Present address and correspondence: A. E. Eycott, Forest Research, Alice Holt Lodge, Farnham, Surrey GU10 4LH, UK (fax + 44 1420 23653; e-mail email@example.com).
1Commercial forests represent an important but often neglected biological resource. This study related the understorey plant species composition of a coniferous plantation forest managed by clearfelling to environmental factors (stand structure, soil pH and previous land use) and ecological patterns (abundance-occupancy relationships, species dispersal and life history).
2Plant species richness and composition were recorded in 326 managed stands of different ages, soil types and land-use histories in a 185-km2 lowland forest planted onto heath and arable land.
3Stands replanted in the last 10 years had the greatest species richness, typically containing in the order of 18 plant species. Stands on soils of high pH had greater plant species richness, as did those on previously arable land.
4Less than a quarter (23%) of all species persisted in the above-ground vegetation throughout the growth cycle. The majority recolonized forest stands during the cycle, by physical dispersal or from the seed bank, largely after canopy opening in mature stands (26%) or after felling (47%). Annual species and species with plumed seeds were most abundant in early growth stages, while shrubs with berries were more abundant in mature stands.
5We found a strong positive interspecific relationship between frequency of stand occupancy and mean abundance within occupied stands. For species not persisting above-ground throughout the forestry cycle (i.e. patch colonizers), the slope of the abundance-occupancy relationship was steeper for those with a long-distance dispersal mechanism than for those lacking such a mechanism.
6Synthesis and applications. Rotational clearfelling of plantations may be an appropriate form of forest and conservation management in forests planted on former open areas such as heaths, where the conservation interest is not in old-growth species but in earlier successional species. Maximizing representation of young growth stages will help maximize plant diversity in such cases. These prescriptions contradict guidance for sustainable forestry; however, it is appropriate to vary guidelines according to land-use history and species composition. Our findings confirm the importance of dispersal to species persistence within landscapes comprising successional patch networks.
Biodiversity conservation is increasingly integral to forest management (UNEP 1994; Norton 1996; UNEP 2002) and modified forests are gaining importance in conservation planning. Only half of the world's remaining closed forests are intact (UNEP 2001), while globally the area of plantation is rapidly increasing, with greater reliance on plantations for wood production (WCFSD 1999; FAO 2001) and carbon management (IPCC 2000; Malhi, Meir & Brown 2003). Sustainable forestry management, including enhancing biodiversity in plantation systems, is encouraged by the international certification process (Upton & Bass 1995) and promoted through the United Nations Forum on Forests (UNEP 2002).
The changes caused by the management and growth of a commercial tree crop have some similarities to natural regeneration of forests after fire (Rees & Juday 2002) or wind throw (Palmer et al. 2000). Clearfell forestry involves intensive disturbance at regular but infrequent intervals, a subsequent reduction in light at the forest floor during the closed canopy phase of tree crop growth (Ferris et al. 2000) and more frequent, less intensive disturbance during thinning. In temperate Europe, coniferous plantations differ greatly from native deciduous woodland, including changed annual light regimes from deciduous to evergreen canopies, even age structures and lack of graded ecotones (Peterken 2001). Ground flora species must either persist through the various stages of the management cycle or disperse spatially or temporally to exploit physically suitable patches, even if these are only available for a relatively short time. In this respect, lessons from clearfell forestry have application to other successional patch management systems such as reedbeds, burning of heather moorland and farmland crop and set-aside rotations.
This study related understorey plant species composition and community dynamics to environmental and ecological factors in a clearfell conifer plantation forest in the UK. We examined the effects of previous land use, soil pH, changes in stand structure over the silvicultural cycle and clearfelling on species richness. The role of life history and dispersal strategies in determining colonization and persistence were explored and we tested the hypothesis (Freckleton et al. 2005) that the relationship between abundance and occupancy depends upon dispersal mechanism (as a proxy for colonization ability).
Thetford Forest, located in the Breckland region of eastern England (0°40′E, 52°27′N), is the largest lowland commercial forest in the UK. It was planted in the early 20th century on extensive heathland and marginal agricultural land. Soils range from rendzinas to podzolized deep sands, with very localized areas of gleys and peat. Scots Pinus sylvestris and Corsican Pinus nigra ssp. laricio pine comprise 83% of the productive area of the forest.
