1 We studied the migration of forest plant species using their percentage cover and frequency in 197 plots distributed over 26 transects across ecotones between ancient and recent deciduous forests in the Meerdaal forest complex in central Belgium. The recent forest stands varied in age between 36 and 132 years, and all occurred on silty, well-drained soils.
2 The total cover, number and diversity of field layer species did not differ significantly between ancient and recent forest stands.
3 The number and cover of the ancient forest plant species and of ant-dispersed species correlated positively with the age of the recent forest and negatively with both the duration of its former agricultural land use and the distance to the ancient forest. This implies a slow colonization of the recent forest stands by these species; all species were, however, able to migrate across the ecotones.
4 The cover of four species (Anemone nemorosa, Lamium galeobdolon, Convallaria majalis and Polygonatum multiflorum) declined along the transect, suggesting that they are limited by seed dispersal. Their colonization rates, calculated from the occurrence of the farthest individual, ranged from < 0.05 to 1.15 m year−1 and for other measures from < 0.05 to 0.65 m year−1. Anemone and Lamium appeared to colonize the recent forest by establishment of isolated individuals, while Polygonatum and Convallaria expanded populations from existing patches on the border between ancient and recent forest.
5 Several forest species were able to colonize the recent forest rapidly, where some of them even reached a higher abundance, due to the increased availability of colonization sites with a higher nutrient content and a thinner organic layer.
Numerous studies point to qualitative differences in plant species composition between ancient and recent forests ( Peterken 1977; Rackham 1980; Hermy & Stieperaere 1981; Peterken & Game 1984; Whitney & Foster 1988; Dzwonko & Loster 1990; Matlack 1994). These differences can be attributed to the limited colonization capacity of species characteristic of ancient forest. Colonization takes place by vegetative propagation and/or seed production, seed dispersal and seedling establishment. Most of these studies focus on the plant species composition of recent woods, which form part of a fragmented landscape ( Dzwonko & Loster 1992; Dzwonko 1993; Matlack 1994; Grashof-Bokdam 1997). They agree that species composition and richness of recent woods depend on the distance from the source of diaspores, the area of the recent forest patch (often a substitute for habitat diversity), and its shape and age. Wind-dispersed (anemochores) and animal-dispersed species (endozoochores) seem to be more abundant in these isolated recent woods than ant-dispersed species (myrmecochores) and species without a specific dispersal agent (barochores), leading to the assumption that the distribution of many forest species is determined by their limited dispersal range.
The extremely low colonization capacity of ancient forest species means that they can be considered to be indicators of the species richness of forests ( Peterken 1974, 1977). The current European Union policy is to promote reforestation, and a better understanding of the colonization of recent forests by ancient forest species is essential if we are to be able to predict future vegetation composition. Where ancient and recent forest are contiguous there is no physical barrier to dispersal and species (including those that rely on vegetative propagation and short distance seed dispersal) can move freely into the recent forest without passing through hostile territory.
Few quantitative studies have discussed colonization patterns of forest species across the border between ancient and recent forest, although this approach allows calculating of species’ migration rates. However, Matlack (1994) studied five different sites with an ancient–recent forest ecotone in northern Delaware and southern Pennsylvania; ages of the recent forest varied between 13 and 54 years (mean = 28) and the colonization rates ranged from 0.08 to 2.5 m year–1. Similarly, for 12 recently established forests in southern Sweden with an age of 30–75 years, colonization rates ranged from 0.0 to 1.0 m year–1 ( Brunet & von Oheimb 1998). In both these studies colonization rates differed significantly between dispersal modes such that ingested zoochores > adhesive zoochores < anemochores ≥ myrmecochores ≥ barochores. Vegetative propagation did not appear to have a significant impact on the migration rate. A question that remains, however, is whether patterns and migration rates will be similar in forests of different ages. One could indeed expect the migration rate to increase with age, as the establishment of plant species leads to larger population sizes and thus greater reproductive capacity and opportunities for dis- persal. A further problem with both studies was that the sites compared probably had different soil properties. Indeed, Dzwonko & Gawrónski (1994) found that, after 70 years of reforestation, vegetation composition in recent forests depended more upon soil conditions, light and the influence of dominants than on the modes of species dispersal. As differences in migration may be partly caused by habitat differences among sites, we limited our study to forest ecotones on a single soil type, but included a much larger age range (36–132 years).
