A comparative study of the seedbanks of heathland and successional habitats in Dorset, Southern England



1 Many areas of lowland heaths are being lost due to invasion by Betula spp., Pinus sylvestris, Pteridium aquilinum, Rhododendron ponticum and Ulex europaeus. One of the factors influencing the success of restoration of heathland on such sites will be the content of their viable seedbanks.

2 Ten heathland areas in the Poole Basin area of Dorset, where succession to one or more of the above species had occurred were studied. The viable seedbanks of the successional sites were compared with those of nearby heathland using Canonical Discriminant Analysis.

3 The seedbanks of all the successional stages were significantly different from the seedbank of the heath.

4 The seedbanks from the Pinus sylvestris and Pteridium aquilinum successional stages contained significantly lower numbers of heathland species than did the heathland seedbank, although few non heathland species were present.

5 The seedbanks from the Betula spp., Rhododendron ponticum and Ulex europaeus successional sites contained both significantly lower numbers of heathland species and significantly higher numbers of non heathland species than the heathland seedbank.

6 The results are discussed in relation to the restoration of heathland on successional sites and the use of the seedbank as a source of propagales for the establishment of heathland species.


Lowland heaths are threatened by successional change to other communities (Marrs, Hicks & Fuller 1986; Webb 1990). As Britain has an international obligation to conserve its lowland heaths (Department of Environment 1995a,Department of Environment 1995b) considerable resources are being spent on restoring successional areas to heathland. The success of this restoration work at least partially depends on the content of the viable seedbank of the managed site (Putwain & Gillham 1990) as this will influence the initial floristic composition of the site after disturbance (Hobbs & Gimingham 1984) as well as future floristic developments (Egler 1954).

The soil seedbank is an important part of the plant community as it acts as a potential pool of propagules for regeneration after disturbance (Hodgson & Grime 1992; Pakeman & Hay 1996). The viable seedbank and the vegetation on a site are dynamically linked, although some species may be present in one of these but not the other (Thompson & Grime 1979).

Calluna vulgaris has long been recognised as having a large buried seedbank (Chippendale & Milton 1934). If this buried seedbank can be made to germinate then restoration will be considerably cheaper than if propagules have to be added as seed, either as such or as a component of litter or topsoil (Gimingham 1992; Pywell, Webb & Putwain 1995). Many seedbank studies of heathland species concern the survival of seeds beneath arable/reclaimed land (Pywell, Putwain & Webb 1997) or beneath plantations (Hill & Stevens 1981). Few studies have looked at the way that heathland seedbanks change during succession (Pakeman & Hay 1996). Miles & Young (1980) noted changes in the species composition of the seedbank on moorland during Betula spp. succession, with the buried seed flora slowly changing to one characteristic of a woodland. Similarly Pakeman & Hay (1996) showed that the seedbank of heathland which has been under Pteridium aquilinum for more than 50 years was very small and propagule introduction was thought necessary for successful restoration. These authors concluded that both the size and quality of the viable seedbank were determinants of the success of heathland restoration on successional sites, which it was important to consider when planning restoration schemes.

For successful restoration it is important to consider the presence of both heathland species and non-heathland ones. If the seedbank contains non-heathland species which are more competitive than the heathland ones then some forms of management may allow these more competitive species to dominate and so prevent establishment of heathland species.

On the Dorset heaths there is a range of species whose invasion commonly initiates succession (Mitchell et al. 1997) and it is possible that some successions may have a greater effect on the seedbank than others. Here we aimed to compare the seedbanks of heathland and different successional communities and to use the results from this study to help develop practical conservation management strategies for heathland restoration on sites where succession has occurred. Throughout this paper the term heathland restoration means the establishment of a dwarf shrub community dominated by Calluna vulgaris.

Materials and Methods


Ten heathland areas within a 20 km radius in the Poole Basin of Dorset (Southern England) were chosen (Table 1). At each area there were still areas of open heath that had not been invaded by any of the successional species as well as areas with a range of successional stages. The heathland portions were all dry heaths and as similar as possible to the vegetation likely to have been present before invasion occurred. The successional stage of the vegetation in each area were classed on the basis of the dominant invasive species as follows: +B, major invader is Betula spp.,+PS, major invader is Pinus sylvestris, +PA, major invader is Pteridium aquilinum,+R, major invader is Rhododendron ponticum,+U, major invader is Ulex europaeus.

