A study of the restoration of heathland on successional sites: changes in vegetation and soil chemical properties


*Present address and correspondence: ITE Banchory Research Station, Hill of Brathens, Glassel, Banchory, Kincardineshire AB31 4BY (fax: 01330 823303; e-mail: r.mitchell@ite.ac.uk).


1. Lowland heaths are high-profile ecosystems for conservation action in Britain, but many areas have been invaded by Betula spp., Pinus sylvestris, Pteridium aquilinum and Rhododendron ponticum. As succession occurs on heaths, changes occur in both the vegetation and the soil chemical properties of the site.

2. Nine heathland sites in the Poole Basin area of Dorset were studied, where management of successional sites to restore heathland had occurred. The efficacy of heathland restoration in terms of both the vegetation and the soil chemical properties was assessed.

3. The management had allowed many heathland species to establish and the majority of sites to start to become similar to the neighbouring heathland. The reversion of increased soil nutrients was found to be more problematic, with levels of ammonium–nitrogen, phosphorus, pH, calcium and magnesium remaining greater than those of the heathland soils.

4. The vegetation and soil data were analysed using canoco (canonical correspondence analysis) and were then used to test four hypothetical models that related changes in biotic factors (vegetation) and abiotic variables (soil nutrients) following management to the success of the restoration of heathland on successional sites.

5. A second canoco analysis was carried out in which the managed sites were treated as passive samples. This model was used to measure the distances between the heath, successional and managed sites. These distances provided measures of management success and the resilience of the treated late-successional ecosystem.

6. The successional species present before management affected the success of reversion; management of Pinus sylvestris sites was generally more successful than management of others sites, especially those invaded by Betula. The most significant effect of different management techniques resulted from litter-stripping, which reduced the nutrients available and improved and accelerated the success of reversion.


Dorset heaths have declined dramatically in area, diversity and structure over the last century (Chapman, Clarke & Webb 1989; Webb 1990). One cause of these losses has been an increase in scrub invasion and succession to woodland (Webb 1990) because of a decline in management practices that inhibit the succession of heath to woodland.

As succession occurs on heathland, the species composition of the site changes (Miles 1981; Mitchell et al. 1997). However, management to reverse succession and restore heathland is usually targeted at a few species, i.e. the major invasive species, which on the Dorset heaths are Betula spp., Pinus sylvestris (L.), Pteridium aquilinum (L. Kuhn) and Rhododendron ponticum (L.). The ability of the invaders to recover or persist after management, i.e. the resilience of the successional stage, will influence the success of conservation management to restore heathland.

Heathlands are generally found on nutrient-poor and infertile soils (Gimingham 1992). As succession occurs soil chemical properties change (Miles 1981, 1985; Mitchell et al. 1997). The greatest changes occur when Betula spp. invade, resulting in increased pH, exchangeable calcium and extractable phosphorus (Miles 1981; Mitchell et al. 1997). Increased concentrations of other soil nutrients are associated with different successional trajectories; extractable ammonium–nitrogen, nitrate/nitrite–nitrogen and exchangeable potassium increased with Pteridium aquilinum and Ulex europaeus (L.) successions, sodium increased with Rhododendron ponticum succession, and organic matter increased with Pinus sylvestris succession on Dorset heaths (Mitchell et al. 1997).

If restoration is to be successful then management to restore heathland must reverse the changes in both vegetation and soil nutrients. There are four hypothetical models that may describe the response of a site to management: (i) good restoration with the managed site indistinguishable from the target; (ii) poor restoration with the managed site remaining close to the ‘start’ site; (iii) partial restoration; and (iv) movement to a different endpoint. This paper describes the study of the efficacy of restoring heathland communities on land where succession had occurred and the ‘starting’ communities were dominated by one of the following species: Betula spp., Pinus sylvestris, Pteridium aquilinum or Rhododendron ponticum. The aims were to answer the following questions.

1. Was management effective in restoring heathland communities and preventing the recovery of late-successional communities?

2. Was management effective in reducing any soil properties that are known to be greater in the late-successional communities?

3. Was it possible to measure the success of management and the resilience of the managed site?

4. Does the species of invasive plant present before management, method of management or the length of time (years) since management affect the success of the restoration?

An attempt was made to answer these questions by contrasting a range of sites where late-successional communities were managed by the Royal Society for the Protection of Birds (RSPB) to restore heathland (‘managed’ sites), with areas where heathland communities were still present (‘target’ communities) and where the successional species were still present (‘start’ communities). An assumption was made that the start community and the managed communities were similar before management was applied.