The purpose of this study was:
1 to determine whether there are species that are confined either to the ancient or to the recent parts of contiguous forest stands;
2 to determine whether changes in vegetation patterns across ancient–recent forest ecotones can be explained by limited seed dispersal or vegetative propagation of ancient forest species or whether they are caused by limited seedling establishment of such species;
3 to model the colonization patterns of species that are characteristically confined to ancient forest stands and to calculate colonization rates across the ancient–recent forest ecotones.
Materials and methods
The Meerdaal forest complex in central Belgium is a 2224 ha remnant of a vast forest that once covered large parts of Belgium ( Tack et al. 1993 ). Soils are generally well drained and are formed from Pleistocene loess (Podzoluvisols, Luvisols) of variable depth ( Rampelberg & Deckers 1995). The total plant species richness here is exceptionally high for a Belgian forest, because most are fragmented or subjected to major external disturbances and only those on sandy or water-logged soils have escaped reclamation for agriculture.
We digitized historical maps (1771, 1859, 1872, 1893, 1908, 1930, 1954, 1970 and 1982) of the Meerdaal forest complex and used a Geographical Information System (GIS) approach (using the Arc/Info software; ESRI 1990) to compare them. Several stands on the edge of the forest complex had been reclaimed between 1771 and 1982, but after a shorter or longer period of arable use they were reforested ( Fig. 1). We define ‘ancient forest’ as areas that have been continually forested since the first systematic maps of De Ferraris (1770–1778, scale 1/18 000). A detailed soil survey, however, suggested that continuous forest cover had been present for a much larger period ( Bossuyt et al. 1999 ). The term ‘recent forest’ refers here to areas where forest has been re-established after 1770.
Recent stands on the southern edge of the complex were characterized. We then selected those with a silty, well-developed and well-drained soil that exhibited a high degree of similarity in forest cover with the contiguous ancient forest. Habitat differences relating to soil type or forest structure and management were therefore minimized. The border between ancient and recent forest stands was clearly recognizable as a small bank, with a height of maximum 30 cm, representing the edge of a former field. Thirteen such recent stands, ranging in age from 36 to 132 years were available ( Fig. 1 and Table 1).
Table 1. Deforestation and reforestation dates of the 13 selected recent forest stands and number of sampled quadrats in the two transects
Number of forest stand
Date of deforestation
Date of reforestation
Number of 3×3 m quadrats
Vegetation survey and historical data collection
Two transects were randomly positioned across the ecotone for each of the 13 stands. Each transect extended from 21 m into the ancient forest stand to 20 m from the edge of the recent forest (to avoid effects from the surrounding land). A variable number of quadrats (each 3 × 3 m) was established along the transects, as shown in Fig. 2, with at least 10 quadrats in the recent forest ( Table 1).
The cover of all individual plant species found in each 3 × 3 m quadrat was estimated in spring and early summer 1997, using a modified decimal scale ( Londo 1976). In total, 197 quadrats were surveyed yielding 50 different species (Appendix 1 on the World Wide Web; for address see the cover of a current issue of Journal of Ecology).
We defined three land-use variables: ‘age of the recent forest’ (the number of years since reforestation; AGE), ‘duration of the agricultural land use’ (the number of years the forest stand was under agriculture between 1771 and 1982; DURAGR) and ‘minimal distance of the quadrat to the ancient forest’. We calculated ‘standardized distance’ (STDDIST) for each quadrat by dividing its distance to the ancient forest by the age of the recent forest This allowed comparison of the vegetation composition of quadrats from forests of different age. Quadrats in the ancient forest were arbitrarily assigned 250 years as AGE (the minimum value based on historical maps) and a negative STDDIST. De- or reforestation was assumed to occur in the middle of the period between successive maps.
Vegetation composition of each quadrat was characterized by the following vegetation variables: sum of the cover of all individual species (COVTOT), total number of species (NUMTOT), Shannon–Wiener diversity index ( Kent & Coker 1995) (DIV), sum of cover and number of ancient forest species (as defined by Honnay et al. 1998 ) (COVANC and NUMANC), sum of cover and number of myrmecochorous species (COVMYR and NUMMYR) and sum of cover and number of ornithochorous species (COVORN and NUMORN) (see Appendix 1). No analysis was executed on the barochorous, epizoochorous and anemochorous species because the cover of these species groups was too small to give any significant results.
Data analysis consisted of three consecutive steps.
1 A vegetation community approach: analysis of differences in vegetation composition between ancient and recent forest stands and with distance within the recent forest stands.
2 An individual species approach: differences between ancient and recent forests in individual species’ patterns were then investigated for all species occurring in at least 10 quadrats.
3 The calculation of colonization rates and modelling of colonization patterns for those species showing decreasing cover along the transects.