Table 1.  The location of heathland areas in Dorset. Successional stages present within each site are represented by the approximate time (years) over which each successional stage has occurred (x); data derived from aerial photographs. Reproduced from Mitchell et al. (1997)
  Successional stages present
SiteGrid referenceHeath+B+PS+PA+R+U
  1. H = Open Heath site.? = stage sampled but time over which successional stage occurred unknown.Where possible we indicate the range of years over which the successional stage has occurred (eg 30 > x > 23 indicates that the site was still heathland 30 years ago but had become dominated by the species indicated by 23 years ago), but for some stages the data was incomplete and only the maximum or minimum of the range is shown.

ArneSY973882H30 > x > 2343 > x > 2349 > x > 4349 > x > 4349 > x > 30
Avon Heath County Park (AHCP)SU128035Hx > 23x > 2323 > x? 23 > x
BlackhillSY840940H23 > x > 923 > x > 948 > x > 2348 > x > 2348 > x > 23
Canford HeathSZ030950H23 > xx > 4948 > x > 2348 > x > 2323 > x
Cranborne CommonSU104112H

x > 49
x > 49
Higher Hyde HeathSY851907H48 > x > 23
23 > x > 948 > x > 2348 > x > 23
St Catherines Hill/ Town CommonSZ142955H48 > x48 > x
48 > x
Sopley & RamsdownSZ133974Hx > 48x > 48
x > 48
48 > x > 2348 > x > 2348 > x > 23
Winfrith HeathSY805865H

23 > x > 9
x > 23

Aerial photographs of the sites taken in 1946/47 1972/73 and 1986 showed that the areas represented by the successional stages were all heathland 20–50 years ago (Table 1). The successional stages sampled were those located as close as possible to the remaining heathland to enable comparisons of the stages to be as valid as possible. The vegetation of these sites has been described in detail in Mitchell et al. (1997).


In late February 1996 six soil cores (diameter 5 cm, depth 6.3 cm), were sampled from each successional stage present within each area (a total of 47 sites). The cores were grouped into three pairs, and the paired samples were then mixed to form a sample from an area of 39 cm2 with a volume of 247 cm3. This gave 30 samples from heath sites, 21 samples from +B, +PS, and +U stages and 24 samples from +PA and +R sites. All samples were stored in a cold room at 4°C until the experiment was set up. Thus although the number of samples per site was small there was a considerable number of samples from each successional stage enabling the analysis of the differences between the successional stages, which was the main focus of this study.

Each sample was sieved through a 4 mm mesh to remove large stones and then spread thinly onto a tray (21 x 35 cm) containing sterile sand. The trays were randomly placed in a polyethylene tunnel and watered regularly. Eleven control trays containing sterile sand were also set up to detect wind blown seed. The seedlings were identified using Muller (1978) and counted as they emerged over the following 15 months. Those species which could not be identified at the seedling stage were potted on until identification was possible. We did not distinguish between the two Erica species (E. tetralix and E. cinerea) and the three Ulex species (U. europaeus, U. minor and U. gallii) which were recorded as Erica spp. and Ulex spp. respectively. This meant that it was impossible to distinguish between the heathland species U. minor and U. gallii and the invasive successional species U. europaeus. Species, mainly glasshouse weeds, that were recorded in both the control trays and the experimental trays were not included in the analysis. Results were expressed as seeds m−2 to a depth of 6.3 cm. Species nomenclature follows Stace (1991).


Differences in densities of species between the successional stages were tested with an unbalanced analysis of variance (anova) (proc glm, SAS Institute Inc. 1988). Differences in densities of the major species between the individual stages and heath sites were compared using LSD values.

