: (i) vegetation in unimproved chalk grassland plots across the English chalk is currently of a more mesotrophic nature than in the early 1950s, (ii) species extinction rates during the last 50 years have been higher for species with low initial frequency, (iii) the magnitude of vegetation change has been greater on flatter areas than on slopes, and (iv) the magnitude of vegetation change is smaller on sites with a greater radiation load.

Materials and methods

the study areas

The locations of Perring's 263 vegetation plots were clustered in four distinct climatic regions of the English chalk and, although not permanently marked in the field, maps, site descriptions and grid references were available for most plots. Some plots, particularly those sited on or around archaeological earthworks, could be accurately re-located with confidence; elsewhere plots for re-survey were located in a homogeneous stand of vegetation with the same slope, aspect and landscape position as the original Perring plot. Fischer & Stöcklin (1997) found that the effects of pseudo-turnover in a calcareous grassland increased with distance between the original and repeated surveys, and that it was independent of the life form and habitat specificity of a species. To minimize pseudo-turnover, re-surveyed plots whose centre was thought likely to be more than 20 m from the original centre were excluded from the analysis. Plots in markedly heterogeneous areas (for example close to a field boundary) were also omitted. Observations were made between May and September over the period 2001–03, as close as was practical to the time of year when Perring originally visited the sites.

survey methods

The main aim of the study was to assess change in vegetation composition within calcareous grassland, rather than land use change. Plots were therefore selected for re-survey only if they remained in 2001–03 as recognizably calcareous grassland of unimproved or semi-improved character, with no evidence of ploughing, reseeding or widespread fertiliser application; in particular, low-diversity improved grasslands dominated by Lolium perenne were omitted. The re-surveyed plots therefore tended to be concentrated in areas where complex topography had hindered agricultural improvement, or on sites with designated conservation status. In the Yorkshire Wolds, in particular, formerly extensive areas of unenclosed rough grazing have been fragmented by the enclosure and conversion of the flatter ground on the Wolds plateau to arable or improved pasture, leaving only the steeper sides of the dry valleys as unimproved grassland. In this region flat or gently sloping plots are under-represented compared to the original survey.

Of the original 263 plots, 38 (14%) could not be located with confidence during the 2002/03 field survey and, for a further 34 (13%), access permission could not be obtained from landowners during the time available. Seventy-two plots (27%) had undergone land-use change to arable or improved pasture; 23 plots (9%) were dominated by scrub communities and 4 (2%) had undergone afforestation. The vegetation surveys of the remaining 92 plots (35% of the original sample) are used in this analysis. Of these, 27 were located in East Anglia, 6 in Kent, 43 in Dorset and 15 in the Yorkshire Wolds. Plot locations, management and nitrogen deposition rates are shown in Table 1.

Table 1.  Location, management and total nitrogen deposition of resurveyed plots. Deposition rates are derived from UK 5-km resolution modelled data for short vegetation (pers. comm., R. Smith, CEH Edinburgh)
Region/areaNumber of surveyed plotsGrid reference (GB grid)Management 1952–53Management 2002–03Total N deposition 2002–03 (kg ha−1 year−1)
East Anglia
Knocking Knoll 9TL 1330Grazed (cattle)Grazed (sheep)23.9
Therfield Heath18TL 34 40Cut (grazed pre 1939)Cut, grazed in parts26.0
Melbury Hill 2ST 87 20Grazed (cattle)Grazed (cattle)29.5
Fontmell Down 7ST 87 16Grazed (cattle, sheep)Grazed (sheep)32.2
Hambledon Hill27ST 84 10Grazed (cattle)Grazed (cattle, sheep)27.9
Hod Hill 1ST 85 10Grazed (cattle)Grazed (cattle)28.6
Ballard Down 6SZ 02 81UngrazedGrazed (sheep)15.1
Clearbury Down (Wilts) 1SU 15 24Grazed (cattle)Grazed (cattle)22.0
Yorkshire Wolds
Brubber Dale 3SE 86 60Grazed (cattle)Grazed (sheep)32.0
Huggate Pastures 4SE 8556Grazed (cattle)Grazed (sheep)31.9
Millington Pastures 6SE 84 53Grazed (cattle, sheep)Grazed (cattle, sheep)31.3
Water Dale 1SE 83 61Grazed (cattle)Grazed (sheep)33.5
Birdsall Brow 1SE 83 63Grazed (cattle)Grazed (cattle)33.5
Broad Downs 3TR 07 45Grazed (cattle, sheep)Grazed22.3
Crundale Downs 1TR 08 47Grazed (cattle, sheep)Grazed22.3
Chilmans Down 2TR 08 52Grazed (cattle, sheep)Grazed19.0
Crundale Pasture 1TR 07 45Grazed (cattle, sheep)Grazed22.3