Materials and methods

Sampling strategy

In 1996 a structured sampling strategy was set up in the Poole Basin area of Dorset; nine heathland areas within the syncline of the Poole Basin were chosen (Table 1). The areas are all in close proximity to each other (within a 20-km radius) and should experience similar climate; they all lie on a similar parent material, the Bagshot beds.

Table 1.  The stages present at each area and the number of sites sampled within each stage.; 10 samples were taken at each site. National Grid references for the areas are also shown. See text for explanation of the stages
AreaNational Grid
Heath +BmB +PSmPS +PAmPA +RmR
ArneSY9738821  151111
Avon Heath Country Park (AHCP)SU12803511112    
BlackhillSY8409401  111211
Cranborne CommonSU1041121    11  
East Holton HeathSY9589171  12    
Grange HeathSY9098351  11    
Merritown HeathSZ11399111111    
Sopley and RamsdownSZ13397411111  12
Trigon HeathSY884908111111111
Total number of sites 9448144545
Total number of samples 9040408014040504050

Within each area there were examples of heathland, successional and managed sites (Table 1). The successional and heath sites provided reference points denoting the start and desired target for the trajectory of the managed sites. The heathland sites were open heath dominated by dwarf ericaceous shrubs. The successional sites were heath until 20–50 years ago when they were invaded by one of four major successional species. Crude estimates of the time over which the successional species had invaded were made using aerial photographs taken in 1946/47, 1972/73 and 1986 (Mitchell 1997). Before succession occurred the vegetation in the successional and heath sites appeared similar. The successional species had been removed on the managed sites by the RSPB as part of management to restore heathland. The successional, managed and heath sites were in close proximity so that comparisons would be as valid as possible.

Four successional stages were sampled: +B, major invader was Betula spp.; +PS, major invader was Pinus sylvestris; +PA, major invader was Pteridium aquilinum; +R, major invader was Rhododendron ponticum. For each of these stages there were corresponding managed sites where the succession had been managed in an attempt to restore heathland, namely: mB, managed +B; mPS, managed +PS; mPA, managed +PA; and mR, managed +R.

The managed sites

For each managed site three variables were known: the stage (the major invader that was present before management occurred), the age (time elapsed since management, in years) and the type of management applied.

Management involved the removal of the dominant species, usually by felling or cutting. However, some PA sites were sprayed with Asulox, sometimes in combination with cutting, and two sites, one mPA and one mR, were bulldozed clear. The cut stumps of Betula spp. and Rhododendron ponticum were treated to stop regrowth, except at one of the mB sites where the site was grazed for 3 years instead. At the mPS sites, some sites were litter-stripped to varying depths.

Vegetation survey

At each site (start, managed and target) 10 1-m2 quadrats were placed randomly and a visual estimate of the cover (%) of all plant species, litter and bare ground recorded. The basal area of trees was estimated in the +B and +PS stages using a relascope (Manx Marker relascope factor ×2; Stanton Hope Ltd, Basildon, UK). A total of 570 quadrats was recorded, 90 from the open heath sites, 280 from managed sites and 200 from successional sites (Table 1). Nomenclature follows Stace (1991) for higher plants, Duncan (1970) for lichens and Smith (1978) for mosses. From the centre of each quadrat a soil sample was taken to a depth of 21 cm using a Bi-partite Edelman auger (Eijelkamp Agrisearch equipment, Giesbeek, the Netherlands). Soil samples were stored in a cold room (4 °C) until analysis.

Soil analysis

Fresh soil (5 g) from each sample was shaken with 30 ml 1 m potassium chloride for 1 h for extraction of available nitrogen. The extract was analysed for ammonium–nitrogen and nitrate/nitrite–nitrogen using colorimetric methods (Allen et al. 1974).