The multivariate technique Canonical Discriminant Analysis (CDA), (the CANDISC procedure; SAS Institute Inc. 1988) was used to detect differences in species composition between groups (Benoit, Derksen & Panneton 1992). CDA finds linear combinations of discriminating variables which maximise the differences between groups (the six stages in this case) and minimises the differences within the groups. It performs this by maximising the ratio of the between-group sum of squares and the within-group sum of squares of the site scores (Jongman, Ter Braak & Van Tongeren 1995). The maximum number of discriminating variables is the number of groups minus one, 5 in this case. Species occurring less than 10 times were removed from the data set which was transformed (square root) prior to analysis.

The Wilks's lambda test, a multivariate measure of group (stage) differences over several variables, was used to test whether the differences explained by the discriminating variables were significant (Klecka 1980). Only discriminating variables that were statistically significant were used in the explanation of the results. The test statistic used to discriminate between groups is the Mahalanobis distance (D2) (Klecka 1980); a large value of D2 indicates good discrimination between groups. The Mahalanobis distance was converted to an F statistic to test if the stages were significantly different from each other.



The results are presented in Table 2, here only the major differences between the stages are discussed.

Table 2.  Species recorded in soil samples at Heath (H), and successional stages (Betula spp. (+B), Pinus sylvestris (+PS), Pteridium aquilinum (+PA), Rhododendron ponticum (+R), Ulex europaeus (+U)) on Dorset Heaths. Means  ±  SE are shown, results are seeds m-2 to a depth of 6.3 cm, the number of samples per successional stage is also shown
 Successional stage
SpeciesH (n = 30)+B (n = 21)+PS (n = 21)+PA (n = 24)+R (n = 24)+U (n = 21)
Agrostis curtisii<hsp sp=2>0 ± 0 85 ± t73   0 ± 0  11 ± 11   0 ± 0 182 ± 157
Betula spp.   59 ± 232534 ± 723 194 ± 93 170 ± 70 753 ± 247  85 ± 32
Calluna vulgaris13647 ± 25464583 ± 14653649 ± 11295432 ± 13382238 ± 10333977 ± 932
Carex pilulifera   17 ± 12 339 ± 255  12 ± 12  21 ± 21   0 ± 0  12 ± 12
Digitalis purpurea<hsp sp=2>0 ± 0 339 ± 327   0 ± 0   0 ± 0   0 ± 0   0 ± 0
Erica spp. 2376 ± 9571988 ± 1566  24 ± 171146 ± 983 202 ± 191 291 ± 134
Geranium robertianum<hsp sp=2>0 ± 0  48 ± 28   0 ± 0  11 ± 11   0 ± 0   0 ± 0
Holcus lanatus   17 ± 17 158 ± 67   0 ± 0  11 ± 1  10 ± 0   0 ± 0
Hypochaeris radicata    8 ± 8   0 ± 0   0 ± 0   0 ± 0  11 ± 1  10 ± 0
Juncus bufonius   59 ± 263104 ± 1211 364 ± 145 509 ± 2381 432 ± 878 291 ± 144
Leucanthemum vulgaris    0 ± 0   0 ± 0   0 ± 0   0 ± 0   0 ± 0  12 ± 12
Luzula campestris    0 ± 0   0 ± 0   0 ± 0  11 ± 1  10 ± 0   0 ± 0
Molinia caerulea   85 ± 61 121 ± 60  12 ± 12  53 ± 26  11 ± 11 352 ± 155
Rubus glandulosus    0 ± 0  24 ± 17   0 ± 0  11 ± 11  11 ± 11  12 ± 12
Rumex acetosella    0 ± 0  85 ± 51   0 ± 0  21 ± 21  21 ± 15   0 ± 0
Sagina procumbens   17 ± 12   0 ± 0  24 ± 24   0 ± 01   1 ± 11   0 ± 0
Taraxacum spp.    0 ± 0   0 ± 0   0 ± 0   0 ± 01   1 ± 11  12 ± 12
Teucrium scorodonia    8 ± 8  12 ± 12   0 ± 0  21 ± 21   0 ± 0   0 ± 0
Tripleurospermum inodorum   25 ± 5   0 ± 0  24 ± 24   0 ± 0  21 ± 21  12 ± 12
Ulex spp.   34 ± 20 158 ± 90  24 ± 17 180 ± 107   0 ± 0 279 ± 80

The samples from the heath sites gave rise to large numbers of Calluna vulgaris seedlings ranging between 2000 and 28 000 m−2. Erica spp. were also present at seven of the ten sites although at lower densities (100–8000 m−2) than Calluna vulgaris. Other species present included Betula spp., Carex pilulifera, Molinia caerulea and Ulex spp., but much lower densities, < 800 m−2 .

