Migration of vascular plants to secondary woodlands in southern Sweden

Abstract

1 We studied the migration of field layer plants across ecotones between ancient woodlands and recent deciduous woods on former arable land varying in age between 30 and 75 years.

2 Number and percentage cover of woodland species in recent woods decreased with increasing distance to the ancient woods, and increased with the age of the recent woods, indicating dispersal limitation during secondary succession.

3 Colonization by typical woodland plants was observed in 183 of 200 species × site combinations. In 72 combinations, a colonization front was characterized by logarithmic or linear decrease in species cover, indicating establishment of isolated individuals and gradual infill of gaps. This pattern was most common in ant-dispersed species and less frequent in species with adhesive or ingested seeds.

4 Migration rates were calculated for 49 woodland species. Mean migration rates based on maximum cover in recent woods varied from 0.00 to 1.00 m year⁻¹ between species, with a median migration rate of 0.30 m year⁻¹. Migration rates calculated on occurrence of the farthest individual ranged from 0.00 to 1.25 m year⁻¹, with a median rate of 0.53 m year⁻¹.

5 Ant-dispersed species had lower migration rates based on maximum cover compared with species with adhesive or ingested seeds. No differences between dispersal modes were found when comparing migration rates based on the farthest individuals. Most of the calculated migration rates (84%) exceeded the rate of possible vegetative spread of woodland species.

6 Establishment of a field layer vegetation in secondary woods comparable to that of the adjacent ancient stands proceeded at a rate of c. 0.3–0.5 m year⁻¹. We conclude that scale and intensity of temperate forest management should be adjusted to the relatively slow migration of the field layer flora in order to enable complete recovery during a management cycle.

Introduction

Under natural conditions, temperate forests are often characterized by small-scale disturbance regimes. Most canopy gaps result from windthrows of single trees or groups of a few trees (Falinski 1986; Peterken 1996). Soil mounds and pits are formed when trees are uprooted, and the activity of digging mammals is another common source of small-scale soil disturbance. Field layer species usually recolonize these disturbed sites within a few years by clonal growth and seedling establishment (Thompson 1980; Beatty 1984; Hughes & Fahey 1991).

Human activity, however, has greatly increased the scale, frequency and intensity of disturbance in many temperate woodlands, and the sensitivity of field layer species to large-scale clear cutting has received considerable attention during recent years (Halpern & Spies 1995; Meier et al. 1995).

The effects of severe disturbance on plant distributions depend partly on the ability of species to survive disturbance and partly on the ability to recolonize disturbed sites after local extinction. Species richness in secondary woods has been shown to be dependent on the distance to ancient woods, indicating dispersal limitation during recolonization (Peterken & Game 1984; Dzwonko & Loster 1992; Dzwonko 1993; Matlack 1994). Seeds of shade-tolerant woodland herbs often lack adaptations to dispersal and simply fall to the ground below the mother plant (Bierzychudek 1982). Furthermore, a relatively large number of species are dispersed by ants, implying dispersal distances of only a few metres (Culver & Beattie 1978; Handel et al. 1981; Kjellsson 1985). Species with ingested or adhesive seeds, adaptations to long-distance dispersal, are generally considered to be the best colonizers among woodland plants (Dzwonko & Loster 1992; Dzwonko 1993; Matlack 1994).

Species that are sensitive to disturbance and lack adaptations for long-distance dispersal may be unable to recolonize larger areas fully in the time between two disturbance events. It is, however, difficult to study the ability of plant species to recolonize disturbed forest areas. Species that are not visible above-ground after disturbance may survive below-ground as rhizomes, bulbs or buried seeds (Dierschke 1988; Hughes & Fahey 1991; Peterken 1993), and in these cases we are unable to determine if occurrence of a species is due to regrowth or recolonization. Abandoned fields initially do not contain any diaspores of typical woodland plants (Brown & Warr 1992) and the occurrence of woodland species in such sites is therefore entirely controlled by their ability to spread vegetatively, together with their capacity for seed dispersal and seedling establishment (Matlack 1994). Plantations on abandoned fields and clear-cut forest areas are both characterized by disturbed soils, initial lack of a tree canopy and presence of light-demanding ruderals in the field layer. Plantations may therefore serve as model areas for the study of recolonization of clear-cut forests, and provide valuable information on the ability of plants to tolerate local extinction after severe disturbance.