The remainder of the soil samples were air dried and sieved through a 2-mm sieve. A 2·5-g sample was extracted in 2·5% v/v acetic acid. The extractable phosphorus was measured using the stannous chloride method (Allen 1989). The same extract was used to measure exchangeable cations (calcium, magnesium, sodium and potassium). Calcium and magnesium were analysed by absorption spectrometry, and sodium and potassium by emission spectrometry (Unicam 1991). Organic matter present was estimated by loss-on-ignition (Allen 1989) and soil pH was recorded in a 1 : 2·5 slurry of soil and deionized water (Allen et al. 1974)

Data analysis

Vegetation and soil data were analysed by canonical correspondence analysis (CCA) (canoco; ter Braak 1990). CCA is a direct ordination technique in which the species/sample data in the ordination are constrained to optimize their linear relationship to the environmental variables, soil data in this case. The samples may be plotted in an ordination diagram with the environmental variables shown by vectors (arrows). The length of the arrows is proportional to their importance and the directions of the arrows show their correlation with the axes. In this paper the vectors are displayed at ×5 their actual length as the scores for the environmental variables, and species/samples are of a different order of magnitude (ter Braak 1988). The statistical validity of the ordination was tested using an unrestricted Monte Carlo permutation test (ter Braak 1990). Here canoco was run without detrending (Palmer 1993) and the vegetation data (% cover) were transformed by a ln(y + 1) transformation. Species occurring in five or less quadrats (less than 1% of the samples) were removed from the data set. This removed 52 species, of which 25 occurred only once. The percentage cover of bare ground and litter was included in the analysis because both were important components at some sites. Two different canoco analyses were carried out.

CCA analysis I

In the first canoco analysis all the samples (heath, managed and successional) were included as active samples to allow new species colonizing managed sites, which were not present on the heath or successional sites, to exert an influence on the direction of change. This was particularly important for the detection of model 4, movement to a different endpoint.

CCA analysis II

In analysis II the managed sites were treated as passive samples. The heath and successional sites were used to produce the model and the positioning of the managed sites within this model were calculated. This model fitted the managed sites onto the trajectories between the heath and successional sites without the managed sites influencing the trajectory. The distances of managed sites from the start and target points could then be calculated and used to measure the success of the management and the resilience of the sites. It is not possible to use analysis I for this as the objects being measured (the managed sites) have influenced the model.

The results from CCA analysis II were used to calculate centroids for each site. As the CCA model has four axes the distances between these centroids were calculated in four dimensions using Euclidean distance (Manly 1986); this uses Pythagoras’ theorem in four dimensions to calculate the distance between two points. As there were nine environmental variables and therefore nine axes, the distances could theoretically be calculated in nine dimensions. However, as over 80% of the explained species–environment relation was explained by the first four axes, the calculations were confined to four dimensions. For any trajectory, the distance between the target and managed site provided a measure of the success of the management to restore heathland, while the distance between the managed and start site provided a measure of the resilience of the managed site.

A measure of the linearity of the trajectory between start, managed and target sites was also made in four dimensions. For a straight line, A should equal zero, where A = (T + S) – D, T is the distance from managed site to the target, S is the distance from the managed site to the start and D is the distance from the start to the target. If the trajectory is a straight line then the managed site is moving towards the heath target, whereas if A is large the trajectory is not a straight line and the managed site is moving in a different direction, towards a different community.


Description of vegetation

The vegetation of the heath and successional sites was similar to that described for the sites in Mitchell et al. (1997); the heaths were dominated by Calluna vulgaris and the successional stages were dominated by the major invader. Here only brief descriptions of the vegetation of the managed sites are given.

On the mB sites, Betula spp. were present with cover ranging from 39% to 1%. Other successional species present at high cover included Pteridium aquilinum and Rubus gladulosus. Calluna vulgaris was present on all sites, with 27% cover at one site and less than 10% cover at the other sites.

On the mPS sites, Pinus sylvestris was present at low cover. Pteridium aquilinum occurred in nine of the 14 sites and was the dominant species at four sites. Betula spp. were present at all sites except Trigon. Calluna vulgaris occurred at all sites, with greater than 20% cover at Arne (a and e) and Trigon.

At the mPA sites, Pteridium aquilinum was the dominant species at Trigon and Blackhill (a) and occurred occasionally at low cover at the other sites. Calluna vulgaris occurred at all the sites, with 84% cover at Arne and 16% at Trigon, but the other sites had less than 5% cover.

Rhododendron ponticum was only recorded at three of the mR sites. Few species occurred at high cover on the mR sites but many species were present at low cover, mostly as young seedlings.

The mosses Campylopus introflexus, Hypnum jutlandicum, Polytrichum juniperinum and Pseudoscleropodium purum occurred frequently on managed sites from all stages, often at greater cover than on either the corresponding successional or heath sites. Typical heathland lichen species (predominantly Cladonia spp.) that were usually absent on the successional sites occurred at low cover on many of the managed sites. Species, such as Agrostis capillaris, A. curtisii, A. gigantea, Holcus lanatus, Juncus effusus and Rumex acetosella, that were often absent on the successional and heath sites occurred frequently on the managed sites.