Thetford Forest is managed by clearfell rotation, and consists of even-aged harvest and replanting units subdivided by a ride network into stands (mean 3·2 ha ± 3·3 SD). Coarse woody debris is removed after clearfelling and, where infection by the fungal pathogen Heterobasidion annosum is indicated, stumps are also removed. Planting lines are ploughed and the stand is replanted, usually the following winter. After initial herbicide application prior to planting, stands receive further treatment with broad-spectrum or grass-specific herbicide (glyphosate or atrazine) during the first 4 years of growth; approximately 90% of stands aged 4–5 years in 2003 had received two to four applications post-planting. Thinning begins at 23–25 years, continuing every 5 years until the stand is clearfelled, usually 60–80 years after planting. The predominant soil type of each planted stand is known from soil series maps (Corbett 1973).
Individual stands of different growth stages were sampled to recreate a chronosequence. Following Hemami, Watkinson & Dolman (2005), we classified stand age as: restock (1–5 years after planting), prethicket (6–10 years), thicket (11–20 years), pole (21–30 years), prefell (31–60 years) and retained (mature stands, greater than 60 years old). All stands up to pole age are second rotation plantings, prefell stands include both first and second rotation and all retained stands are the original planting.
Three-hundred and twenty-six stands planted with P. sylvestris (89 stands) or P. nigra (237) were surveyed between 30 April and 15 July 2001. Stands were selected across the planting–growth–harvest cycle, and included 15 older stands established before formal afforestation commenced in 1922. The week of sampling was stratified across stand age, geographical forest block (n = 13), soil type and tree crop species, to include all available combinations and avoid bias from seasonal changes in vegetation (the minimum number of stands of any one combination was one, because of isolated soil types, and the maximum was eight).
Within each stand, four evenly distributed sample points were selected, with one located at the estimated centre of each quarter of the stand. At each sample point, the composition of the ground vegetation (below one metre height) was recorded in one 3 × 3-m quadrat, by visually estimating percentage cover of all plant species as well as litter and bare ground. Where vegetation was multilayered (e.g. with a Pteridium aquilinum overstorey), total cover recorded exceeded 100%. Nomenclature follows Stace (1997) for vascular plants, Paton (1999) and Smith (2004) for bryophytes. Canopy cover was measured using a hemispherical densiometer, and litter depth to the decomposed humus layer was measured to the nearest centimetre, at each of the four sample points.
A single combined measure of soil pH was taken for each stand. A cylindrical soil core of 4·75 cm diameter and 5 cm depth, not including undecomposed leaf and needle litter or superficial root mat material, was taken from the centre of each quadrat. The four samples from each stand were combined and the pH of 100 cm3 mixed with 100 cm3 of distilled water was measured using an electronic pH meter.
Analyses of environmental variables and species composition were performed using mean values for each stand, while for species richness we used the total number of species recorded in each stand. Structural variables (bare ground, litter cover and depth and canopy cover) covaried and were combined by unrotated principal component analysis using SPSS version 11, with arcsine square-root transformation of percentage values. PCA1 was the only axis with an eigenvalue greater than one (2·77) and described 69% of the total variation. PCA1 scores were therefore used as a proxy for stand structure. Correlations with PCA1 were: litter depth 0·89, canopy cover 0·91, litter cover 0·86 and bare ground −0·65. Species composition was analysed by correspondence analysis (CA) using canoco version 4·5, with square-root transformation and downweighting of rare species.
Relationships of stand species richness and composition to soil pH, PCA1 score, previous land-use, stand size (ha) and crop type (P. sylvestris or P. nigra) were considered using general linear models (type III SS). Species richness was square-root transformed to satisfy conditions of normality and homoscedasticity. pH and PCA1 scores were included as independent variables as they were not correlated (R = −0·03, n= 326, P= 0·63). Minimal models were obtained by sequential variable removal, with retention determined by the change in residual variance (F-test, P < 0·05).