Matlack (1994) calculated migration rates ranging from 0 to 2.5 m year⁻¹ across ancient–recent woodland ecotones for north-east American forest plant species. Comparable data for European species are lacking and information is available for only a few species, such as Anemone nemorosa, Hyacinthoides nonscripta, Mercurialis perennis and Primula elatior (Rackham 1980; Falinski & Canullo 1985).

The aim of our study was to analyse migration of European woodland species to recent woodlands. The following questions were addressed. How quickly do woodland plants disperse and establish at the local scale? What is the relative importance of seedling establishment compared with clonal growth for the migration pattern observed? Are differences in migration rates explained by adaptations of species to seed dispersal and vegetative spread? What are the consequences of modern land management for local scale population dynamics of woodland plants?

Materials and methods

Study area

The study was conducted in deciduous woodlands in the southern Swedish provinces of Skaåne, Smaåland, Blekinge and Öland(Table 1). Important early successional canopy trees in this area are silver birch (Betula pendula) and aspen (Populus tremula). Ash (Fraxinus excelsior), pedunculate oak (Quercus robur) and hornbeam (Carpinus betulus) may also establish during early phases of secondary forest succession, but have a longer individual life span. Late successional forests on intermediate soils are often dominated by beech (Fagus sylvatica) in the southern part of the study area and by Norway spruce (Picea abies) in the northern part. Mature stands on richer soils are characterized by wych elm (Ulmus glabra) and small-leaved lime (Tilia cordata), while pedunculate oak and Scots pine (Pinus sylvestris) usually dominate on dry soils (Sjörs 1967). In spring, the field layer on richer soils includes profusely flowering geophytes, e.g. Anemone spp., Gagea ssp. and Corydalis spp. The field layer is characterized by other geophytes, e.g. Convallaria majalis and Mercurialis perennis, and by hemicryptophytic herbs and grasses during the summer (Diekmann 1994).

Table 1. Characteristics of the studied woods. Dominant trees: A, Alnus glutinosa; B, Betula pendula; F, Fraxinus excelsior; Q, Quercus robur; U, Ulmus glabra. Stand origin: P, planted; N, natural

Study site	Geographical positionlatitude longitude	Age(years)	Area(ha)	Dominanttrees	Standorigin	Length per transect in recent wood (m) (no. of transects)
Skaåne 1	55°32′N 13°11′E	75	1.0	Q	P	100 (5)
Skaåne 2	55°32′N 13°11′E	75	1.1	Q	P	50 (6)
Skaåne 3	55°32′N 13°11′E	70	0.7	Q	P	50 (4)
Skaåne 4	55°55′N 13°55′E	35	1.1	B, A	P	50 (5)
Skaåne 5	55°55′N 13°55′E	35	1.0	B	P	50 (5)
Blekinge 1	56°11′N 15°19′E	30	2.4	Q	P	35 (6)
Blekinge 2	56°09′N 15°19′E	45	0.9	Q, F	P	50 (4)
Smaåland 1	56°38′N 14°10′E	35	0.9	Q, U	N	35 (6)
Smaåland 2	56°38′N 14°10′E	35	0.9	Q, U	N	35 (6)
Öland1	56°37′N 16°27′E	40	0.5	Q, F	N	50 (3)
Öland2	56°34′N 16°27′E	55	0.7	B, F	N	50 (3)
Öland3	56°34′N 16°26′E	57	1.5	B, F	P	60 (5)

Field work

Twelve recently established deciduous woods, for which we could determine the year of origin by interviewing the landowners, were studied. Old land survey maps revealed that all the sites had been arable land prior to woodland establishment. The size of the woods varied from 0.5 to 2.4 hectares and the age of their stands, calculated from the year of tree planting (eight sites) or the beginning of secondary succession (four sites), varied from 30 to 75 years (Table 1). The tree layer consisted of planted pedunculate oak at five sites, of which one site also contained planted ash and another site naturally established ash. Three sites were planted with silver birch, of which one contained naturally established black alder (Alnus glutinosa), and four sites were natural mixed stands with mainly pedunculate oak, silver birch, ash and wych elm (Table 1). All were situated directly adjacent to ancient deciduous woods, as apparent from old land survey maps. The sites Smaåland 1 and 2 and Öland3 are parts of nature reserves and have been unmanaged during the past few decades; all other sites are managed by repeated canopy thinning. In each case, the border (ecotone) between the ancient and recent woods was straight and consisted of earth walls, low boulder walls, ditches or distinct old ploughing lines that were in accordance with evidence from old survey maps or aerial photographs.

