Recreation of lowland heathland on ex-arable land: assessing the limiting processes on two sites with contrasting soil fertility and pH


Kevin J. Walker, NERC Centre for Ecology and Hydrology, Monks Wood, Abbots Ripton, Huntingdon, Cambridgeshire PE28 2LS, UK (fax +44 1487773467; e-mail


  • 1European heathlands are of high conservation value but have declined as a result of agricultural intensification. Heathland recreation is currently being undertaken on ex-arable land throughout north-west Europe to reverse these losses, but seed limitation, elevated soil pH and fertility and competition from ruderal species have been shown to limit community re-assembly.
  • 2We examined the relative importance of these constraints over 9 years (1994–2003) on two ex-arable sites with contrasting soil fertility and pH. Recreation treatments comprised seed addition (including Calluna vulgaris) combined with (i) applications of elemental sulphur (S) at two rates, to reduce the pH to 4·5 and 5·5, respectively; (ii) sowing a nurse crop, to facilitate the establishment of sown species; and (iii) topsoil removal, to deplete nutrients. The importance of sowing time was assessed by (iv) comparing autumn with spring seed addition.
  • 3By 2003 the pH of the S-addition plots had stabilized close to pH 4·5 and 5·5, but levels of extractable phosphorus (P) had increased dramatically at both sites. Topsoil removal reduced soil P and organic matter but increased pH because of exposure of underlying mineral horizons. These conditions were unsuitable for the establishment of Calluna vulgaris.
  • 4Applications of S (3–6 t S ha−1) were sufficient to recreate Calluna vulgaris-dominated heathland at the less productive site. Acid grassland developed on all other treatments, including S-amended plots at the more productive site.
  • 5The establishment of desirable species was not significantly enhanced by sowing seed mixtures with a nurse crop or in the spring, whereas S addition and topsoil removal reduced the abundance of potentially competitive species.
  • 6Synthesis and applications. The application of S can be used to create conditions suitable for the re-assembly of Calluna vulgaris-dominated heathland on unproductive ex-arable soils in north-west Europe. However, interventionist approaches such as S addition may not be practical for large-scale heathland recreation schemes. Competition with more nutrient-demanding species will limit the establishment of slow-growing heathland species under more fertile conditions, even where the pH has been reduced to appropriate levels. Therefore the recreation of acid grassland by seed addition alone is likely to be a more realistic target for the majority of agricultural sites included in agri-environment schemes in north-west Europe.


The extent and quality of lowland heathland has declined dramatically throughout north-west Europe in recent centuries (Gimingham 1972). In the UK this has led to heathland recreation on former agricultural land as part of agri-environment schemes (HMSO 1995). However, recent research suggests that changes brought about by intensive agriculture will make community re-assembly difficult to achieve on ex-arable sites (Walker et al. 2004). Arable soils typically have higher soil fertility and pH than heathlands because of repeated applications of fertilizers and lime (Pywell, Webb & Putwain 1994; Aerts et al. 1995). Soil pH in particular is usually too high for the establishment of ericaceous dwarf shrubs such as Calluna vulgaris, which is usually confined to soils with pH 3·4–6·5 (Clarke 1997). Frequent ploughing can also impoverish heathland seed banks because of repeated soil inversion and aeration (Pywell, Putwain & Webb 1997), especially where fragmentation has reduced the number of potential seed sources in the surrounding landscape (Piessens, Honnay & Hermy 2005). In most cases these are replaced by ruderal species, which can inhibit the establishment of slow-growing heathland species during the early stages of recreation (Owen & Marrs 2000a). Consequently, natural recolonization is unlikely to take place unless heathland seed banks are long-lived or potential seed sources are locally abundant (Smith, Webb & Clarke 1991).

On ex-arable soils pH has been reduced by applications of elemental sulphur (S) (Owen et al. 1999) and acidic peat (Dunsford, Free & Davy 1998); acidic plant materials (e.g. pine chippings and bracken litter) have generally proved less effective (Owen et al. 1999; Allison & Ausden 2004; Lawson et al. 2004). Techniques shown to reduce nutrient availability [primarily levels of extractable phosphorus (P)] have included shallow cultivation (Smith, Webb & Clarke 1991), deep ploughing (Allison & Ausden 2004), topsoil removal (Aerts et al. 1995; de Graaf et al. 1998; Allison & Ausden 2004) and applications of S (Owen et al. 1999). At grassland and heathland sites, P and soil pH have been reduced by additions of iron and aluminium sulphate (Tibbett & Diaz 2005). No appreciable reductions have been achieved by arable cropping (Marrs et al. 1998; McCrea, Trueman & Fullen 2001).

Competition with tall, nutrient-demanding species has been shown to inhibit the establishment of Calluna vulgaris on ex-arable soils (Dunsford, Free & Davy 1998; Owen & Marrs 2000a). However, established plants can act as ‘nurse crops’ by protecting slow-growing seedlings from the harsh conditions encountered on bare agricultural soils. Historically, under-sowing grasses in spring barley has been used to establish grass leys on sandy soils susceptible to wind erosion and, more recently, annual grasses (e.g. Lolium multiflorum) have been used to facilitate the recreation of species-rich grassland (Pywell et al. 2002). Sowing in the spring may also reduce seedling mortality for slow-growing heathland species that are susceptible to winter frosts (e.g. Calluna vulgaris).