Vascular plant species were classified by life history and dispersal mechanism, from Salisbury (1961), Grime, Hodgson & Hunt (1988) and the Ecoflora database (Fitter & Peat 1994). Dispersal mechanisms were classified by seed structure, first between two groups according to whether the species had long-distance dispersal (plumed, winged or animal-dispersed) vs. either short distance (e.g. ants) or no dispersal mechanism, then among nine more detailed groups. Bryophytes were divided between acrocarps and pleurocarps in life-history analyses: acrocarps tend to be short-lived species of disturbed habitats and pleurocarps longer lived species from less-disturbed habitats. Species Ellenberg scores adapted for the UK (Hill et al. 1999) were used to indicate associations with soil fertility and pH and were analysed by Spearman rank correlation as scores were ordinal.
The land-use of each stand prior to afforestation was determined using tithe records for 1836–1840 and the Ordnance Survey 1st edition (1838). We distinguished between arable, woodland and heath (including heathland, scrub and other grassland). Within Breckland, the extent of cultivation was maximal in the first half of the 19th century, followed by widespread arable abandonment during the latter half of the 1800s because of agricultural depression (Sheail 1979; Sussams 1996). It is therefore likely that most areas identified as heath in 1836–1840 remained heathland prior to afforestation in the early 20th century, although many areas recorded as arable at 1840 were probably abandoned to regenerate as grass heath prior to afforestation (Sheail 1979). The extent to which species occurred in a greater number of previously heath than arable stands was considered as the ratio of the proportion of previously heath stands a species occupied to the proportion of previously arable stands occupied.
species richness, persistence and colonization through the management cycle
A total of 217 plant species was recorded in the survey of forest stands (see Table S1 in the supplementary material), with the minimum and maximum number of species recorded in any one stand being two and 51, respectively. Species richness was highest and most variable (seven to 51 species) in stands 3–15 years old (Fig. 1a). After this period, richness declined to a mean of eight (SD 3·7) at around 25 years. There was a subsequent increase coinciding with the onset of thinning after 25 years; species richness at 40–49 years was approximately two-thirds of that at 5–6 years. The apparent slight decline from 45 to 142 years was non-significant (r = −0·164, P= 0·114, n= 94).
Canopy cover, litter cover and litter depth all increased rapidly from 7 to 8 years, reaching a maximum (95%, 80% and 3·8 cm, respectively) at approximately 25 years (Fig. 1b,d). All three variables then decreased concurrently with the increase in species richness (Fig. 1a). The percentage of bare ground was greatest (40%) in stands aged 1–2 years, but declined rapidly to zero at 7–8 years, remaining zero for the remainder of the growth cycle (Fig. 1c).
The minimal model (Table 1a) showed that species richness decreased with increasing PCA1 score and was significantly higher in stands with higher soil pH. Soil pH was reduced under older Pinus stands, particularly on the two calcareous soil types (rendzina and calcareous brown earth, with stand age as covariate, = 0·224, soil type F1,111 = 22·37, P < 0·001; age F1,111 = 11·88, P = 0·001; B=−0·017 ± 0·005 SE) but less so on the acidic soils (podzol, lithosol, brown earth, = 0·078, soil type F2,184 = 7·90, P= 0·001; age F1,184 = 3·70, P= 0·056; B=−0·005 ± 0·003).
Table 1. Minimal general linear models for species richness and vegetation composition, as defined by the first two axes of a CA performed on all plant species. B shows the direction and magnitude of the effect of each retained variable (original units, unstandardized). Species richness counts were square-root transformed. Removal of the land-use parameter did not significantly affect the pH coefficient in species richness (t322 = 0·20, P= 0·421) or CA1 (t322 = 0·28, P= 0·389) models
B (± SE)
(a) Richness ( = 0·38)
−0·46 ± 0·04
0·20 ± 0·03
Previous land use
0·05 ± 0·19
−0·18 ± 0·19
Without land use: pH
0·23 ± 0·03
(b) CA1 ( = 0·35)
−0·24 ± 0·05
−0·28 ± 0·04
Previous land use
−0·30 ± 0·20
0·21 ± 0·20
Without land use: pH
0·35 ± 0·04
(c) CA2 ( = 0·37)
0·40 ± 0·05
−0·34 ± 0·03
Stands that were previously arable had more species, and stands that were previously heath fewer species, relative to stands that were previously woodland (Table 1a). The pH of previously arable stands (mean 5·12 ± 1·52 SD, n= 150) was higher than that of previously heath stands (mean 4·00 ± 0·98, n = 158; t306 = 7·76, P < 0·001). However, the coefficient of pH in the model of species richness did not change significantly on removal of land use, indicating that effects of previous land use were largely independent of pH.