Transects were laid out from the ancient wood into the recent stands perpendicular to the ecotone. Each point on the transects in the recent stands was situated closer to the ancient wood of the transect origin than to any other ancient wood. Between three and six parallel transects were analysed in each of the 12 study sites. The distance between the transects was 5 or 10 m depending on the length of the ecotone. Ten metres of each transect were situated in the ancient wood and, depending on stand size, 35 m to 100 m in the recent wood (Table 1). Circular sample plots of 5 m² were laid out along the transects at intervals of 5 m. Thus, two sample plots were situated in the ancient wood, one plot on the ecotone and all other plots in the recent wood. We studied a total of 716 plots on 58 transects, 113 plots in the ancient woods, 58 plots on the ecotone and 545 plots in the recent woodland stands.

Percentage cover was estimated visually for all vascular plants of the field layer in each plot. Vernal species were recorded between 24 April and 9 May 1995; all other species were studied between 26 June and 24 July 1995.

Data analysis

Typical species of the woodland interior that are rarely found outside woodlands were identified with the aid of Oberdorfer (1990), who provides a detailed ecological characterization of species. Nomenclature of species was also according to Oberdorfer (1990). Additional information was compiled from two local Swedish floras (Weimarck & Weimarck 1985; Sterner 1985). Fifty-five species in the field layer at our study sites were classified as typical woodland species. Colonization was analysed for each site where a species occurred at four or more sample plots. This resulted in 200 species × site combinations, involving 49 species (Table 2), for which we measured the distance from the ecotone to the plot in the recent wood with the farthest individual and the distance to the plot with maximum cover (Matlack 1994). Migration rates for a species along each transect were calculated by dividing these distances by the age of the wood, and the average migration rate at each study site by calculating the mean value for all transects (excluding transects lacking the species).

Table 2. Means and standard errors of migration rates (m year⁻¹) of woodland field layer species in recently established deciduous woods in southern Sweden. Only data from where a species occurred in four or more sample plots are included. Species are ranked according to migration rates based on maximum cover, with data based on farthest individual also given. Data on maximum vegetative spread and on seed dispersal vector are according to the literature. The number of sites where migration fronts showed a significant linear or logarithmic decrease in cover (best fit) is indicated. The first number in parentheses indicates the number of sites where a species had reached maximum cover at one of the furthest plots. The second number indicates the number of sites with occurrence of a species at one of the furthest plots