Relationships between the land-use variables and the vegetation variables were calculated using a Spearman rank correlation coefficient ( Siegel & Castellan 1988). Trendlines were fitted with Microsoft Excel 5.0 ( Microsoft Corporation 1993) to establish mathematical relationships between vegetation variables and standardized distance.
In the individual species approach, we calculated the mean cover and the frequency of each species occurring in at least 10 quadrats throughout the ancient and recent forest. A Mann–Whitney and a chi-square test were used to detect species that occur relatively more either in ancient or in recent forest stands ( SPSS 1995). Because the matrix of species × quadrats contains a lot of empty cells, a point-biserial correlation coefficient ( Kent & Coker 1995) was calculated between species cover and land-use variables (AGE, STDIST and DURAGR). Ellenberg values for these species were also noted ( Ellenberg et al. 1991 ).
To determine migration rates, the standardized distance (STDIST) was divided into classes (class width 0.1 m year–1) and mean cover and frequency of the negatively correlated species were calculated for each class. The colonization rate of these species was assessed using four importance parameters: (i) occurrence of the furthest individual; (ii) furthest occurrence of half the peak frequency ( Matlack 1994); (iii) furthest occurrence of half the peak mean cover; and (iv) furthest occurrence of half the peak mean cover multiplied by the frequency (weighted cover). The first two parameters are based on the presence of species, but because the colonization process includes both arrival of individuals and establishment of populations, we also decided to include parameters that reflect cover (iii and iv). Moreover, a comparison of colonization rates based on different parameters helps to ascertain the colonization pattern. In order to describe the relationships between importance parameters and standardized distance, trendlines were fitted with Microsoft Excel 5.0 ( Microsoft Corporation 1993) and this derived a colonization model for the individual species.
A significant positive correlation was observed between the age of the recent forest (AGE) and the number and cover of the ancient forest species (NUMANC and COVANC) and of ant-dispersed species (myrmecochores) (NUMMYR and COVMYR) ( Table 2), suggesting that their importance increases with the age of the recent forest. Duration of agricultural land use (DURAGR) and standardized distance (STDIST) were negatively correlated with the same species groups ( Table 2), implying a slow colonization of the recent forest stands by such species. The total cover (COVTOT), number of species (NUMTOT), cover and number of bird-dispersed species (ornithochores) (COVORN and NUMORN) and total species diversity (DIV) did not vary with any of the land-use variables ( Table 2). The form of the typical relationships between the cover of myrmecochores (COVMYRM) and ornithochores (COVORN) species and standardized distance is shown in Fig. 3.
Table 2. Spearman rank correlation coefficient between land-use variables and vegetation variables (n = 197)
Twenty-one species were included in the second analysis. Significant differences in distribution pattern favouring ancient (a) and recent (r) forest stands ( Table 3) were found for relatively few species: Acer pseudoplatanus L. (a), Anemone nemorosa L. (a), Convallaria majalis L. (a), Polygonatum multiflorum L. (a), Rubus fruticosus L. (r) and Urtica dioica L. (r). There were no species where migration did not take place across the ecotone ( Table 3).
Table 3. Mean cover and frequency of the species occurring in at least 10 quadrats in ancient (n = 52) and recent forest plots (n = 145) and in the total sample (n = 197)
Standardized distance was significantly negatively correlated with presence of five species [the species favoured in ancient forest ( Table 3) plus Lamium galeobdolon], while nine species were significantly positively correlated ( Table 4). Curves were fitted for those species having a significantly negative point-biserial correlation coefficient ( Fig. 4), except for Acer pseudoplatanus which was omitted because adults had been planted in the recent forest stand. Best fits were negative logarithmic (Anemone nemorosa, Lamium galeobdolon) or exponential (Convallaria majalis, Polygonatum multiflorum) functions. Colonization rates of these four species, which are assumed to be dispersed limited, ranged from 0.25 m to 1.15 m year–1 when calculated from the furthest individual, and from < 0.05 m to 0.65 m year–1 based on the other parameters ( Table 5).