Stand size and crop species were not retained (P > 0·05) in the minimal model. Crop species did not affect the PCA1 score (t324 = 0·12, P = 0·907) and was not confounded with previous land use (Fisher exact test P = 0·359). Plant species richness at years 4 and 5 was not related to the number of broad-spectrum herbicide applications received (partial correlation controlling for pH, R= 0·031, d.f. = 29, P= 0·869).
Only 50 species persisted throughout the management cycle as vegetative plants (Fig. 2) to be recorded in all growth stages. These included pleurocarpous bryophytes (e.g. Hypnum jutlandicum and Kindbergia praelonga) and shade-tolerant perennials (e.g. Dryopteris dilatata and Deschampsia flexuosa; Hill et al. 1999). Of the remaining 167 species, 101 recolonized after felling in stands aged 1–20 years, nine at the pole stage (20–30 years) and 57 in mature (> 30 years) stands after the canopy reopened.
Those species that attained the highest cover in the stands in which they occurred were widespread (i.e. found in many stands) and those found in fewer stands had lower cover in those stands (Fig. 3). A notable outlier was box Buxus sempervirens, which only occurred in two stands but dominated one of those (66% cover). As B. sempervirens is planted locally for game cover, subsequent analyses of abundance and occupancy excluded this species. The most frequent three species were Holcus lanatus (205 stands), Kindbergia praelonga (195 stands) and Rubus fruticosus agg. (187 stands). The species attaining the highest mean cover in stands where present were Buxus sempervirens (35%), Deschampsia flexuosa (17%) and Holcus lanatus (11%). Thirty-eight species occurred in only one stand, and 132 species in 10 or fewer stands.
Seed-bearing vascular plant species that possess a mechanism for long-distance dispersal were not found in a greater number of stands to those lacking such a mechanism, although the difference was close to significant (with, mean 21·0 stands ± 33·6 SD, n= 102; without, 15·0 ± 28·8, n= 76; t-test on log frequency, t1,176 = 1·74, P= 0·083). The slope of the relationship between log frequency and log abundance did not differ between seeding species with (B = 0·74 ± 0·11 SE) and without (B = 0·58 ± 0·13) a long-distance dispersal mechanism (t176 = 0·92, P= 0·18). Within a generalized linear model that assumed a common slope, an additive categorical variable for long-distance dispersal mechanism was again close to significance ( = 0·28, log abundance B= 0·67 ± 0·08, F1,174 = 64·52, P < 0·001; long-distance dispersal B= 0·15 ± 0·08, F1,174 = 3·61, P= 0·059). For those species not present throughout the growth cycle, for which colonization ability was expected to more strongly affect regional frequency, the slope of the relationship between log frequency and log abundance differed significantly between seeding species with (B = 0·59 ± 0·12, n= 83) and without (B = 0·18 ± 0·14, n= 64) a long-distance dispersal mechanism (t146 = 2·21, P= 0·014). No such difference was found for species that persisted throughout the management cycle (with long-distance dispersal mechanism, B= 0·30 ± 0·09, n= 18; without, B= 0·38 ± 0·18, n= 12, t29 = 0·40, P= 0·346).
Sample scores from the first two axes of the CA ordination were used in analyses of species composition. CA axes one and two had eigenvalues of 0·51 and 0·34, respectively, and accounted for 9·7% and 6·4% of the total species variation. Stands with lower CA1 scores had higher pH and higher PCA1 values (Table 1b), while stands with higher CA2 scores had higher PCA1 values and lower pH (Table 1c).
All growth stages had a wide range of stand scores on CA1 (Fig. 4), although stands in older age classes tended towards lower scores (Table 2a), in accordance with the negative relation between CA1 and PCA1 (Table 1c). Restock and prethicket stands occurred on the lower part of CA2 (Fig. 4 and Table 2a), with CA2 scores increasing through thicket and pole stages to prefell and retained stages, in accordance with the positive relation between CA2 and PCA1 (Table 1c).