Viable seeds of Betula spp. were present at all +B sites at densities of between 400-8000 m−2. Calluna vulgaris was also present in all +B sites, although, except at AHCP, usually at much lower densities than the heath sites. The heathland species Agrostis curtisii, Erica spp., Molinia caerulea and Carex pilulifera were also present in some samples. Other species recorded included Digitalis purpurea, Geranium robertianum, Holcus lanatus, Juncus bufonius, Rubus glandulosus Rumex acetosella and Ulex spp., with J. bufonius being the most common.

The +PS stage contained fewer seedbank species than the +B stage. These were mainly heathland species, such as Calluna vulgaris, Carex pilulifera, Erica spp. and Molinia caerulea. The Calluna vulgaris content of the samples was variable with some sites having very high densities, similar to the corresponding heathland (AHCP and Blackhill) and others having much lower densities (Canford). Betula spp. and Juncus bufonius were also recorded although at lower densities than in the +B stage. Sagina procumbens and Teucrium scorodonia were also present.

The +PA stage was similar to the +B stage with both stages containing a greater variety of non-heathland species than the other stages, including Betula spp., Geranium robertianum, Holcus lanatus, Luzula campestris, Rubus glandulosus, Rumex acetosella, Teucrium scorodonia and Ulex spp. The heath species Agrostis curtisii, Calluna vulgaris, Carex pilulifera and Erica spp. were also present.

The +R stage generally contained few viable seeds compared to the other stages. Calluna vulgaris was present at all sites except St Catherines Hill. Betula spp. and Juncus bufonius were the most frequent species other than Calluna vulgaris. Other species which occurred occasionally were Erica spp., Hypochaeris radicata, Tripleurospermum inodorum, Molinia caerulea, Rubus glandulosus, Rumex acetosella and Sagina procumbens.

At the +U sites the most common species were heathland species, Agrostis curtisii, Calluna vulgaris, Erica spp. and Molinia caerulea. Other species which occurred were Betula spp., Juncus bufonius, Leucanthemum vulgaris, Tripleurospermum inodorum, Rubus glandulosus, Taraxacum spp. and Ulex spp.

The densities of eight of the species (Carex pilulifera, Betula spp., Calluna vulgaris, Erica spp., Holcus lanatus, Juncus bufonius, Molinia caerulea and Ulex spp.) varied significantly between the stages (Table 3a). The densities of the species in each successional stages were then compared with the heathland stage (Table 3b). The +PA stage had lower densities of Calluna vulgaris seeds than the heathland, while the +PS stage had significantly lower densities of both Calluna vulgaris and Erica spp. The other stages also had lower densities of Calluna vulgaris seedlings when compared with the heath, however they also had significantly greater densities of other species: Carex pilulifera, Betula spp., Digitalis purpurea, Holcus lanatus, Juncus bufonius and Rumex acetosella in the +B stage; Betula spp. and Juncus bufonius in the +R stage and Agrostis curtisii, Molinia caerulea and Ulex spp. in the +U stage.

Table 3.  Densities of species present in the seedbank which are significantly different between stages (heathland and 5 successional stages) on the Dorset heaths
(a) Significance of differences across all stages as tested by an unbalanced analysis of variance
Agrostis curtisiiNS
Carex pilulifera*
Betula spp.***
Calluna vulgaris***
Digitalis purpureaNS
Erica spp.*
Holcus lanatus***
Juncus bufonius***
Molinia caerulea*
Rumex acetosellaNS
Ulex spp.***
(b) Significances of differences between individual stages and heathland sites using t-test (LSD) from the GLM procedure in SAS

  1. NS = not significant; * = significant at P = 0.05; ** = significant at P = 0.01; *** = significant at P = 0.001.

  2. NS = not significant; * = significant at P =  0.05.+, density of seedlings significantly greater than in the heathland stage; −, density of seedlings significantly lower than in the heathland stage.