The vegetation survey methods described by Perring (1956) were followed as closely as possible, and are summarized briefly here. Plots were approximately 50 m2 in area, usually a square measuring 7 × 7 m. These plots were selected as representative of a stand of homogeneous vegetation with a given slope and aspect; in each region plots were selected to represent a range of slopes and aspects. In each plot 20 quadrats measuring 10 × 10 cm were thrown at random. Within each quadrat, all vascular plants and bryophytes were identified and recorded on a cover scale of 1–5, where 1 represents between 1% and 20% cover and 5 represents between 81% and 100%.

data analysis

Data from Perring's summary record cards were entered into a computer data base. Nomenclature for vascular plants was adjusted to follow Stace (1997) for comparison with the present survey. All of Perring's records of Thymus species were recorded as Thymus drucei (=Thymus polytrichus); in several of the plots in the present survey Thymus pulegioides was recorded and the two species may not have been distinguished by Perring. Therefore, for this analysis Thymus spp. were aggregated, as were the complex taxa Euphrasia, Rosa, Taraxacum and Rubus. Twenty-three of Perring's records were of Carex nigra, a species that is normally absent from dry calcareous grasslands. We have treated these as transcription errors for Carex caryophyllea, which was adjacent in the numbering system used on his record cards. Ten species names from the record cards were unclear or could not be allocated modern species names and were omitted; all of these were uncommon species found in less than five plots.

Species frequency values were used in the analyses to minimize errors associated with the subjective assessment of cover. Frequencies were calculated as the proportion of quadrats within a plot which contained the species, and were expressed as a percentage.

The sampling method, although entailing 20 replicate quadrats, was not exhaustive, and on several occasions species that did not fall into quadrats were noted within plots. Few conclusions may therefore be drawn from the occurrence or apparent disappearance of uncommon species present at low frequencies in the data. Out of 153 vascular plant species recorded during the surveys, 83 were either found in less than 10 plots overall or were at less than 10% average frequency within occupied plots. These were classed as uncommon.

species trends

To characterize changes in the abundance of individual species, n1 was defined as the total number of plots occupied by a species at the first survey, n2 as the number of plots occupied at the second survey, ncol as the number of new plots colonized by the species between surveys, next as the number of plots from which the species was lost. The number of plots in which the species was present on both dates was defined as n1,2 =n1next and the change in frequency scores between surveys in these plots was assessed using paired t-tests on arcsine transformed frequency data. Three trends could therefore be identified for each species: (i) the colonization of previously unoccupied plots, (ii) rates of extinction at previously occupied plots, and (iii) changes in frequency within continuously occupied plots. Multiple linear regression was used to test whether the initial population size and/or species habitat preference affects the probability of extinction for a species, with arc-transformed extinction rate next/n1 as a dependent variable and average recorded arc-transformed frequency at the first survey date and species Ellenberg N (fertility) index (Ellenberg 1988; Hill et al. 1999) as independent variables. To test whether the probability of colonization and changes in within-plot frequency were significantly influenced by sensitivity to nutrient availability, linear regressions were undertaken with the number of colonized plots ncol and the normalized change in frequency for continuously occupied plots as dependent variables and fertility index as independent variable. To minimize the effect of pseudo-turnover, uncommon species were excluded from these analyses.

Classification of plant strategy follows the Competitor–Stress tolerator–Ruderal (CSR) system with values obtained from Grime (1988).

vegetation change

Data from both surveys were used to classify plots into the UK National Vegetation Classification (NVC) communities (Rodwell 1992), and mean Ellenberg Indicator Values were calculated for each plot using mavis Plot analyser vs. 1.0 (Stuart 2000). Plot scores represent weighted mean species scores from the Ellenberg indicator species data set (Ellenberg 1988) extended to Great Britain (Hill et al. 1999; Hill et al. 2000). The indices used are L (light), F (moisture), R (reaction) and N (nitrogen); the latter is probably best interpreted as a productivity or soil fertility indicator, rather than a measure of nitrogen available to plants (Hill & Carey 1997; Schaffers & Sýkora 2000). Detrended correspondence analysis (DCA) was carried out on the frequency values for each plot using decorana (Hill 1979), with default values for all parameters and detrending by segments.

Six indices of vegetation composition were used to quantify vegetation change in this analysis: the plot scores along DCA axes 1 and 2, and the weighted mean Ellenberg L, F, R and N indices.