Table 2. Mean CA axis 1 and 2 scores (±1 SD) of (a) age classes (sample ordination), (b) life histories (species ordination) and (c) dispersal strategies (species ordination). Means of groups differ on both CA axis one and two for age classes (CA1 F5·320 = 13·46, P < 0·001; CA2 F5,320 = 24·49, P < 0·001), life-history categories (CA1 F6,212 = 5·38, P < 0·001; CA2 F6,212 = 13·15, P < 0·001) and dispersal strategies (CA1 F8,214 = 4·30, P < 0·001; CA2 F8,214 = 7·20, P < 0·001). Tukey HSD post-hoc homogeneous subsets (P < 0·05) are shown as superscripts
Mean CA1 score (± 1 SD)
Mean CA2 score (± 1 SD)
(a) Stand age class (sample ordination)
Restock (1–5 years, n= 63)
0·60 (± 1·01)a
−0·55 (± 1·11)a
Prethicket (6–10 years, n= 36)
0·45 (± 0·93)a
−0·68 (± 0·66)a
Thicket (11–20 years, n= 45)
0·13 (± 0·91)ab
Pole (21–30 years, n= 35)
−0·23 (± 0·86)bc
0·21 (± 0·56)bc
Prefell (31–60 years, n= 95)
−0·43 (± 0·79)c
0·69 (± 1·05)c
Retained (= 61 years, n= 52)
−0·31 (± 0·98)bc
0·50 (± 1·06)c
(b) Life-history group (species ordination)
Acrocarpous bryophytes (n = 23)
0·22 (± 0·79)a
−0·73 (± 0·70)ac
Pleurocarpous bryophytes (n = 13)
0·07 (± 0·64)ab
0·10 (± 0·56)d
Woody plants (n = 30)
−0·24 (± 0·83)ab
0·18 (± 0·71)d
Winter annual (n = 20)
−0·50 (± 0·51)b
−1·55 (± 0·63)ab
Iteroparous herbs (n = 93)
−0·51 (± 0·67)b
−0·61 (± 0·80)ac
Summer annual (n = 34)
−0·56 (± 0·67)b
−0·99 (± 1·05)abc
Semelparous perennial herbs (n = 4)
−0·83 (± 0·50)ab
−0·77 (± 0·54)abcd
(c) Dispersal strategy (species ordination)
Bryophytes and pteridophytes (spores) (n = 39)
0·12 (± 0·73)ns
−0·29 (± 0·88)ab
Wind, winged (n = 22)
−0·18 (± 0·96)ns
−0·24 (± 0·69)ab
Ants (n = 13)
−0·22 (± 0·80)ns
−0·70 (± 1·33)ab
Wind, capsules held aloft (n = 16)
−0·36 (± 0·65)ns
−1·01 (± 0·65)a
Wind, plumed (n = 18)
−0·37 (± 0·45)ns
−1·18 (± 0·46)a
Explosive (n = 3)
−0·43 (± 0·643)ns
−0·42 (± 1·06)ab
Animals, externally (n = 39)
−0·55 (± 0·73)ns
−0·72 (± 0·85)ab
No apparent mechanism (n = 47)
−0·59 (± 0·55)ns
−1·01 (± 0·76)a
Animals, internally (n = 26)
−0·65 (± 0·61)ns
0·16 (± 0 .89)b
Most annual and winter annual species were only present at high cover values in the first 5 years after replanting, with the exception of Ceratocapnos claviculata and Galium aparine. The number of annual species per stand decreased slightly (B = 0·06 ± 0·01 SE) but significantly with stand age in the first 20 years ( = 0·09, F1,143 = 14·42, P < 0·001). Herbaceous perennials and trees and shrubs achieved their highest cover after the pole stage, but this was dominated by a few species, and many species had relatively high cover before the pole stage.
Life-history categories differed in their location on CA axes one and two (Fig. 5 and Table 2b). Annuals, acrocarpous bryophytes and semelparous perennials tended to have lower CA1 and CA2 scores, consistent with the low CA2 scores of restock and prethicket stands, while iteroparous herbs were widely scattered. In contrast, bryophytes had relatively high CA1 scores, acrocarpous bryophytes had low CA2 scores and pleurocarpous bryophytes high CA2 scores. Woody plants also had high CA2 scores.
Species locations within the ordination were also related to dispersal mechanism (Table 2c). Species dispersed by animals externally (ectozoochorous), wind-capsule dispersed species and those with no physical dispersal mechanism occurred lower on CA1 and CA2. Species with plumed seeds had low CA2 scores, occurring in younger growth stages. Species dispersed by animals internally (endozoochorous) had low CA1 and high CA2 scores (Table 2c), consistent with the location of prefell and retained stands within the ordination.