Table 4. Point-biserial correlation coefficient between the presence/absence of species occurring in at least 10 quadrats and the land-use variables (n = 197) and the Ellenberg value for light and nitrogen ( Ellenberg et al. 1991 )
P < 0.0001
0.0001 < P < 0.001
0.001 < P < 0.01
0.01 < P < 0.05 (
) 0.05 < P < 0.1 NS: not significant. x = indifferent ? = unknown
Table 5. Colonization rates (m year–1) of four ancient forest species, calculated based on the furthest individual, the furthest occurrence of half the peak cover, the furthest occurrence of half the peak frequency and the furthest occurrence of half the peak weighted cover
Half peak cover
Half peak frequency
Half peak weighted cover
DIFFERENCES BETWEEN ANCIENT AND RECENT FOREST
The vegetative differences between ancient and adjacent recent forest stands were more qualitative than quantitative. The total number of species and the total cover did not differ along the transect. This result broadly agrees with the findings of Peterken (1974) and Hermy (1994), but contrasts with those of Matlack (1994) and Brunet & von Oheimb (1998). Moreover, all individual species were found in both types of stand and values for the Shannon–Wiener diversity index were comparable. On the other hand, myrmecochorous species and species that had previously been identified as ancient woodland species had a higher cover and frequency in ancient forest. Thus, up to 132 years after restoration of forest cover, forest stands developed on old field still differed from forest that had never been cleared.
Vegetative differences between ancient and adjacent recent forest could be explained by continuing soil differences generated during the period of cultivation, which could influence the germination, establishment and growth of herbaceous species. Our soil survey showed that recent forest soils were significantly less acid (pH KCl 3.3, cf. 3.1) and were covered by less litter (4.27 cm depth, cf. 5.07 cm) than ancient forest soils ( Bossuyt et al. 1999 ). Some species typical of ancient forest (Lamium galeobdolon, Anemone nemorosa) tolerate deep litter ( Persson et al. 1987 ; Eriksson 1995; Holderegger 1996) while other competing species respond to the reduced litter and implied higher fertility of recent forest. The ancient forest species will therefore be even less favoured species, and the differences between forest types reinforced ( Persson et al. 1987 ).
Species having a bias to ancient forest are characterized by a limited colonization capacity due to limited dispersal, either by seed or vegetative propagation, or hampered seedling establishment. Although colonization of the recent forest could originate from a population in the interior of the ancient forest, we assumed that it started at the border and thus, as in the study of Brunet & von Oheimb (1998), actual migration rates may be underestimated. As in other studies ( Peterken & Game 1984; Matlack 1994), there were no species that could not grow in the recent forest. If a species is more common in the ancient forest, (i) the soils could be unsuitable or there may be strong competition so that species survive but do not grow well, or (ii) the species is colonizing, although at a very slow pace. If such a species is mainly confined to the part of the recent forest that is closest to the ancient forest, it could again be that colonization takes place very slowly, but variation in soil can be discounted due to the uniform nature of the former agricultural land use of the recent forest. Declining cover along the transects therefore suggests dispersal limitation, although colonization of these species may also be influenced by hampered seedling establishment because many species are both seed and microsite limited ( Eriksson & Ehrlën 1992) . The remaining species, whose occurrence is scattered over the recent forest stand, are considered to be limited mainly by seedling establishment.
The significant differences between myrmecochorous and ornithochorous species in their correlation with the land-use variables ( Table 2) indicate that differences in vegetation patterns of contiguous ancient–recent forest stands can be at least partly explained by dispersal. In temperate deciduous forests, myrmecochorous species have a lower dispersal capacity than ornithochorous, endozoochorous or anemochorous species ( Willson et al. 1990 ; Matlack 1994; Brunet & von Oheimb 1998).
When Ellenberg values for nitrogen ( Ellenberg et al. 1991 ) of species that were significantly correlated were plotted against the correlation coefficient, those with negative relationships had the lowest indicator values ( Table 4 and Fig. 5). This suggests that recent forest species have a stronger growth response to increased nitrogen availability compared to ancient forest species, which are more stress tolerant ( Grime 1979). The newly established forest is more likely to be invaded by ruderal or competitive species through germination of buried or newly dispersed seeds than by slowly migrating ancient forest species. As can be seen in Table 4, several of the species occurring more at larger distances from the ancient forest have high Ellenberg values for light (Sambucus nigra, Adoxa moschatellina, Glechoma hederacea, Ajuga reptans). These species have difficulty in expanding their population in ancient forests due to high cover by ancient forest species, which are better adapted to low light, poor nutrient levels and a thick organic layer. Nevertheless, they have a capacity for quick colonization of the newly established forest, where the light stress imposed by the developing recent forests is partly compensated by significantly higher nutrient levels and a faster decomposition of organic matter compared with the ancient forest. These species also have high Ellenberg values for nitrogen and are less well adapted to germination and seedling establishment in thicker organic layers than species confined to the ancient forest stands ( Persson et al. 1987 ; Eriksson 1995; Holderegger 1996). However, our data did not enable us to distinguish between the effects of changed inherent chemical soil characteristics and increased competition on seed germination and seedling establishment of ancient forest species.
