Kris Verheyen, Laboratory for Forest, Nature and Landscape Research, Catholic University of Leuven, V. Decosterstraat 102, B–3000 Leuven, Belgium (fax +32 16 329760; e-mail email@example.com).
1 Distribution patterns (frequency and percentage cover) of 18 forest plant species were studied in 34 ha of mixed deciduous forest (Muizen Forest, north Belgium). Stands varied in age between 6 and more than 223 years and both slow and fast colonizing species were studied.
2 Detailed land use history data were combined with the species distribution maps to identify species-specific colonization sources and calculate colonization distances.
3 A multiple logistic regression model was constructed with four covariables: pH (which can impose limits on the potential species-distribution), secondary forest age, distance from nearest colonization source and age–distance interaction, to allow us to account for the gradual completion of colonization over time.
4 We could distinguish species which are limited by both dispersal and recruitment (Primula elatior, Arum maculatum and Lamium galeobdolon), mainly by dispersal (Anemone nemorosa, Deschampsia cespitosa), mainly by recruitment (Paris quadrifolia and Polygonatum multiflorum) and by neither (Geum urbanum, Ranunculus ficaria, Glechoma hederacea, Aegopodium podagraria, Ajuga reptans, Adoxa moschatellina and Oxalis acetosella).
5 The low colonizing capacity of ancient forest plants cannot be attributed to a single cause; rather both dispersal and recruitment are limiting but the relative importance varies.
Their low colonizing capacity may be due to these species being unable to disperse to recent forests or being unable to recruit because of low habitat quality. The traditional view of secondary forest succession (Clements 1916) suggests that there will be a relationship between habitat quality and secondary forest age, dependent on the nature and duration of the former land use and on processes related to aggrading forest ecosystems. Former arable land use for instance, generally results in increased mineral nutrient levels, especially of phosphate (Koerner et al. 1997; Wilson et al. 1997; Honnay et al. 1999), while increasing secondary forest age results in decreasing pH and base saturation of the topsoil (Froment & Tanghe 1967; Goovaerts et al. 1990; Bossuyt et al. 1999a) and in increases in a range of factors, including shading, litter accumulation (Muys et al. 1992), topsoil organic matter content (Catt 1994; Verheyen et al. 1999) and saturated hydraulic conductivity (Bossuyt et al. 1999c).
The relative importance of dispersal vs. recruitment limitation has previously been assessed by labour-intensive, long-term experimental introduction of diaspores to unoccupied sites (Eriksson & Ehrlén 1992; Primack & Miao 1992). However, we argue that it can be inferred from species distribution patterns in secondary forests, by taking both distance from diaspore source and secondary forest age into account. Dispersal limitation will generate species distribution patterns that are negatively correlated with the distance from the colonization source and positively correlated with secondary forest age. Furthermore, the importance of distance is likely to decrease over time, as colonization continues within the secondary forest. Recruitment limitation, however, will generate patterns which are positively correlated with age but independent of distance from the colonization source.
Studies have mapped the distribution of forest plant species as a function of distance from the colonization source (Matlack 1994; Brunet & von Oheimb 1998; Bossuyt et al. 1999b), but did not take account of differences in secondary forest age between sites, despite the considerable variation (13–54 years, 30–78 years and 36–132 years, respectively).
Such analyses require areas composed of a mosaic of stands of different ages, such as those found in Flemish forests in general and the study area – the Muizen Forest – in particular. Continued intensive human impact means that Flemish forests are no more stable than any other land use in the region and encompass the whole age range from ancient to very young forests (Tack et al. 1993; Tack & Hermy 1998).
Our aims were to assess the degree to which species are confined to older forests, to develop a method for analysing spatio-temporal distribution patterns by taking into account species-specific colonization sources and to determine the relative importance of dispersal limitation for colonization.
Material and methods
The study was carried out in the 34 ha Muizen Forest (Fig. 1), which is situated 15 km east of Antwerp in a flat and low lying (10 m above sea level) region. Soils are developed in quarternary niveo-eolian sandy loam and silt loam deposits. The presence of a sandy clay layer of tertiary marine origin at approximately 1 m depth limits the drainage and consequently pedogenetic processes are very slow (Gleysols). The local presence of fossil shell lime in the sandy clay layer causes soil acidity to be highly variable with pH (KCl) ranging from 3 in the northern parts of the forest to 7 in the south-west (Verheyen et al. 2001).