Maximum cover		Farthest individual		Migration front
Species	Mean	SE	Mean	SE	Clonal growth	Seed dispersal	log	lin	n
Lathyrus niger	1.00	-	1.00	-	-	Auto			1 (0, 0)
Elymus caninus	0.94	-	1.25	-	-	Adhesive			1 (1, 1)
Rubus saxatilis	0.79	-	0.88	-	−300 cm	Ingested			1 (0, 1)
Brachypodium sylvaticum	0.77	0.04	0.93	0.16	-	Adhesive			3 (2, 3)
Hedera helix	0.72	0.05	0.72	0.05	>100 cm	Ingested			2 (1, 1)
Epipactis helleborine	0.67	-	0.67	-	-	Wind		1	1 (0, 0)
Scrophularia nodosa	0.58	-	0.58	-	-	None			1 (1, 1)
Polygonatum multiflorum	0.56	0.05	0.63	0.06	−10 cm	Ingested	1		5 (1, 3)
Viola mirabilis	0.53	0.07	0.68	0.13	-	Ants			2 (0, 0)
Festuca gigantea	0.51	0.17	0.75	0.19	-	Adhesive	1		3 (1, 3)
Poa nemoralis	0.47	0.09	0.63	0.12	-	None			11 (7, 7)
Stellaria holostea	0.46	0.10	0.91	0.08	−30 cm	None	4		8 (3, 8)
Viola riviniana/reichenb.	0.44	0.07	0.67	0.09	-	Auto/ants	1	1	9 (3, 7)
Circaea lutetiana	0.43	0.10	0.73	0.10	−30 cm	Adhesive			3 (1, 3)
Adoxa moschatellina	0.42	0.07	0.59	0.18	−30 cm	Ingested		1	4 (1, 2)
Melica nutans	0.42	0.24	0.42	0.24	−10 cm	Ants			3 (1, 1)
Milium effusum	0.40	0.18	0.55	0.18	−20 cm	None			5 (3, 3)
Pulmonaria obscura	0.39	0.09	0.68	0.15	−5 cm	Ants	2	1	7 (1, 4)
Dryopteris carthusiana	0.38	0.06	0.44	0.08	-	Wind	1	1	5 (0, 2)
Mercurialis perennis	0.36	0.08	0.73	0.15	−15 cm	Auto/ants	4	1	7 (2, 4)
Listera ovata	0.34	0.08	0.57	0.31	-	Wind			2 (0, 1)
Maianthemum bifolium	0.33	0.08	0.34	0.08	−30 cm	Ingested	1		6 (0, 0)
Paris quadrifolia	0.33	0.07	0.37	0.05	−15 cm	Ingested	1		3 (0, 0)
Polygonatum verticillatum	0.33	-	0.62	-	−5 cm	Ingested			1 (0, 0)
Carex sylvatica	0.30	-	0.48	-	−5 cm	Ants		1	1 (0, 0)
Stachys sylvatica	0.30	0.23	0.30	0.23	−30 cm	None			2 (0, 0)
Convallaria majalis	0.28	0.09	0.43	0.14	−30 cm	Ingested	1	1	8 (1, 3)
Ficaria verna	0.28	0.06	0.63	0.11	-	Ants	6	1	10 (2, 3)
Lathyrus vernus	0.26	0.13	0.39	0.26	-	Auto			2 (1, 1)
Melica uniflora	0.26	0.06	0.48	0.11	−10 cm	Ants	1	1	6 (1, 2)
Oxalis acetosella	0.26	0.06	0.43	0.07	−15 cm	Auto	2	1	7 (1, 2)
Athyrium filix-femina	0.25	0.06	0.42	0.10	-	Wind	1		5 (1, 2)
Corydalis intermedia	0.25	0.05	0.53	0.13	-	Ants	3		6 (1, 2)
Gagea lutea	0.25	0.10	0.34	0.10	-	Ants		2	5 (1, 1)
Dentaria bulbifera	0.22	0.08	0.34	0.10	−10 cm	Ants	1		6 (0, 0)
Gagea spathacea	0.22	0.07	0.23	0.08	-	Ants	2		4 (0, 0)
Lamium galeobdolon	0.22	0.05	0.50	0.20	−160 cm	Ants	2		3 (0, 0)
Veronica montana	0.21	0.07	0.35	0.21	−10 cm	Ants			2 (0, 0)
Anemone nemorosa	0.20	0.04	0.85	0.09	−5 cm	Ants	9	2	12 (1, 9)
Galium odoratum	0.20	0.05	0.29	0.11	−30 cm	Adhesive	1		2 (0, 0)
Hepatica nobilis	0.19	0.03	0.69	0.16	-	Ants	4		4 (0, 3)
Dryopteris filix-mas	0.18	0.07	0.30	0.12	-	Wind	1		6 (0, 2)
Actaea spicata	0.15	0.15	0.15	0.15	-	Ingested			2 (0, 0)
Campanula latifolia	0.13	0.13	0.13	0.13	-	None	1		3 (0, 0)
Stellaria nemorum	0.12	0.02	0.23	0.08	>100 cm	None	3		4 (0, 0)
Anemone ranunculoides	0.11	0.02	0.72	0.24	−5 cm	Ants	3		3 (0, 3)
Circaea intermedia	0.04	-	0.07	-	−30 cm	Adhesive			1 (0, 0)
Carex digitata	0.00	-	0.00	-	-	Ants			1 (0, 0)
Gymnocarpium dryopteris	0.00	-	0.00	-	−10 cm	Wind			1 (0, 0)

We assumed that the ecotone was the starting point for species migration into a recent wood, although it is unlikely that all species were present at the forest edge when the wood established. It is also unlikely that all species colonized from the nearest part of the ancient wood. The method used in our study therefore gives migration rates that may be lower than the actual migration rates of a particular species.