In this study we evaluated the effectiveness of techniques to recreate Calluna vulgaris-dominated heathland and acid grassland communities on ex-arable soils. Treatments were designed to test the following hypotheses: the re-assembly of heathland communities conforming to British types (i.e. Rodwell 1991–2000) will be facilitated by (i) sowing seed of desirable species, (ii) reducing soil pH and fertility and (iii) amelioration of extreme abiotic and biotic conditions during the establishment phase. The study was carried out over 9 years (1994–2003) at two sites with contrasting soil pH and fertility. It therefore provides a unique assessment of the importance of initial site conditions and the long-term sustainability of heathland recreation treatments.

Materials and methods

experimental sites

The two experimental sites were located on sandy soils in the Breckland region of East Anglia, UK. Both fields had a long history of cereal cropping. The field at Euston (52°36′N 0°79′E) had been ploughed-up from heathland in the 1940s, whereas the field at Honnington (52°33′N 0°81′E), 3 km to the south, had been under arable cultivation since at least 1900. As a consequence, soil pH and concentrations of NPK were significantly higher at Honnington (see Appendix S1 in the supplementary material). In the decade prior to recreation, both fields had been used to grow cereals and root crops with unspecified amounts of fertilizer. Following arable cultivation in 1993, both fields were entered into agri-environment schemes and sown with seed mixtures; at Euston these included two ‘turf grasses’ (Festuca brevipila and Agrostis castellana) and at Honnington these included Lolium multiflorum, Lolium perenne and Trifolium repens. Vegetation composition was therefore markedly different at the start of the experiment: Euston most closely resembled an Agrostis capillaris acid grassland and Honnington a species-poor Lolium perenne ley (see Appendix S2 in the supplementary material).

targets for recreation

The British National Vegetation Classification (NVC) was used to derive target heathland communities that were appropriate to the location, soil and climate of the experimental sites (Rodwell 1991–2000). These included the typical heathland community Calluna vulgaris–Festuca ovina, Teucrium scorodonia subcommunity (H1c), of infertile, sandy soils in eastern England, and species-rich subcommunities of the Festuca ovina–Agrostis capillaris–Rumex acetosella acid grassland (U1c and U1d). Both are related to the Genisto–Callunion and Plantagini–Festucion ovinae of central Europe and occur as mosaics in eastern England, where grazing and disturbance reduces the dominance of ericaceous subshrubs.

recreation treatments

In 1994, a randomized experiment with four replicate blocks was set up at both sites, with 11 treatments arranged in 20 × 4-m plots. The potential for colonization of heathland species from the seed bank and seed rain was assessed by autumn cultivation followed by natural regeneration (treatment 1) and natural regeneration with the vegetation left intact (treatment 2). These were compared with the autumn addition of seed of heathland species (treatment 3) following cultivation. A seed mixture designed to create the composition and structure of the target NVC communities was made up of British provenance seed of five native grasses and eight herbs, with sowing rates proportional to their abundance in the target communities (see Appendix S3 in the supplementary material). The seed mixture was sown on rotavated plots at a rate of 34 kg ha−1, with a 4:1 grass to herb ratio. In the following spring, heathland topsoil/litter containing Calluna vulgaris (70·4 ± 9·6 seeds m−2) collected from Dunwich Heath, Suffolk (52°25′N 1°62′E), UK, was applied at a rate of 750 kg ha−1.

In order to facilitate the establishment of heathland species, a nurse crop of Italian ryegrass Lolium multiflorum was sown with the seed mixture in September 1994 at a rate of 30 kg ha−1 (treatment 4). This treatment was combined multifactorially with applications of elemental S at two rates to reduce soil pH (treatments 5–8). Application rates were based on the assumption that 2·7 t ha−1 S were required to reduce soil pH by a single unit (MAFF 1983). At Euston, where the soil reaction was less than pH 6, S was applied at 3 and 6 t ha−1 in order to achieve the target levels of pH 5·5 without and with a nurse crop treatments 5 and 6) and 4·5 (treatments 7 and 8), respectively. Target values were the same at Honnington, although the presence of free CaCO3 meant that the application rates were much higher (9 and 18 t ha−1, respectively). Elemental S was applied by hand and rotavated to a depth of 20 cm. Because of practical difficulties, treatment 7 was not sown until 1995 and was therefore excluded from the analyses reported in this paper.

To assess the interaction between the season of sowing and cultivation, the seed mixture was also sown in the following spring (March 1995) with a spring-germinating nurse crop (spring barley Hordeum vulgare cv. ‘Chariot’) sown at a rate of 180 kg ha−1 (treatment 9). This was compared with autumn cultivation with a herbicide application to reduce weed competition (treatment 10).