Soil results

The differences in soil nutrient concentrations between the heath and successional stages were similar to those found in the earlier study (Mitchell et al. 1997). Here, we concentrate on the nutrient levels in the managed sites compared with the heath and successional sites; Table 2 summarizes the results. The managed sites typically were less acidic than the corresponding successional site. The estimated organic matter content of the soil (percentage loss-on-ignition) was lower on mPS sites that had been litter-stripped (4·5–7·2%) than those that had not been litter-stripped (5·0–10·5%). Concentrations of extractable ammonium–nitrogen, nitrate/nitrite–nitrogen and exchangeable phosphorus were usually greater on the managed sites than the successional sites, except where management had removed the litter layer to the mineral soil. Calcium concentrations were greater in the mB, mPA and mR sites than the corresponding heath. The majority of the managed sites had lower sodium concentrations than the successional sites and greater concentrations of exchangeable magnesium than the corresponding heath sites. Exchangeable potassium concentrations were lower in the mPA samples than the +PA samples, although still higher than in the heath samples.

Table 2.  Chemical properties of soil samples from heath, successional and managed sites in Dorset; the range of the means from the different sites are presented (ions are expressed as µg g–1)
StageCodespH% loss-on-
Heath 3·9–5·02·5–10·30·9–2·30·5–1·11·1–3·812·4–1167·312·0–65·07·6–36·37·0–44·4
Betula successional sitesB4·0–5·63·7–18·03·3–5·30·7–2·90·8–7·394·1–1429·239·3–132·620·2–75·528·2–87·4
Betula managed sitesmB4·1–5·95·4–21·22·9–13·70·8–3·81·9–4·688·8–1367·636·1–150·19·2–65·116·5–90·3
Pinus sylvestris successional sitesPS3·7–4·24·0–18·11·9–14·10·6–2·10·8–4·325·0–424·815·6–86·313·3–101·711·0–66·5
Pinus sylvestris managed sitesmPS3·9–4·44·5–10·51·6–23·00·7–1·11·3–7·632·1–194·420·5–68·510·8–38·112·9–66·2
Pteridium aquilinum successional sitesPA3·7–4·110·3–13·85·1–76·20·7–13·50·9–9·650·8–274·842·5–79·051·6–110·960·7–319·2
Pteridium aquilinum managed sitesmPA3·9–4·28·4–20·21·7–34·80·6–3·80·9–18·7119·8–472·026·0–92·731·3–62·242·0–146·6
Rhododendron ponticum successional sitesR3·6–4·57·0–12·71·1–8·50·5–0·81·9–4·260·8–848·425·3–51·632·5–48·418·0–50·6
Rhododendron ponticum managed sitesmR3·8–4·93·8–7·81·6–14·10·6–2·90·6–3·811·8–396·38·3–54·17·3–42·06·1–70·3

The cca model

CCA analysis I: managed sites as active samples

As expected, the data set showed large variation and therefore the CCA results had low eigenvalues (Table 3); however, such models can still be quite informative (Gauch 1982). The first axis accounted for 33·9% of the explained species–environment relationships, and axis 2 accounted for a further 20·8%.

Table 3.  Eigenvalues and intraset correlations of soil variables for the four CCA axes for analysis I (managed sites as active samples) and analysis II (managed sites as passive samples). The cumulative percentage variances and the rank order of the correlations are given in brackets
Analysis IAnalysis II
Axis1 2 3 4 1 2 3 4
Eigenvalue0·21 0·13 0·09 0·08 0·25 0·22 0·08 0·06
Species–environment correlations0·77 0·57 0·59 0·53 0·77 0·66 0·62 0·57
Percentage variance of species data3·4 2·1(5·5)1·5(7·0)1·0(8·0)5·3 4·8(10·1)1·7(11·8)1·2
Percentage variance of species–environment relation33·9 21·0(54·9)14·2(69·1)9·7(69·1)36·0 31·6(67·6)12·0(79·6)8·1
pH0·86(1)–0·12 0·05 –0·34(3)0·73(1)0·36 –0·24 0·05
Loss-on-ignition0·33 0·47(2)0·21 0·48(1)–0·09 0·44 –0·20 0·52
NH4-N0·15 0·31 0·49(3)0·23 –0·26(3)0·47 0·06 0·33
NO3-N0·33 0·17 0·56(2)–0·04 –0·12 0·47 0·11 0·46
P0·26 –0·16 0·77(1)0·01 0·02 0·46 0·72(1)0·24
Ca0·73(3)0·17 0·07 0·08 0·26 0·27 –0·33(2)0·43
Mg0·78(2)0·40(3)0·06 0·28 0·23 0·73(2)–0·26 0·45
K0·32 0·39 0·37 0·46(2)–0·12 0·57(3)0·11 0·51
Na0·10 0·92(1)0·26 0·03 –0·43(2)0·80(1)–0·26(3)0·12