Trends in indicator scores between and within regions were tested using repeated measures analysis of variance (Zar 1996). Plot scores were used as the dependent variable, repeated for the two survey dates, and the region of the plot was used as a between-subjects factor. This method tests simultaneously for a significant change in the plot scores over time, for consistent effects of region on index over the entire data set, and, through the interaction terms, for differences in the magnitude of change in plot scores in different regions.

It was established by Perring (1956, 1959) that in the original data, vegetation varies systematically with region, slope and aspect. In order to test for any effects of slope and aspect on the magnitude and/or direction of vegetation change, it was necessary to control for the known initial difference in vegetation composition. In order to do this, a mixed model analysis of covariance (ancova) was used with the 2003 plot scores as the dependent variable and with 1953 plot scores, slope angle and a radiation index as covariates. The radiation index was calculated as Ri = cos β cos Z + sin β sin Z cos(Ω  Ωs), where β is the angle of slope, Z is the solar zenith (angle of the sun from vertical), Ω is the solar azimuth (angle of the sun from due north), Ωs is the slope aspect, and Ri is a radiation index estimating the proportion of potential direct irradiance intercepted by the slope for a given solar zenith and azimuth (Oke 1987). Values of Ω and Z representative of 15.00 hours in August were used, to represent maximum conditions of sward temperature and evaporative demand, and to allow for east–west asymmetry in topographic microclimate. A maximum Ri value of 1 is achieved when a slope is perpendicular to the direct radiation beam, and a value of zero when the slope is parallel to the beam. Region was included as a fixed factor in the analysis. The inclusion of the 1953 plot scores as a covariate aims to control for variation in the initial composition of the plots. Any significant effect of other factors implies an effect on the magnitude of change, independent of initial conditions.


species turnover

Overall, the mean number of species recorded in plots decreased significantly from 25.0 to 21.4 (paired t-test, P < 0.001). Table 2 shows the changes in plot occupancy and frequency of common species recorded at both survey dates. Multiple linear regression showed a significant negative relationship between the extinction rates of species (next/n1) and their mean initial frequency (r2 = 0.460, P < 0.001). There was no significant relationship between species extinction rate and Ellenberg N index. The number of colonizations per species, in contrast, was significantly positively related to Ellenberg fertility values (r2 = 0.272, P < 0.001). The percentage change in frequency for each species was also significantly positively related to Ellenberg fertility values (r2 = 0.120, P = 0.008). Table 3 lists species found at only one survey date. Thirty-six of the species present in the original survey were not recorded at the second survey. Thirty-five of these were uncommon species found at low frequencies in five or fewer plots; many of these are still known to exist at low frequencies in grasslands from the survey area. Thirty-three species found in the second survey were not recorded in the original plots; all of these were uncommon species.