Previous land use significantly affected vegetation composition, as represented by CA1 score. In minimal models also including PCA1 and soil pH (Table 1b), stands that were previously arable had lower CA1 scores, and stands that were previously heath had higher CA1 scores, than stands that were previously woodland. The coefficient of pH did not change significantly on removal of the land-use term, suggesting that effects of previous land use were not artefacts of covariance with pH.
Considering just the more frequent species (found in 20 or more stands), many showed a significant bias in their distributions towards stands that were previously either arable or heathland (Table 3), although none was entirely restricted to a single land-use type. Nineteen of the 28 frequent species that showed a significant bias were present at all stages of the life cycle. Considering vascular species for which Ellenberg values were available, the tendency to occur in more stands that were previously heath than were arable was strongly negatively related to Ellenberg values for soil reaction (i.e. having lower pH) and soil fertility (less nitrogen-rich) [R (reaction) Rs = −0·353, P < 0·001, N (fertility) Rs = −0·375, P < 0·001, n= (181)].
Table 3. Species that were recorded in 20 or more stands that occurred in a significantly different (greater or fewer) proportion of heath than arable stands (chi-square test with Bonferroni correction to P≤ 0·002). Species are arranged in order of the ratio of the proportion of stands occupied on previously heath land to the proportion of stands occupied on previously arable land (163 heath stands, 151 arable stands). n, total number of stands occupied; *, species found in all stand age classes. Ellenberg scores (Hill et al. 1999): N (fertility) ranges from 1, indicating the most infertile soils, to 9, the most fertile; R (reaction) ranges from 1, very acidic, to 9, very base-rich soils
The species found more often on previously arable stands did not include many arable weeds. Galium aparine and Reseda lutea are arable weeds, but are also associated with a range of open habitats, and Galium aparine with hedgerows. Instead, most species were associated with higher soil fertility (e.g. Stachys sylvatica and Urtica dioica), and some were associated with woodland or hedgerows (e.g. Acer pseudoplatanus, Crataegus monogyna, Hedera helix and Geranium robertianum). The species found more often on stands that were previously heath included species associated with infertile and acidic soils, such as Galium saxatile and Deschampsia flexuosa.
A total of 217 plant species was recorded in the managed stands of Thetford Forest, of which only 50 persisted throughout the management cycle. The maintenance and pattern of species diversity within the forest are subject to a wide range of influences, including the nature of the regional species pool, environmental filters to colonization and species interactions (Lawton 1999).
species richness and environment
The clearfell management cycle and the development of the tree crop exerted strong influences on the ground flora of stands, through their impact on stand structure and conditions. The most species-rich stands were those that had been recently felled, were of high soil pH and had been arable land prior to the planting of the forest 80 years previously; they typically contained in the order of 25 species (maximum 57).
Mean canopy cover was lowest and bare ground greatest in restock stands. After the restock and prethicket stages, mean total plant species richness decreased by 60% as litter accumulation and canopy cover increased to a maximum during the late thicket and early pole stages, 15–25 years after planting. Effects of litter could not be distinguished from synchronous changes in canopy cover; however, litter affects species richness and composition directly in British conifer plantation (Ferris et al. 2000), boreal spruce (Økland et al. 2003) and temperate deciduous (Sydes & Grime 1981; Fredericksen et al. 1999) forests.
After thinning began at 25 years, canopy cover decreased to around 64%, litter cover and depth decreased and species richness increased, with 57 plant species colonizing. However, species richness in prefell and mature retained stands only reached two-thirds of the species richness of restock and prethicket stands, with vegetation of mature stands dominated by just a few perennial species. The vernal herbs that form an important component of both ancient and secondary deciduous woodland in Britain (Rose 1999) are infrequent in Thetford Forest, as phenological escape is not possible in the year-round shade of older stands of Pinus monoculture. Rather, the vegetation in stands with a high canopy cover include shade-tolerant species such as pleurocarpous bryophytes (e.g. Kindbergia praelonga and Pseudoscleropodium purum) and pteridophytes (e.g. Dryopteris dilatata), as seen elsewhere in British coniferous plantations (Hill 1979; Ferris et al. 2000). The most abundant angiosperm species in prefell stands were perennial grasses (e.g. Deschampsia flexuosa and Holcus lanatus), which attained up to 75% cover in some stands.