The ordination diagram (Fig. 1) for the sample scores from the first two axes shows the heath samples clustered together at the lower end of both axes, while the samples from managed sites from all stages are clustered near the centre of the graph. Samples from the same successional stage are clustered together and are positioned in an arch around the managed sites and further away from the heath samples than the managed sites. The lengths of the vectors represent their relative importance; thus sodium, pH, magnesium and calcium are the four most important vectors. The positioning of the samples relative to the vectors relates the samples to different chemical properties of the soil. Thus sodium concentrations were higher on the +PA sites, pH, magnesium and calcium on the +B sites, ammonium–nitrogen, nitrite/nitrate–nitrogen and loss-on-ignition on the +B and managed sites, and phosphorus on the managed sites. The heath samples were found at the low end of all the soil vectors.

Figure 1.

Ordination diagram from canoco for analysis I (managed sites as active samples) for the first two axes showing the relationship between the different stages and the soil nutrients. The soil nutrient vectors are shown by arrows and their length is multiplied by a factor of five. Heath, open heathland; +B, Betula spp. is the major invader; +PS, Pinus sylvestris is the major invader; +PA, Pteridium aquilinum is the major invader, +R, Rhododendron ponticum is the major invader; mB, a managed +B site; mPS, a managed +PS site; mPA, a managed +PA site; mR, a managed +R site; Ca, exchangeable calcium; K, exchangeable potassium; LOI, percentage loss-on-ignition; Mg, exchangeable magnesium; Na, exchangeable sodium; NH4, extractable ammonium–nitrogen; NO3, extractable nitrate/nitrite–nitrogen; P, extractable phosphorus; pH, pH.

An unrestricted Monte Carlo test was carried out on both axes and the result was significant at P = 0·001, showing that the relationship between the vegetation and soil data was not random and that this analysis gave the best possible result.

CCA analysis II: managed sites as passive samples

The eigenvalues for this analysis were similar to that for analysis I (Table 3). The first two axes explained a total of 67·6% of the species–environment relationships. The ordination diagram (Fig. 2) for the sample scores from the first two axes also shows the samples from the managed site in the centre of the graph, with the successional samples in an arch around them and the heath samples positioned at the opposite end of the second axis from the successional and managed sites. As with analysis I, sodium increased with the +PA samples, and pH, magnesium and calcium with the +B samples. However, in this analysis, ammonium–nitrogen, nitrite/nitrate–nitrogen and potassium and loss-on-ignition all increased with the +PA rather than with the +B and the managed sites. These differences are due to this ordination being constructed using data only from the heath and successional sites.

Figure 2.

Ordination diagram from canoco for analysis II (managed sites as passive samples) for the first two axes showing the relationship between the different stages and the soil nutrients. The soil nutrient vectors are shown by arrows and their length is multiplied by a factor of five. Abbreviations as for Fig. 1.

Managed sites

The centroids of each managed site and the overall centroids for the successional stages and heath were plotted (Fig. 3) using the data from analysis I. The mB sites are shown scattered, with Merritown and AHCP close to the heath, Trigon close to the +B, and Sopley and Ramsdown moving in a different direction. The mPS sites generally cluster between the +PS and heath sites, with some sites [Trigon and Arne (a)] closer to the heath than others (Blackhill and Grange). The mPA sites range from Trigon, close to the start site, to Blackhill (a), which appears to be moving away from both the start and target sites. Indeed, the mPA sites may be moving in a common direction from the +PA start site towards Blackhill (a). The mR sites at Sopley and Ramsdown (a and b), Trigon and Blackhill are positioned between the +R and the heath, with Blackhill closer to the heath than the others; Arne appears to be moving in a different direction.

Figure 3.

Simplified CCA ordination diagrams from analysis I for: B, Betula; PS, Pinus sylvestris; PA, Pteridium aquilinum; R, Rhododendron ponticum sites. The centroids and SE bars for the heath (TARGET), successional (START) and managed sites are shown. The centroids for the START and TARGET sites are average centroids for all the START and TARGET sites.