Colonization rates and pattern
Seed dispersal curves, unlike colonization curves, are well characterized and are in most cases best described as inverse power or negative exponential functions ( Okubo & Levin 1989; Harper 1977; Willson 1993). The colonization pattern of the recent forest stands by myrmecochores was negatively logarithmic ( Fig. 3), as it was for the two myrmecochores shown to be mainly limited by seed dispersal (Anemone nemorosa and Lamium galeobdolon) (cf. Brunet & von Oheimb 1998). The other two dispersal-limited species (Convallaria majalis and Polygonatum multiflorum) were best fit by a negative exponential fit ( Fig. 4). As most fits had high R2-values, we assume these models are suitable for describing our data.
Colonization rates were very low ( Table 5), with a maximum of 1.15 m year–1 for Lamium galeobdolon. For this particular species, the colonization rates calculated by presence parameters (individual and frequency) were much higher than the rates based on cover, suggesting that the colonization process occurs by establishment of isolated individuals. In contrast, the colonization rate of Polygonatum multiflorum based on cover exceeded the one based on frequency, so we consider this species to be colonizing the recent forest by expansion of patches on the border between ancient and recent forest. For Anemone nemorosa and Convallaria majalis, the colonization rates were highest based on the furthest individual due to the presence of individual outliers. Colonization rates were in the same range but differed to some extent from those determined for the particular species in the study of Brunet & von Oheimb (1998). Matlack (1994) suggests that species with a negative logarithmic colonization model (in this case Lamium galeobdolon and Anemone nemorosa) expand their population by establishment of isolated individuals, and those with a negative exponential fit (in this case Convallaria majalis and Polygonatum multiflorum) would indicate colonization by the advance of a wave front. This interpretation confirms that from colonization rates for Lamium galeobdolon and Polygonatum multiflorum.
The higher colonization rate based on weighted cover of Lamium galeobdolon compared with other species probably reflects increased importance of vegetative propagation following initial colonization. However, clonal propagation and seed banks have often been considered to have no great effect on colonization capacity, in comparison with seed size (influencing seedling establishment) and dispersal mode ( Matlack 1994; Fröborg & Eriksson 1997), even for forest plant species (e.g. Anemone nemorosa) ( Eriksson 1995; Holderegger 1996). Two ornithochorous species (Convallaria and Polygonatum) seem not to have the scattered pattern that is typical of most such species ( Figs 3 and 4), because of the scattered availability of perching and storage microsites for birds in a forest stand ( McClanahan & Wolfe 1987; Hoppes 1988; Masaki et al. 1994 ). Most ornithochorous species are tree or shrub species and the modest size of the herbaceous Convallaria and Polygonatum may reduce their availability to dispersal vectors and thus give them a colonization pattern more like a barochorous species. Brunet & von Oheimb (1998) suggest that clonal species may show an exponential decrease in cover, indicating colonization in a wave front, and such propagation may be another explanation for the lack of a scattered pattern in Convallaria and Polygonatum.
The ancient–recent forest ecotone indeed appears to be a starting point for colonization by forest species. However, the colonization of ancient forest species, which is of great importance in nature conservation ( Peterken 1977), is extremely slow (less than 0.1 m year–1). Qualitative differences in the vegetation can still be found after more than a century of colonization, even when ancient and recent stands are contiguous. Furthermore, a considerable number of species is able to colonize the recent forest quickly and may even reach a higher abundance in recent forest stands (Stachys sylvatica, Circaea lutetiana, Geum urbanum, Glechoma hederacea, Ranunculus ficaria, Adoxa moschatellina, Ajuga reptans). This is probably due to the increased availability of suitable colonization sites with higher nutrient levels in newly established forests. Colonization of more isolated fragments of recent forest may depend, however, on whether these species can bridge the longer distances involved ( van Ruremonde & Kalkhoven 1991; Grashof-Bokdam 1997).
We thank Dr G.F. Peterken, O. Honnay and two anonymous referees for their valuable comments on an earlier version of the manuscript, and A. Verheyen and P. Campling for improvement of the language. The field work would not have been possible without the assistance of B. Degroote, E. Swinnen and R. Wallays and the kind permission of the Forestry Administration Leuven. The research was supported financially by a Research Assistant grant from the Fund for Scientific Research, Flanders (FWO).
Received 2 July 1998revision accepted 7 January 1999