Table 2.  Changes in plot occupancy and frequency of common vascular plant species in the surveyed plots. n1, number of plots in which species was found, 1952–53; n2, number of plots in which species was found 2001–03; n1,2, number of plots in which species was found on both dates. p2), significance of change in number of occupied sites (chi-square test); p(t) significance and direction of change in frequency within continuously occupied sites (paired two-tailed t-tests of arcsine-transformed frequency at sites at which species was found on both dates). Due to multiple comparisons, three species in each analysis are expected to show significance at P < 0.05 by chance. Perring's nomenclature in brackets, plant strategy follows Grime (1988). Perennation from Hill et al. (2004): P, perennial; B, biennial; A, annual
Species nameStrategyPerennationn1n2p2)n1,2p(t)
Decreased species:
Carex caryophylleaSP6621< 0.00112n.s.
Koeleria macrantha (K. gracilis)SP8846< 0.00145< 0.001 (–)
Thymus spp.SP6441< 0.00136< 0.01 (–)
Succisa pratensisSP23 5< 0.001 3n.s.
Leontodon taraxacoidesSR/CSRP20 5< 0.001 0
Briza mediaSP7346< 0.0140< 0.001 (–)
Pilosella officinarum (Hieracium pilosella)S/CSRP6035< 0.0133< 0.01 (–)
Festuca ovinaSP8259< 0.0154< 0.05 (–)
Campanula rotundifoliaSP6947< 0.0138< 0.01 (–)
Helianthemum nummularium (H. chamaecistus)SP4221< 0.0119< 0.01 (–)
Ranunculus bulbosusSRP5633< 0.0124n.s.
Cirsium acauleSP6141< 0.0538< 0.01 (–)
Linum catharticumSRB5738< 0.0529< 0.01 (–)
Sanguisorba minor (Poterium sanguisorba)SP8061< 0.0557< 0.001 (–)
Lotus corniculatusS/CSRP8162< 0.0558< 0.05 (–)
Helictotrichon pratenseS/CSRP7957< 0.0551< 0.05 (–)
Plantago media P2914< 0.05 9n.s.
Hippocrepis comosa P4027< 0.0520n.s.
Asperula cynanchicaSP4229< 0.0520< 0.001 (–)
Polygala calcarea P2413< 0.0512n.s.
Medicago lupulinaR/SRA3221< 0.0513n.s.
Little or no change:
Plantago lanceolataCSRP8175n.s.68n.s.
Trifolium repensCR/CSRP2824n.s.10n.s.
Primula verisS/CSRP1614n.s. 6n.s.
Brachypodium sylvaticumS/SCP 7 6n.s. 1
Centaurea nigraCSRP3029n.s.16n.s.
Galium mollugo (inc. G. erecta) P 8 9n.s. 0
Trisetum flavescensCSRP2829n.s.12n.s.
Bromopsis erecta (Bromus erectus)SC/CSRP2528n.s.22< 0.05 (–)
Galium verumSC/CSRP4449n.s.26n.s.
Carex flaccaSP8063n.s.59n.s.
Agrostis capillaris (Agrostis tenuis)CSRP3727n.s.18< 0.05 (–)
Scabiosa columbariaS/SRP4434n.s.19n.s.
Prunella vulgarisCSRP4737n.s.26n.s.
Luzula campestrisS/CSRP2012n.s. 6n.s.
Polygala vulgarisSP16 9n.s. 4n.s.
Bellis perennisR/CSRP1710n.s. 2
Euphrasia nemorosaSRA2821n.s.10n.s.
Taraxacum officinaleR/CSRP1610n.s. 2
Senecio jacobaeaR/CRP1611n.s. 1
Crepis capillarisR/SRA1417n.s. 0
Filipendula vulgaris P1316n.s. 5n.s.
Centaurea scabiosaS/CSRP 711n.s. 4< 0.01 (–)
Pimpinella saxifragaS/SRP2934n.s. 9n.s.
Rumex acetosaCSRP1622n.s.11n.s.
Helictototrichon pubescensS/CSRP3340n.s.17n.s.
Festuca rubraCSRP5164n.s.38n.s.
Increased species:
Brachypodium pinnatumSCP1221< 0.0512< 0.05 (–)
Achillea millefoliumCR/CSRP2540< 0.0114n.s.
Gentianella amarellaSRP 512< 0.01 0
Campanula glomerata P 920< 0.001 7< 0.05 (–)
Lolium perenneCR/CSRP 517< 0.001 1
Trifolium pratenseCSRP1532< 0.001 6n.s.
Leontodon hispidusSP2746< 0.00118n.s.
Veronica chamaedrysCSRP1130< 0.001 5n.s.
Arrhenatherum elatiusCSRP1131< 0.001 4n.s.
Holcus lanatusCSRP1738< 0.00111n.s.
Phleum bertolonii (P. nodosum) P1337< 0.001 7n.s.
Cynosurus cristatusCSRP1544< 0.001 9n.s.
Dactylis glomerataC/CSRP2462< 0.00120n.s.
Table 3.  Vascular plant species found in 1952–53 but not in 2002–03 (extinct species) and in 2003–03 but not in 1952–53 (new species), number of plots in which the species was found and mean frequency within those plots. Perring's nomenclature in brackets. Plant strategy following Grime (1988). Perennation from Hill et al. (2004): P, perennial; B, biennial; A, annual
Species nameStrategyPerennationNumber of plotsMean frequency (%)
Extinct species:
Aira caryophylleaSRA111.7
Anagallis tenella P110
Arenaria serpyllifoliaSRA1 5
Astragalus danicus P313.3
Carex humilis P170
Carlina vulgarisSRB5 7.8
Centaurium pulchellum A1 7.5
Cerastium arvense P125
Convolvulus arvensisCRP1 5
Danthonia decumbens (Sieglingia decumbens)SP1034.5
Desmazeria rigida (Catapodium rigidum) A225
Desmazeria marina (Catapodium marinum) A1 5
Festuca arundinaceaCSRP245
Fragaria vescaCSRP217.5
Fraxinus excelsiorCP1 5
Gentianella anglica B3 8.3
Inula conyzaS/SRP1 5
Knautia arvensisCSRP215
Ononis repensS/CSRP110
Orchis masculaS/SRP110
Plantago coronopusSRB1 5
Poa trivialis P415
Prunus spinosaCSP2 5
Serratula tinctoria P1 5
Sonchus arvensisCRP115
Sonchus oleraceusR/CRA110
Stachys officinalisSP110
Thalictrum minus P1 5
Thesium humifusum P420
Torilis nodosa A115
Tragopogon pratensisCRB220
Trifolium mediumCS/CSRP125
Trinia glauca B215
Veronica officinalisS/CSRP411.3
Viola odorataCSRP420
Viola rivinianaSP3 5
New species:
Agrimonia eupatoriaCSRP311.7
Anacamptis pyramidalisS/SRP4 6.3
Anthyllis vulnerariaS/SRP325
Blackstonia perfoliata A1 5
Bromus hordeaceusR/CRA413.75
Carex montana P5 6.9
Cerastium semidecandrumSRA1 5
Cirsium palustreCRB1 5
Crepis biennis B615.8
Galium cruciata P511
Geranium molleS/SRA215
Lathyrus montanus P415
Linum perenne P110
Odontites vernaRA5 6
Origanum vulgareCSP1 5
Picris hieracioides P1 5
Poa angustifolia P3 6.7
Ranunculus repensCRP513
Reseda lutea P2 5
Rhinanthus minorR/SRA510
Rosa spp. P1 5
Rubus spp. P1 5
Salvia pratensis P1 5
Senecio squalidus A1 5
Seseli libanotis B368.3
Silene vulgaris P1 5
Stellaria holosteaCSRP310
Teucrium scorodoniaS/CSRP115
Trifolium campestreSRA517
Ulex europaeusCSP155
Urtica dioica P140
Vicia hirsutaR/CSRA116.25
Vicia lathyroides A115