A population was considered to be advancing if regression analysis showed a significant decrease between observed maximum cover and the farthest individual in the recent wood. The advancing edge of a population was then characterized by fitting linear, exponential or logarithmic functions to the relationship between the maximum cover distance and the farthest individual in the recent wood (Matlack 1994). Negative logarithmic distributions would indicate that populations advance by the establishment of isolated plants, with later plants filling in between them.

Since comprehensive sources are lacking, data on the theoretical dispersal mode and on the vegetative spread of species were compiled from a large number of studies in the literature (Table 2). If required, a list of the studies upon which this information is based can be obtained from the authors. Modes of seed dispersal included adaptations to short-distance dispersal (ants, ballistic, combined ants/ballistic, no known dispersal vector) and adaptations to long-distance dispersal (wind, adhesive, ingested). The influence of different adaptations to seed dispersal and the effect of vegetative spread on mean migration rates (n = 49) were tested with the two-sample t-test, as distributions of migration rates were similar to normal distributions (Zar 1996). Due to small sample size, species with ballistic dispersal (n = 3) and with combined ballistic and ant dispersal (n = 2) were excluded in these comparisons. The effects of distance from the ecotone and of stand age on total number and cover of woodland species were analysed by linear regression analysis. These analyses included only the sample plots at the ecotone and in the recent wood.

Results

Species richness and abundance

The number of woodland species in the recent woods decreased with increasing distance from the ancient woods at all study sites (Table 3 and Fig. 1). Distance explained between 13% and 67% of the observed variation in woodland species numbers between plots. At eight sites, species richness of woodland plants was highest on the ecotone (three sites) or in parts of the recent woods (five sites) mostly close to the ecotone. In only two of the study sites was the highest mean number of woodland species found in plots of the ancient woods, whereas another two sites had maxima at several distances.

Table 3. Coefficients of determination (r²) according to linear regression between the number and the total cover of typical woodland species in recent woods (including ecotone) and the distance (m) from the ecotone. The slopes of the regression lines are shown in parentheses

Study site	r ^2	P	n
Number of species vs. distance
Skaåne 1	0.13 (−)	0.002	74
Skaåne 2	0.37 (−)	<0.001	66
Skaåne 3	0.52 (−)	<0.001	44
Skaåne 4	0.67 (−)	<0.001	55
Skaåne 5	0.67 (−)	<0.001	55
Blekinge 1	0.53 (−)	<0.001	48
Blekinge 2	0.26 (−)	<0.001	44
Smaåland 1	0.55 (−)	<0.001	48
Smaåland 2	0.28 (−)	<0.001	47
Öland1	0.28 (−)	0.002	33
Öland2	0.39 (−)	<0.001	33
Öland3	0.34 (−)	<0.001	56
Total cover of species vs. distance
Skaåne 1	0.01 (+)	0.787	74
Skaåne 2	0.25 (−)	<0.001	66
Skaåne 3	0.35 (−)	<0.001	44
Skaåne 4	0.57 (−)	<0.001	55
Skaåne 5	0.44 (−)	<0.001	55
Blekinge 1	0.42 (−)	<0.001	48
Blekinge 2	0.01 (+)	0.626	44
Smaåland 1	0.15 (−)	0.007	48
Smaåland 2	0.17 (−)	0.004	47
Öland1	0.68 (−)	<0.001	33
Öland2	0.28 (−)	0.002	33
Öland3	0.58 (−)	<0.001	56

Figure 1

Mean species richness and mean species abundance of typical woodland plants as related to distance from the ancient–recent woodland ecotone at the sites Skaåne 4, Blekinge 2 and Öland1.

The total cover of woodland species also decreased with the distance from the ancient woods at all sites except Skaåne 1 and Blekinge 2 (Table 3 and Fig. 1). Between 15% and 68% of the observed variation in woodland species cover was related to distance from the ecotone at these sites. At only three sites was the observed maximum cover percentage found in plots of the ancient woods. Species cover was highest at the ecotone at two sites and in part of the recent wood at the remaining seven sites.

Data from 30 to 35 m into the recent wood, the furthest point available for all sites, were compared. The number of species relative to the adjacent ancient wood increases with age of the recent wood. The total cover of typical woodland species also increased with age (Fig. 2). The results indicate that approximately 70 years were needed for an area 30–35 m into the recent wood to attain the species richness of the adjacent ancient wood.