In addition to the effects of clearfelling and thinning, stands with higher soil pH had greater and more variable species richness. Sites with a higher pH had more winter annuals, and fewer bryophytes, trees and shrubs. Species richness is also greater on more calcareous soils in plantations on ancient woodland sites (Kirby 1988). In contrast, Ferris et al. (2000) found plant species diversity decreased slightly but non-significantly with increasing pH in conifer plantations, with species richness more closely related to soil fertility (NH4, Mg and Ca). We found that the pH of calcareous soils was reduced underneath older trees, where accumulation of acidic litter is greater and leaching may have occurred (Hornung 1985; Reich et al. 2005).
Certain species were associated with stands that were previously either arable or heath. Previous land use affects forest vegetation composition (Motzkin et al. 1999; Wulf 2004) but less so in upland British plantations (Hill & Jones 1978) because of similarities between open and woodland ground vegetation. Effects of previous land use occur via species persistence in the buried seed pool (Bossuyt & Hermy 2001), persistence in the above-ground flora and impacts upon environmental conditions (Walker et al. 2004). Some of the species most strongly associated with a specific land-use may have persisted through the growth cycle (19 out of 28; Table 3). However, others are unlikely to have persisted in situ in either the above-ground flora or the buried seed pool (see below). For such species, the most likely explanation for the association with previous land use is a long-lasting effect of cultivation on soil pH or fertility. This is especially true for highly dispersive species. For example, the bryophyte Campylopus introflexus was only introduced to the British Isles in the 1940s (Richards & Smith 1975) but was, nevertheless, widespread in stands that were previously heath.
The degree of association of species with previously arable vs. heath land use was related to Ellenberg scores for soil fertility and reaction. Differences in forest plant species composition between historic arable and pasture land has been related to persistent effects of arable land use on soil fertility (Verheyen & Hermy 2001). In Breckland, from the 16th century planting of nitrogen-fixing fodder crops, livestock grazing and marling (applying chalk) were used in arablization of even the most acidic podzols (Martelli 1952; Sheail 1979; Sussams 1996).
persistence and dispersal strategies
Less than a quarter of the species recorded persisted in the vegetation throughout the entire management cycle, while 167 (77%) recolonized at some stage, either from the buried pool of seeds and other propagules or via dispersal. Persistent species included pleurocarpous bryophytes (e.g. Hypnum jutlandicum and Kindbergia praelonga) and shade-tolerant perennials (e.g. Dryopteris dilatata and Deschampsia flexuosa; Hill et al. 1999).
Of those 101 species that colonized young stands, 82 were angiosperms. Of these, 36% had physical structures associated with wind dispersal, 27% had structures to attract animals (e.g. berries) or attach to fur (e.g. awns) and 10% were dispersed by ants. However, 27% (22 species) had no physical dispersal adaptation, despite the expectation that dispersal-limited plants should be disadvantaged in clearfell forestry (Matlack & Monde 2004). While 67 of the species that colonized young stands are recorded as having a persistent seed bank (Thompson, Bakker & Bekker 1997), only 12 have been shown to have buried seed longevities exceeding 40 years (the minimum time between canopy closure and clearfelling). Ten of these 12 were among the 22 colonists without a physical dispersal adaptation. However, establishment from the original seed pool following the first cycle of felling and replanting does not guarantee that the seed pool will then be replenished sufficiently to persist through the next cycle. For species restricted to young stands, there is only a short window of opportunity for establishment and reproduction, so over repeated rotations overall population size could decrease (Hill & Stevens 1981).
Bryophytes and pteridophytes accounted for 18% of the total species richness of young stands, with 16 species persisting from retained stands and a further 18 species colonizing. There can be a persistent diaspore bank of bryophytes under conifer forest, which behaves in a similar way to an angiosperm seed bank (Jonsson 1993).
macro-ecological patterns of abundance-occupancy
The regional species pool has a strong influence on local species richness (Lawton 1999). About 400 species have been recorded in systematic surveys of heathland and arable habitats in the Breckland region (Eycott 2005), including a large complement of stress-tolerant species dependent on disturbed ground (Dolman & Sutherland 1992). Of these 400 species, 282 were recorded in the forest (including ride verges; Eycott 2005) and 217 in managed stands. There is very little ancient woodland in Breckland; approximately half of the 123 plant species recorded in Wayland Wood (2 km from the forest boundary) were recorded in managed stands of Thetford Forest, but these were mostly common species and the regional deciduous woodland specialist flora is still poorly represented.