dca ordination

Plots were classified into the calcareous grassland communities CG1 Festuca ovina–Carlina vulgaris, CG2 Festuca ovina–Helictotrichon pratensis, CG3 Bromus erectus, CG4 Brachypodium pinnatum, CG6 Helictotrichon pubescens and mesotrophic grassland communities (MG1, MG8, MG7) following Rodwell (1992). These communities were separated along the first two DCA axes (Fig. 1a), with CG2 plots scoring low values on both axes, and the mesotrophic communities scoring higher values on axis 1; CG3, CG4 and CG6 were intermediate between the two (Fig. 1a). Axis 1 (eigenvalue = 0.428) was found to be highly positively correlated with plot Ellenberg fertility (r2 = 0.92), and moisture scores (r2 = 0.85) and negatively correlated with soil reaction (r2 = −0.65) and light (r2 = −0.59, P < 0.001 in all cases). Axis 2 (eigenvalue = 0.232) was less strongly correlated with reaction (r2 = 0.49, P < 0.001) and fertility (r2 = 0.18, P < 0.05). During the 50 years interval between surveys, there was a clear shift in species composition along both DCA axes 1 and 2 (paired t-tests; significant at P < 0.001) moving away from species-rich CG2-type communities towards those dominated by competitive/ruderal grass species (Fig. 1b). This increase in DCA axis 1 scores is consistent across all regions (Fig. 2) with a significant effect of region (variation in scores between different regions) but no significant interaction effect between region and date (no difference in magnitude and/or direction of change over time between regions; Table 4). Axis 2 shows a more complex pattern of change, with both a significant contrast between survey dates and a significant interaction term with region, implying that the direction and/or magnitude of change differs between regions (Fig. 2b, Table 4).

Figure 1.

Plots of first two DCA axes (a) site scores from initial survey (open circles) and second survey (filled circles); polygons enclosing sites by NVC community (b) polygons enclosing sites by date of survey. Vector shows mean direction and magnitude of change between surveys.

Figure 2.

Mean (± 1 SE) plot values of vegetation indices for all regions at both survey dates. (a) DCA axis 1, (b) DCA axis 2, (c) Ellenberg F, (d) Ellenberg L, (e) Ellenberg R, and (f) Ellenberg N.

Table 4.  Summary of repeated measures anova statistics for plot DCA axes and Ellenberg indicator values. Date of survey is the within-subject factor; region is the between-subject factor
 Sum of squaresd.f.F-ratioP
Axis 1
Date126823 147.358< 0.001
Date × region  1417 3 0.176  0.912
Error (date)  141788  
Region200816 3 8.534< 0.001
Axis 2
Date 21296 119.830< 0.001
Date × region246294 359.538< 0.001
Error (date)22169388  
Region 72280 3 2.158  0.099
Error 6106188  
Ellenberg L
Date 0.108 1 7.806  0.006
Date × region 0.158 3 3.793  0.013
Error (date) 1.21888  
Region 0.469 3 7.037< 0.001
Error 1.95688  
Ellenberg F
Date 0.446 114.696< 0.001
Date × region 0.647 3 8.989< 0.001
Error (date) 2.67188  
Region 2.113 3 7.106< 0.001
Error 6.89488  
Ellenberg R
Date 0.157 1 2.944  0.090
Date × region 0.852 3 5.340  0.002
Error (date) 4.67988  
Region 9.200 328.601< 0.001
Error 9.43588  
Ellenberg N
Date10.0 130.805< 0.001
Date × region 0.471 3 0.481  0.696
Error (date)28.73088  
Region 4.259 3 2.241  0.089