Finally, topsoil removal to reduce soil fertility was investigated by removing the top 35 cm of soil at Euston and 45 cm at Honnington in September 1994 (treatment 11). Both sites were fenced to exclude grazing animals and the vegetation was cut to a height of 7·5 cm in the late summer.

ecological assessments

In April of each year, five 50 × 50-cm quadrats (divided into 25 10 × 10-cm grid cells) were placed at random within the core 2 × 18-m area of each plot. Species frequency (out of 25) and percentage cover were recorded in each quadrat. At Euston Agrostis castellana/capillaris and Festuca ovina/brevipila were difficult to differentiate and were therefore treated as single-species aggregates. Nomenclature for vascular plants follows Stace (1997). Topsoil samples (0–15 cm) were collected from each plot and analysed for pH and soil nutrients using standard analytical techniques (MAFF 1986). In this paper we report the changes in soil pH and Olsen extractable P (mg L−1).

statistical analysis

Five response variables were calculated to assess the effectiveness of the recreation techniques: the number of sown and unsown species, percentage cover of Calluna vulgaris and the similarity between the recreated and target vegetation assemblages (U1c/d and H1c) calculated using the Tablefit computer program (Hill 1996).

The effect of recreation treatment, year and interactions between treatment × year were examined at both sites separately for each response variable using analysis of variance (anova) with repeated measures (SAS for Windows Version 8, 1990). The anova model included block and treatment as random and fixed factors, respectively, and had repeated measures by including data from 9 consecutive years (1994–2003). In addition, factorial combinations of low S addition and sowing a nurse crop (treatments 3–6) were examined for response variables and the percentage cover of individual species. The effects of recreation treatments intended to reduce soil pH and P (treatments 1, 5, 8 and 11) were examined for both sites separately using an anova model that included block and treatment with repeated measures (1995, 1999 and 2003). Finally, a two-sampled t-test was performed to test the significance of the differences between sites in 2003.

Univariate and multivariate repeated-measure anovas (Maxwell & Delaney 1990) were carried out using proc glm in SAS (1990) and both gave identical qualitative results. Within the univariate analyses, ɛ-values for adjustment of degrees of freedom according to the amount by which the population covariance matrix departs from homogeneity were calculated using the Geisser–Greenhouse method and the less conservative Huynh–Feldt method (Maxwell & Delaney 1990). Both gave the same qualitative results, and so the results from the Geisser–Greenhouse adjustment are reported here. Arcsin transformation of percentage cover values was undertaken to achieve normality of residuals as required, and Tukey pairwise comparisons were used to determine the significance of differences between means.


effects of recreation treatments on vegetation composition

At both sites, significantly more sown species were recorded on the seed addition (treatment 3) than natural regeneration plots throughout the course of the experiment (Table 1; see Appendix S4 in the supplementary material). In comparison, more unsown species persisted on natural regeneration plots (treatment 2), particularly on the uncultivated plots at Honnington, which had significantly more unsown species than Euston (Table 2). The combined effects of seed addition and other recreations treatments are presented below.

Table 1.  Results of the repeated-measures anova showing the effects of recreation treatments on the number of sown and unsown species, percentage cover of Calluna vulgaris and similarity between the recreated vegetation and target Agrostis capillaris acid grassland (U1c/d) and Calluna vulgaris (H1c) heathland in 1995 and 2003. Means with the same letter are not significantly different from one another. ***P < 0·001. Data for all years are given in Appendix S4 (see the supplementary material)
Year and treatmentEustonHonnington
SownUnsown%Calluna% fit U1% fit H1SownUnsown%Calluna% fit U1% fit H1
  1. Treatment 7 excluded (see text).