Agrostis curtisiiNSNSNSNS*+
Carex pilulifera*+NSNSNSNS
Betula spp.*+NSNS*+NS
Calluna vulgaris*****
Digitalis purpurea*+NSNSNSNS
Erica spp.NS*NS**
Holcus lanatus*+NSNSNSNS
Juncus bufonius*+NSNS*+NS
Molinia caeruleaNSNSNSNS*+
Rumex acetosella*+NSNSNSNS
Ulex spp.NSNSNSNS*+


The eigenvalues for the first two discriminating variables show that these two variables have the greatest discriminating power; 87% of the total (Table 4a). The squared canonical correlation indicates that the first two discriminating variables were highly correlated with the stages and that the stages were different when analysed according to the species present in the seedbank (Table 4a). The Wilks's lambda test (Table 4a) showed that the first two discriminant variables were significant (P<0.001) and that the remaining three were not. Thus, here the interpretation of results are confined to the first two discriminant variables.

Table 4.  Canonical Discriminant Analysis of the species present in seedbanks on Dorset heaths and successional stages
(a) Results for the five discriminating variables
Discriminating variableEigenvalueEigenvalue proportionEigenvalue cumulative proportionSquared canonical correlationWilk's lambdaProbability derived from Wilk's lambda
(b) Mahalanobis squared distances (D2) (upper right portion) and level of significance of differences between stages (lower left portion)

  1. NS = not significant; *significant at P < 0.05; **significant at P < 0.01; ***significant at P < 0.001.

Heath13.782.161.975.28 4.10
+PS****0.480.93 1.82
+PA****NS1.67 1.00
+R******NSNS 3.27

The discriminating variables (which may be treated as axes) are composed of discriminating species. Betula spp was positively correlated with the first axis and Calluna vulgaris was negatively correlated with it. Calluna vulgaris and Erica spp. were positively correlated with the second axis and Ulex spp. was negatively correlated with this axis.

The distribution of samples on the first two canonical axes (Fig. 1) separated out the +B stage from other samples at the positive end of Axis 1. In the middle of this axis the +PS, +PA and +R samples were intermixed with some of the +B samples. The heath and +U samples were found at the negative end of this axis. There is a considerable amount of intermixing between groups on this axis. The heath samples were positioned at the positive end of the second axis, and were intermixed with the +PA and +PS samples in the middle of this axis while the +U samples occurred at the negative end of this axis.

Figure 1.

Samples plotted by their scores for the first two canonical variables as calculated by CDA. The centroids for each stage are also shown.

The differences between the Mahalanobis distances (Table 4b) show that the Heath stage is significantly different from all the successional stages, although at different levels of significance, P = 0.05 for +PS and +PA, P = 0.01 for +U and P = 0.001 for +R and +B. The +B stage is significantly different from all the other successional stages and the +R stage is significantly different from the +U stage.



Seedbank studies are beset by a number of limitations which need to be considered when discussing the results of this experiment. Insufficient replication and sampling size (Benoit, Kenkel & Cavers 1989) may give rarer species little chance to be present as a seed in the sample and may thus restrict the precision of the other estimates. However these rarer species are unlikely to be important in terms of restoration. Although the number of samples per site was small (n = 3), the aim here was to compare successional stages for each of which we had a considerably greater number of samples (n>20). It is also possible that new germinations would have occurred several years after the start of the germination trials. However a period of more than a year provided opportunity for both spring and autumn germinating seeds to germinate. In spite of these uncertainties the values obtained are adequate for the purposes of this comparative study; they provide a general indication of seed density and allowed comparisons between the heath and successional stages to be made.