ellenberg indicator values

Ellenberg N values follow the pattern of change of Axis 1, increasing consistently across all sites during the time period, although any overall difference in scores between regions falls short of significance (Fig. 2f, Table 4). The other Ellenberg indices show more complex patterns of change, with significant within-subject contrasts between survey dates and significant interaction terms with region (L, F, N) or significant interaction terms only (R), implying that the direction and/or magnitude of change differs between regions. Thus the Ellenberg L scores have remained relatively high in the East Anglian plots, decreased in Dorset and Kent, and remained consistently low in Yorkshire; F scores have increased in East Anglia and Kent but remained roughly constant in Dorset and Yorkshire, and R scores have decreased in East Anglia and remained more or less constant in the other regions (Fig. 2c–e, Table 4).

influence of slope and aspect

Both slope and radiation index are significant factors in the ancova of axis 1 (Table 5), while region is not significant at P < 0.05. The direction of the relationship with topography is shown in Fig. 3(b); the average magnitude of change is greatest on flat plots, and least on south-facing slopes; north-facing slopes are intermediate. In other words, the magnitude of change decreases with angle of slope and with radiation index. The significant effect of the original score as a covariate shows that the scores in 2002/03 along this axis are correlated with those from the previous survey. The ancova for axis 2, by contrast, shows a highly significant effect of region, but non-significant effects of either slope or radiation index, or of the original survey scores. It may be surmised that the position of a plot along this axis is influenced by regional or site-specific factors but is more or less independent of topography and the previous state of the vegetation.

Table 5.  Summary of mixed-model ancova statistics for plot DCA axes and Ellenberg indicator values for the 2002–04 survey. In each case the original 1952–53 plot score, angle of slope and radiation index are treated as covariates, region as a fixed factor
 Sum of squaresd.f.F-ratioP
DCA axis 1
Original score 28880 1 7.026  0.010
Region 32596 3 2.643  0.054
Angle of slope 69049 116.798< 0.001
Radiation index 38734 1 9.423  0.003
DCA axis 2
Original score  1524 1 1.116  0.294
Region 55183 313.469< 0.001
Angle of slope  1212 1 0.887  0.349
Radiation index  1337 1 0.979  0.325
Ellenberg L (light)
Original score 0.036 1 2.092  0.152
Region 0.253 3 4.863  0.004
Slope 0.164 1 9.481  0.003
Radiation index 0.253 1 4.684  0.033
Error 1.47385  
Ellenberg F (moisture)
Original score 0.414 1 9.096  0.003
Region 0.038 3 0.277  0.842
Slope 0.923 120.251< 0.001
Radiation index 0.778 117.078< 0.001
Error 3.87385  
Ellenberg R (soil pH)
Original score 0.280 1 5.527  0.021
Region 0.626 3 4.121  0.009
Slope 0.381 1 7.526  0.007
Radiation index 0.345 1 6.819  0.011
Error 4.30485  
Ellenberg N
Original score 0.430 1 0.879  0.351
Region 3.706 3 2.528  0.063
Slope 9.548 119.540< 0.001
Radiation index 6.129 112.544  0.001
Figure 3.

Mean (± 1 SE) change in (a) Ellenberg indices, and (b) DCA axis score between surveys. Plots are classified by topography. Flat: slope angle is less than or equal to 15°; s/w-facing: slope angle is greater than 15°, facing SE, S, SW or W; N-facing: slope angle is greater than 15°, facing NW, N, NE or E.

Significant effects of both slope and radiation index are found for Ellenberg L, N, F and R scores (Table 5). Figure 3(a) shows that the magnitude of change for L, N and F scores is again lower for steeper angles of slope and for south-facing aspects.