1. Autumn cultivation + natural regeneration2·2b2·2de0·051·3a19·3a0·2b2·6b0·00·0c0·0b
2. Natural regeneration1·4b9·4a0·0 8·8c0·5bc0·1b8·0a0·00·5c0·0b
3. Autumn seed/cultivation5·4a8·1ab0·033·0ab5·8b6·9a7·4a0·039·5a7·8ab
4. Seed + nurse crop6·0a10·2a0·024·8b5·0bc5·9a7·2a0·023·5a6·5ab
5. Seed + low S2·1b0·6e0·0 6·3c0·5bc4·8a4·3b0·017·0ab2·5ab
6. Seed + low S + nurse crop1·9b0·5e0·0 9·5c3·5bc5·2a2·8b0·022·0a8·3a
8. Seed + high S + nurse crop0·4b0·1e0·0 4·3c0·3bc4·6a2·0b0·016·3ab8·5a
9. Spring seed/cultivation2·2b3·5dc0·0 9·0c3·8c1·0b3·4b0·02·3c2·8ab
10. Autumn cultivation + spring seed2·0b5·9bc0·0 7·5c2·5bc1·0b3·7b0·03·5bc4·0ab
11. Topsoil removal6·2a3·2d0·024·8b3·8bc5·5a2·0b0·025·8a4·0ab
1. Autumn cultivation + natural regeneration3·1c2·8ab0·0c62·8ab19·5bc3·4b3·7ab0·052·5ab13·0
2. Natural regeneration2·8c2·5abc0·0c52·8ab14·5c4·0b4·5a0·057·5ab12·0
3. Autumn seed/cultivation5·9ab2·2abc0·0c76·5a14·5c6·8a1·3cde0·080·0a12·3
4. Seed + nurse crop6·3ab2·4abc0·0c80·3a15·0bc7·2a1·6cde0·083·0a11·3
5. Seed + low S4·0bc0·8bc12·9bc34·5bc24·8bc7·4a1·0cde0·063·0ab10·0
6. Seed + low S + nurse crop4·2bc0·6bc22·7b38·3bc37·0ab7·8a0·3e0·072·0ab15·8
8. Seed + high S + nurse crop3·3c0·2c50·2a15·0c51·3a3·7b0·3e0·040·8b16·8
9. Spring seed/cultivation6·2ab3·4a0·0c65·3ab12·3c6·8a2·7bc0·077·5a12·8
10. Autumn cultivation + spring seed6·0ab2·5abc0·0c64·8ab12·3c7·0a2·4bcd0·083·5a13·0
11. Topsoil removal6·9a2·4abc0·0c81·8a14·0c6·9a0·7de0·078·0a15·8
anova F-value
Year × treatment2·7***8·4***6·0*** 3·2***5·2***5·8***5·6***5·8***4·7***
Table 2.  The significance of differences between sites for soil and vegetation variables in 2003 calculated using a two-sampled t-test. †P < 0·10; *P < 0·05; **P < 0·01; ***P < 0·001
TreatmentpHPSownUnsown% fit to U1% fit to H1
1. Autumn cultivation + natural regeneration−9·1**−3·3*−0·4 NS−0·9 NS1·2 NS2·3*
2. Natural regeneration−13·7**−2·5−1·7 NS−3·0*−0·4 NS0·5 NS
3. Autumn seed/cultivation−28·7***−4·4*−1·6 NS1·3 NS−0·6 NS2·0 NS
4. Seed + nurse crop−3·2*−5·9**−1·3 NS0·9 NS−0·5 NS2·4
5. Seed + low S−1·5 NS−3·8*−3·2*−0·5 NS−1·8 NS2·4
6. Seed + low S + nurse crop−1·1 NS−7·1***−3·2*0·5 NS−2·8*2·4
8. Seed + high S + nurse crop−0·3 NS−4·8*−0·8 NS−0·3 NS−2·46·0**
9. Spring seed/cultivation−2·9−3·6*−0·9 NS0·5 NS−3·1*−0·2 NS
10. Autumn cultivation + spring seed−5·2*−3·4*−1·2 NS< 0·1 NS−2·4−0·4 NS
11. Topsoil removal−6·2**−1·3 NS< 0·1 NS2·8−0·6 NS−1·5 NS
All treatments−5·5***−4·3***−3·0**0·3 NS−2·4*3·5***

sowing a nurse crop

There were few significant effects of sowing a nurse crop at either site (Table 1). At Honnington, plots sown with a nurse crop were more similar to the target heathland (percentage fit to H1), particularly when combined with S addition (Table 3). The presence of a nurse crop also reduced the cover of the sown grass Anthoxanthum odoratum at Euston (Table 4) and the unsown grass Lolium perenne at Honnington (Table 5), whereas the unsown herb Geranium molle was more abundant on plots sown with a nurse crop during the initial phase of the experiment (Table 5).