Most forest stands consist of planted Populus x canadensis (Moench.). The ground flora of the acid part of the forest belongs to Violo-Quercetum roboris (Oberdorfer 1957) with Deschampsia cespitosa, Pteridium aquilinum and Oxalis acetosella as differential species. The carbonate rich part of the forest belongs to Primulo-Fraxinetum excelsioris (Hermy 1985) characterized by Anemone nemorosa, Adoxa moschatellina, Primula elatior, Ranunculus ficaria, Cardamine pratensis and Filipendula ulmaria. Nomenclature follows Lambinon et al. (1998).
Historical land use, vegetation and soils
Land use was reconstructed using historical maps (1775, 1850, 1865, 1892, 1912, 1922, 1939), aerial photographs (1952, 1961, 1970, 1984, 1990, 1995), the primitive land register (from 1834 onwards), old notarial acts and an historical work (van Berchem 1971). Three land use types were distinguished: grassland, arable land and forest and changes between 1775 and 1998 are shown in Fig. 2. The forested area was at its minimum in the first half of the 19th century but has since increased gradually.
We then delineated parcels which had a common land use throughout the survey period and, for each, calculated the number of years since reforestation, assuming that any land use changes occurred midway between the dates of two maps (cfr. Verheyen et al. 1999). Six classes were distinguished (Fig. 1), with ages of 6–8 years, 17–41 years, 52–67 years, 81–96 years, 119–140 years and more than 223 years. Written records going back to the 15th century (see van Berchem 1971) indicate that at least some of the oldest class were forested much earlier than 1775, and this class is therefore referred to as ancient forest. Furthermore, to account for possible relict populations of forest plant species along former hedgerows and ditches, we identified those boundaries that have not been disturbed since 1775 (Fig. 1, hereafter, ancient forest margins).
All the historical land use maps were digitized using Arc/Info 7.1.2 (ESRI 1997).
During April and May 1998, plant species were mapped using a semiregular 20 × 20 m grid, covering approximately 22 ha, to visualize their spatial distribution across the forest (Fig. 1). The northern part of the forest was not mapped as it floods every winter and spring, resulting in a patchy and limited occurrence of forest plant species. Cells were subdivided if historical land use boundaries crossed them, to give a total of 729 mapped cells. The semiregular grid was digitized using Arc/Info 7.1.2 (ESRI 1997).
For practical reasons, only a selection of 18 forest species were mapped (Table 1). These were selected on the basis of former historical ecological research in the Muizen Forest and elsewhere in Belgium (De Keersmaeker & Muys 1995; Honnay et al. 1998) to include both fast colonizers and slow colonizing or ‘ancient forest’ species. Cover within each cell was estimated using a pragmatic ranked dominance index with the following eight categories: 0, < 1/16, < 1/8, < 3/16, < 1/4, < 1/2, < 3/4, < 1.
Table 1. The mapped plant species, their dispersal mode according to Grime et al. (1988) and their number of occurences. The number of true colonization sources and the number of colonized cells are shown (percentage of each type in parentheses); 729 cells surveyed in total
Number of true colonization sources
Number of colonized cells
Only clonal growth in Belgium (J. Van Assche, personal communication).
As the highly variable pH can impose limits on the potential species-distribution throughout the forest it was necessary to control for this variable during the analysis. Other site conditions (e.g. texture, soil moisture regime) are, however, fairly homogenous throughout the surveyed forest (Verheyen et al. 2001) and were not included. Soil samples were collected during August and September 1998 at 40 by 40 m intervals (156 points on the semiregular grid). The pH(KCl) (1 N, 1 : 5 suspension) was determined on oven-dried samples of the A horizon. In order to assign a pH value to each of the 729 cells, the point data were interpolated by means of kriging (for more details see Verheyen et al. 2001), and cells were classified as acid (pH < 4.5), mildly acid (pH 4.5–5.5) or neutral to alkaline (pH > 5.5).
Species-specific colonization sources were identified by examining all grid cells either contiguous to ancient forest margins or lying in ancient forest parcels (potential colonization sources) for the presence of each species (its true colonization source, Table 1 and as shown for Anemone nemorosa and Ranunculus ficaria in Fig. 3).
Then, for each species, the distance from all other cells to the nearest true colonization source was calculated, using the centre coordinates of each cell. In concordance with the mapping resolution, distances were grouped in six consecutive classes of 20 m: 0–20 m, 21–40 m, 41–60 m, 61–80 m, 81–100 m and > 100 m.