changes in species composition

This study clearly identifies a marked and consistent shift in species composition in unimproved chalk grassland sites across different regions of the English chalk during the second half of the 20th century. The observed shift towards mesotrophic grassland communities and the associated increase in fertility values is similar to the increases in fertility values from long-term studies of other semi-natural vegetation types in the UK (McCollin et al. 2000; Haines-Young et al. 2003; Smart et al. 2003; Smart et al. 2005). However we know of no other study finding evidence of a widespread pattern of change of this magnitude in unimproved calcareous grassland in Britain. The consistency of this effect across regions suggests that the causes of change are widespread; however, the relative importance of different processes are not easily determined from the available data. The survey methodology does, however, allow species trends to be identified at two separate spatial scales, and there are notable differences in responses at the two scales. Species may have colonized or gone extinct in plots (at a sampling scale of 50 m2), and may have increased or decreased in frequency within occupied plots (at a sampling scale of 0.01 m2). At the plot-level, the negative relationship between initial frequency and extinction rate is similar to that observed in calcareous grasslands by Fischer & Stöcklin (1997) in Switzerland, and indicates that, as expected in a fragmented landscape, the risk of extinction is greater for species with initially small populations. Furthermore, the lack of a significant relationship between species fertility index and extinction rate implies that, as in the former study, the risk of extinction at the plot scale for a species is independent of whether it is associated with fertile or infertile habitats. However, in the present study the number of colonization events per species was found to be significantly positively correlated with fertility index, implying that the chances of successful colonization are greater for species associated with more productive habitats. The group of species showing highly significant increases in plot occupancy are predominantly CSR or similar strategists; several are grasses more typical of mesotrophic than calcareous grassland, such as Arrhenatherum elatius, Dactylis glomerata, Cynosurus cristatus, Holcus lanatus and Lolium perenne. These species are common within the agricultural landscapes surrounding chalk grassland fragments. Although these species increased strikingly in plot occupancy, at the subplot scale none of them increased significantly in frequency within occupied plots. This may be attributed in part to a relatively low sample size of plots in which these species were found at both survey dates.

With the notable exception of the grass Danthonia decumbens, a species generally associated with mildly acidic soils which appears to occur much less frequently on chalk sites now than previously, the species that were found at only one survey date tended to be uncommon species, typically found in a single plot on one occasion. As these species could have been overlooked or fallen outside quadrats, it is difficult to draw general conclusions on individual species. It is noted that many of these species have short life-cycles and a more or less ruderal habit, and may have been only transient constituents of the chalk grassland swards (Table 3). In contrast, the species that were present in some plots on both survey dates but showed significant decreases in plot occupancy are predominantly perennial stress-tolerators, and include both matrix-forming grasses such as Festuca ovina, Helictotrichon pratense and Koeleria macrantha, and forb species typical of short chalk turf.

These observed patterns in plot-scale extinction and colonization are thus compatible with population-level processes whereby vegetation change occurs as calcareous grassland specialists are less likely to recolonize plots after extinctions, while common mesotrophic species have a high probability of colonization from elsewhere in the landscape. However, most of the group of stress-tolerant species showing significant decreases in occupancy also decreased significantly in frequency within occupied plots. Significant decreases in frequency have also occurred in the calcicoles Bromopsis erecta, Brachypodium pinnatum, Centaurea scabiosa and Campanula glomerata, all of which have colonized new plots. This decrease in frequency of many of the more abundant chalk grassland perennial species implies that factors operating at spatial scales below that of the local population have put these species at a competitive disadvantage.

The gradient in species composition from calcareous grassland communities to mesotrophic grassland communities identified by DCA axis 1 appears to be analogous to gradients identified by both direct and indirect ordinations of calcareous grasslands at a range of spatial scales (Gittens 1965; Austin 1968; Duckworth et al. 2000; Critchley et al. 2002b). The strong correlation of this axis with weighted plot Ellenberg N values suggests that soil fertility is strongly associated with this vegetation gradient; however, correlations with Ellenberg F, R and L values suggest further possible associations with environmental gradients in soil moisture, pH and sward height. Direct causation is difficult to establish, as within the chalk grassland flora Ellenberg indicator scores and environmental variables are often highly correlated in the field; for example high soil pH is associated with low phosphorus availability (Rorison 1987; Tyler 2002), and low soil moisture with low nutrient uptake (Grime & Curtis 1976; Misra & Tyler 2000). This gradient may be best interpreted as a compound environmental gradient associated with both soil fertility and soil moisture favouring stress-tolerant calcicole species at one extreme and competitive-ruderal species at the other.