Measuring management success

Centroids for all the sites were also calculated using analysis II, and the distances between these centroids are shown in Table 4. The sites closest to the heath target are mB at Merritown and the mPS sites at Arne (a) and Trigon; the management on these sites can therefore be judged to be most successful. Those furthest from the target are the mB sites at Sopley and Ramsdown and Trigon, and the mPA sites at Blackhill; these sites appear to be least successfully managed.

Table 4.  Distances in four dimensions of managed sites from heath (target, T) and successional (start, S) sites and of start site from target (distances calculated from CCA analysis II, managed sites treated as passive samples). A measure of whether this trajectory (start to managed to target) is a straight line is provided by the A value, which is calculated as ((distance from managed site to target) + (distance from managed site to start) – (distance from start to target (D)); the closer this value is to zero the closer the trajectory is to a straight line. The managed sites are ranked by distance from the heath target and by how close they are to a straight line trajectory
Distance of managed site fromRanking, closest to heathRanking, closest to straight line
StageAreaTarget (T)Start (S)Distance of start from target (D)A valueWithin stageOverallWithin stageOverall
BSopley and Ramsdown7·328·115·969·47427427
Arne (d)2·914·294·992·2269712
E. Holton (a)2·793·774·332·2358813
AHCP (a)2·642·284·570·354612
E. Holton (b)3·671·764·321·09111737
Sopley and Ramsdown3·463·794·652·60913915
Arne (a)1·714·674·991·401149
Arne (b)4·316·524·995·9213211324
Arne (c)3·495·004·993·5110141222
AHCP (b)3·082·534·571·047925
Trigon2·363·963·632·69231017 =
Arne (e)4·627·454·997·0914221425
Blackhill (a)7·124·673·688·10426426
Blackhill (b)7·965·793·6810·07528528
Trigon3·261·584·160·68212217 =
Sopley and Ramsdown (a)3·632·283·392·53316214
Sopley and Ramsdown (b)3·702·313·392·62519315

Measuring resilience

Those sites that were closest to the start sites were the mB sites at Trigon and AHCP and the mPA sites at Trigon. These sites may have the highest resilience to management as they remained close to the corresponding start site.

Measuring the trajectory

The managed sites that were closest to the target tended to have straighter trajectories than those sites further away from the target. Sites with a large A value (values of 4 or more) were mB at Sopley and Ramsdown, mPS at Arne (b), mPA at Blackhill (a and b) and mR at Arne, indicating that they may be moving to a different endpoint.


Two assumptions were made in this study. The first was that the managed and start sites were similar before management was applied. Most managed sites were once part of the successional site, part of which had been managed and the rest left as the start site. Therefore the variation between the two sites before management is likely to have been only as great as the variation within the start site. Secondly, we assumed that the heath target does not move over time. As the heath is a steady-state assemblage maintained by management but influenced by climate and increasing atmospheric nitrogen deposition, the heath may be a moving target. However, management aims to restore sites to a community similar to the heath that is present today. Therefore the assumption that the heath that was present before succession occurred is similar to the heath that is present today is not critical. If changes have occurred, this means that management is not strictly reversing successional changes but rather restoring a different community. If the study is repeated in the future and the heath targets resurveyed then any movement of the target could be incorporated within the model.

Restoration of typical heathland vegetation

Management did not remove all of the successional species, just most of the major invaders. This disturbance allows the establishment of heathland and other species in addition to the recovery of the successional species already present. Many heathland species are able to establish on the managed sites, including Cladonia spp., typical of heathland but not present in the successional stages (Mitchell et al. 1997). Many of the successional species are very resilient and able to recover after management. The success of management is therefore dependent on the balance between colonization and/or persistence of species typical of heathlands or successional stages, along with other species that might colonize. All of this may be influenced by the continued management of the site, for example by grazing, to continue the movement of the site towards the target.