Figure 2

(a) Relation between mean woodland species richness of recent woodland plots at 30–35 m distance from the ecotone and recent woodland age. Species richness is given as a percentage of mean species richness in the adjacent ancient woodland plots (r² = 0.66; P = 0.002; n = 12). (b) Relation between mean woodland species cover of recent woodland plots at 30–35 m distance from the ecotone and recent woodland age. Species cover is given as a percentage of mean species cover in the adjacent ancient woodland plots (r² = 0.54; P = 0.010; n = 12).

The total number of species (woodland species and other species) did not change significantly along the transects in nine recent woods and increased (P < 0.05) with the distance from the ancient wood at three study sites (Smaåland 1 and 2, Öland1).

Species distributions

Colonization by typical woodland species was observed in 183 of the 200 species × site combinations analysed. Carex digitata and Gymnocarpium dryopteris were the only species where we did not observe any colonization, but both these species only occurred in one of the ancient woods and, even then, at low abundance (Table 2). Other species occasionally showed a lack of colonization, including Actaea spicata, Athyrium filix-femina, Campanula latifolia, Convallaria majalis (one of the study sites shown in Fig. 3a), Dentaria bulbifera, Dryopteris filix-mas, Gagea lutea, Melica nutans, M. uniflora, Milium effusum and Poa nemoralis.

Figure 3

Examples of species abundance as related to distance from the ancient–recent woodland ecotone. Mean cover values of all transects of a site with occurrence of a particular species are given: Convallaria majalis at site Skaåne 4; Hepatica nobilis at Öland1; Lamium galeobdolon at Skaåne 4; Pulmonaria obscura at Öland1; Paris quadrifolia at Blekinge 2; and Milium effusum at Skaåne 3.

Populations were determined to be advancing in 72 of the 183 species × site combinations (39%) where we observed colonization. The advancing edge was best characterized by negative logarithmic functions in 57 distributions and by linear functions in 15 distributions. Negative exponential models did not provide a best fit in any species. When only more abundant species were considered, having mean ground cover of >5% at one or more distances from the ecotone (79 of 183 distributions), a higher proportion (61%) was advancing. The advancing edge of 40 out of 79 populations (51%) showed a negative logarithmic decrease and a further eight populations (10%) decreased linearly. The remaining distributions included mainly species that had colonized the entire recent wood, and species with low abundance and frequency for which we could not determine significant colonization fronts.

Negative logarithmic and, to a lesser extent, linear distributions indicated initial colonization by isolated individuals, with subsequent population increase filling in between them. Species that often show a colonization of this kind are Anemone nemorosa, A. ranunculoides, Corydalis intermedia, Gagea spathacea, Galium odoratum, Hepatica nobilis, Lamium galeobdolon, Mercurialis perennis, Ficaria verna, Stellaria holostea and S. nemorum (Table 2). An example of this pattern is the distribution of the ant-dispersed Hepatica nobilis at the site Öland1 (Fig. 3b). There is a relatively distinct decrease in cover in the recent wood, but scattered individuals were found at greater distances from the ecotone. A similar pattern was observed in Lamium galeobdolon, a species that predominantly spreads by long runners, at the Skaåne 4 site (Fig. 3c). Pulmonaria obscura, while still showing a logarithmic decrease, may exemplify woodland species that colonize from the ecotone and gradually attain considerably higher cover in the recent wood than in the ancient stand (Fig. 3d). The pattern shown by Paris quadrifolia at site Blekinge 2 is probably due to dispersal of ingested seeds from a more distant part of the ancient wood that was not covered by our transects (Fig. 3e). The distribution curve of Milium effusum at site Skaåne 3 provides an example of colonization along the entire transect length with an abundance similar to that of the ancient wood (Fig. 3f).

Migration rates

The mean migration rates of the species studied ranged from no observed colonization to 1 m year⁻¹ when based on distance to maximum cover within the recent wood (Table 2). The three lowest and highest values, however, refer to species found at only one study site. Among the species that occurred at two or more sites, the vernal geophyte Anemone ranunculoides showed the lowest migration rate (0.11 m year⁻¹) and the grass Brachypodium sylvaticum the highest (0.77 m year⁻¹). The median migration rate of all 49 species was 0.30 m year⁻¹.