For the statistical analyses, cells within ancient forest parcels as well as cells along ancient forest margins were omitted, as they do not give any information about the colonization process. All statistical analyses were performed with SPSS 9.0 (1999).
The degree to which species are associated with older forest was assessed by means of a logistic regression of the individual species presence/absence scores on secondary forest age. The resulting ‘odds ratios’ that indicate how much more likely it is that a species is present when the secondary forest age is increased by one class (Hosmer & Lemeshow 1989), allowed us to rank the species according to their affinity with older forest. The relative importance of dispersal as a limiting factor for colonization was assessed from a multiple logistic regression of the individual species presence/absence scores against secondary forest age, distance from the true colonization source, the age–distance interaction and pH(KCl). Inclusion of the interaction allowed us to assess whether the importance of distance from the colonization source decreases with increasing secondary forest age. Additionally, for the six species present at at least four different cover classes, Kendall partial rank-order correlation coefficients (Siegel & Castellan 1988) were calculated between cover and distance from colonization sources with the effect of pH(KCl) held constant. This analysis was done seperately for the five age classes.
Two species (Oxalis acetosella and Stellaria holostea) occurred in less than 5% of the grid cells and were not analysed.
Comparison of the distribution maps (Fig. 3) with the secondary forest age map (Fig. 1) indicates that species such as R. ficaria and A. nemorosa differ in their response and this is confirmed by the logistic regression of the species presence/absence scores on secondary forest age (Table 2). Although most species have some affinity with older parcels, their ranking according to the ‘odds ratio’ reveals a clear trend (see also Fig. 4), from those that have only colonized the oldest parcels (high ratios, e.g. Listera ovata) to those also able to colonize younger parcels (e.g. Arum maculatum). Those species with the lowest ‘odds ratio’ are either indifferent to age (e.g. Geum urbanum) or may even be more frequent in younger parcels (e.g. Adoxa moschatellina).
Table 2. Logistic regression of individual species presence/absence on secondary forest age (five classes: 6–8 years; 17–41 years; 52–67 years; 81–96 years and 119–140 years). The species are ranked by decreasing odds ratio
Wald test statistic with 2-sided P-values. ****P < 0.0001, **P < 0.01, *P < 0.05, NS = not significant.
The results of the multiple logistic regression (Table 3, Fig. 4) show that the species distributions of 13 out of 16 species are significantly correlated with pH. Indeed, for Listera ovata, Ornithogalum umbellatum and Brachypodium sylvaticum, inclusion of pH removes the age effect; the affinity with older forest is caused by the covariation between secondary forest age and pH (age–distance and pH–distance relations are identical in L. ovata, Fig. 4). Distance to species-specific colonization sources is the next most important factor for most of the other species with a strong affinity for older forest (Primula elatior, Anemone nemorosa, Deschampsia cespitosa, Arum maculatum and Lamium galeobdolon, but not Paris quadrifolia and Polygonatum multiflorum;Table 3, Fig. 4). For A. nemorosa and D. cespitosa colonization patterns are determined by distance from the colonization source and its interaction with secondary forest age (age per se not significant). The significant interaction between distance and secondary forest age is due to more-or-less complete colonization of older forest parcels (Fig. 4).
Table 3. Multiple logistic regression of the species presence/absence scores on secondary forest age, distance from the true colonization source, the interaction between these variables and the pH(KCl). The species are ranked according to Table 2
Odds ratio in multiple logistic regression
Distance × age
Wald test statistic with 2-sided P-values. ****P ≤ 0.0001, ***P ≤ 0.001, **P ≤ 0.01, *P ≤ 0.05, (*)P ≤ 0.1, NS = not significant.
Among the species that are relatively indifferent to secondary forest age, only Geum urbanum keeps having its affinity for older forest parcels after accounting for distance and pH. Of the others (Aegopodium podagraria and, to a lesser extent, Ranunculus ficaria), the age effect is due to a covariation with pH or a strong interaction between age and distance (Glechoma hederacea and Ajuga reptans, where distance effects are limited to the youngest age classes, Fig. 4). The only species with a negative relationship with age, Adoxa moschatellina, exhibits an overall distance effect.
Although not statistically tested, both Stellaria holostea and Oxalis acetosella show weak age effects, prefer lower pH values and tend to occur close to colonization sources (Fig. 4).
For the species where it is possible to take cover into account (Table 4), results were similar to those in Table 3 for Anemone nemorosa, Deschampsia cespitosa and Lamium galeobdolon and Glechoma hederacea (a much better colonizer). While the frequency of Ranunculus ficaria is not distance dependent, its abundance clearly is, with the cover decreasing with increasing distance, particularly in young woodland. Unexpectedly, the cover of Aegopodium podagraria decreases significantly in the oldest age class.