Table 3.  Results of the repeated-measures anova showing the effects of factorial combinations of sulphur addition and sowing a nurse crop on vegetation response variables. Mean values for all 9 years are presented. NS, no significant difference; *P < 0·05; **P < 0·01; ***P < 0·001. Significant trends are indicated as follows: declined (–); increased (+); declined/increased in a non-linear fashion (≅)
 Nurse crop (d.f. = 1, 9)S addition (d.f. = 1, 9)anova F-value
NoYesNurseNurse × yearNoYesSS × yearNurse × SYearTrend
Sown species4·14·20·2 NS0·6 NS5·43·0197·9***3·5**< 0·1 NS13·2***+
Unsown species1·51·51·7 NS1·0 NS3·70·5140·7***45·0***2·0 NS44·6***
% fit to U139·937·90·6 NS1·9 NS57·620·2213·7***9·0***< 0·1 NS53·3***+
% fit to H112·613·00·4 NS1·0 NS12·213·33·3 NS4·9**2·0 NS14·1***+
Sown species6·15·80·4 NS1·1 NS6·25·71·4 NS4·7**0·5 NS21·8***+
Unsown species1·51·50·2 NS1·0 NS2·10·964·9***11·4***6·6*75·6***
% fit to U163·962·90·2 NS1·1 NS66·960·01·9 NS1·8 NS< 0·1 NS39·5***+
% fit to H112·814·59·2*2·1 NS12·814·63·5 NS2·6 NS14·6**28·4***
Table 4.  Results of the repeated-measures anova showing the effects of factorial combinations of sulphur addition and sowing a nurse crop on the percentage cover of sown and unsown species (present in > 10% of quadrats) at Euston. Mean values for all 9 years are presented. NS, no significant difference; *P < 0·05; **P < 0·01; ***P < 0·001. Overall trends indicate whether species declined (–), increased (+) or declined/increased in a non-linear fashion (≅) during the course of the experiment
 Nurse crop (d.f. = 1, 9)S addition (d.f. = 1, 9)anova F-value
NoYesNurseNurse × yearNoYesSS × yearNurse × SYearTrend
Sown grasses
Agrostis capillaris agg.41·337·50·4 NS0·5 NS36·442·40·9 NS13·1***< 0·1 NS11·9***+
Anthoxanthum odoratum0·60·36·0*0·4 NS0·60·213·9**3·0 NS4·5 NS8·5***+
Deschampsia flexuosa5·45·00·2 NS1·6 NS0·110·391·4***5·7**< 0·1 NS3·6*+
Festuca ovina agg.14·714·10·1 NS1·6 NS24·93·9146·0***9·9***0·3 NS29·9***+
Koeleria macrantha1·31·30·01 NS1·4 NS2·30·238·0***10·0***0·4 NS18·0***+
Sown herbs
Achillea millefolium2·53·70·0 NS0·6 NS5·50·715·3**1·7 NS0·3 NS5·9***+
Erodium cicutarium0·00·11·0 NS0·4 NS0·10·04·6 NS2·0 NS< 0·1 NS4·9*
Calluna vulgaris2·83·80·4 NS0·8 NS0·06·617·6**6·6**0·4 NS6·6**+
Galium verum0·60·40·03 NS1·3 NS1·00·112·2**6·6**0·7 NS11·1***+
Lotus corniculatus0·20·10·01 NS0·4 NS0·30·06·1 NS3·9*< 0·1 NS3·9*
Plantago lanceolata3·63·30·2 NS2·2 NS6·60·388·0***13·7***< 0·1 NS21·3***+
Rumex acetosella0·20·40·2 NS0·4 NS0·20·42·0 NS0·8 NS< 0·1 NS23·2***
Unsown species
Arenaria serpyllifolia< 0·1< 0·1< 0·1 NS0·2 NS< 0·10·010·6**5·4*< 0·1 NS5·4*
Crepis capillaris0·30·81·7 NS3·7*1·10·051·8***13·2***2·7 NS15·6***
Geranium molle0·30·20·6 NS0·5 NS0·4< 0·128·1***15·4***0·3 NS28·0***
Ornithopus perpusillus0·10·060·3 NS0·6 NS0·1< 0·141·8***10·0***0·7 NS17·3***
Senecio jacobaea0·40·41·1 NS0·6 NS0·60·34·0 NS3·7*0·1 NS14·1***+
Stellaria media0·70·33·5 NS3·9*0·80·170·8***46·0***4·7 NS91·3***
Taraxacum officinale0·20·2< 0·1 NS0·4 NS0·3< 0·119·6**1·8 NS0·3 NS2·4 NS 
Tripleurospermum inodorum0·70·40·9 NS2·1 NS1·1< 0·135·0***32·6***< 0·1 NS45·8***
Table 5.  Results of the repeated-measures anova showing the effects of factorial combinations of sulphur addition and sowing a nurse crop on the percentage cover of sown and unsown species (present in > 10% of quadrats) at Honnington. Mean values for all 9 years are presented. NS, no significant difference; *P < 0·05; **P < 0·01; ***P < 0·001. Overall trends indicate whether species declined (–), increased (+),or declined/increased in a non-linear fashion (≅) during the course of the experiment
 Nurse crop (d.f. = 1, 9)S addition (d.f. = 1, 9)anova F-value
NoYesNurseNurse ×  yearNoYesSS ×  yearNurse ×  SYearTrend
Sown grasses
Agrostis capillaris8·06·01·4 NS2·5 NS2·211·847·1***10·9***1·0 NS9·8***+
Anthoxanthum odoratum0·30·61·4 NS0·6 NS0·40·40·3 NS0·1 NS2·0 NS8·4***+
Deschampsia flexuosa3·72·9< 0·1 NS0·6 NS0·56·16·7*4·9*0·0 NS6·9*+
Festuca ovina33·435·10·3 NS1·1 NS43·025·525·5***0·4 NS< 0·1 NS13·6***+
Koeleria macrantha2·82·43·2 NS0·4 NS4·90·3188·6***10·1***1·2 NS16·2***+
Sown herbs
Achillea millefolium9·17·50·8 NS1·2 NS9·37·22·5 NS1·3 NS0·918·5***+
Erodium cicutarium< 0·1< 0·10·9 NS0·7 NS< 0·1< 0·10·9 NS0·7 NS1·0 NS2·1 NS 
Galium verum0·70·70·1 NS0·3 NS0·90·50·4 NS0·9 NS1·2 NS11·8*** 
Lotus corniculatus6·65·10·7 NS0·3 NS7·44·42·5 NS3·5*0·8 NS31·5***+
Plantago lanceolata6·66·20·4 NS0·6 NS8·64·210·1*2·9*3·3 NS17·7***+
Rumex acetosella1·01·93·7 NS1·1 NS0·12·766·4***10·3***6·0*11·1***+
Unsown species
Arenaria serpyllifolia0·30·11·5 NS1·5 NS0·4< 0·113·4**7·1*0·5 NS23·8***
Capsella bursa-pastoris0·70·42·8 NS2·8 NS0·90·218·9***13·2**0·6 NS70·7***
Geranium molle0·0< 0·16·4*1·8 NS< 0·1< 0·12·4 NS1·9 NS1·3 NS2·6 NS 
Lolium perenne0·50·111·2**3·9*0·30·30·0 NS1·5 NS0·8 NS13·4***
Stellaria media0·3< 0·11·5 NS1·1 NS0·4< 0·14·8 NS3·3 NS1·2 NS10·7**
Trifolium repens0·40·60·2 NS0·4 NS0·50·41·0 NS1·7 NS4·9 NS6·5**
Tripleurospermum inodorum0·20·10·1 NS0·7 NS0·3< 0·116·7**11·3**0·6 NS32·2***