effects of slope and aspect

The marked effects of slope and aspect on microclimate and soils in calcareous grasslands (Perring 1959; Geiger 1965; Påhlsson 1974; Rorison et al. 1986b; Amezaga et al. 2004) and on the small-scale distribution of species (Boyko 1947; Perring 1959, 1960; Pigott 1978; Weiss et al. 1988) have been noted in previous studies. Likewise, the effect of low soil phosphorus and moisture content on the resistance of vegetation to invasion by more competitive species has been stressed. Chapman et al. (1989) suggest that low phosphorus adsorption capacity in heathland soils may prevent invasion by scrub; several studies have shown that low available soil phosphorus content is associated with high diversity in grassland (Janssens et al. 1998; Critchley et al. 2002a). Rodwell et al. (1993) predicted that vegetation change due to nitrogen enrichment would be more likely to occur on calcicolous grasslands over calcareous brown earths than those on shallow rendzinas as the latter are lower in phosphorus. However, the nature of an indirect link between topography and the resistance of vegetation to change has been considered less often; Pykäläet al. (2005) have shown that slopes with a greater solar radiation load maintained a greater species richness following the abandonment of grazing. The present study shows that the magnitude of vegetation change in British chalk grassland plots during the second half of the 20th century, as measured by changes in Ellenberg values and indirect ordination, is strongly associated with the slope and aspect of the plot. Plots on sloping ground, and particularly those facing south, showed less change, and maintained a more stress-tolerant and light-demanding flora. These results give support to the two hypotheses proposed here, that steeper slopes are buffered to some extent against invasion by more competitive species, probably due to edaphic factors including low phosphorus availability, and that slopes receiving higher radiation loads are further buffered against invasion by more frequent and severe drought events.


This study provides evidence of vegetation change in British chalk grasslands through the invasion of common competitive species, the local extinction of infrequently occurring species and both local extinction and decrease in frequency of stress-tolerant calcareous grassland specialists. These changes have led to a significant decrease in total species number and a shift towards mesotrophic vegetation communities of lower conservation value. Several processes may have contributed to this change. While patterns of extinction and colonization at the plot-scale are consistent with the effects of decreasing size and increasing isolation of local populations following fragmentation of the habitat, the significant decrease in subplot-scale frequency of the more abundant stress-tolerators implies processes operating at a finer spatial scale than that of the local population, presumably through changes in the environmental conditions within the grassland fragments themselves. Furthermore, there is evidence that these changes are mediated through topographic variations in soil and microclimate.

The evidence presented here does not demonstrate the cause of vegetation change; indeed several factors have certainly influenced the composition of vegetation at these sites over the past 50 years. It is noted that similar widespread increases in fertility indices in other habitats have been ascribed to the effects of atmospheric nitrogen deposition (McCollin et al. 2000; Haines-Young et al. 2003; Smart et al. 2003), and that the observed pattern of vegetation change in chalk grassland is very similar to that predicted to occur under enhanced nitrogen deposition on the calcareous brown earths typical of plateau sites (Rodwell et al. 1993), a pattern consistent with the correlation of vegetation change with topography found in this study. However, similar increases in competitive grasses and decreases in stress-tolerant species may occur through local population processes, and are also known to be long-term vegetation responses to declining rabbit populations after 1954 (Thomas 1960, 1963), succession after abandonment of grazing (Köhler et al. 2001; Pykäläet al. 2005), and intensive grassland management (Fuller 1987; Blackstock 1999). More subtle changes in management over the past half century, such as changes in stocking rates, breed of stock or winter feeding of stock, may also have been locally important drivers of vegetation change at these sites. Detailed studies of individual site management histories may be required to disentangle the relative effects of these factors; however, this information was not available in a quantitative form allowing comparison between sites in this study.

Over the last 50 years there has clearly been a high risk of local species extinctions and loss of conservation value in British chalk grasslands, even at sites where traditional management has maintained a semi-natural character in the vegetation. During this period, conservation of biodiversity has increased as a priority in managing chalk grasslands; meanwhile, nitrogen emissions have peaked and are expected to decline in the future (NEGTAP 2001). It is not clear whether these trends are sufficient to halt or reverse the observed changes in chalk grassland vegetation. Climate change scenarios for the UK suggest an increase in summer drought events in the south-east (Hulme et al. 2002), which, following the findings of this study, might be expected to increase the resistance of grasslands to invasion by competitive species and act to maintain stress-tolerant vegetation in the future. However, preventing further losses may depend on reversing the fragmentation of these habitats through restoration of sites. Many of the remaining chalk grassland sites occur on areas of steeply sloping ground or complex topography which are difficult to reach with agricultural machinery. The restriction of several plant species of calcareous grassland to south and south-west facing slopes towards the northern limit of their range is well known (Perring 1959, 1960; Pigott 1975; Hennenberg & Bruelheide 2003). The greater resistance of the vegetation of these slopes to change should increase the importance of sites with sloping, south-facing or complex topography as a priority for both conservation and ecological restoration.


We are indebted to the late F. Perring for advice and the use of his survey data, and would like to thank two anonymous referees for helpful suggestions for improving this paper. J.J.B. was funded by NERC PhD studentship grant number GT04/1999/TS/0056 and British Ecological Society Small Ecological Project Grant no. 2165.