Restoration of typical heathland soil chemical properties

The increased levels of soil nutrients that occur during succession do not automatically decrease following management. Levels of ammonium–nitrogen, nitrate/nitrite–nitrogen, phosphorus and calcium were generally greater in the managed sites than the successional sites. This increase may be due to a loss of biotic control following management (Bormann & Liken 1979); with time these levels may decline and eventually stabilize [the transition and steady-state phases in Bormann & Liken's (1979) model]. The sites that were similar to the heath soils were either those that were litter-stripped to the mineral surface [Arne mPS (a and b) and mR] or had corresponding successional sites that were not very different from the heath. Unless management involves litter-stripping, soil nutrient levels are unlikely to decrease to heathland level. Indeed, the major soil nutrients are more likely to increase. The relative lack of change in these abiotic parameters at some sites may be due to the time scale involved. If the abiotic changes occur at a slower rate than the biotic changes then declines in abiotic factors may occur in the future. More sampling would be needed to test this.

Raised soil nutrient levels will make the restoration of heathland on successional sites more difficult (Marrs & Gough 1989). It has been shown that some soil nutrients may increase following management. However, whether the level of increase is sufficient to be detrimental to the restoration process is unknown and further studies need to be done.

Testing of models

Of the four possible responses to management, no site had reverted perfectly to coincide with the heath target and none of the centroids of the managed sites overlapped with the centroids of the corresponding successional stage, the start point (Fig. 3). Most of the sites were positioned between the successional stage and heath sites, indicating that partial restoration is the most frequent result. However, some managed sites appear to be moving towards a different target other than the heath.

Measuring management success and the resilience of the sites

The distances between the managed and target sites allow a measure of the success of reversion to be made, and these provide a method of judging restoration success in terms of both vegetation and soil nutrients. The distance between the managed site and the start allows a measure of the combined impact of treatment and the resilience of the site. However, this measure is based on only one sampling time for analysis of the post-management trajectory. To measure resilience accurately the survey would have to be repeated over time. It would then be possible to monitor the managed sites and see if they continued to move towards the heath or if they showed high resilience and started to move back towards the start point.

Ecosystems may have more than one equilibrium (Holling 1986), so management may not always revert the site to the desired steady-state assemblage (heath); this is happening for sites that follow model 4 and have a high A value. Sites with a high A value are on a trajectory towards a different equilibrium, whereas sites with a low A value are moving in the desired direction, even if they have a long way to go. The sites with high A values are those that appear to be reverting towards a community other than heathland. In this study, sites with an A value of greater than 4 have been found. However, we do not know how much greater than zero A has to be before it becomes a serious problem.

Conservation relevance: factors affecting the success of management

In order to increase management success it is important to understand why some sites are further along the trajectory from start to target than others. Three factors may influence this: age, stage and management.

Age (time elapsed since management implementation) will affect the success of reversion; some time is obviously needed for the establishment of heathland species, but too long an interval may allow the recovery of the successional species. Within each stage in Table 4 the sites are ordered by increasing age. The young sites (less than 1·5 years) are furthest from the heath as they are dominated by bare ground and litter and contain little vegetation; the age of the other sites has no impact on how similar they are to the heath sites.

Mitchell et al. (1997) ranked successional stages in order of the magnitude of the correlation of the vegetation and soil nutrients with that of the heath, and suggested that sites invaded by Pinus sylvestris were the easiest and those invaded by Betula the most difficult to restore to heathland. The site closest to the heath target was a managed site from which Pinus sylvestris had been removed [Arne (a)] (Table 4). However, the second site that had moved most towards its target was a managed Betula site (Merritown), indicating that the +B stage may not be as difficult to restore as Mitchell et al. (1997) suggested. The other managed Betula sites, together with the managed Pteridium aquilinum, sites were the sites that were most dissimilar to the target heathland. Generally, the restoration of heathland on sites invaded by Pinus sylvestris was most successful, while management work on Betula and Pteridium aquilinum was least successful; however, some Betula and Pteridium aquilinum sites may revert better than some Pinus sylvestris sites. Managed sites that were close to either their target [Arne (a), a managed Pinus sylvestris site, and the managed Betula site at Merritown] or start point (the managed Betula site at Trigon) had low A values. Compared with the number of managed Pinus sylvestris sites studied, very few had high A values. This may indicate that sites invaded by Pinus sylvestris are more likely than other stages to move back along a straight line towards their target and not to veer off in a different direction.