Migration rates calculated from occurrence of the farthest individual were similar to rates based on maximum cover in slow-colonizing species with generally low abundance. Examples included Actaea spicata, Campanula latifolia and Gagea spathacea (Table 2). Differences were also relatively small in some rapidly colonizing species, where large parts of the recent woods had been colonized at the time of the study. Large differences between the two methods of calculation were often found in species whose distributions showed a negative logarithmic decrease.

As shown by Fig. 1, many woodland species were recorded from the farthest sample plots within the recent woods. The migration rates calculated on the occurrence of the farthest individual are therefore underestimates of the potential migration of many species. For 88 of the 200 species × site combinations, involving 30 species, a woodland species was recorded on at least one of the farthest sample plots of a site (Table 2). The remaining 19 species had not yet reached any of the farthest sample plots. Species that occurred at the farthest plots in a majority of their sites included Anemone nemorosa, A. ranunculoides, Brachypodium sylvaticum, Circaea lutetiana, Festuca gigantea, Hepatica nobilis, Mercurialis perennis, Milium effusum, Poa nemoralis, Pulmonaria obscura, Stellaria holostea and Viola riviniana/reichenbachiana (Table 2). The degree of underestimation is much lower for migration rates based on maximum cover: maximum cover was recorded at one of the farthest sample plots for only 39 of 200 species × site combinations, involving 24 species, and 26 of these cases referred to maximum cover at only one of the farthest plots (Table 2).

Dispersal mode

The number of woodland plants at the study sites was greatest for species with adaptations to local dispersal. Ant-dispersed species were the largest group, with an average of 37.5% of all woodland species at a study site. On the whole, 67.5% of the woodland flora of a study site consisted of species with adaptations to short-distance dispersal. Species with adaptations to long-distance seed dispersal (adhesive, ingested, wind) made up the remaining 32.5%. Twelve of the 16 species with ant-dispersed diaspores had migration rates below the median value (based on maximum cover). Seven of nine species with ingested seeds and four of six species with adhesive seeds had mean migration rates above the median value. Ant-dispersed species had lower migration rates calculated on maximum cover (0.25 m year⁻¹) than species with ingested seeds (0.43 m year⁻¹; P = 0.012) and species with adhesive seeds (0.48 m year⁻¹; P = 0.025). No other significant differences were found between species with different adaptations for seed dispersal.

There were no significant differences in migration rates between species with clonal growth and species lacking means of vegetative dispersal. Among clonal species, migration rates were not related to potential vegetative propagation. Out of 183 calculated migration rates, only 30 rates (16%) were lower than the potential vegetative propagation of the species concerned. However, Galium odoratum, Hedera helix, Lamium galeobdolon, Rubus saxatilis and Stellaria nemorum showed annual migrations that throughout were lower than the potential length of their procumbent stems (Table 2).

Discussion

The general decrease in species richness and abundance with increasing distance from the ancient wood indicates dispersal limitation during colonization of secondary woods. Our results suggest that migration is limited by both slow vegetative propagation and by seed dispersal. Only four of the studied species possess the ability to spread more than c. 30 cm year⁻¹ by lateral growth. Therefore, in most species, rapid colonization depends on effective seed dispersal.

Seed shadows usually have a peak, with a majority of seeds deposited relatively close to the parental plant, and a tail with a few seeds dispersed over longer distances (Willson 1993). Most advancing populations in our study were characterized by the occurrence of isolated individuals beyond a rather distinct colonization front. The establishment of isolated individuals followed by gradual infill between the initial colonists seems to be a common colonization pattern in field layer species in temperate forests (Matlack 1994). The similarity between seed shadows and species distributions suggests that achieved colonization is partly controlled by the seed shadow of an advancing population.

Different adaptations to seed dispersal may influence the location of the peak and the slope and shape of the tail of the seed shadow (Willson 1993). We would expect that species with adaptations to long-distance dispersal migrate faster than species lacking those adaptations. This was partly confirmed when comparing migration rates based on maximum cover, as ant-dispersed species had lower migration rates than species with adhesive and ingested seeds. However, occasional dispersal over longer distances probably occurred to a similar extent in species with different dispersal modes as no differences were found when migration rates were based on the farthest individual.