The CDA analysis showed that all the stages were significantly different from the heath. The seedbanks changed significantly as succession occurred, the greatest changes occur during the +R and +B stages while +PA and +PS changed the least. There are two main ways in which the successional stages differed from the heath (1) a lower density of heathland species, especially Calluna vulgaris in all stages; (2) a higher density of non heathland species especially in the +B, +R and +U stages (Table 3b)

Although all the successional stages were significantly different from the heath these differences are only important in terms of conservation if the density of heathland species is too low for adequate restoration and/or the density of non heathland species is high enough to allow establishment of these species to the detriment of the heathland species. From this it may be argued that restoration of heathland is more likely to be successful on +PS and +PA sites as they only differ from the heath by having lower densities of Calluna vulgaris and Erica spp. than the +B, +R and +U stages which not only have lower densities of Calluna vulgaris and Erica spp. but also have higher densities of non-heathland species.


The five most common species in the seedbank were Calluna, Erica, Betula, Juncus and Ulex and discussion here is limited to these species as these are the ones that will have the greatest influence on which species first colonise a cleared site and thus the success of restoration.

Calluna vulgaris was present in the seedbanks of all the stages and the numbers recorded were fairly typical of heathland seedbanks present beneath plantations or Pteridium aquilinum on former heathland (Hill & Stevens 1981; Pakeman & Hay 1996) confirming that Calluna vulgaris seeds can survive buried beneath successional stages for up to 50 years. However the total density of seeds is significantly different between the stages, with +B and +R stages having the lowest densities. The decline in the +B stage may be due to an increase in earthworm and other soil fauna and flora activity as the soil beneath a Betula spp. wood slowly changes from a mor to a mull soil (Miles 1981). The decline in numbers in the +R stage may be due to toxins released by Rhododendron ponticum (Cross 1975) decreasing the viability of the Calluna vulgaris seeds.

This study is based on the assumption that before invasion occurred the seedbanks of the successional stages were the same as the present heathland. Recent changes in management (grazing and burning) could have altered the density of Calluna vulgaris seeds. However there has been little management on these heaths for several decades and the changes in the seedbank with regard to the density of the Calluna vulgaris seeds are likely to be real rather than management induced.

Calluna vulgaris is present at reduced densities in the seedbanks of the successional stages, but for successful restoration it is important to determine whether there are still enough Calluna vulgaris seeds to establish and to prevent the invasion of more competitive species. A density of 40 000 seeds m-1 is recommended in commercial reseeding (Putwain pers com.) which is considerably greater than most of the densities recorded here. This would imply that for these stages the densities may not be great enough and propagules may need to be added for successful restoration (Pakeman & Hay 1996), but this will have to be determined by experiment.

Erica spp. were also present at many sites. Although Erica cinerea and Erica tetralix were not distinguished, this is unlikely to be important for heathland restoration as both species would be important contributors to a heathland community. It is also unlikely that any Erica cilliaris was present as this species was not recorded at any of these sites (Mitchell et al. 1997).

After Calluna vulgaris, Juncus bufonius was the most common species in the seedbank. The difference in densities of Juncus bufonius seedlings was significantly different between the different stages, with the +R and +B stages having significantly higher densities of Juncus bufonius compared to the heath seedbanks. Juncus spp. are often present in large densities in the seedbanks of heathland and former heathland sites (Putwain & Gillham 1990) and may be present in sufficient numbers in the top soil of managed sites to become an undesirable component of the developing heathland plant community (Putwain & Gillham 1990; Mitchell unpublished data).

Of the major successional species Betula spp. and Ulex spp. were the only ones occurring in the seedbanks, the densities of both were significantly different between the stages. The density of Betula seedlings was significantly different from the heath in both the +B and the +R stages. These seeds would become a problem in restoration work if a large seedling population emerged (Putwain & Gillham 1990). Betula spp. were frequently present as seedlings on successional sites in Dorset recently managed to restore heathland (Mitchell unpublished data.). Whether these plants become a problem and limit the success of the restoration depends on the survival rate of the seedlings relative to the development of the heathland species.