The type of management used will influence the success of the reversion. This study shows the effect of litter-stripping following the clear felling of Pinus sylvestris. At Arne (a), where litter-stripping exposed the mineral soil, the restoration of heathland was more successful than at the other two areas managed at the same time [Arne (b) and (c)]. Indeed, the litter-stripped site was more similar to the heath than Arne (e), a site managed in the 1980s and therefore 10 years older, indicating that litter-stripping improves and accelerates the success of reversion. The quality of the litter-stripping also appears important; it is better if the litter is removed to expose the mineral soil [compare sites Arne (a), (b) and (c), total, partial and no litter-stripping, respectively]. Litter-stripping will remove more nutrients from the site and uncover the buried Calluna vulgaris seed, which requires light to germinate (Grime, Hodgson & Hunt 1988). The most effective management at sites dominated by Pteridium aquilinum was bulldozing (Arne); this removes the fronds, rhizomes and litter, thus removing a potential source of propagules and nutrients. Reversion of sites originally dominated by Rhododendron may be slow and only partially successful because of high concentrations of toxins and the deep root mat (Cross 1975). Bulldozing is therefore an attractive option as it solves these problems. However, management at the bulldozed site (Arne) was too recent to compare with the other sites. There was no clear difference between the two types of management on the sites originally dominated by Betula.

Although this study is limited, it does indicate that the type of management used is very important. More detailed comparisons between the success of the different management options outlined above would be worthwhile, especially on the effects of total litter-stripping to remove nutrients and release the available Calluna vulgaris seed bank, in the management of sites invaded by not only Pinus sylvestris but other successional species as well.

Comparing the two types of model

In this study two models have been used, treating the managed sites as (I) active and (II) passive samples. Analysis I treats the managed sites as active samples and shows the relationship between the heath, successional and managed sites. This model is especially useful for testing if the managed site is moving in a direction other than towards the target, e.g. towards grassland. Analysis II calculates the position of the managed sites from the abundance of species occurring in successional and heath sites. Therefore if the site is dominated by species not present in the heath or successional communities then the true positioning of the site may not be shown. However, analysis II enables the managed sites to be positioned along the trajectory between the heath and successional sites and a measurement of the closeness of the managed site to the target to be made. Both models are therefore needed for a realistic judgement of management success to be made.

Future developments of the approach

This study and the previous study (Mitchell et al. 1997) have shown that the analytical approach used here produces models that can be used to relate the managed sites to successional processes on nature reserves. The models can be used to place managed sites in relation to the start and target sites and provide a measure of the effectiveness of management, a measure of the resilience of the site and an early warning system. The advantage of this method to assess the success of management (the distance of the managed site from the target) is that it is not based on one factor alone, vegetation or soil, but combines abiotic and biotic factors to produce a combined measure of success. Such models may be used for a more detailed study of the effectiveness of different management treatments. Models such as these could also be developed for other habitats, where other environmental driving variables may be important.

If the same site was surveyed repeatedly over time the resilience of the site could be measured quantitatively and could help to provide information on basic ecological processes. Having moved along the successional trajectory towards the heath, does the site remain there or does it ‘bounce’ back towards the starting point? Thus, for the first time, we have a multivariate method using biotic and abiotic process for measuring ecosystem resilience.

The measurements of the distances between the start, target and managed sites and the calculation of the A value are based on a single point in time and, for simplicity, we have assumed that the trajectory is linear. However, it is possible that some trajectories will be curvilinear. More samples taken through time would be needed to test this. If curvilinear trajectories are found to occur, this type of model allows the manager to detect this relationship and make a judgement as to whether it is acceptable for conservation purposes or whether other management must be applied to correct the direction of change.


The management of successional sites allows many heathland species to re-establish and the site to start to become similar to the neighbouring heath. The reversion of increased soil nutrients and pH is more problematic, with ammonium–nitrogen, pH, phosphorus, calcium and magnesium frequently remaining higher in the managed sites than the heaths and sometimes being even higher than the successional stages. The stage and type of management of the site will affect the success of reversion. Specifically, management of sites originally dominated by Pinus sylvestris is generally more successful at restoring heathland than management on other sites. Litter-stripping reduces the nutrients available and may also release the dormant Calluna vulgaris seed bank. The study of successional, managed and plagioclimax habitats is of great value for the development of models that may then be used to monitor the success of the restoration of the managed site.


We would like to thank the RSPB for funding this study, Dr Ceri Evans (RSPB) and Dr Phil Putwain (The University of Liverpool) for their support of this work and Dr Cajo ter Braak for his helpful comments on this manuscript and statistical advice. We are also grateful to the landowners for allowing us access to the heathlands used in this study and Professor Nigel Webb (Institute of Terrestrial Ecology at Furzebrook) for storage of soil samples during fieldwork. This work would not have been possible without the help of the RSPB staff in Dorset, especially Mr Bryan Pickess and Mr Nigel Symes.