sulphur addition

At Euston, the addition of S had a strong negative effect on the establishment of sown species, particularly on the high S-addition plots, although these differences became less marked through time (Table 1). These plots, and in particular the high S-addition plots, were typically co-dominated by Agrostis capillaris and the calcifuges Deschampsia flexuosa and Calluna vulgaris (Fig. 1a,b). In comparison, Calluna vulgaris failed to establish on any of the plots at Honnington or unamended plots at Euston (Tables 1 and 3). Generalist herbs such as Achillea millefolium and Plantago lanceolata were significantly more abundant on unamended plots at Euston (Table 4). Overall, significantly more sown species established on the low S-addition plots at Honnington (Table 2). However, the differences between the S-addition and unamended plots were less marked, with few significant differences in the diversity of sown species (Table 1) or cover of generalists (e.g. Achillea millefolium and Lotus corniculatus). Nevertheless, some species were significantly less abundant on the low S-addition plots (e.g. Festuca ovina, Koeleria macrantha and Plantago lanceolata) whereas some calcifuge species (e.g. Deschampsia flexuosa and Rumex acetosella) were significantly more abundant (Table 5).

Figure 1.

Changes in the cover of (a) Calluna vulgaris and (b) Deschampsia flexuosa on the low and high sulphur (S)-addition plots (treatments 5 and 8) at Euston.

Although there was a marked decline in the number of unsown species on all plots during the experiment, the addition of S had a highly significant negative effect at both sites (Tables 1 and 3). This was more evident at Euston, where seven of the eight most frequent unsown species were significantly less abundant on the S-addition plots (e.g. Crepis capillaris, Ornithopus perpusillus and Stellaria media) (Table 4). Only three of the most frequent unsown species were significantly less abundant on S plots at Honnington (Arenaria serpyllifolia, Capsella bursa-pastoris and Tripleurospermum inodorum) (Table 5). At both sites, the number of unsown species was consistently lower on the higher S-addition plots, although never significantly so (Table 1; see Appendix S4 in the supplementary material).

sowing time

There were few significant effects of sowing time at either site: spring-sown plots had fewer sown and unsown species in the first year, as many species had not germinated before the plots were monitored in the spring (Table 1). From 1996 onwards, however, the number of unsown species was generally higher, albeit not significantly so, on spring-sown plots at both sites, but more so at Honnington (Table 1; see Appendix S4 in the supplementary material). In general, the number of unsown species was higher on the plots cultivated in the spring (treatment 9) as opposed to the autumn (treatment 10), although again the differences were not significant. Consequently, the number of sown species was higher on the autumn-cultivated plots in some years, but not significantly so and only at the Honnington site.

topsoil removal

The effects of topsoil removal were more site-specific than most other recreation treatments (Table 1). At Euston, topsoil removal resulted in the greatest number of sown species in all years, although not significantly more than either autumn- or spring-seed addition plots. This was not the case at Honnington, where there were few significant differences between the diversity of sown species on any of the seed-addition plots. At both sites topsoil removal significantly reduced the number of unsown species during the initial phase of the experiment. Nevertheless, these differences became less marked through time as the numbers of unsown species declined on the seed-addition plots.

effects of recreation treatments on soil ph and phosphorus

The addition of S reduced the soil pH to c. 3·5 and c. 2·5 on the low and high S-addition plots, respectively, although the pH had increased to close to the target levels (pH 4·5 and 5·5) at both sites by 2003 (Fig. 2a,b). The only exception was on the low S-addition plots at Euston, where the pH was lower than expected (pH 4·7 ± 0·2) and not significantly different from the high S-addition plots (pH 4·2 ± 0·1). Levels of soil P showed the opposite trend, increasing significantly on the S-addition plots, particularly at the more productive (Honnington) site (Fig. 2c,d). Soil P declined after the first year at both sites, but only significantly so at Euston (F = 20·9, P= 0·0002). Topsoil removal reduced concentrations of soil P, although levels were not significantly lower than natural regeneration plots (Fig. 2c,d). However, the removal of topsoil resulted in an increase in soil pH at both sites, although only significantly so at the less productive (Euston) site (Fig. 2a,b). By the end of the experiment, levels of soil pH and P were still significantly higher on all the unamended treatments at Honnington, whereas there were no significant differences between the pH of the S-addition and P levels of the topsoil-removed plots between the two sites (Table 2).

Figure 2.