It was not possible to distinguish between the different Ulex species at the seedling stage. As only Ulex europaeus is a symptom of lack of management and the other two species (Ulex minor and Ulex gallii) are normal constituents of Dorset heaths it is impossible to say that the large number of Ulex seeds is likely to be detrimental to the restoration process. However as only the +U stage had a significantly different density of Ulex seedlings from the heath and this stage was dominated by Ulex europaeus it is likely that the majority of the Ulex seeds were Ulex europaeus. Ulex europaeus is known to have large persistent seedbanks (Grime, Hodgson & Hunt 1988) and may become a problem if large seedling populations emerge from the topsoil in managed sites (Putwain & Gillham 1990).

Of the other major successional species studied, Pinus sylvestris and Rhododendron ponticum have transient seedbanks with no dormancy (Granström 1987; Cross 1975) and so the seeds of these species are unlikely to be problematic in terms of restoration. Pteridium aquilinum can establish from spores, although the significance of this as a means of regeneration is unknown (Dyer 1990), and no sporelings were found in this study. Pteridium aquilinum also spreads by means of rhizomes, so whilst not present in the seedbank it may still be difficult to control (Marrs, Pakeman & Lowday 1993).

Other potentially problematic species, from a restoration viewpoint, that were detected in the seedbanks included Holcus lanatus, Rubus glandulosus, Rumex acetosella, Tripleurospermum inodorum, Geranium robertrianum, Teucrium scorodonia, Hypochaeris radicata and Leucanthemum vulgaris. Whether these species cause problems for the restoration of heathland depends on their density in the seedbank and whether increases in soil nutrients that occur during succession are reversed by management intervention (Mitchell et al. 1997). The increase in nutrients may give these competitive species an advantage over the heathland species (Marrs & Gough 1989).


This study has shown that a viable population of Calluna vulgaris seeds is present in the soil under all successional sites, albeit at a lower density than in the heath seedbank. Thus for the time being there is a reasonable hope of successful restoration of heathland on these stages. However with time the viable Calluna vulgaris seed population will decrease further, the population of successional species will increase and the capacity of the seedbank to act as a source of heathland propagules will slowly decline. Restoration should therefore be undertaken sooner rather than later and is likely to be more successful on the younger sites.

Despite heathland species being present in all stages, some stages (+B, +R, +U) also contain significantly higher densities of non-heathland species when compared to the heath. Restoration from these stages is likely to be more problematic than for +PS or +PA. When deciding where to target resources the results from this study should be combined with information on the changes that occur in both vegetation and in soil nutrient status (Mitchell et al. 1997), the ease with which the invasive species may be controlled (Marrs 1984; Marrs & Lowday 1992; Squires 1991), and the cost involved (Woodrow, Symes & Auld 1996).

Although many heathland species are obviously present in the buried seed flora of successional sites the seeds may need to be brought to the surface to stimulate germination. Areas where the soil has been disturbed by scarifying, turf or litter-stripping have a greater Calluna vulgaris establishment than where the soil/litter is undisturbed (Putwain & Gillham 1990; Mitchell unpublished data). Litter-stripping of successional sites is therefore strongly advised as this releases the dormant seedbank and removes nutrients from the system (Mitchell unpublished data) both of which will help to increase the probability of successful restoration. However, as 96% of Calluna vulgaris seeds may be in the top 50 mm of the mineral soil (Putwain & Gillham 1990) this disturbance must be done with care. Where litter is removed the aim should be to leave the surface mineral soil intact.


We thank the RSPB for funding this study and the many the landowners and conservation bodies for allowing access to their land. This study would not have been possible without the help of the RSPB staff in Dorset and the horticultural staff at Ness Botanic Gardens, University of Liverpool. Dorset County Council kindly allowed access to their aerial photograph collection, and Dr Hugh McAllister (University of Liverpool) assisted with the identification of the seedlings.


  1. Correspondence: R.J. Mitchell, The Royal Society for the Protection of Birds, The Lodge, Sandy, Beds, SG19 2DL,UK (fax 01767 692365; e-mail ruth.mitchell@RSPB.org.uk).

Received 4 September 1997revision accepted 17 December 1997