A large number of species occupied very few stands and were found with low cover values where present, while the most widespread species had the highest mean abundance within stands. Similar positive interspecific relationships between abundance and occupancy have been reported for a wide range of taxa, including plants (Gaston et al. 2000). Although explanations include statistical artefacts, range position, niche breadth or resource use (Gaston, Blackburn & Lawton 1997), much recent attention has focused on patterns arising from dispersal within spatially structured populations (Hanski 2000; Kean & Barlow 2004; Freckleton et al. 2005). Overall, we did not find a significant difference in the abundance-occupancy relationship between species with and without a mechanism for long-distance seed dispersal, although dispersive species tended to be more frequent. However, when those species capable of persisting through the growth cycle were excluded to consider just those with temporally discontinuous local populations, the slope of the abundance-occupancy relationship differed between good and poor dispersers, with species possessing a mechanism for long-distance dispersal increasing in frequency more for a given increase in local abundance. This supports the importance of dispersal in generating positive abundance-occupancy relationships in spatially structured populations. For persistent species the observed abundance-occupancy pattern may instead be the result of habitat quality (Thompson, Hodgson & Gaston 1998; Freckleton et al. 2005).
Previous studies have focused on the negative impacts of coniferization in ancient semi-natural woodlands (Thomas, Kirby & Reid 1997; Peterken 2001) or enhancing biodiversity in woodlands created for amenity and conservation (Harmer et al. 2001; Dolman & Fuller 2003). In contrast, the biodiversity potential of non-native conifer forests managed for timber production has received less attention, although see Hill (1979) and Ferris et al. (2000), who found similar patterns of species richness with stand age to those reported here. We have shown that a clear-fell plantation system planted on land with a prolonged history of open ground can support substantial plant diversity, but not necessarily of woodland specialist species. No species recorded in the mature stands were of conservation concern in Britain (Church et al. 2001; Cheffings & Farrell 2005), while the three red data-listed species recorded were open-ground species found in stands less than 10 years old (Arabis glabra, Viola tricolour and Filago vulgaris).
Native plant species of open and infertile habitats have declined throughout the managed countryside of the UK (Preston et al. 2002). Maximizing plant species richness is therefore an appropriate conservation aim in commercial forestry systems. This will be achieved by maximizing the area of managed forest in growth stages that reflect non-woodland habitats in terms of physical environment (shade, open ground, disturbance). Increasing the area of forest in young growth stages by shortening rotation length may conflict with timber production objectives but be compatible with carbon management. Similarly, reducing retention of mature stands would help maintain plant diversity in this system but conflicts with management guidance to increase structural diversity and provide conditions for species associated with mature forest (Humphrey et al. 2002; Forestry Commission 2004; Parlane et al. 2006).
More than three-quarters of the species recorded are excluded from shaded stands during the growth cycle. That dispersal is important for local persistence of such a large part of the flora has implications for conservation in other managed landscapes comprising shifting successional patch networks. Juxtaposition of young stands is currently excluded in UK forestry guidance (Forestry Commission 2004), where the main provision for open-ground species is through permanent open space. However, many ruderal species are actually restricted to planted stands as the trackway and open space network is not adequately disturbed (Eycott 2005). Increasing contagion between restock and prethicket patches could therefore facilitate dispersal, colonization and the maintenance of diversity.
Our recommendations conflict with presumptions in sustainable forest management towards greater use of continuous cover silviculture and mimicking natural disturbance regimes (MCPFE 1993; Forestry Commission 2004). However, for biodiversity conservation it may be appropriate to vary guidelines according to regional land-use history and species composition.
We thank Forest Enterprise (East Anglia District) for permission to work in Thetford Forest and access to the forest management GIS database and are grateful to Tom Williamson and Kate Skipper for the land-use data. A. E. Eycott was funded by NERC (NER/S/A/2000/03322).