Effects of sulphur addition (treatments 5 and 8) and topsoil removal (treatment 11) on soil pH and Olsen extractable phosphorus (ext. P mg L−1) compared with natural regeneration (treatment 1). For each year means (± 1 SE) with the same letter are not significantly different from one another.

recreation of target communities

With the exception of the high S-addition plots at Euston, the vegetation at both sites showed greater similarity to Agrostis capillaris acid grassland (U1c/d) than Calluna vulgaris heathland (H1c) (Table 1). At both sites there were few significant differences between the unamended seed-addition treatments with topsoil removal and seed-addition (autumn and spring) plots showing the greatest similarity to the target acid grassland (U1c/d). In contrast, the similarity was much lower on S-addition plots. Nevertheless, seed addition treatments at Honnington showed significantly greater similarity than at Euston, particularly on the S-addition and spring-sown plots (Table 2). Conversely, similarity with the target heathland (H1c) was significantly higher on the S-addition plots at Euston (Table 2). All other treatments had very poor fits to H1c (generally < 15%), and by 2003 there were no significant differences at Honnington or between unamended treatments at Euston (Table 1).


is the re-assembly of heathland communities seed limited?

The vegetation communities resulting from natural regeneration at both sites had significantly fewer sown heathland species than treatments sown with the NVC seed mixture. Sown species that colonized the natural regeneration plots were either present in the seed bank (e.g. Erodium cicutarium and Rumex acetosella), had already been sown (e.g. Agrostis capillaris and Festuca ovina) or colonized from adjacent plots (e.g. Achillea millefolium). This confirms that the recreation of heathland communities is seed-limited and that there is little potential for natural colonization from the seed bank (Pywell, Putwain & Webb 1997) or the surrounding landscape (Piessens, Honnay & Hermy 2005). Recreation on ex-arable land will therefore be dependent on the introduction of propagules as commercial seed mixtures or within plant materials harvested from semi-natural sites (Pywell, Webb & Putwain 1995).

Seed addition was sufficient to establish generalist species at both sites. Unamended plots were co-dominated by Agrostis capillaris and/or Festuca ovina and generalist herbs able to compete (e.g. Achillea millefolium, Lotus corniculatus and Galium verum) or regenerate rapidly on disturbed microsites within closed swards (e.g. Plantago lanceolata). In comparison, slow-growing and/or calcifuge species (Calluna vulgaris, Deschampsia flexuosa, Galium saxatile and Thymus polytrichus) failed to establish on unamended plots, even at Euston where Calluna vulgaris was locally abundant and Galium saxatile and Thymus polytrichus occurred in the surrounding grassland (K. J. Walker, unpublished data). Competition with Agrostis capillaris is known to suppress the growth of Calluna vulgaris during heathland recreation (Dunsford, Free & Davy 1998) and this was clearly one of the key limiting factors on the establishment of slow-growing heathland species in this study.

are high levels of soil ph and p a constraint to community re-assembly?

This experiment demonstrated that applications of S can reduce the pH of agricultural soils, thereby confirming the findings of previous studies (Owen & Marrs 2000b; Lawson et al. 2004): the pH declined to pH < 3·0 on the high S-addition plots before stabilizing close to the target levels of pH 4·5 and 5·5 (Fig. 2a,b). In contrast with previous studies (Owen & Marrs 2000b), these changes were associated with a significant increase in the availability of soil P as calcium phosphate complexes were broken down (Fig. 2c,d). In this study, applications rates of 3–6 t S ha−1 were sufficient to restore the target heathland at the less productive (Euston) site, whereas no heathland regeneration took place at the higher rates at Honnington (9–18 t S ha−1). These findings are in broad agreement with earlier studies where 4 t S ha−1 was the most successful rate for the establishment of Calluna vulgaris (Owen & Marrs 2000a). In the same study, very poor establishment was recorded where rates exceeded 8 t S ha−1.

In this study, the recreation of heathland (H1c) was only successful on the acidified plots at Euston, where soil pH was similar to Honnington but soil fertility, in particular extractable P, was significantly lower (Table 2). Therefore soil fertility appears to have been a limiting factor at the more productive (Honnington) site. Lawson et al. (2004) showed that Calluna vulgaris was able to establish and grow well on nutrient-rich arable soils in the absence of other species: its failure to establish on the more fertile S-addition plots at Honnington can therefore be attributed to competition from sown grasses (Agrostis capillaris and Festuca ovina), which dominated many of the plots during the establishment phase. Germination of Calluna vulgaris is known to be inhibited by shade (Gimingham 1972) and, even if established, slow-growing seedlings are unlikely to compete well with the growth of productive grasses. Similar effects have been found on plots acidified with acidic peat in Suffolk (Dunsford, Free & Davy 1998), whereas tall ruderal species have limited Calluna vulgaris establishment at other studies (Owen & Marrs 2000a). Alternatively, the very low pH values experienced on the high S-addition plots may account for the lack of establishment at Honnington. The germination of Calluna vulgaris is pH dependent, with very little germination occurring at pH < 3 (Dunsford, Free & Davy 1998; Owen & Marrs 2000a) because of the toxic effects of ions produced during S oxidation (Lawson et al. 2004). In this study, the highest S rate (18 t ha−1) produced a very acid soil (pH < 2·5) that was probably unsuitable for the germination of Calluna vulgaris in the first year. Furthermore, the bare patches created may have also reduced the abundance of heathland species, such as Calluna vulgaris, which require high humidity for the establishment of seedlings (de Hullu & Gimimgham 1984). Soil fertility and pH were less important at the less productive (Euston) site. Greater establishment of strict calcifuges (Calluna vulgaris and Deschampsia flexuosa) occurred on the high S-addition plots despite the fact that they were significantly more fertile than the low S-addition plots during the initial phase of the experiment. Therefore the application of S appears to have been beneficial to the establishment of Calluna vulgaris by reducing, but not completely removing, the cover of more competitive species (Lawson et al. 2004). This was confirmed by the greater cover of Calluna vulgaris on the high S-addition plots at Euston.

does topsoil removal enhance the assembly of heathland communities?