Table 4. Kendall partial rank-order correlation coefficients between plant species cover and distance from the colonization sources with the effect of pH(KCl) held constant. Correlation coefficients are calculated separately for the five age classes. The species are ranked according to Table 2
Secondary woodland age
< 10 years
2-sided P-values. ****P ≤ 0.0001, ***P ≤ 0.001, **P ≤ 0.01, *P ≤ 0.05, (*)P ≤ 0.1, NS = not significant, / = species not present.
The mapping of the plant species and the identification of species-specific colonization sources
The use of complete vegetation maps in combination with detailed historical land use data has some important advantages compared with the use of individual plots scattered throughout the forest. When individual plots are used and there is no information about the overall distribution of the species throughout the forest, one has to assume that colonization started at the ancient-recent forest ecotone (Matlack 1994; Brunet & von Oheimb 1998; Bossuyt et al. 1999b; Singleton et al. 2001). Furthermore, it is impossible to account for relict populations found, for instance, along ditches.
We do assume, however, that the studied species persisted along ancient forest margins during the period of adjacent agricultural land use. This seems justified since (i) many species are still associated with such margins and (ii) other colonization sources either in the ancient forest parcels or in the vicinity of the Muizen forest may be absent (Listera ovata, Brachypodium sylvaticum) or rare (Arum maculatum, Ornithogalum umbellatum, Paris quadrifolia). Furthermore, it has already been demonstrated that hedgerows, and especially those that are remnants of former forest, are important for harbouring relict populations of forest plants (Pollard 1973; Rackham 1980; Peterken & Game 1981; Burel & Baudry 1990; Corbit et al. 1999). A further assumption is that the species distributions in ancient forest parcels and along ancient forest margins have not changed dramatically over the past 200 years.
Whilst forest plant species are typically long-lived (Bierzychudek 1982; Falinski 1986), individual ramets will certainly not live that long. However, long-term population studies of forest plants (Tamm 1956) demonstrate their long half-life (see also Harper 1977) and field experience and comparisons with old publications (Peterken 1981) also suggests that the distribution of forest species remains fairly stable. Plants are therefore likely to have been continuously present in a particular 20 × 20 m grid cell. Finally, it is assumed that the studied species did not survive in the parcels during the period of agricultural use. Although it has been found that some forest species can survive periods of extensive grassland use (Rackham 1980; Peterken 1981), it is less likely under arable use. All but one of our parcels had a period of arable use and most of the arable to grassland conversion took place between 1930 and 1950 (see Fig. 2), when the introduction of intensive farming practices (e.g. the application of chemical fertilizers) would have impeded colonization of grassland by forest species.
The species affinity with older forest
Our results correlate very well with those of Bossuyt et al. (1999b) from the Meerdaal forest complex (Leuven, Belgium). Species common to the two studies all show the same response to secondary forest age (Polygonatum multiflorum, Lamium galeobdolon and Anemone nemorasa confined to ancient forest, Ranunculus ficaria, Adoxa moschatellina, Ajuga reptans, Geum urbanum, Glechoma hederacea and Oxalis acetosella were more widely distributed). In contrast, considerable differences are found between our age-sensitive species and European ancient forest species listed by Hermy et al. (1999). Rothera & Davy (1986), however, found that diploid populations of Deschampsia cespitosa were largely confined to ancient woodland, while tetraploid populations, which occurred in a wide range of habitats, were reported to be good colonizers. Since polyploidy is a very common phenomenon, especially in perennial plants (Grant 1981), this might be one of the causes for the regional differences in response of species to secondary forest age. Another reason for divergence from the European list is that there is a gradual transition from species having a very high affinity for older forests towards species exhibiting little or no preference at all (Tables 2, 3, 4 and Fig. 4), rather than the clear threshold used to define an ancient forest plant.
The relative importance of dispersal limitation
The observed decline in frequency and abundance of species with increasing distance from the colonization source can be attributed to dispersal limitation since (i) it is generally accepted that diaspore number declines with increasing distance from the diaspore source (Willson 1993) and (ii) studies that examined both seed shadows and recruitment patterns demonstrated a coincidence between the seed shadow and the spatial pattern of recruitment (Debussche & Lepart 1992).