At the local dispersal scale of our study, with maximum dispersal distances of 50 m in 30–75 years, the effects of adaptations to long-distance seed dispersal were small compared with those related to seed dispersal by ants. Seeds are usually transported up to a few metres by ants (Culver & Beattie 1978; Beattie 1985; Kjellsson 1985), and a minor proportion may be transported up to 10 m (Andersen 1988). The potential of adhesive and ingested seeds to travel farther than ant-dispersed seeds is more important in the colonization of isolated woodlands at greater distances from source populations (Peterken & Game 1984; Dzwonko & Loster 1992; Matlack 1994). Several ant-dispersed woodland species that rarely colonize isolated secondary woods, often described as ancient woodland indicators (Hermy 1994), showed relatively rapid spread in at least some of the recent woodlands studied by us, e.g. Mercurialis perennis, Pulmonaria obscura and Viola mirabilis. We conclude that the relative importance of dispersal mode for successful colonization increases with increased habitat fragmentation.

The frequency of long-distance dispersal events and the fate of these seeds ultimately determines the distribution range of a species. All species studied have migrated to south Sweden from remote refugia after the last glaciation. This recolonization implies average migration rates of at least several hundreds of metres per year. Long-distance dispersal has thus frequently occurred during the past for all the species studied. Despite these high potential migration rates across natural landscapes, local development of a typical woodland field layer vegetation is a slow process, advancing at a rate of some dm year⁻¹. Important factors, besides seed dispersal, that may limit the local advance of populations are the availability of seeds (seed production relative to seed predation), the time period until first seed set, the availability of suitable microsites for establishment, and the vigour of clonal growth (Grubb 1977; Eriksson & Ehrlén 1992). Clonal growth often plays an important role in population development after initial seedling establishment (Oinonen 1971). Clonal species may advance in a wave front described by a negative exponential decrease in cover. Such patterns probably occurred in clones at our sites but were not detected within the study scale.

Implications for woodland management

Disturbed sites in natural deciduous forests, such as mounds and pits after fallen trees or areas dug up by animals, are rarely larger than a few square metres. The migration rates calculated in our study and similar rates calculated by Matlack (1994) for North American species theoretically enable most species to recolonize such disturbed sites within a few years. The long-term maintenance of plant species diversity in temperate forests may depend on periodical small-scale disturbance creating new sites for colonization and increasing habitat heterogeneity (Thompson 1980; Beatty 1984). This view is supported by our results, as species richness of woodland plants showed a peak in the recent wood close to the ecotone at many of the study sites.

Disturbed sites after selective tree cutting are of a size similar to naturally disturbed sites and are readily recolonized by herbaceous woodland plants (Collins & Pickett 1988; Hughes & Fahey 1991). Large-scale clear-cutting and soil scarification may have more negative effects on shade-tolerant species (Meier et al. 1995; Hannerz & Haånell 1997). Tree plantation after clear-cutting produces even-aged stands with high canopy cover and stem density. Severe root competition and very low light flux to the ground in such stands may cause local extinction of field layer species that survived the clear-cutting. Mid-aged conifer plantations in particular are often completely lacking in ground vegetation (Kirby 1988; Peterken 1993). Many typical woodland species do not accumulate persistent soil seed banks (Brown & Oosterhuis 1981; Staaf et al. 1987; Kjellsson 1992). Their re-establishment in periods when growth conditions for the field layer are more favourable is dependent on migration from adjacent areas.

Our data suggest that a century or more may be needed until species that had become locally extinct after a clear-cut have recovered to the original population size at a distance of 50 m from the edge of a clear-cut area. Short management cycles and large clear-cuts imply that sensitive species will not recolonize to predisturbance limits. In a long-term perspective, this may lead to regional impoverishment of the woodland flora in intensively managed forest landscapes.

Management regimes that are close to natural disturbance dynamics may be the most promising way to ensure long-term survival of the field layer flora in managed forests. Reducing the area of logging units may be one way to achieve this in temperate forests. Areas with shelterwood cutting or single stem management usually provide a net of relatively undisturbed microsites where sensitive species survive better than in clear-cut areas (Reader 1987; Brunet et al. 1996; Hannerz & Haånell 1997).

Acknowledgements

This study was financially supported by WWF Sweden. We thank the landowners for valuable information on site history and providing access to the study sites. Martin Diekmann, Ove Eriksson, Lindsay Haddon and two referees provided many helpful suggestions on how to improve the manuscript.

Received 17 January 1997revision accepted 3 November 1997