In this study, topsoil removal was very effective in reducing concentrations of NPK (A. Bhogal, unpublished data). Nevertheless, P status remained relatively high, at 20–25 mg L−1, and soil pH increased to high levels, pH 6·5, because of exposure of underlying mineral-rich horizons (Aerts et al. 1995; de Graaf et al. 1998; Allison & Ausden 2004). Furthermore, topsoil removal created a very shallow, stony soil with low levels of organic matter (A. Bhogal, unpublished data). These conditions were probably unsuitable for the establishment of Calluna vulgaris, which is known to establish poorly on mineral soils with poor moisture retention capacities (Bannister 1964). Conversely, the relatively infertile conditions reduced competition from grasses and promoted the growth of nitrogen-fixing grassland species (e.g. Cladonia spp., Lotus corniculatus and Trifolium arvense). As a consequence, an open and diverse Agrostis capillaris acid grassland developed on the topsoil-removed plots at both sites.

Topsoil removal also reduced the frequency of unsown herbs, particularly during the critical establishment phase. This effect was sustained for 9 years at Honnington but not at Euston because of the more open nature of the grassland and abundance of small annuals (e.g. Crepis capillaris and Ornithopus perpusillus) in the seed bank. This confirms that topsoil removal can reduce competition from more nutrient-demanding species by removing the large seed banks of ruderal species that usually occur in the top 40 cm of former agricultural soils (Pywell, Putwain & Webb 1997).

do nurse crops or spring sowing facilitate the establishment of heathland species?

Sowing a nurse crop in the spring had few significant effects on the number of sown and unsown species recorded, and therefore was of little benefit to the recreation process. This is contrary to the findings from studies on the recreation of species-rich grassland on ex-arable land where a nurse crop of Lolium multiflorum had a negative effect on sown species because of increased competition for resources during the establishment phase (Pywell et al. 2002). More unsown species were recorded on spring- than autumn-sown plots throughout the course of the experiment, particularly on spring-cultivated plots at the more productive site. Therefore, contrary to the aim of spring sowing, this treatment was of little benefit to the recreation process and could potentially have negative effects because of increased competition with ruderal species during the establishment phase.

implications for heathland recreation

The results of this study clearly demonstrate that the recreation of Calluna vulgaris-dominated heathland is feasible on relatively infertile ex-arable soils (extractable P was < 50 mg L−1) with low pH (4·5–5·5). Treatments were less successful at the more productive site, even where soil pH had been reduced to levels suitable for the establishment of Calluna vulgaris. Here, colonization by heathland species was limited by the potentially toxic effects of S addition (9–18 t S ha−1) and competition with productive grasses during the critical establishment phase. In comparison, seed addition to unamended soils was sufficient to recreate the target Agrostis capillaris acid grassland on plots with pH > 5·5 at both sites. These findings suggest that heathland recreation should be targeted at former agricultural sites with a short history of cultivation (Smith, Webb & Clarke 1991), or low soil pH or nutrient levels (Marrs et al. 1998). The recreation of acid grassland communities is likely to be a more realistic and cost-effective alternative on more fertile soils.

Despite expectations, there were few beneficial effects of sowing heathland propagules with a nurse crop or in the spring. In comparison, soil acidification and topsoil removal were effective in reducing pH and soil fertility, respectively, as well as limiting the growth of potentially competitive species. However, these interventionist approaches have a number of disadvantages. S addition may not be desirable on or close to sensitive sites because of potentially toxic effects on fauna and flora. It is also relatively expensive, difficult to apply and can provide considerable variability in pH within single plots (Owen & Marrs 2000a). As shown in this study, topsoil removal can be detrimental to the establishment of Calluna vulgaris because of the low moisture retention of the soil and increase in pH as a result of exposure of underlying mineral horizons. Furthermore, even small-scale treatments would require the removal of large amounts of soil, potentially damaging underlying archaeological features or the agricultural value of the land. Therefore interventionist techniques such as S addition and topsoil removal are unlikely to be practical for large-scale heathland recreation schemes. Recent research suggests that the removal of conifer plantations from afforested heathland may provide a much more reliable and cost-effective alternative (Walker et al. 2004).


This study was funded by the Department for Environment, Food and Rural Affairs (BD0801, BD0802 and BD1506). The authors are grateful to Euston Estates for permission to undertake the experiment, and David Barratt, Claire Carvell, Rachel Hirst, Niall Moore, Bill Nickson and Phil Lambdon for assistance with fieldwork. We would also like to thank James Bullock for statistical advice and two anonymous referees for their valuable comments.