Our results therefore indicate that many species, even if they are not restricted to older forest and thus are better able to colonize, are to a certain extent dispersal-limited. However, distance-effects for species such as Ranunculus ficaria and Glechoma hederacea are only present in the youngest forest parcels and dispersal limitation is therefore unimportant at least at a local scale. Many species exhibiting an affinity for older forest are, however, severely dispersal limited. Anemone nemorosa, Lamium galeobdolon and Polygonatum multiflorum are dispersal limited in both southern Sweden (Brunet & von Oheimb 1998) and central Belgium (Bossuyt et al. 1999b), whereas Adoxa moschatellina and Oxalis acetosella showed this pattern only in Sweden, together with a further four species (Stellaria holostea, Listera ovata, Brachypodium sylvaticum and Paris quadrifolia) not present in the Belgian study.
Even after taking dispersal limitation into account, some species still exhibit a significant affinity with older forest. As many habitat quality factors covary with secondary forest age, it is likely that recruitment limitation also impedes the colonization of some forest species. Species can be allocated to the following four ecological groups according to the relative importance of the two factors.
1 Species for which both dispersal and recruitment limitation is important include Primula elatior, Arum maculatum and Lamium galeobdolon where distance to colonization source cannot fully account for association with forest age. Dispersal limitation is rather unexpected for the endozoochorous Arum maculatum.
2 Species for which dispersal limitation is predominant (age becomes redundant when distance is included in the model) include Anemone nemorosa and Deschampsia cespitosa, which both lack adaptations for long-distance dispersal. Both have a relatively wide ecological amplitude: Anemone nemorosa grows on a wide range of soil types (Shirreffs 1985) and is tolerant for litter (Eriksson 1995; Holderegger 1996; but see Hermy 1985), as does Deschampsia cespitosa, but it is more dominant on wet soils (Davy 1980). Recruitment limitation is therefore not expected.
3 Species for which recruitment limitation is predominant, such as Paris quadrifolia and Polygonatum multiflorum, exhibit no distance effect, but a significant age effect. Although Paris quadrifolia is known to have specific habitat requirements (Rackham 1980; Hermy 1985), the reasons for the recruitment limitation in Polygonatum multiflorum are less clear since this species has a wide ecological amplitude, comparable with that of Anemone nemorosa (Hermy 1985). The absence of dispersal limitation may be due to their endozoochorous dispersal mode.
4 Species for which neither dispersal limitation nor recruitment limitation is of major importance. In our study area, Geum urbanum, Ranunculus ficaria, Glechoma hederacea, Ajuga reptans, Aegopodium podagraria, Adoxa moschatellina and probably also Oxalis acetosella are not confined to older forests and exhibit little or no distance-effect. Surprisingly, none except Geum urbanum have adaptations for long distance dispersal, but all have a wide ecological amplitude (for Geum urbanum see Taylor 1997; for Oxalis acetosella see Packham 1978; and for Ranunculus ficaria see Taylor & Markham 1978).
Such groupings must nevertheless be regarded as a rather arbitrary break-up of a continuum, as all species are to some extent dispersal and recruitment limited although the degree varies between species (Eriksson & Ehrlén 1992).
Listera ovata, Brachypodium sylvaticum, Ornithogalum umbellatum could not be classified since the age effect was only due to the covariation between pH and forest age. Stellaria holostea is mainly restricted to the parcel margins (81% of occurrences) and its distribution, as well as the significant correlations between the cover of Aegopodium podagraria and distance, can be explained by high light fluxes along forest tracks.
The results presented here may not be easily translated to long-distance dispersal events. Cain et al. (1998) suggested that different dispersal processes might operate at larger spatial and longer temporal scales. Hence, distance to the diaspore source may then predict patterns of colonization poorly (but see Jacquemyn et al. 2001 and Butaye et al. 2001). However, long distance dispersal processes are still poorly understood due to methodological constraints. Therefore, collection of data at larger spatial and longer temporal scales and linking them to the underlying mechanisms is a challenge for future research and will facilitate generalization to unstudied sites and species (Nathan & Muller-Landau 2000).
The first author thanks his father, brother, Dries Adriaens and Wouter Van Muyzen for their assistance with the field work and Bea Bossuyt, Olivier Honnay, George Peterken, Anthony Davy, Lindsay Haddon and an anonymous referee for their useful comments on the manuscript. The paper was written while the first author held a grant from the Flemish Institute for the encouragement of Scientific and Technological Research (I.W.T.).
Received 14 September 2000 revision accepted 4 March 2001